Advances in Group Processes 9781784410773, 9781784410780

Citation preview

ADVANCES IN GROUP PROCESSES

ADVANCES IN GROUP PROCESSES Series Editors: Edward J. Lawler and Shane R. Thye Recent Volumes: Volumes 1
17: Edited by Edward J. Lawler Volume 18:

Edited by Edward J. Lawler and Shane R. Thye

Volume 19:

Group Cohesion, Trust and Solidarity
Edited by Edward J. Lawler and Shane R. Thye

Volume 20:

Power and Status
Edited by Shane R. Thye and John Skvoretz

Volume 21:

Theory and Research on Human Emotions
Edited by Jonathan H. Turner

Volume 22:

Social Identification in Groups
Edited by Shane R. Thye and Edward J. Lawler

Volume 23:

Social Psychology of the Workplace
Edited by Shane R. Thye and Edward J. Lawler

Volume 24:

Social Psychology of Gender
Edited by Shelley J. Correll

Volume 25:

Justice
Edited by Karen A. Hegtvedt and Jody Clay-Warner

Volume 26:

Altruism and Prosocial Behavior in Groups
Edited by Shane R. Thye and Edward J. Lawler

Volume 27:

Edited by Shane R. Thye and Edward J. Lawler

Volume 28:

Edited by Shane R. Thye and Edward J. Lawler

Volume 29:

Edited by Will Kalkhoff, Shane R. Thye and Edward J. Lawler

Volume 30:

Thirtieth Anniversary Edition
Edited by Shane R. Thye and Edward J. Lawler

ADVANCES IN GROUP PROCESSES VOLUME 31

ADVANCES IN GROUP PROCESSES EDITED BY

SHANE R. THYE Department of Sociology, University of South Carolina, SC, USA

EDWARD J. LAWLER School of Industrial and Labor Relations, and Department of Sociology, Cornell University, NY, USA

United Kingdom
North America
Japan India
Malaysia
China

Emerald Group Publishing Limited Howard House, Wagon Lane, Bingley BD16 1WA, UK First edition 2014 Copyright r 2014 Emerald Group Publishing Limited Reprints and permission service Contact: [email protected] No part of this book may be reproduced, stored in a retrieval system, transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without either the prior written permission of the publisher or a licence permitting restricted copying issued in the UK by The Copyright Licensing Agency and in the USA by The Copyright Clearance Center. Any opinions expressed in the chapters are those of the authors. Whilst Emerald makes every effort to ensure the quality and accuracy of its content, Emerald makes no representation implied or otherwise, as to the chapters’ suitability and application and disclaims any warranties, express or implied, to their use. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: 978-1-78441-078-0 ISSN: 0882-6145 (Series)

ISOQAR certified Management System, awarded to Emerald for adherence to Environmental standard ISO 14001:2004. Certificate Number 1985 ISO 14001

CONTENTS LIST OF CONTRIBUTORS

vii

PREFACE

xi

TWENTY-FIVE YEARS OF THE GROUP PROCESSES CONFERENCE: A REVIEW ESSAY Morris Zelditch Jr.

1

EXPECTATION STATES THEORY: GROWTH, OPPORTUNITIES AND CHALLENGES Joseph Berger, David G. Wagner and Murray Webster Jr.

19

THE DEVELOPMENT OF IDENTITY THEORY Jan E. Stets and Peter J. Burke

57

RELATIONAL COHESION, SOCIAL COMMITMENTS, AND PERSON-TO-GROUP TIES: TWENTY-FIVE YEARS OF A THEORETICAL RESEARCH PROGRAM Shane R. Thye, Aaron Vincent, Edward J. Lawler and Jeongkoo Yoon BACK TO THE FUTURE: 25 YEARS OF RESEARCH IN AFFECT CONTROL THEORY Neil J. MacKinnon and Dawn T. Robinson ELEMENTARY THEORY: 25 YEARS OF EXPANDING SCOPE AND INCREASING PRECISION David Willer, Pamela Emanuelson, Michael J. Lovaglia, Brent Simpson, Shane R. Thye, Henry Walker, Mamadi Corra, Steven Gilham, Danielle Lewis, Travis Patton, Yamilette Chacon and Richard Chacon v

99

139

175

vi

CONTENTS

PERCEPTIONS OF ABILITY AND ADHERENCE TO RULES, GUIDELINES, AND TRADITION Jeffrey W. Lucas, Wesley S. Huey, Marek N. Posard and Michael J. Lovaglia REFERENT NETWORKS AND DISTRIBUTIVE JUSTICE David Melamed, Hyomin Park, Jingwen Zhong and Yue Liu BEYOND NETWORKS IN STRUCTURAL THEORIES OF EXCHANGE: PROMISES FROM COMPUTATIONAL SOCIAL SCIENCE James A. Kitts

219

241

263

LIST OF CONTRIBUTORS Joseph Berger

Department of Sociology, Stanford University, CA, USA

Peter J. Burke

Department of Sociology, University of California, Riverside, CA, USA

Richard Chacon

Department of Sociology and Anthropology, Winthrop University, SC, USA

Yamilette Chacon

Department of Sociology and Anthropology, Winthrop University, SC, USA

Mamadi Corra

Department of Sociology, East Carolina University, NC, USA

Pamela Emanuelson

Department of Sociology and Anthropology, North Dakota State University, ND, USA

Steven Gilham

Independent Scholar

Wesley S. Huey

Division of Leadership, Education and Development, United States Naval Academy, MD, USA

James A. Kitts

University of Massachusetts, Amherst, Department of Sociology, MA, USA

Edward J. Lawler

School of Industrial and Labor Relations and Department of Sociology, Cornell University, NY, USA

Danielle Lewis

Department of Sociology, University of South Carolina, SC, USA

Yue Liu

Department of Sociology, University of South Carolina, SC, USA

vii

viii

LIST OF CONTRIBUTORS

Michael J. Lovaglia

Department of Sociology, University of Iowa, IA, USA

Jeffrey W. Lucas

Department of Sociology, University of Maryland, MD, USA

Neil J. MacKinnon

Department of Sociology and Anthropology, University of Guelph, ON, Canada

David Melamed

Department of Sociology, University of South Carolina, SC, USA

Hyomin Park

Department of Sociology, University of South Carolina, SC, USA

Travis Patton

Department of Sociology, Morehouse College, GA, USA

Marek N. Posard

Department of Sociology, University of Maryland, MD, USA

Dawn T. Robinson

Department of Sociology, University of Georgia, GA, USA

Brent Simpson

Department of Sociology, University of South Carolina, SC, USA

Jan E. Stets

Department of Sociology, University of California, Riverside, CA, USA

Shane R. Thye

Department of Sociology, University of South Carolina, SC, USA

Aaron Vincent

Department of Sociology, University of South Carolina, SC, USA

David G. Wagner

Department of Sociology, University at Albany, SUNY, NY, USA

Henry Walker

Department of Sociology, University of Arizona, AZ, USA

Murray Webster Jr.

Department of Sociology, University of North Carolina, Charlotte, NC, USA

David Willer

Department of Sociology, University of South Carolina, SC, USA

ix

List of Contributors

Jeongkoo Yoon

School of Business Administration, Ewha Womans University, South Korea

Morris Zelditch Jr.

Department of Sociology, Stanford University, Stanford, CA, USA

Jingwen Zhong

Department of Sociology, University of South Carolina, SC, USA

PREFACE Advances in Group Processes publishes theoretical analyses, reviews, and theory based 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, trust, justice, social influence, identity, decision-making, intergroup relations, and social networks. Previous contributors have included scholars from diverse fields including sociology, psychology, political science, economics, business, philosophy, computer science, mathematics, and organizational behavior. Volume 31 represents somewhat of a first for the series. In August 2013, group processes scholars from around the world gathered in New York City to celebrate the 25th anniversary of the Annual Conference for Theory and Research on Group Processes. That conference was first held at Emory University in 1988, and part of the goal of the 2013 event was to celebrate several of the mainstream theoretical research programs in the group processes arena. This volume brings together a subset of the papers that were presented at that conference. The volume begins with “Twenty-Five Years of the Group Processes Conference: A Review Essay” by Morris Zelditch Jr. This paper reviews and analyzes the impact that the group processes conference (and to a lesser extent the Advances in Group Processes series) has had on small groups research. “Buzz” has been an important part of the formalization and growth of the group processes domain and a mainstay at the Annual Group Processes Conference. We are extremely pleased to have his insights to open the volume. The next five papers represent what are likely the most complete and comprehensive reviews of some of the most influential theoretical research programs in the group processes domain. First, in “Expectation States Theory: Growth, Opportunities and Challenges,” Joseph Berger, David G. Wagner, and Murray Webster Jr. review the Expectation States theoretical research program. After identifying the contemporary branches of the program and specifying the important concepts, the authors explicate the main theoretical branches prior to 1988 and those that have emerged xi

xii

PREFACE

since then. They close by identifying a number of experimental and theoretical challenges that lie ahead. Next, Jan E. Stets and Peter J. Burke outline the current state of the affairs in “The Development of Identity Theory.” The authors trace the development of identity theory from 1988 to the present day, detailing the development of the perceptual control system, the role of resources and symbols, and the fundamental bases of identity stability and change. The authors also show how identity theory has been applied in the areas of crime and law, education, race and ethnicity, gender, family, and the environment. The chapter closes by examining future areas of research and development. Shane R. Thye, Aaron Vincent, Edward J. Lawler, and Jeongkoo Yoon review the past two and a half decades of research on relational and group ties in “Relational Cohesion, Social Commitments and Person-to-Group Ties: Twenty-Five Years of a Theoretical Research Program.” Beginning with research on commitment that emerged in the early 1990s, the authors describe and explain the evolution of three inter-related theories about the role of emotions in social exchange: the theory of relational cohesion, the affect theory of social exchange, and the theory of social commitments. Empirical tests and applications are discussed and new directions are identified. The following chapter entitled “Back to the Future: 25 Years of Research in Affect Control Theory” by Neil J. MacKinnon and Dawn T. Robinson track the development of affect control theory from the late 1960s to today. They not only explore the questions, theoretical advances, and empirical base for the theory, they also examine how changes in technology have allowed the theory to answer new questions and forge into new domains. The set of review chapters closes with “Elementary Theory: 25 Years of Expanding Scope and Increasing Precision” (by David Willer, Pamela Emanuelson, Michael J. Lovaglia, Brent Simpson, Shane R. Thye, Henry Walker, Mamadi Corra, Steven Gilham, Danielle Lewis, Travis Patton, Yamilette Chacon, and Richard Chacon). This contribution explains how Elementary Theory works (i.e., how it simplifies the world into basic elements and builds more complex models from there) and how the theory has experienced both increased scope and precision over its life course. Covered are the many ways that elementary theory has guided prediction and explanation in both laboratory and field applications. The final three papers are not reviews in the same sense as the preceding papers but offer new insights into other well established realms of the group processes tradition. First, in “Perceptions of Ability and Adherence to Rules, Guidelines, and Tradition,” Jeffrey W. Lucas, Wesley S. Huey,

Preface

xiii

Marek N. Posard, and Michael J. Lovaglia offer a new theory linking status processes in groups and the proclivity to follow rules. Testing the theory, a new experiment shows that individuals who perceive they have low abilities are stricter in following the rules relative to those believed to have high ability. The authors discuss both the theoretical and practical implications of the work. The final two papers explore issues relating to social networks. In “Referent Networks and Distributive Justice,” David Melamed, Hyomin Park, Jingwen Zhong, and Yue Liu theorize how simply knowing others’ reward levels (i.e., knowledge of some referent network) can impact perceptions of justice. Whereas prior work in the area tends to examine referent individuals they recast the problem in terms of referent networks, and thus move to a more complex referent unit. A series of theoretically derived hypotheses are tested. The final paper in the volume is by James A. Kitts entitled “Beyond Networks in Structural Theories of Exchange: Promises from Computational Social Science.” Social networks are typically viewed as a set of role relations, interpersonal sentiments, social interactions, or opportunities to exchange. Kitts examines the interplay across these four dimensions and argues that with emerging technologies comes the opportunity to more carefully analyze dynamic structural interdependencies and advance theories of exchange. Overall, this paper nicely integrates several conceptions of social networks and suggests new avenues for the development of theory and research. Shane R. Thye Edward J. Lawler Series and Volume Co-Editors

TWENTY-FIVE YEARS OF THE GROUP PROCESSES CONFERENCE: A REVIEW ESSAY Morris Zelditch Jr. ABSTRACT Purpose
The primary purpose of this chapter is to assess the effects of twenty-five years of the Group Processes Conference on advances in the study of group processes that have taken place between 1988 and 2014. Design/methodology/approach
This chapter places the twenty-five years of the Group Processes Conference in the context of the changes that have taken place between small groups research in the 1950s and group processes research in the 1980s and beyond. Findings
Between the 1950s and 1980s small groups research reinvented, reconceptualized, and reinvigorated itself as group processes research. In this period, small groups research, its applied research, and its research programs became increasingly theory-driven, and its concept of the group and its levels increasingly abstract, general, and analytic. As a consequence of these changes, the concept of the field itself became increasingly analytic. The Group Processes Conference was at once a reflection of these changes and a driving force in the subsequent advances in group processes research. It both quickened and amplified the effects

Advances in Group Processes, Volume 31, 1
17 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0882-6145/doi:10.1108/S0882-614520140000031001

1

2

MORRIS ZELDITCH JR.

of individual-level factors and of thirty years of Advances in Group Processes on the transformation of the field and was also, like Advances in Group Processes, a driving force in the subsequent advances in group processes research. The present chapter concludes with an analysis of the mechanisms of the effects of the Group Processes Conference on group processes research. Originality/value
The program for the twenty-fifth year of the Group Processes Conference celebrates its effects on the field of group processes research. Keywords: Small groups; group processes; programs of theoretical research; theory and research; theory and application; experimental methods

INTRODUCTION The year 2013 was a year of two milestones: the thirtieth year of Advances in Group Processes (Thye & Lawler, 2013) and the twenty-fifth year of the Group Processes Conference. Both reflected changes that had taken place between the 1950s and the 1980s in the study of small groups. “Thirty Years of Advances in Group Processes: A Review Essay” (Zelditch, 2013) was a comment on the part played by the first thirty years of Advances in Group Processes in the growth of group processes research. The present essay is a comment on the part played by the first twenty-five years of the Group Processes Conference. With apologies to those who have read the review essay on Advances in Group Processes, part of the present essay recapitulates for those who have not read it some of what it had to say about the changes that took place between the two periods, starting with the fact that the 1950s were a period of rapid, prolific growth of small groups research and the 1980s were a period of equally rapid, prolific growth of group processes research; but between the two periods, while group processes research was waxing, small groups research appeared to be waning. In fact, between the two periods many were asking “whatever happened to small groups research?” (e.g., Hare, 1976, appendix 2; Steiner, 1974, 1983, 1986).1 The founding of the first Group Processes Conference was in itself a strikingly apt metaphor for the years between the two periods. The 1988 annual meeting of the American Sociological Association had simply

Twenty-Five Years of the Group Processes Conference

3

forgotten small groups research! Many, in fact, thought the field had declined, some even thought that it had disappeared (Mullins, 1973). On the other hand, some thought its disappearance was illusory, and that it was in fact much the same as it had been in the 1950s but had simply fragmented into compartmentalized specialty areas (Davis & Stasson, 1988). Its disappearance was in fact illusory, but speciation was only one side of the story. The other side of the story was that between 1950 and 1980 small groups research had climbed a substantial number of rungs up the ladder of abstraction: Its research, applications, and programs became increasingly theory-driven. More theory led to a more abstract, more general, more analytic conception of the group and its levels, which transformed the way the field itself was conceptualized from small groups research into group processes research. What happened to small groups research in the 1960s and 1970s was less a disappearance of the field than its reconceptualization, its reinvention, and, with it, its reinvigoration. The change in the way the field was conceptualized led to more theory, more growth of theory, and a greater impact of theory on applications and interventions. The first volume of Advances in Group Processes and the first conference on group processes were at once reflections of the way the field had been transformed, crucial factors in its institutionalization, and, in consequence, driving forces in the advances that have since been made in the field. The entire story may at first sight appear unlikely. The 1950s were awash in theory, particularly field theory, interactionism, and functionalism, and the period’s signal accomplishment was the explosive growth of laboratory observation and experiment, supposedly the most analytic methods known to science. What, then, made the 1980s any different?

SMALL GROUPS RESEARCH Not all theory is testable and not all laboratories test, refine, or extend it. With some exceptions (e.g., Festinger, 1950), the 1950s were more awash in approaches to theory, such as field theory, symbolic interactionism, or functionalism, than in empirically grounded theories. The period was, in fact, mostly notable for the gap between its theory and its empirical research. The explosive growth of its laboratory observation and experiment, again with some exceptions (e.g., Back, 1951), was more oriented to demonstrating effects than testing, refining, or extending theory.

4

MORRIS ZELDITCH JR.

For example, such classic experiments as the Asch experiment (1951) demonstrated the existence of a conformity effect, but were not much concerned with the process or processes that gave rise to it. Such classic laboratory observations as Bales’ interaction process analysis (Bales, 1950) demonstrated the emergence of interaction hierarchies (Bales, Strodtbeck, Mills, & Roseborough, 1951) and role differentiation (Bales & Slater, 1955), but interaction process analysis was induced from observation, not deduced from theory. The applied research of the period was as effect-oriented as its basic research. Up to and into the 1950s, “applied” research simply meant research oriented to social problems. With some exceptions (e.g., Williams, 1947), it did not in fact apply any theory. The period was as notable for the gap between its theory and its applied research as it was for the gap between its theory and its empirical research. Its “applications” were mostly ad hoc, its interventions mostly trial and error. They were grounded either in no theory at all (e.g., the relay assembly test room of Roethlisberger & Dickson, 1939) or only after the fact (e.g., the relative deprivation hypothesis of Stouffer, Suchman, DeVinney, Star, & Williams, 1949). That both its pure and applied research were largely effect-oriented meant that, again with some exceptions (e.g., Berger, 1958; Cohen, 1958),2 the dominant research programs of the period were also largely effectoriented. Some were simply replications of an effect to demonstrate its regularity (e.g., replication of Bales’ interaction hierarchy by Strodtbeck, James, & Hawkins, 1957, and role differentiation by Strodtbeck, 1951; Strodtbeck & Mann, 1954).3 But no attempt was made to explain them. Others were attempts to explain the effect by elaborating its initial conditions (as distinct from its mechanisms). But the effects of the period, although often fundamental, were also often complex and underanalyzed. They were complex in the sense that often two or more processes were involved in them, and underanalyzed in the sense that their conceptualization often did not distinguish one process from another. Thus, Deutsch and Gerard (1955) showed that the conformity effect demonstrated by Asch’s experiment (1951) confounded normative with informational social influence. Burke (1967, 1968) showed that the role differentiation effect observed by Bales (Bales & Slater, 1955) confounded the emergence of interaction hierarchies with the legitimation of power. One consequence was that programs focused on elaborating the initial conditions of an effect tended to multiply ad hoc, disparate, unrelated factors ad infinitum: for example, in the case of the Asch experiment, stimulus ambiguity, the size of the majority, the subject’s status, power dependence,

Twenty-Five Years of the Group Processes Conference

5

and personality all had effects on conformity (Allen, 1965, 1977; Hare, 1976, p. 394). In the case of Bales’ interaction hierarchies, group size, group consensus, the seating position of the subject, another host of personality characteristics, and the personality composition of the group all had effects on the emergence of inequality (Hare, 1976, p. 393; 1989). In consequence, the growth of effect research programs tended to be voluminous but incoherent. The concept of a group and its levels was as complex and underanalyzed in the 1950s as its effects. Any aggregate, whether taken individually or collectively, was a group. A collective actor of any sort was of course a group, but even simple togetherness, any interaction, relation, or network was also a group. The field was in fact held together largely by the looseness with which it used the term “group.” In addition, its levels, typically taken concretely, were often assumed to be ontologically real. Reification made for much heated but largely fruitless controversy, lasting well into the 1980s, over whether to dichotomize them (e.g., Blau, 1977; Mayhew, 1980), reduce them (e.g., Homans, 1958, 1964), or interrelate them (e.g., Berger, Eyre, & Zelditch, 1989). Nevertheless, no matter how loosely it used the term “group,” no matter how underanalyzed its “levels,” the field’s most prominent approaches
again with some exceptions (e.g., cf. Goffman, 1956, 1961, 1967)
assumed, in one way or another, that there was “a” theory of “the” group, either “a” fundamental process of “any” group or an invariant confluence of such processes.

GROUP PROCESSES RESEARCH But the seeds of change that would eventually transform it were already being sown in the 1950s. The 1950s were the apogee of small groups research, but they were also a watershed of change. Empirically grounded theories were already being constructed (e.g., not only by Festinger, 1950, but also by others such as Goffman, 1956), tested (e.g., not only by Back, 1951, but also by others such as Schachter, 1951), and applied (e.g., not only by Williams, 1947, but also by others such as Stouffer et al.’s, 1949, test of Williams). Some explanations of complex effects were already explicating the processes that gave rise to them (e.g., not only Deutsch & Gerard, 1955, but also others such as Cartwright & Harary, 1956). One consequence of explicating the complex effects of the 1950s was that theory-driven programs began to spin off of effect-driven programs, such as Berger’s expectation states theory of Bales’ interaction hierarchies

6

MORRIS ZELDITCH JR.

(Berger, 1958; see also Berger et al., 2014) and Cohen’s conflict model of Asch’s conformity effect (Cohen, 1958). The explication of a complex effect had this consequence because complex effects are, by definition, the effect of two or more processes. That the complex effects of the 1950s were often not only complex but also underanalyzed meant that the causes, conditions, and consequences of the two or more processes underlying them were often conflated, neglecting their differences. For example, Deutsch and Gerard’s (1955) explication of Asch’s conformity effect found that normative social influence depended on the presence of a group and the observability of subject’s responses, but informational social influence did not. Hence, the causes, conditions, and consequences of the two processes were different: Normative social influence was due to social control, occurred only if responses were public, and had a coercive effect. Informational social influence was due to the effect of others as sense evidence, occurred whether public or private, and had a persuasive effect. Similarly, Burke’s (1967, 1968) explication of Bales’ role-differentiation effect found that it depended on the legitimation problems of homogeneous groups, not found in heterogeneous groups. Again, the causes, conditions, and consequences of inequality and its legitimacy differed. The problem with conflating their differences was that what appeared, at least at first sight, to be the same causes, under what appeared to be the same conditions, did not always have the same effects. Driven by their irregularity, explication analyzed such underanalyzed complex effects into their constituent processes, differentiating their causes, conditions, and consequences. Simplifying a complex process by separating it into its more elementary unit processes facilitated isolating them, that is, investigating each of them independently of the effects of any other processes. Thus, the kind of research programs that spun off of the explication of Asch’s conformity effect (Cohen, 1958) and Bales’ interaction hierarchies (Berger, 1958) were programs that tested, refined, and extended a theory of a process. The 1960s and 1970s were a period in which these trends began to gather momentum: There was more theory (e.g., Emerson, 1962), more theorydriven research (e.g., Emerson, 1964), more theory-driven applications (e.g., Cohen, 1972), more deduction of policy from theory (e.g., Cohen & Roper, 1972), and more theory-driven programs (e.g., Emerson, 1972; Heise, 1979; for the latter program, affect control theory, see also MacKinnon & Robinson, 2014). By the 1980s, small groups research looked more like the program of the Group Processes Conference of 2013 (Shelly, Wagner, Welser, & Shelly, 2013) than the small groups research of the 1950s. Its theory was largely

Twenty-Five Years of the Group Processes Conference

7

empirically grounded, and its research, applications, policy, and programs largely theory-driven. Where the rapid, prolific growth of laboratory observation and experiment had been the signature of the 1950s, the rapid, prolific growth of theoretical research programs has been the signature of the 1980s and beyond
programs such as Expectation States Theory (Berger et al., 2014), Elementary Theory (Willer et al., 2014), Affect Control Theory (MacKinnon & Robinson, 2014), Identity Control theory (Stets & Burke, 2014), and the Theory of Relational Cohesion (Thye, Vincent, Lawler, & Yoon, 2014). They grew not only in number but also in the growth of their theory, that is, growth of their confirmation status, precision, rigor, scope (hence their generality), and domain (i.e., the number of effects they explained, hence their explanatory power). Extension of their scope or domain made some of them grow as voluminous as any effect research program. Extension of either scope or domain often led to modification of the basic concepts and principles of a program by auxiliary concepts and principles for its special cases, often resulting in families of interrelated theories rather than a theory of a process. But, unlike effect research programs, although they were often as voluminous, they were coherent, held together by a common core of basic concepts, principles, and methods. More theory-driven research, applications, and programs implied a concept of the group and its levels that, although just as elastic as it had been in the 1950s, was more abstract, more general, and more analytic than the “group” of the 1950s. Although it was also just as interdisciplinary as it had been in the 1950s, a more abstract and more general concept of the group and its levels also led to a more sociological conception of the social psychology of group processes. Social processes were abstracted from differences in individual-level predispositions, focused on more situationally conditioned motives, more situationally conditioned interactor processes, and more mutually contingent relations between the levels of a system. Levels were linked rather than dichotomized or reduced: Differences in size, differences between structure and action, differences between actor and system were seen to be differences within rather than between different theories; differences between individual and group actors were seen to be differences in the interpretations of a theory, not different theories (see, e.g., Lawler, Ridgeway, & Markovsky, 1993). But a more analytic concept of the group and its levels led to the breakdown of the concept of “the” group sui generis, characterized by one or more fundamental processes to be found in “any” of its instances. The concept of a group and its levels in the 1950s was just as complex and

8

MORRIS ZELDITCH JR.

underanalyzed as its effects. For example, Harrington and Fine (2000), attempting to revive it, saw the importance of the concept of “the” group as “an entity in its own right” in the invariant confluence of five fundamental processes, important in any group but most easily observed in small groups. But one of its processes was the creation of culture. Most encounters are products of pre-given culture but few encounters create it. Another process was the formation of networks. Most groups are networks, but not all networks are groups. The more analytic strategy of the 1980s broke such intrinsic, invariant confluences down into their constituent elements: Unit processes, such as the group formation, expectation state, exchange, and identity processes of this year’s Group Processes Conference displaced them (Berger et al., 2014; MacKinnon & Robinson, 2014; Stets & Burke, 2014; Thye et al., 2014, Willer et al., 2014, all in this volume). But, in the process, a more analytic approach to the group and its levels did in fact lead to the disappearance of “the” group. That “the” group disappeared did not mean that small groups research did. But the increase in its level of abstraction and the decrease in the complexity of its processes did fundamentally transform it. A more abstract, more general, and more analytic approach to it reconceptualized the field, reinventing it and transforming it from small groups research into group processes research. It cannot be said that the small groups research of the 1950s was forgotten in the process. Its effects, regularities, and findings were not so much forgotten as subsumed, reconceptualized as the foundations on the shoulders of which group processes research was built. In the process, reinvention of the field revitalized and reinvigorated it.

THE GROUP PROCESSES CONFERENCE Like Advances in Group Processes, the Group Processes Conference was at once a reflection of the changes that took place in small groups research between the 1950s and 1980s, an institutionalization of them, and a driving force in both the transformation of the field and the advances in group processes research that have taken place since. But unlike Advances in Group Processes, it was more decentered, and its effects less mandated. The mechanisms that drove its effects were therefore a good deal different. Any academic conference systematically organizes the exchange of information among a community of scholars engaged in the same sort of enterprise. It is an institution that formalizes, routinizes, and often, like the Group Processes Conference, annualizes the exchange of information

Twenty-Five Years of the Group Processes Conference

9

between them. Its participants are individuals, but they span the organizations, departments, centers, and programs that make up a field. But conferences are not only platforms enabling intergroup interaction and communication, their practices and procedures are themselves a means of communication. Their groups are exchanging information, the channels of communication that are the essential structure of a conference matter to who exchanges what and with whom. But their name, mission statement, and program are themselves a media of communication between their organizers and their participants. Thus, Group Processes Conferences have reflected the changes that took place between the 1950s and 1980s in two ways. Their name, in itself, reflected the substantial climb up the ladder of abstraction and the more abstract, more general, and more analytic approach to the group and its levels that had taken place between the 1950s and 1980s. Both their mission and the programs of the conferences embodied its generalizing aims and its rigorously generalizing but also rigorously empirical strategy of achieving them. Their mission was “studying group phenomena employing rigorous, scientific principles of theory, data collection, and analysis” (e.g., Shelly et al., 2013). The programs of the conferences embodied their mission in empirically testable theories, theory-driven research, theory-driven applications, and theory-driven research programs (e.g., ibid.). But their interaction processes also reflected the changes that had taken place in the field. Enactment of the programs of the conferences exemplified them, and the informal social interaction that is a by-product of a conference’s program enabled not only all the gossip and networking that make up a meeting but also, because the innovators of the changes in the field and many of those who had adopted them were among the participants, both enacted and debated them. By its fifth conference, what the Group Processes Conference was reflecting had in fact become a “tradition” (Foschi & Lawler, 1994).

THE GROUP PROCESSES CONFERENCE AND GROUP PROCESSES RESEARCH The Group Processes Conference had important consequences for the formation of the field. Group processes research was already flourishing by the 1980s. But up to that point, the processes transforming small groups research into group processes research were, Advances in Group Processes aside, largely individual-level effects: partly the effect of competition among

10

MORRIS ZELDITCH JR.

research programs for resources, hence mimesis of the efficiency and effectiveness of the innovations in its aims, strategies, and methods that had been made between the 1950s and 1980s; and partly, in time, mimesis simply of their increasing legitimacy as they became increasingly prevalent (DiMaggio & Powell, 1983). But the founding of the Group Processes Conference had an effect at the collective level that both quickened and amplified the effects of Advances in Group Processes and these individuallevel effects on the formation of the field. This effect was due partly to its increased diffusion of the innovations in aims, strategies, standards, and forms of organization that were transforming the field between the 1950s and 1980s; partly to the increased effect of its diffusion of them on those of its participants who had not, up to that point, adopted them; and partly to the effect of its increased diffusion of them on the formation of intergroup relations. Partly, its effect on the adoption of innovations was due simply to increasing exposure to them. Adoption itself was often driven by much the same factors that had driven adoption of them before its founding. A Group Processes Conference’s formal and informal channels of communication exposed its participants to the innovations taking place in the field in four ways: One was simply that repeated reflection of them in its name, mission, and program implied that they were, in fact, a tradition; second, both its formal and informal channels of communication enabled the exchange of information about them between participants; third, both also enabled enactment of them by those of its participants who had innovated or adopted them; finally, both enabled those who had not adopted them to observe responses by other participants to enactment of them. Some of those newly exposed to them adopted them, as many had before the founding of the Group Processes Conference, simply because they were efficient and effective or, as they became increasingly prevalent, simply because that was how things were done. But to increase exposure to them the Group Processes Conference also added more reasons for adopting them. Tradition, unchallenged enactment, and observation of it, all of which implied their validity, added more ways of legitimating them. More ways of legitimating them had a further effect of considerable importance: The innovations that had taken place between the 1950s and 1980s, unlike those that had taken place in the 1950s, tended to persist once adopted. They were not eclipsed by any of the further innovations in the field. Finally, if paradoxically, while the innovations that had taken place in the field between the 1950s and 1980s were by the 1980s widely understood to be efficient and effective, much of the exchange of information between

Twenty-Five Years of the Group Processes Conference

11

groups in any given conference was about innovations in ongoing theory, ongoing research, ongoing applications, and programs. Borrowing Kollock’s analogy, their quality was more like rubber than rice (Kollock, 1994). But uncertainty increases commitment to the partners exchanging them (ibid.). Embedding (Granovetter, 1985), in turn, increased intergroup relations not only within but also between conferences. Thus, founding the Group Processes Conference both quickened and amplified the effects of the factors transforming the field, affecting all of the characteristics that define an organizational field (DiMaggio & Powell, 1983). It increased the density of its interpersonal and intergroup interaction and relations, between as well as within its annual meetings. It transformed both the cognitive base of the field and the common enterprise it shared. It incorporated small groups research into group processes research in the process of redefining its cognitive base and redefined the common enterprise it shared as theory construction. By the mid-1990s, its programs had achieved considerable isomorphism, and its aims, strategy, and standards considerable consensus. Its research programs had settled on an agreed form, theoretical research programs (such as the five of this volume) displacing the effect-driven programs of the 1950s. Despite the considerable heterogeneity of its orienting strategies, a working consensus had evolved at the level of the actual practice of its aims, strategies, and standards. It remained a heterogeneous mix of orienting strategies
field theory, behavioral exchange theory, rational choice theory, the theory of action, functionalism, symbolic interactionism, ethnomethodology, plus hybrid mixes of any two, and sometimes more, of them. But it was at peace, not war, over whether you can really study an army in the laboratory (e.g., Lucas, Huey, Posard, & Lovaglia, 2014; Melamed, Park, Zhong, & Liu, 2014, both in this volume) or really link micro- with macrosociology (e.g., Kitts, 2014). Its aims, strategy, and standards were at once more rigorously generalizing and more rigorously empirical than in the 1950s, the group and its levels more abstract, more general, and more analytic, and its social psychology more sociological.

THE GROUP PROCESSES CONFERENCE AND ADVANCES IN GROUP PROCESSES RESEARCH The Group Processes Conference was a driving force not only in the transformation of the field but also in the subsequent advances in group processes research. By the 1980s, advances in group processes research were

12

MORRIS ZELDITCH JR.

also already flourishing, in part, again, because of the individual-level effects of innovations in the field’s strategy and programs: Its generalizing strategy was more fecund; assuming a fruitful theory, its theoretical research programs had a greater capacity for theoretical growth than effect programs (Zelditch, 2007). From the mid-1980s on, they also flourished because of the editorial policies of Advances in Group Processes and, by the end of the 1980s, there was the added effect of a more connected and connecting organizational field. But the collective effects of the Group Processes Conference were nevertheless themselves a driving factor in the subsequent advances in group processes research. For one thing, it had a considerable effect on the effectiveness of intergroup communication in the field. The Group Processes Conference intensified the intergroup exchange of information among its participants. The emerging similarity of its aims, strategies, standards, and programs enhanced the comprehensibility of its enhanced intergroup communication. Its comprehensibility encouraged participants to believe others understood what they were doing and why they were doing it. It encouraged more feedback, more of it constructive, from more people. Expectations of effective communication, in turn, had a recursive effect, further increasing the level of intergroup communication. Similarities in aims, strategies, standards, and programs enhanced not only the comprehensibility but also the acceptability of what one was doing and why one was doing it. Within any particular research program, of course, one might expect others to support what one was doing. But acceptability spanning the local communities brought together by a Group Processes Conference implied a much broader legitimation of, hence support for, theory, theory-driven research, theory-driven applications, and theory-driven programs. All of which, in turn, led to more theory, more theory-driven research, more theory-driven applications, and more theory-driven programs, which further implied more theory growth. That more intergroup exchange of information exposed the research groups of the field to more, and more diverse, points of view was also a source of more, and more innovative, theory. Research programs, of course, are typically in competition with each other for resources. But awareness of a common enterprise that unites them motivates a kind of partnership of rivals, a partnership that increases cooperation among competitors that is among the most important contributions of any well-run academic conference. One effect of the partnership of rivals is awareness of, a shared understanding of, and a common concern for the fundamental problems facing the field, the puzzles that have to be solved for the field as

Twenty-Five Years of the Group Processes Conference

13

a whole to advance, such as how ties form (e.g., Thye et al., 2014), how structures emerge (e.g., Berger et al., 2014; Willer et al., 2014), and how they enable and constrain behavior (MacKinnon & Robinson, 2014; Stets & Burke, 2014). The sense of a common body of problems to be solved, mobilizing the competition of the groups in the field to address them, is in itself a fundamental factor in theory growth in any field. But the other side of the intergroup exchange of information enhanced by a conference is exposure to more diverse points of view. Exposure to more diverse points of view increased the rate of innovation in the theory of the field (e.g., Lucas et al., 2014; Melamed et al., 2014), and enhanced intergroup exchange of information increased the rate of its diffusion. Exposure to more diverse points of view also enhanced the success of its applications and the power of the field’s interventions. An analytic approach to complexity simplifies complex processes. But application of a theory often applies a theory of a unit process to a complex process. The other side of an analytic approach to complex processes is synthesis. The interorganizational exchange of information exposes programs of theoretical research on any one process to awareness of other programs of theoretical research on other processes, enabling the modeling of interrelations between processes (e.g., the interrelation of status, power, and legitimacy processes by Willer et al., 2014)
not that we do not read journals, but there is a lot to be said for what can be learned about ongoing research at the meetings. The modeling it enabled of the interrelations between processes enhanced both the application of theories to complex processes and the power of the interventions deduced from them. The partnership of rivals is more problematic if the rivals compete to explain the same phenomena. Had the competition been between orienting strategies it might have been crisis, not growth. But, although conflict between its theories has been contentious, it has not been irreconcilable. At the level of its actual working strategy, the aims of group processes research, its means of attaining them, and its standards of assessing them have all been more consensual than might be expected from the heterogeneity of its orienting strategies; its competition has been between testable, empirically grounded theories, not orienting strategies; and its standards of assessment have been appeals to reason and evidence (study, e.g., Elementary Theory’s comparison of alternative exchange theories of power, Willer et al., 2014). If standards are consensual, and if what they appeal to is reason and evidence, and what they assess is testable theory, conflict between theories is reconcilable. More of it is growth, not crisis.

14

MORRIS ZELDITCH JR.

Thus, like Advances in Group Processes, the Group Processes Conference has been a driving force in the advances that have taken place in group processes research beyond the 1980s. The mechanisms of its effect have been different but the effect has been much the same: There is more theory. More of it is empirically grounded, hence testable. More of its research is theory-driven, more of its theory tested. More of its theory is applied and more of its application, more of its interventions, is theory-driven. Finally, more of its theory has grown; it has grown in more ways
in confirmation, precision, rigor, scope, and explanatory power
and more of its growth has been cumulative than it had been in the 1950s.

NOTES 1. Some continued to ask the question even into the 1990s (Manson, 1993) and beyond (Harrington & Fine, 2000), and to answer it into the 1990s (Moreland, Hogg, & Hains, 1994; Sanna & Parks, 1997) and for another decade after that (Harrod, Welch, & Kushkowski, 2009). 2. Berger (1958) and Cohen (1958) were the initial theoretical formulators of programs that subsequently emerged testing, revising, and extending them. For Berger’s program, for example, see Berger, Wagner, and Webster (2014). 3. Some were simply rediscoveries of the same phenomenon in concretely different settings. For example, the effects of army rank on the interaction hierarchy of air crews or of race and ethnicity in the classroom were rediscovered for gender, education, and occupation as well as rank, race, and ethnicity in juries, hospitals, and universities as well as in the army and schools (Cohen, Berger, & Zelditch, 1972). But, absent theory, they remained unrelated, scattered, rather than evolving into programs.

REFERENCES Allen, V. I. (1965). Situational factors in conformity. Advances in Experimental Social Psychology, 2, 133
175. Allen, V. I. (1977). Social support for nonconformity. Advances in Experimental Social Psychology, 8, 1
43. Asch, S. E. (1951). Effects of group pressure on the modification and distortion of judgments. In H. Guetzkow (Ed.), Groups, leadership, and men (pp. 177
190). Pittsburgh, PA: Carnegie University Press. Back, K. W. (1951). Influence through social communication. Journal of Abnormal and Social Psychology, 46, 9
23. Bales, R. F. (1950). Interaction process analysis: A method for the study of small groups. Cambridge, MA: Addison-Wesley.

Twenty-Five Years of the Group Processes Conference

15

Bales, R. F., & Slater, P. E. (1955). Role differentiation. In T. Parsons & R. F. Bales (Eds.), Family, socialization, and interaction process (pp. 259
306). Glencoe, IL: Free Press. Bales, R. F., Strodtbeck, F. L., Mills, T. M., & Roseborough, M. E. (1951). Channels of communication in small groups. American Sociological Review, 16, 461
468. Berger, J. (1958). Relations between performance, rewards, and action-opportunities in small groups. Ph. D. Dissertation, Harvard University, Cambridge, MA. Berger, J., Eyre, D. P., & Zelditch, M. (1989). Theoretical structures and the micro-macro problem. In J. Berger, M. Zelditch, & B. Anderson (Eds.), Sociological theories in progress: New formulations (pp. 11
32). Newbury Park, CA: Sage. Berger, J., Wagner, D. G., & Webster, M., Jr. (2014). Expectation states theory: Growth, opportunities and challenges. In S. Thye & E. Lawler (Eds.), Advances in group processes (Vol. 31). Advances in Group Processes. Bingley, UK: Emerald Group Publishing Limited. Blau, P. (1977). Inequality and heterogeneity: A primitive theory of social structure. New York, NY: Free Press. Burke, P. J. (1967). The development of task and social-emotional role differentiation. Sociometry, 30, 379
392. Burke, P. J. (1968). Role differentiation and the legitimation of task activity. Sociometry, 31, 404
411. Cartwright, D., & Harary, F. (1956). Structural balance: A generalization of heider’s theory. Psychological Review, 63, 277
293. Cohen, B. P. (1958). A probability model for conformity. Sociometry, 21, 69
81. Cohen, B. P., Berger, J., & Zelditch, M. (1972). Status characteristics and social interaction: A case study of the problem of developing cumulative knowledge. In C. McClintock (Ed.), Experimental social psychology (pp. 449
483). New York, NY: Holt, Rinehart, Winston. Cohen, E. G. (1972). Interracial interaction disability. Human Relations, 25, 9
24. Cohen, E. G., & Roper, S. S. (1972). Modification of interracial interaction disability: An application of status characteristic theory. American Sociological Review, 37, 643
657. Davis, J. H., & Stasson, M. F. (1988). Small group performance: Past and present research trends. Advances in Group Processes, 5, 245
277. Deutsch, M., & Gerard, H. B. (1955). A study of normative and informational social influences upon individual judgment. Journal of Abnormal and Social Psychology, 51, 629
636. DiMaggio, P. J., & Powell, W. W. (1983). The iron cage revisited: Institutional isomorphism and collective rationality in organizational fields. American Sociological Review, 48, 147
160. Emerson, R. M. (1962). Power-dependence relations. American Sociological Review, 27, 31
41. Emerson, R. M. (1964). Power-dependence relations: Two experiments. Sociometry, 27, 282
298. Emerson, R. M. (1972). Exchange theory, Part II. Exchange relations and network structures. In J. Berger, M. Zelditch, & B. Anderson (Eds.), Sociological theories in progress (Vol. II, pp. 58
87). Boston, MA: Houghton Mifflin. Festinger, L. (1950). Informal social communication. Psychological Review, 57, 271
292. Foschi, M., & Lawler, E. J. (Eds.). (1994). Group processes: Sociological analyses. Chicago, IL: Nelson-Hall.

16

MORRIS ZELDITCH JR.

Goffman, E. (1956). Presentation of self in everyday life. Edinburgh: University of Edinburgh Press. Goffman, E. (1961). Encounters: Two studies in the sociology of interaction. Indianapolis, IN: Bobbs-Merrill. Goffman, E. (1967). Interaction ritual: Essays on face-to-face behavior. New York, NY: Doubleday. Granovetter, M. (1985). Economic action and social structure: The problem of embeddedness. American Journal of Sociology, 91, 481
510. Hare, A. P. (Ed.). (1976). Handbook of small group research (2nd ed.). New York, NY: Free Press. Hare, A. P. (1989). New field theory: Symlog research, 1960
1988. Advances in Group Processes, 6, 229
257. Harrington, B., & Fine, G. A. (2000). Opening the “black box”: Small groups and twenty-first century sociology. Social Psychology Quarterly, 63, 312
323. Harrod, W. J., Welch, B. K., & Kushkowski, J. (2009). Thirty-one years of group research in social psychology quarterly (1975
2005). Current Research in Social Psychology, 14, 75
103. Heise, D. (1979). Understanding events: Affect and the construction of social action. New York, NY: Cambridge University Press. Homans, G. C. (1958). Social behavior as exchange. American Journal of Sociology, 63, 597
606. Homans, G. C. (1964). Bringing men back in. American Sociological Review, 29, 809
818. Kitts, J. (2014). Beyond networks in structural theories of exchange: Promises from computational social science. In S. Thye & E. Lawler (Eds.), Advances in group processes (Vol. 31). Advances in Group Processes. Bingley, UK: Emerald Group Publishing Limited. Kollock, P. (1994). The emergence of exchange structures: An experimental study of uncertainty, commitment, and trust. American Journal of Sociology, 100, 313
345. Lawler, E. J., Ridgeway, C., & Markovsky, B. (1993). Structural social psychology and the micro-macro problem. Sociological Theory, 11, 268
290. Lucas, J. W., Huey, W. S., Posard, M. N., & Lovaglia, M. J. (2014). Perceptions of ability and adherence to rules, guidelines, and tradition. In S. Thye & E. Lawler (Eds.), Advances in group processes (Vol. 31). Advances in Group Processes. Bingley, UK: Emerald Group Publishing Limited. MacKinnon, N. J., & Robinson, D. T. (2014). Back to the future: 25 years of research in affect control theory. In S. Thye & E. Lawler (Eds.), Advances in group processes (Vol. 31). Advances in Group Processes. Bingley, UK: Emerald Group Publishing Limited. Manson, P. (1993). What is a group? A multi-level analysis. Advances in Group Processes, 10, 253
281. Mayhew, B. (1980). Structuralism and individualism: Part I, Shadow boxing in the dark. Social Forces, 59, 335
375. Melamed, D., Park, H., Zhong, J., & Liu, Y. (2014). Referent networks and distributive justice. In S. Thye & E. Lawler (Eds.), Advances in group processes (Vol. 31). Advances in Group Processes. Bingley, UK: Emerald Group Publishing Limited. Moreland, R. L., Hogg, M. A., & Hains, S. C. (1994). Back to the future: Social psychological research on groups. Journal of Experimental Social Psychology, 30, 527
555. Mullins, N. (1973). Theories and theory groups in contemporary American sociology. New York, NY: Harper and Row.

Twenty-Five Years of the Group Processes Conference

17

Roethlisberger, F. J., & Dickson, W. J. (1939). Management and the worker. Cambridge, MA: Harvard University Press. Sanna, L. J., & Parks, C. D. (1997). Group research trends in social and organizational psychology: Whatever happened to intragroup research? Psychological Science, 8, 261
267. Schachter, S. (1951). Deviation, rejection, and communication. Journal of Abnormal and Social Psychology, 46, 190
207. Shelly, R., Wagner, D. G., Welser, H., & Shelly, A. C. (2013). Unpublished program notes and program of the 25th Group Processes Conference. New York City, NY. August 9, 2013. Steiner, I. D. (1974). Whatever happened to the group in social psychology? Journal of Experimental Social Psychology, 10, 94
108. Steiner, I. D. (1983). Whatever happened to the touted revival of the group? In H. H. Blumberg, A. P. Hare, V. Kent, & M. Davies (Eds.), Small groups and social interaction (Vol. 2). Chichester: Wiley. Steiner, I. D. (1986). Paradigms and groups. Advances in Experimental Social Psychology, 19, 251
289. Stets, J. E., & Burke, P. J. (2014). The development of identity theory. In S. Thye & E. Lawler (Eds.), Advances in group processes (Vol. 31). Advances in Group Processes. Bingley, UK: Emerald Group Publishing Limited. Stouffer, S. A., Suchman, E. A., DeVinney, L. C., Star, S. A., & Williams, R. M. Jr. (1949). The American soldier: Studies in social psychology in world war II (Vol. 1). Adjustment during army life. Princeton, NJ: Princeton University Press. Strodtbeck, F. L. (1951). Husband-wife interaction over revealed differences. American Sociological Review, 16, 468
473. Strodtbeck, F. L., James, R. M., & Hawkins, C. (1957). Social status in jury deliberations. American Sociological Review, 22, 713
719. Strodtbeck, F. L., & Mann, R. D. (1954). Sex role differentiation in jury deliberations. Sociometry, 19, 3
11. Thye, S. R., & Lawler, E. J. (Eds.). (2013). Advances in group processes (Vol. 30). Bingley, UK: Emerald. Thye, S. R., Vincent, A., Lawler, E. J., & Yoon, J. (2014). Relational cohesion, social commitments, and person-to-group ties: Twenty-five years of a theoretical research program. In S. Thye & E. Lawler (Eds.), Advances in group processes (Vol. 31). Advances in Group Processes. Bingley, UK: Emerald Group Publishing Limited. Willer, D., Emanuelson, P., Lovaglia, M., Simpson, B., Thye, S. R., Walker, H., ... Chacon, R. (2014). Elementary theory: 25 years of expanding scope and increasing precision. In S. Thye & E. Lawler (Eds.), Advances in group processes (Vol. 31). Advances in Group Processes. Bingley, UK: Emerald Group Publishing Limited. Williams, R. M., Jr. (1947). The reduction of intergroup tensions. New York, NY: Social Science Research Council. Zelditch, M. (2007). Laboratory experiments in sociology. In M. Webster & J. Sell (Eds.), Laboratory experiments in the social sciences (pp. 517
531). New York, NY: Elsevier. Zelditch, M. (2013). Thirty years of advances in group processes: A review essay. In S. R. Thye & E. J. Lawler (Eds.), Advances in group processes: Thirtieth anniversary edition (Vol. 30, pp. 1
19). Advances in Group Processes. Bingley, UK: Emerald Group Publishing Limited.

EXPECTATION STATES THEORY: GROWTH, OPPORTUNITIES AND CHALLENGES Joseph Berger, David G. Wagner and Murray Webster Jr. ABSTRACT Purpose
We survey and organize over fifty years of theoretical research on status and expectation state processes. After defining some key terms in this theoretical approach, we briefly describe theories and branches in the program. Methodology/approach
We also focus on a few theories that illustrate distinct patterns of theory growth, using them to show the variety of ways in which the research program has grown. Findings
The program structure developed from a single set of theories on development and maintenance of group inequality in the 1960s to six interrelated branches by 1988. Between 1988 and today, the overall structure has grown to total 19 different branches. We briefly describe each branch, identifying over 200 resources for the further study of these branches.

Advances in Group Processes, Volume 31, 19
55 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0882-6145/doi:10.1108/S0882-614520140000031000

19

20

JOSEPH BERGER ET AL.

Research implications
Although the various branches share key concepts and processes, they have been developed by different researchers, in a variety of settings from laboratories to schools to business organizations. Second, we outline some important issues for further research in some of the branches. Third, we emphasize the value of developing new research methods for testing and applying the theories. Practical implications
These theories have been used to explain phenomena of gender, racial, and ethnic inequality among others, and for understanding some cases of personality attributions, deviance and control processes, and application of double standards in hiring. Social implications
Status and expectation state processes often operate to produce invidious social inequalities. Understanding these processes can enable social scientists to devise more effective interventions to reduce these inequalities. Originality/value of the chapter
Status and expectation state processes occupy a significant segment of research into group processes. This chapter provides an authoritative overview of ideas in the program, what is known, and what remains to be discovered. Keywords: Expectation states; status processes; theory growth; group inequality; experimental issues; new theoretical challenges

This chapter is written as part of activities surrounding the celebration of the 25th anniversary of the annual Group Processes conferences. These conferences, which started in 1988, as well as the series in Advances in Group Processes, have contributed to institutionalizing what Berger once referred to as the “new group process movement” (Berger, 1992). In what sense is this a new movement? First, it is new in the sense that there is deep commitment among many of the scholars in this area to construct theories for different types of group processes. Zelditch (2013) has argued that this commitment to theory was not always true of early small groups researchers. Second, it is new in the sense that the concepts in these theories are formulated in abstract and general terms. Third, it is new in the sense that the structures of these theories tend to be rigorous
and in some cases they are also formalized. Fourth, it is new in the sense that these theories are being tested, refined, and applied through an ongoing interchange with empirical research. The result of this new movement has been the emergence of theoretical research programs in the study of social

Expectation States Theory: Growth, Opportunities and Challenges

21

identity processes, affect control processes, power and exchange processes, and expectation state and status processes, among others. These features of the new movement may not be unique to group process research, but they are certainly not universal in sociology as a whole. It therefore seems appropriate on this occasion to review some of the research programs currently being developed in this group process area. The objective of this chapter is to undertake one such review for the Expectation States theory program. In section “Diffuse and Specific Status Characteristics and Instantiation,” there is a brief description of the contemporary branches of the program and introduction of some important concepts. In section “The Initial Period,” we present an overview of the different branches of the program that existed prior to 1988, and in section “From 1988 to the Present,” an overview of the branches that were primarily developed since 1988. In section “Elaboration, Proliferation, Integration,” we describe three major types of theoretical growth that have taken place in the program by examining three actual cases. In section “Expectation States Theory: Overall Conceptual Structure,” we describe the overall conceptual structure of the program at this time. In section, “Opportunities And Challenges,” we outline some of the experimental and theoretical challenges that lie ahead.

THE BRANCHES OF THE PROGRAM In this section we briefly examine the different branches of the program as they exist at present. First, we review the early branches, namely those that were developed between the initial expectation states research by Berger (1958) and 1988 when the group process conferences were initiated. Then we examine the branches that were primarily developed from 1988 to the present. Before turning to these branches, we introduce some basic concepts and the idea of instantiation. This will be of use in our examination of the program’s branches.

Diffuse and Specific Status Characteristics and Instantiation Two concepts and one analytic idea are particularly important in Expectation States theory. They are the concepts of diffuse and specific status characteristics and the notion of instantiation.

22

JOSEPH BERGER ET AL.

A diffuse status characteristic involves at least the following features: (1) a socially important characteristic such as race, gender, educational attainment, or immigrant status; (2) distinct states of the characteristic such as male
female, white
black, high
low educational attainment, native
immigrant, which can partition the concerned populations; (3) different status evaluations of these states in terms of honor, esteem, prestige, respect, and, in general, social worth; (4) high and low expectations of the generalized performance capacities of individuals who possess these different states, where these expectations are consistent with the status evaluations; high expectations associated with high-status evaluations, and low expectations with low-status evaluations.1 Specific status characteristics share many of the features of diffuse status characteristics. The concept involves the idea of a characteristic or ability with typically high or low states that carry differential status values. Associated with these states are consistently evaluated expectations. However, the expectations in this case are for performance capacities relevant to particular tasks. For instance, the specific status characteristic mathematical ability carries expectations for skill at solving differential equations or logic puzzles. We distinguish different levels of a specific characteristic, we associate differential status values with those levels, and we associate consistent beliefs about the specific performance capacities of individuals possessing the different states of the characteristic. These concepts of diffuse and specific status characteristics represent abstract theoretical constructions of theoretically important features of status distinctions. To apply these abstract constructions to a particular concrete social distinction in a society or relevant social unit involves instantiation. Instantiation is the decision by a researcher, based on assumptions or evidence or both, to treat some concrete social distinction as an instance, say, of a diffuse status characteristic for a particular population at a particular time. In doing so the researcher has decided to treat the states of the concrete social distinction as embodying differences in status value. Furthermore, the researcher has decided to treat the states of the concrete social distinction as being associated with high and low differences in performance expectations (anticipated capacities) that are consistent with their relative differences in status values.

The Initial Period Expectation States theory is a theoretical research program. It consists of a set of substantive and methodological working strategies that are embodied

Expectation States Theory: Growth, Opportunities and Challenges

23

in a set of interrelated theories, a body of empirical research relevant to these theories, and a body of applied and intervention research that is grounded in these theories.2 Building on the work of Bales (1950) and his colleagues, Berger (1958) initiated research in the program and in what became known as the power and prestige branch by comparing the structure of the interaction hierarchies in groups that were instrumentally oriented to their tasks with groups that were collectively oriented.3 Subsequently, Berger and Conner (1969, 1974), continuing research in this area, sought to explain the emergence of interaction hierarchies (power and prestige orders), the alignment of different behaviors in these hierarchies, and the stability of these hierarchies through time and tasks. In that research they were concerned with face-toface task groups whose members were initially similar in status. In the early 1960s, Berger, Cohen, and Zelditch (1966, 1972) constructed the initial version of the Status Characteristics theory to explain the impact of external status distinctions, using concepts of performance expectations. This theory was elaborated by Berger and Fisek (1974) and was significantly  Norman, and Zelditch (1977) reformulated and formalized in Berger, Fisek,  as the graph version of Status Characteristics theory. As research was taking place on the power and prestige and Status Characteristics theories, Berger, later joined by Cohen, Zelditch, and graduate students at Stanford, worked on developing a standardized experimental situation (Berger, 2007). The goals were to facilitate the task of developing a body of cumulative research and to provide researchers with experimental control of the interaction process in a group operating in a collectively oriented and valued task situation. The initial tests of predictions from the power and prestige and Status Characteristic theories were conducted in this situation (Berger & Conner, 1966, 1969; Berger et al., 1972; Moore, 1968). In the late 1960s, Murray Webster (1969) developed what has become known as Source theory: Given that an individual has the right to evaluate others (e.g., a teacher or a supervisor), what conditions and what processes make the evaluator’s evaluations effective in determining the expectations of others? In brief, Webster’s answer was that one condition for an effective evaluator was that the person being evaluated held high expectations for the evaluator’s skill. Barbara Sobieszek (1970) expanded the theory to deal with multiple evaluators who might agree or disagree on the evaluations they communicate, and Webster and Sobieszek (1974) carried out a program of experiments to test other implications of their formulation, including cases where an evaluator’s ability is unknown but his or her status is known. Also in the late 1960s, Berger, Zelditch, Anderson, and Cohen constructed the Status Value theory of Distributive Justice (1972). This theory

24

JOSEPH BERGER ET AL.

was developed as an alternative to exchange theories of justice being formulated at that time by Homans (1961) and Adams (1965). Subsequently, the Reward Expectation States theory was developed, which elaborates concepts introduced in the justice theory, such as the notion of referential structures, socially validated beliefs that connect status characteristics to goal objects or reward outcome states in a society or relevant social unit. This theory was constructed using the conceptual structure of the graph version of Status Characteristic theory (Berger, Fisek, Norman, & Wagner, 1985). Thus, in the period from the initiation of the program to 1988, when the first Group Processes Conference was held, six branches of the Expectation States theory program had emerged as active areas of research. Table 1 represents this situation. For readers interested in studying additional research related to a given topic, we have included the author names and

Table 1.

Expectation States Program Theories and Areas of Research Initiated Prior to 1988.

I. Power and Prestige Theory (Bales, 1950; Berger, 1958; Berger & Conner, 1966; Berger & Conner, 1969; Berger & Conner, 1974; Bienenstock & Bianchi, 2004; Conner, 1977; Conner, 1985) II. Status Characteristics Theory (Balkwell, 1991a, 1991b, 1994; Balkwell, Berger, Webster, Nelson-Kilger, & Cashen, 1992; Berger, 1974b; Berger, 1988; Berger, 1992; Berger & Fisek, 1974; Berger et al., 1966; Berger et al., 1972; Berger, Norman, Balkwell, & Smith, 1992; Berger, Fisek, Norman, & Zelditch, 1977; Berger, Rosenholtz, & Zelditch, 1980; Dovidio,  Brown, Heltmann, Ellyson, & Keating, 1988; Driskell & Mullen, 1990; Fisek, 1991; Fisek  Harkness,  et al., 1992; Fox & Moore, 1979; Gallagher, Gregory, Bianchi, Hartung, & 2005; Humphreys & Berger, 1981; Kuokkanen, 1993; Melamed, 2011, 2013a, 2013b; Moore, 1968; Norman, Smith, & Berger, 1988; Ridgeway & Johnson, 1990; Shelly, 1998; Wagner, Larshus, & Vogel, 2013; Walker & Cohen, 1985; Webster & Driskell, 1978; Whitmeyer, 2003; Zelditch, 1980; Zelditch, 1985) III. Source Theory (Savage & Webster, 1972; Sobieszek, 1970; Webster, 1969; Webster & Sobieszek, 1973; Webster & Sobieszek, 1974) IV. Status Value Theory of Distributive Justice (Berger, Zelditch, Anderson, & Cohen, 1972; Norman & Roberts, 1972; Webster, 1984; Webster & Smith, 1978) V. Reward Expectation States Theory (Berger, Fisek, Norman, & Wagner, 1985; Bierhoff,  Buck, & Klein, 1986; Cook, 1975; Fisek & Hysom, 2008; Fisek & Wagner, 2003; Hysom &  Fisek, 2011; Melamed, 2012; Stewart & Moore, 1992)  VI. The Standardized Experimental Situation (Berger, 2007; Cohen, Kiker, & Kruse, 1969; Conner, 1964; Dippong, 2012; Foschi et al., 1990; Gregory & Kalkhoff, 2013; Kalkhoff & Thye, 2006; Kalkhoff, Younts, & Troyer, 2008; Moore, 1965; Thye & Kalkhoff, 2009; Troyer, 2000; Troyer, 2001; Wagner & Harris, 1995)

Expectation States Theory: Growth, Opportunities and Challenges

25

the year of publication of the articles, reports, and books that are relevant to each branch.

From 1988 to the Present We have witnessed an extensive expansion of the expectation states program within the last twenty-five years. New theories and new branches have been developed along with much empirical research. Among the new branches primarily developed in this period is one concerned with the interrelation of Expectation States theory with other social science theories. In addition, while applications and intervention research originated early in the program with the work of E. G. Cohen (1970) and Entwisle and Webster (1974), the main empirical and theoretical research in this area of applications and interventions has developed since 1988, particularly the research treating gender as a diffuse status characteristic (Ridgeway, 2011). Also, this period has seen the introduction of two new and important formulations, the theory of Double-Multiple Standards by Martha Foschi (1989) and the theory of Status Construction by Cecilia L. Ridgeway (1991). Foschi’s theory is concerned with processes by which existing status distinctions are maintained. One way this occurs is by activating lenient standards for high-status individuals and strict standards for low-status individuals, given inconsistencies between status and performances. Ridgeway’s theory deals with a process by which initially nonvalued states of significant characteristics can acquire differential status values and high and low performance expectations consistent with those status values. Many of the other theories developed in this period build on the conceptual foundations of the graph formulation of the Status Characteristics theory. They include • Status-Evolution theory, which describes changes in status organizing activities as the process moves through a sequences of tasks (Berger, Fisek, & Norman, 1989)  theory, which is an integration of the Power and Prestige • Behavior-Status theory and the graph version of Status Characteristics theory (Fisek,  Berger, & Norman, 1991) • Evaluations and the Formation of Expectations theory (Fisek, Berger, &  Norman, 1995), which is an integration of the Source theory and the graph version of the status theory

26

JOSEPH BERGER ET AL.

In addition, there are • Status Legitimation theory (Berger, Ridgeway, Fisek, & Norman, 1998),  which describes the legitimation of a group’s interaction hierarchy • Status Cues theory (Berger, Webster, Ridgeway, & Rosenholtz, 1986; Fisek, Berger, & Norman, 2005), which describes how status cues can  create expectations and also how the expression of cues can be a function of expectation differences • Spread of Status Value theory (Berger & Fisek, 2006), which describes the creation of status characteristics as a result of the association of a significant characteristic with existing and well-established status characteristics in a society Finally, in this period we have seen the construction of variant models for second-order expectations by Troyer, Younts, and Kalkhoff (2001), Fisek, Berger, and Moore (2002), and Webster, Whitmeyer, and Rashotte  (2004). These models are concerned with how the expectations of the other person affect the focal actor’s expectations. In addition, there is a model constructed to describe the effect of positive and negative sentiments on status processes (Fisek & Berger, 1998), and there are probabilistic models  to describe the expectation state processes by which interaction hierarchies are formed (Fisek, 1974; Skvoretz & Fararo, 1996).  topics of interest among the theories in the program may While specific differ, what most of them have in common is a concern for the processes by which individuals form performance expectations for themselves and others, the maintenance of those expectations, and their consequences for status structures and behaviors. Table 2 provides a representation of the theoretical branches that were primarily developed in the last twenty-five years and listings of research relevant to each branch. Table 2.

Expectation States Program Theories and Areas of Research Primarily Developed from 1988 to Present.

VII. Theory of Double-Multiple Standards (Correll, 2001; Correll, 2004; Foschi, 1989; Foschi, 2000, 2009; Foschi, Lai, & Sigerson, 1994; Foschi & Valenzuela, 2012; Foschi, Warriner, & Hart, 1985; Gorman, 2006; Leahey, 2004) VIII. Theory on Evolution of Status Processes (Berger et al., 1989; Lockheed & Hall, 1976; Markovsky, Smith, & Berger, 1984; Prescott, 1986; Pugh & Wahrman, 1983) IX. Status Construction Theory (Ridgeway, 1991; Ridgeway, 2000; Ridgeway, Backor, Li, Tinkler, & Erickson, 2009; Ridgeway & Balkwell, 1997; Ridgeway, Boyle, Kuipers, & Robinson, 1998; Ridgeway & Correll, 2006; Ridgeway & Erickson, 2000; Webster & Hysom, 1998)

Expectation States Theory: Growth, Opportunities and Challenges

Table 2.

27

(Continued )

X. Behavior-Status Theory (Bienenstock & Bianchi, 2004; Fisek, Berger, & Norman, 1991;  1997; Robinson & Fisek, Berger, & Norman, 1995; Fisek, Berger, & Norman,   Balkwell, 1995; Webster & Rashotte, 2010) XI. Evaluations, Second-Order Expectations (Berger, Conner, & McKeown, 1969; Fisek,  Berger, & Norman, 1995; Fisek, Berger, & Moore, 2002; Kalkhoff, Younts, & Troyer,  Younts, 1997; Troyer et al., 2001; Webster, Rashotte, & 2011; Moore, 1985; Troyer & Whitmeyer, 2008; Webster & Whitmeyer, 1999; Webster & Whitmeyer, 2002; Webster, Whitmeyer, & Rashotte, 2004; Whitmeyer, Webster, & Rashotte, 2005) XII. Sentiments and Task Performance Expectations (Bianchi, 2004; Bianchi & Lancianese, 2007; Driskell & Webster, 1997; Fisek & Berger, 1998; Lovaglia & Houser, 1996;  1997) Shelly, 1993, 2001; Shelly & Webster, XIII. Status Legitimation Theory (Berger et al., 1998; Johansson & Sell, 2004; Kalkhoff, 2005; Munroe, 2001; Ridgeway & Berger, 1986; Ridgeway, Johnson, & Diekema, 1994) XIV. Status Cues and the Formation of Expectations (Berger et al., 1986; Driskell, Olmstead, & Salas, 1993; Fisek, 2009; Fisek et al., 2005; Foddy & Riches, 2000; Mohr,  1986; Ridgeway, Berger, & Smith, 1985; Tuzlak & Moore, 1984; Willard & Strodtbeck, 1972) XV. Spread of Status Value Theory (Berger & Fisek, 2006, 2008, 2013; Hysom, 2009; Thye,  2000; Thye, 2010; Walker et al., 2011) XVI. Expectation State Models of Interaction Hierarchies (Fararo & Skvoretz, 1988; Fisek, 1974; Skvoretz & Fararo, 1996; Skvoretz, Webster, & Whitmeyer, 1999)  XVII. Interrelation of Expectation States Theories with Other Theories (Barnum & Markovsky, 2007; Kalkhoff & Barnum, 2000; Kalkhoff et al., 2010; Lucas & Phelan, 2012; Melamed & Savage, 2013; Melamed & Abromaviciute, 2013; Oldmeadow, 2007; Oldmeadow, Platow, Foddy, & Anderson, 2003; Oldmeadow, Platow, & Foddy, 2005; Randel, Chay-Hoon, & Earley, 2005; Simpson, Willer, & Ridgeway, 2012; Thye, Willer, Markovsky, 2006; Wagner, 1988; Wagner, 2007; Willer, Lovaglia, & Markovsky, 1999) XVIII. Applications and Social Interventions (Bianchi, Kang, & Stewart, 2012; Bunderson, 2003; Cohen & Zhou, 1991; Cohen, 1970; Cohen, 1982; Cohen, 1993; Cohen & Lotan, 1997; Freese, 1974; Goar & Sell, 2005; Jackson, Hunger, and Hodge (1995); Lovaglia, Lucas, Houser, Thye, & Markovsky, 1998; Sauer, Thomas-Hunt, & Morris, 2010; Webster & Driskell, 1983; Entwisle & Webster, 1974; Webster, Hysom, & Fullmer, 1998; Webster & Whitmeyer, 2001; Yuchtman-Yaar & Semyonov, 1979) XIX. Applications of Expectation States Theory to Gender Research (Balkwell & Berger, 1996; Correll, Benard, & Paik, 2007; Foschi, 1996; Gerber, 2001; Hogue & Yoder, 2003; Lockheed & Hall, 1976; Lucas, 2003; Rashotte, 2006; Rashotte & Webster, 2005; Ridgeway, 2011; Ridgeway & Correll, 2004a, 2004b; Ridgeway & Diekema, 1994; Shelly & Munroe, 1999; Thomas-Hunt & Phillips, 2004; Wagner, 1995; Wagner & Berger, 1998; Wagner et al., 1986; Wood & Karten, 1986)

28

JOSEPH BERGER ET AL.

THEORETICAL GROWTH IN THE PROGRAM Elaboration, Proliferation, Integration What kind of growth has occurred in the program? To answer this question, we describe the construction of three of the theories formulated over the history of the program. These theories provide examples of the primary types of growth that have occurred: theory elaboration, theory proliferation, and theory integration (Wagner & Berger, 1985). Our descriptions of the theories involved are partial and informal. (For complete and formal descriptions of these theories see the references cited in the following sections). Elaboration The first of these examples is the graph formulation of the Status Characteristics theory (Berger et al., 1977). The principal objective was to elaborate the original Status Characteristics theory (Berger et al., 1966, 1972)
that is, to expand its domain and to increase the precision of its predictions.4 The original Status Characteristics theory (Berger et al., 1966, 1972) was formulated to provide a theoretical explanation of how external status differences strongly influenced the power and prestige order that emerged in groups irrespective of whether the status distinction was directly connected with the group’s goal. While restricting itself to two-person task and collectively oriented situations, this theory introduced the concept of a diffuse status characteristic, and conceptualized a status organizing process that involved the operation of just a single diffuse status characteristic at any given time. In content, the theory consisted of a set of assumptions that described how a structure of performance expectations emerged in collective task groups. It described conditions under which diffuse status characteristics possessed by actors would become significant to them (activation or salience), how status characteristics would become relevant to the task (burden of proof), how task expectations would be attributed to group members (assignment of competence), and how the behavior of actors would be determined by their expectations (basic expectation assumption). This early theory, as is also true of subsequent versions, was applicable to a broad range of concrete social distinctions including gender, race, ethnicity, educational attainment, occupational position, physical attractiveness, and sexual orientation.

Expectation States Theory: Growth, Opportunities and Challenges

29

The graph formulation (Berger et al., 1977) generalized and formalized the original Status Characteristics theory. It also provided a theoretical account of the results of experiments that were outside the domain of the original theory. With respect to generalization, the 1977 theory increased the number of actors (within limits) that could be dealt with in a status situation. It distinguished interactants from referent actors. (A referent actor is one not in the situation whose status information nevertheless affects the behavior of interactants.) Furthermore, the new theory increased the number of status characteristics that could be operating in the situation, whether consistently or inconsistently allocated. It also introduced the concept of Specific Status Characteristics to the theory. With respect to formalization, the new theory introduced extensive changes. A formal language was developed that involves signed graphs, which enables a theorist to describe a broad range of different types of status situations. In these descriptions, actors possessing different types of status elements are represented with concepts consistent with those that are also used to describe their group’s immediate task environment. In other words, actors, status elements, and some new theoretical constructs are represented by points in the graphs. Relations among all those elements are represented by signed lines and multiline paths connecting the theoretically significant elements. The concept paths of relevance describes theoretical linkages between actors and task components of these graphs. The paths describe how status elements possessed by actors become connected to the group’s task outcomes as, for instance, a particular status characteristic comes to be treated as relevant to task success. The sign of the path indicates whether the possessed status element links to task success or task failure, while the length of the path indicates how extended that link is. Finally, the strength of a path (its task relevance) provides a measure of the weight or significance of a possessed status element in contributing to the group’s success or failure. With these paths of relevance one can also distinguish and assess the relative strength of different status-biasing structures: those where actors from different social categories are culturally tied to tasks (relevance structures) and those in which actors, through the operation of a status generalization process, become connected to tasks (associational or burden of proof structures) (Humphreys & Berger, 1981). The elaborated theory also introduced a model, the principle of organized subsets, that describes how actors process multiple items of positive and

30

JOSEPH BERGER ET AL.

negative status information. We assume that these information-organizing processes, for the most part, occur outside of an actor’s awareness. Given multiple items of status information, an actor segregates this information into positive and negative subsets, and within these subsets combines information, taking into account the task relevance of the various items. These univalent subsets are, in turn, combined to determine the actor’s aggregated performance expectations. An elaborated version of the basic expectation assumption then specifies how the actor’s behavior in relation to a given other is a direct function of the difference in his/her aggregated performance expectations and the aggregated performance expectations of the other. Further, because of the formalization, the deductive capacities of the Status Characteristics theory are significantly enhanced. Given the estimation of parameters, we can (a) describe and make predictions for a wide range of specific status situations and (b) describe and make general predictions for classes of status situations, for example, the effects of status consistency or inconsistency on the power and prestige order that emerges in a group. We also note that there has been research that is strictly concerned, per se, with the nature of the formalization of the Status Characteristics theory. Whitmeyer (2003) has constructed an equation version of the graph theory, which can facilitate relating the Status Characteristics theory to other formal group process theories (Kalkhoff, Friedkin, & Johnsen, 2010). Kuokkanen (1993) has developed a set-theoretic axiomatization of the theory that provides other techniques for deriving consequences from the formulation, and Humphreys and Berger (1981) present an axiomatization of the theory with major implications that include the “inconsistencyequality” result. This is an implication that, under specified conditions, increasing the inconsistency among a set of equally relevant status characteristics possessed by members of a group will increase the equality of the group’s power and prestige hierarchy. The graph formulation of the Status Characteristics theory represents growth by theory elaboration. The elaborated theory increases the domain of the initial Status Characteristics theory and increases the precision of its predictions. It was also found to be consistent with existing experimental data relevant to the Status Characteristics theory (Berger et al., 1977). Major results predicted by the original theory can be predicted from the elaborated theory. In addition, a broad range of predictions that are outside the domain of the original formulation are now predicted by the graph formulation.

Expectation States Theory: Growth, Opportunities and Challenges

31

Much research conducted since presentation of the 1977 theory provides a body of empirical support. For summaries of relevant empirical research, see Fisek, Norman, and Nelson-Kilger (1992) and Balkwell (1994).  Proliferation A second example of growth is represented in the Spread of Status Value theory (Berger & Fisek, 2006). The objective here was to create a theory using the spread of status value mechanism to describe how new diffuse status characteristics can be created as a result of their relations to established status distinctions in a society or relevant social unit. One of the interesting properties of status values
high or low
is that under appropriate conditions they spread from valued elements to a wide range of associated and initially nonvalued elements such as states of social objects, states of roles, states of different groups, and from status-valued groups to individuals who are members of those groups.5 On an informal level, we can say that the theory has four major arguments. In this discussion, we shall consider only the situation where all positive status values are spreading to one state of a new characteristic, and all negative status values to the second state of a new characteristic. The first argument is that status values spread from the states of established status characteristics to associated states of an initially nonvalued new characteristic, call it N. An evaluated status characteristic and a nonvalued element N are associated if they are both possessed by the same actor in the immediate situation of action. This spread of status value is a function, first, of the task relevance of those diffuse and specific status characteristics associated with the states of the nonvalued characteristic N. The greater the relevance of the characteristic is to the task at hand, the greater its contribution to the spread of status value. Second, it is a function of the degree of consistency of the status characteristics. The greater the consistency of equally relevant status characteristics, the greater their contribution to the spread of status value. For instance, for a set of characteristics possessed by p and o, if there is inconsistency, some of the characteristics are contributing positive status value to N(a) and negative status value to N(b), while other characteristics are contributing positive status value to N(b) and negative to N(a). Third, given that they are consistent and associated with states of N, the greater the number of status characteristics that are involved, the greater their contribution to the status-spreading process. The second argument says, as status values spread to the states of N, high and low generalized performance expectations become attached to

32

JOSEPH BERGER ET AL.

those states of N in a manner consistent with the positive and negative signs of the accumulating status values. The third argument says, as status values accumulate on states of N, the bonds linking these states of N to high and low generalized performance expectations are strengthened until they approach a strength at least as strong as that represented by an already established diffuse status characteristic. At this point, if the above process has occurred as described, the new characteristic has attained a graph structure very similar to the one we have used in the past to represent existing and already well-established diffuse status characteristics. Therefore, the outcome of this creation process is predicted to produce a situation explained by the preexisting theory. There is one additional argument. Assume that the bonds between states of N and generalized performance expectations have approached at least the strength represented by established diffuse characteristics, and that actors act in accord with the status beliefs and expectations associated with states of N, then if the behavior of those actors is socially validated (supported, reinforced, confirmed) by others, the newly constructed characteristic can become a stable diffuse status characteristic for the group. In the language of the theory, “[T]he validation process confirms the expectations of the newly evaluated characteristic …; it translates these expectations into publicly observed behaviors …; and it demonstrates that there is a consensus on the newly created characteristic within the immediate group.” (Berger & Fisek, 2006, pp. 1050
1051).6 Recently, Walker, Webster, and Bianchi (2011) carried out an initial empirical test of some of the predictions of this theory. Among other things they were able to show that beliefs about high and low-status value and related beliefs about differences in performance expectations come to be attributed to the states of N as predicted, and second, that differences in the exercise of influence, as a function of these performance expectation differences, were also found to hold as predicted by the theory. The Spread of Status Value theory represents theory growth by proliferation. The theory introduces concepts and principles that are auxiliary to the graph formulation in order to deal with new substantive problems of interest. At the same time, the theory rests on core concepts and principles of the graph formulation. These core concepts and principles enable making precise predictions from that theory. We see similar theoretical structures in the theories that have been recently developed for second-order expectations (Fisek et al., 2002; Troyer  et al., 2001; Webster et al., 2004). As in the above case, these theories rest

Expectation States Theory: Growth, Opportunities and Challenges

33

on core concepts and principles of the graph formulation of the Status Characteristics theory while introducing new concepts and principles to deal with the issue of how second-order expectations affect first-order expectations. But from these theories considered by themselves, as with the Spread of Status Value theory, it is not possible to derive some of the major derivations of the Status Characteristics theory (Humphreys & Berger, 1981). Proliferations, nevertheless, represent an important form of growth within theoretical research programs. Integration The third example we consider is that of the Behavior-Status theory (Fisek  et al., 1991). The objective in this case was to construct a theory that integrated key arguments from the Power and Prestige theory with arguments in the graph version of the Status Characteristics theory. The Power and Prestige theory was constructed to explain the emergence of interaction hierarchies (what later became known as the group’s power and prestige order). The theory applied to status homogeneous groups working in task and collectively oriented situations (Berger & Conner, 1969, 1974). An important idea in the Power and Prestige theory is the concept of unit sequences or cycles of behavior that result in status advantage of one actor relative to a second. Examples of such sequences are when A gives action opportunities to B and B performs and A positively evaluates and then accepts the performance of B, or when B performs and A negatively evaluates that performance, but B persists with additional performances and B eventually influences A to accept his performance. The theory develops a typology of such action sequences. A key argument in the Power and Prestige theory is that if sequences occur that favor one actor over a second actor, then a consistent expectation ranking will emerge. This expectation ranking will give rise to corresponding inequalities in the actor’s subsequent power and prestige behaviors, for example, the likelihood of initiating performances or of having those performances positively evaluated or of being able to influence the other. The notion of a basic interchange pattern (or a bip as it is called) introduced in the Behavior-Status theory translates these sequence ideas into the graph language of the Status Characteristics theory. “A behavior interchange pattern … is a set of interaction cycles or unit sequences such that all actors involved in the interaction accept the acts, and all cycles have the same power-and-prestige significance with respect to the actors” (Fisek  et al., 1991, p. 117). The ideas about sequences in this context primarily

34

JOSEPH BERGER ET AL.

refer to the different behavioral patterns as these are conceptualized in the original Power and Prestige theory. The argument is that if a behavior interchange pattern becomes salient, it will create an expectation ranking, and thus, an actor’s subsequent power and prestige behavior will be a function of that expectation ranking. With the concept of a behavior interchange pattern, the Behavior-Status theory can be applied to status homogeneous situations. It allows us to derive an assertion that, for example, if an expectation ordering emerges, then the subsequent behaviors will in general maintain that ordering, which is one of the oldest results in the Power and Prestige theory (Berger & Conner, 1974).7 The Behavior-Status theory also applies to status heterogeneous situations. In addition, the theory can now be applied to an important class of new situations
those where both status characteristics and behavior interchange patterns are salient. In such cases, the theory tells us, for example, what happens when there are inconsistencies between salient status characteristics and activated behavioral interchange patterns. Thus, growth in this case is a result of integrating arguments from two theories in the program. The integrated theory can make some major predictions when the two initial theories are considered separately. It can also make new predictions that are outside the domains of each of the initial theories. Webster and Rashotte (2010), who carried out the first independent test of this theory, rather ingeniously solved the problem of how to create behavior interchange patterns in the standardized experimental situation. Their experimental results were mixed. With respect to female subjects, they found an extraordinarily close fit of their model predictions to observations. With respect to male subjects, there was an inadequate fit of model to data. They continue their empirical research on the Behavior-Status theory.8

Expectation States Theory: Overall Conceptual Structure As the program has grown over its history, new theories and new branches have emerged and its conceptual structure has undergone change. Can we represent, in some direct and simple form, the overall conceptual structure of the program as it exists today? The program has theories that involve or are directly connected to the graph version of the Status Characteristics theory, and other active and important theories that are not so directly connected. If, for the present, we restrict ourselves to those theories that are connected to the graph version of the Status Characteristics theory, we can

Expectation States Theory: Growth, Opportunities and Challenges

35

construct a representation of the conceptual structure of the program. Eleven theories including the Status Characteristics theory itself and three theories on second-order expectations fit this category.9 Fig. 1 is a representation of the overall conceptual structure of the graph-connected theories in the program. On the first panel of this figure are the 11 distinct unit theories that are connected with the graph formulation. From each of these theories there is an arrow linking it to the Core, which is on the second panel. The Core consists of the concepts and principles listed in the figure: status characteristics; the basic relations of the theory
possession, relevance, and dimensionality; the concept of performance expectations; the concept of paths of relevance; the principle of organized subsets; and the basic expectations assumption. Each of the unit theories operates within the framework of these core concepts and principles. “Within the framework” is used in the strong sense that each of these theories must use some concepts and principles from the Core to generate predictions. Many of these theories use concepts and principles in addition to those from the Core, but they all use at least some of those from the Core. The third panel distinguishes the order of expectations. First-order expectations are those that the focal actor holds for self and the other actor, while second-order expectations are those that the focal actor anticipates that the other holds for self and the focal actor. Most theories in the program deal with first-order expectations. However, as already observed, there are theories that deal with second-order expectations and their effect on first-order expectations. The fourth panel distinguishes the types of expectations that are dealt with within the program: expectations for performances, expectations for rewards, and expectations for valued status positions, where the latter is a concept that is developed in the Status Legitimation theory (Berger et al., 1998). The two types of measures that have been developed to assess expectation differences between actors are presented on the fifth panel. Expectation advantage refers to the difference between the aggregated performance expectations of the focal actor and the aggregated performance expectations of the other. Expectation standing, which was developed for groups larger than a dyad, is the proportion of positive expectations each member of a group possesses (Fisek et al., 1991).  list the three mathematical functions that have On the sixth panel, we been constructed to translate expectation differences to outcomes: the linear, logistic, and the identity function.

36

JOSEPH BERGER ET AL.

UNIT THEORIES

CORE CONCEPTS AND PRINCIPLES

ORDER OF EXPECTATIONS

TYPES OF EXPECTATIONS

MEASURES OF EXPECTATIONS

TRANSLATION FUNCTIONS TO OUTCOMES

TYPES OF BEHAVIORAL OUTCOMES

Fig. 1.

Unit Theory2

Unit Theory1

,...,

Unit Theoryn

Status Characteristics (Categorical & Graded) Basic Relations Performance Expectations Expectations’ Order Paths of Relevance Principle of Organized Subsets Basic Expectations Assumption

Second

First

Rewards

Performance

Expectation Standing

Expectation Advantage

Linear

Power and Prestige Behaviors and Status Beliefs

Valued Status Positions

Identity

Logistic

Reward Allocation Behaviors

Status Cues Behaviors

Dominating and Compliance Behaviors

Conceptual Structure of Program: Graph-Connected Components.

Expectation States Theory: Growth, Opportunities and Challenges

37

On the seventh panel, we distinguish the different types of behavioral outcomes that are dealt with in the different unit theories that use the graph formulation. These include status beliefs and power and prestige behaviors, reward allocation behaviors, status cue behaviors, and dominance and compliance behaviors. It is important to bear in mind that as the research program has developed, its conceptual structure has undergone change. New ideas and arguments have emerged as important (e.g., the principle of organized subsets), while some older ideas and arguments have become less important to the program (e.g., the spread of inequality). Fig. 1 is a picture of the current conceptual structure. But as the program continues to evolve, we can surely expect that its conceptual structure will undergo further change and elaboration. Finally, we can sum it all up. Looking at the program as a whole, we can say that over its history there has been a steady increase in the explanatory domain of the program, an increase in the precision of its claims, and an increase in its empirical support. In summary, the expectation states program has grown. However, we must quickly add that there are major experimental and theoretical tasks and challenges that still lie ahead. We turn next to a consideration of some of these tasks and challenges.

OPPORTUNITIES AND CHALLENGES While progress has been made over the years in various parts of the program, many opportunities and challenges remain. Some of these are primarily experimental in character; they have to do with the evidence generated and the nature of the research situations used to test Expectation States theories. Other issues are more theoretical in character; they deal with resolving differences between existing theories and models in the program, expanding the domain of some of those theories, and developing theories of social processes not already considered in the program.

Some Experimental Issues One of the most obvious ways in which the evidence relevant to theories in the program can be improved is through replication of the existing work. Both theoretical and empirical replications are possible. Theoretical

38

JOSEPH BERGER ET AL.

replications involve testing similar predictions under similar theoretically derived conditions, but with different concrete variables involved. For example, Wagner, Ford, and Ford’s (1986) predictions regarding the greater effect of disconfirmation on status behavior as compared with confirmation, originally tested with gender, could be replicated with race or ethnic distinctions as the salient diffuse status. Empirical replications would require that the same theory and the same concrete variables be involved. Theoretical replications are particularly valuable; they “empirically stretch” the application of the tested theory more effectively. Other kinds of “empirically stretching the theories” should be considered. There are important predictions of Expectation States theories that are still to be tested, or that have been tested only under limited theoretical conditions. Perhaps more important, some kinds of predictions may be quite difficult to test. For example, Status-Evolution theory (Berger et al., 1989) makes a pair of interesting long-term predictions: First, given that a specific characteristic is introduced as an intervention that is inconsistent with a diffuse status characteristic in a closed evolution process (one in which no new information is introduced to the process), then the actor’s expectations will eventually reach an equilibrium as the actor moves from task to task. Second, any intervention introduced in a closed evolution process will have long-lasting effects. So far, no simple procedures have been devised to create such sequences requiring evolving tasks that are meaningful to subjects. These are experiments still to be devised. Theory-stretching experiments would also be appropriate for helping to determine the effect of scope conditions on important theoretical principles within Expectation States theories. For example, the evidence to date strongly supports subset-combining as the means by which actors organize status information in multiple characteristic status situations, even when the implications of those statuses are inconsistent. To what extent is this due to the fact that we have been concerned with scope-defined situations where achieving success on a valued task is strongly emphasized? Suppose we were dealing with a nontask focused situation
that is, a situation involving a high level of emotional-expressive behaviors, where no valued task goal is defined, but status information has become salient. Would subsetcombining occur in such a situation? Or would we find a form of balancing with one subset dominating a second? Or would we find instability with vacillation of which subset dominates as the interaction evolves? Such research would give us an understanding of the generality as well as the explanatory power of the principle of organized subsets. It would also be of

Expectation States Theory: Growth, Opportunities and Challenges

39

considerable value in developing expectation state theories of other types of social processes. Challenges arise with respect to the standardized experimental situation, first developed in the early 1960s. As noted previously, a major objective was to construct a standardized experimental situation so as to facilitate the task of producing a body of comparable and cumulative empirical research. As an example, the protocols developed for the situation were constructed to detail a problem-solving environment for subjects and to operationalize the scope conditions of task and collective orientation, which were already being regarded as sufficient conditions to activate a status organizing process. These protocols were assumed to define background variables, as distinct from experimental variables, that would be constant conditions in studies using this situation. This assumption of constancy was true of other features of the situation, for example, the decision-making process, the tasks employed, the debriefing of subjects, and eliciting information on their reactions to the experiment (Berger, 2007). However, as one might imagine, over the years all sorts of changes have been introduced into the standardized situation. Some of these changes are technological advances such as the use of closed circuit TV in the setting (Cohen, Kiker, & Kruse, 1969), and the subsequent computerization of the experimental situation (Foschi, Sigerson, Lai, & Foschi, 1990; Troyer, 2000; Webster et al., 2004). Still other changes were due to use of new types experimental manipulations as, for example, in the changes introduced by Webster and Rashotte (2010) in order to create “behavior interchange patterns” in the standardized situation. But the motivations of others who have introduced still more changes in the decision-making process, in the ways of eliciting relevant information from subjects, and in the protocols used in experiments, are less clear. The first challenge we face in this area is to strengthen and to create greater uniformity and standardization in all aspects of the experimental procedures that are used in this setting and in the reporting of experimental results. We now have resources for this task. We can draw on the experiences of those who have worked with this experimental situation, also on the recent research of Kalkhoff and Thye (2006; Thye & Kalkhoff, 2009) and Dippong (2012) on variations in behavior in this setting, and on the handbook created by Wagner and Harris (1995) that deals with doing experiments in the standardized experimental situation. There is a second challenge in this area, namely, to develop new experimental situations to provide us with additional information on expectation

40

JOSEPH BERGER ET AL.

state and status processes. In this task, we can take advantage of what is known, for example, about the relation of latency processes to status processes (Conner, 1977, 1985; Willard & Strodtbeck, 1972), or the differences in emotional responses to individuals of different statuses (Ridgeway & Johnson, 1990), or, most promising, the explorations on speech wave frequency by Stanford Gregory and Will Kalkhoff (2013). At the very least, such experimental situations should supply empirical information that is valuable in addition to that from the standardized experimental situation (SES). It is also just possible that one or more of such settings might become the bases of a new standardized experimental situation.

Some Theoretical Challenges One of the most interesting challenges arising in the program at the theoretical level concerns our understanding of the role of second-order expectations. As noted, second-order expectations are those that a focal actor anticipates the other holds for the focal actor and self. An important theoretical question is “what is the impact of the focal actor’s second-order expectations on his/her first-order expectations
those the focal actor holds for self and the other”? Three different models on the impact of second-order expectations have been developed. Fisek et al. (2002) based  Source theory with their model on the work by Fisek et al. (1995) linking  Status Characteristics theory. Among other things, they argue that if the focal actor behaves in accord with others’ expectations, that will determine the focal actor’s first-order expectations. In contrast, Troyer et al. (2001) base their model on identification of three motives for an actor’s task behavior. These motives concern task performance, avoidance of status loss, and facilitation of interaction, which they connect to the task and collective scope conditions of Status Characteristics theory. They argue that the greater the weighting of the latter two motives, the more an actor’s behavior will reflect second-order expectations over first-order expectations. Finally, Webster and Whitmeyer (1999) suggest that the effects of secondorder expectations are an increasing function of the expectations associated with the status of the other actor. Data has been generated that supports each model. It is probably not possible at this point to judge any one of them as generally more adequate. Two interesting theoretical questions are worth posing at this stage. First, are there conditions under which one or another of these models is more appropriate? If so, what are those conditions? Second, can ideas from these

Expectation States Theory: Growth, Opportunities and Challenges

41

models be combined to create a unified formulation of some or all of these models? There is a real challenge here to find answers to these theoretical puzzles. Another area for additional theoretical investigation involves the Status Construction theory and the Spread of Status Value theory. It seems plausible to argue, on the basis of research that has already been done, that we can experimentally embed differential status values and differential performance beliefs and expectations in initially unevaluated states of a nominal characteristic. In other words, we can create what we might call “proto-status characteristics” under laboratory conditions. But big questions remain: By what theoretical processes do we go from the level of “proto-status characteristics” to the level of long-enduring, well-established, and institutionalized social distinctions such as those associated with gender, race, and ethnicity? As a start, we can assume that processes of diffusion (Ridgeway & Balkwell, 1997) and social validation (Berger & Fisek, 2006) play important roles in the emergence of these stable structures.  These suggest that a very slow evolutionary process may be involved in the creation of enduring status distinctions. Yet, Berger and Fisek (2008)  may have identified a well-established societal distinction whose emergence not have involved a very slow evolutionary process. Turkish immigration into Germany after World War II generated a strong, stable ethnic-like distinction between “native German” and “immigrant Turkish” populations. How this societally established status distinction emerged is explained by an application of Berger and Fisek’s (2006)  theory of the spread of status value. The Turkish immigration into Germany in the early 1960s was part of the Gastarbeiter program introduced by the Federal Republic in agreement with the Turkish government. The German
Turkish differentiation became a recognized and important social distinction in German society. This social distinction could be considered a diffuse status characteristic for the populations involved if these populations held differential status evaluations for Germans and Turks, with, say, Germans being more highly socially valued than Turks, and with members of these populations commonly assuming that Germans are in general superior and more capable than Turks on a wide (and typically unspecified) range of valued tasks and activities (Berger & Fisek, 2008).  Basically, the indigenous German population acquired high-status values and high performance expectations relative to the low-status values and low performance expectations for the immigrant Turkish population, and those differences in status values and performance expectations were related to existing differences in occupational and educational attainments

42

JOSEPH BERGER ET AL.

and in cultural competencies (command of German language and norms, values, rituals) of the two populations. It may be that the emergence of stable and well-established status characteristics depends on particular social conditions in which the process occurs as well as the basic creation mechanism that is involved (the interaction hierarchy in Status Construction theory, or the spread of status value in the Spread of Status Value theory). It may be that other processes are also involved. The theoretical challenge, as already noted, is to obtain a deeper theoretical understanding of the process of moving from “proto-status characteristics” to the level of the enduring and major diffuse status characteristics in societal life. Finally, there are opportunities to isolate and construct theories of other social processes as state-organizing processes. One process that has generated significant interest is the social control process. Talley and Berger (see Berger, 1988) have developed a general conception of five steps in the application of social control to groups: (1) the establishment of a set of normative expectation conditions; (2) activation of the social control process through unexpected events that violate those normative expectations; (3) emergence of control states among members of the group
“Violator” for those who have deviated from expectations, and “Carrier” for those who have followed and maintain expectations; (4) development of behaviors based on those states (e.g., expressions of moral indignation among Carriers, and guilt and remorse among Violators); and (5) establishment of stable outcomes structures based on the process, such as the attribution of different personality characteristics or different individual level stereotypes to Carriers versus Violators.10 Using much of the conceptual structure of the Berger and Talley work, Wagner (1988) began developing a more precise theory of status violations. This theory spells out specific conditions for identifying a status violation, including, for example, legitimation of the expectations in a differentiated power and prestige order and a regular or extreme deviation from those expectations by an actor in the group. Once a violation is identified, the group exerts behavioral control toward the violating actor, attempting to convince him or her to behave in accord with normative expectations. If behavioral control does not work, the group switches to expectation control, assigning the Carrier and Violator roles to themselves and to the deviant actor, respectively. Since these roles represent differential status states, the expectations toward the Violator are lowered, and behavior toward that actor changes to reflect the lowered expectations.

Expectation States Theory: Growth, Opportunities and Challenges

43

Wagner then develops derivations from those principles. First, behavioral control is likely to be successful and rather rapid in returning a deviant actor to conforming behavior if the deviant actor’s behavior was based simply on a lack of knowledge of the group’s expectations; behavioral control is likely to be less successful, with a gradual transition to expectation control if the deviant actor’s behavior was based on already established expectations from another group; behavioral control is likely to be least successful, and the transition to expectation control quite rapid if the deviant actor’s behavior was based on legitimated expectations from another group. Second, expectation control will have differential impacts on Violators with prior high status in the group than for the ones with prior low status. High-status Violators will definitely lose some status, but they are likely to remain valued members of the group. Low-status Violators may not even be able to retain their membership in the group. Wagner has recently developed a more formal version of this theory and is in the process of testing some of its implications. These are only some of the theoretical challenges and opportunities faced by the program. New challenges surely will arise with new theoretical formulations and with new specific problems to be solved. The growth of the program in the future will be shaped, in good measure, by the ways that theorists and researchers respond to such challenges.

NOTES 1. Diffuse status characteristics can have an additional important feature (Berger, Cohen, & Zelditch, 1966): associated with the states of the characteristic there may be specific characteristics or beliefs, such as a belief that men are more mechanical than women or that women are more nurturing. Currently, neither the Status Construction nor the Spread of Status Value theories deal with the emergence of such stereotypes. However, it is a subject of research interest in the program (Berger & Webster, 2006; Wagner et al., 2013). 2. For previous reviews of the Expectation States theory as a theoretical research program, see Berger, Conner, and Fisek (1974), Wagner and Berger (2002),  and Berger and Webster (2006). 3. In the language of that time, these were labeled, respectively, the “instrumental-solution oriented” and the “integrative-evaluative oriented” conditions (Berger, 1958). 4. An earlier formulation by Berger and Fisek (1974) was an initial attempt to  elaborate the original theory. The graph formulation discussed here was the second attempt to elaborate the original Status Characteristics theory.

44

JOSEPH BERGER ET AL.

5. In her presidential address to the 108th meeting of American Sociological Association on “Why Status Matters,” Cecilia L. Ridgeway (2014) argues that one of the important consequences of status is an “associational bias”
individuals want to be associated with high-status groups but have difficulty in being associated with low-status groups (pp. 6
7). We suggest that the underlying mechanism producing this bias is a spread of status value process. 6. The Spread of Status Value theory recently has been extended to apply to the social construction of status objects and the embedding of differential status values in roles (Berger & Fisek, 2013).  to note that not all claims derivable from the Power and 7. It is important Prestige theory are also derivable from the Behavior-Status theory. For example, the Power and Prestige theory argues that inequalities in the group “spread” or increase, which is an argument not implied by the Behavior-Status theory. This is a consequence of differences in the conceptualization of some of the underlying processes in the two theories. 8. For further research on the nature of theoretical growth in sociology that has influenced the development of the expectation states program, see Berger (1974a), Wagner and Berger (1985) and Berger and Zelditch (1993). 9. The theories directly involved or directly connected to the graph version of status theory are as follows: the Reward Expectation States theory, the theory on the Evolution of Status Processes, the Behavior-Status theory, the theory on Evaluations and the Formation of Expectations, the three theories on second-order expectations, the Status Legitimation theory, the theory on Status Cues and the Formation of Expectations, the Spread of Status Value theory, and the graph version of the Status Characteristics theory itself. 10. Berger (1988) has also sketched a theory of dynamic affect states. He assumes that these states are activated by events producing a high level of interpersonal emotional arousal. Aside from being based on positive and negative emotions, these affect states are associated with univalent cognitions. Individuals appear to behave toward each other in purely positive or purely negative terms with affiliating behaviors engendered by positive states and rejection behaviors engendered by negative states. These affect states are conceptualized as situational and transitory, and in its evolution, the process is seen as being capable of oscillating between bonding and rejection states.

ACKNOWLEDGMENTS This is an expanded version of a paper presented at the 25th Group Processes Conference at the annual meetings of the American Sociological Association in New York City. Joseph Berger is pleased to acknowledge the research support of the Hoover Institution at Stanford University. David G. Wagner’s participation was partially supported by the Department of Sociology, University at Albany, SUNY. Murray Webster’s participation was partially supported by NSF grants SES-1269342 and SES-131093.

Expectation States Theory: Growth, Opportunities and Challenges

45

REFERENCES Adams, J. S. (1965). Inequity in social exchange. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 2, pp. 267
299). New York, NY: Academic Press. Bales, R. F. (1950). Interaction process analysis. Cambridge, MA: Addison-Wesley. Balkwell, J. W. (1991a). From expectations to behavior: An improved postulate for expectation states theory. American Sociological Review, 56, 355
369. Balkwell, J. W. (1991b). Status characteristics and social interaction: An assessment of theoretical variants. In E. J. Lawler, B. Markovsky, C. Ridgeway, & H. A. Walker (Eds.), Advances in group processes (Vol. 8, pp. 135
176). Greenwich, CT: JAI Press. Balkwell, J. W. (1994). Status. In M. Foschi & E. J. Lawler (Eds.), Group processes: Sociological analyses (pp. 119
148). Chicago, IL: Nelson-Hall. Balkwell, J. W., & Berger, J. (1996). Gender, status, and behavior in task situations. Social Psychology Quarterly, 59, 273
283. Balkwell, J. W., Berger, J., Webster, M., Jr., Nelson-Kilger, M., & Cashen, J. (1992). Processing status information: Some tests of competing theoretical arguments. In E. J. Lawler, B. Markovsky, C. Ridgeway, & H. A. Walker (Eds.), Advances in group processes (Vol. 9, pp. 1
20). Greenwich, CT: JAI Press. Barnum, C., & Markovsky, B. (2007). Group membership and social influence. Current Research in Social Psychology, 13. Retrieved from http://www.uiowa.edu/∼grpproc/ crisp/crisp13_3.pdf Berger, J. (1958). Relations between performance, rewards, and action-opportunities in small groups. Unpublished Ph.D. Dissertation, Harvard University. Berger, J. (1974a). Expectation states theory: A theoretical research program. In J. Berger, T. L. Conner, & M. H. Fisek (Eds.), Expectation states theory: A theoretical research  program (pp. 3
22). Cambridge, MA: Winthrop. Berger, J. (1974b). Scope-defined formulations. In J. Berger, T. L. Conner, & M. H. Fisek  (Eds.), Expectation states theory: A theoretical research program (pp. 15
16). Cambridge, MA: Winthrop. Berger, J. (1988). Directions in expectation states research. In M. Webster, Jr. & M. Foschi (Eds.), Status generalization: New theory and research (pp. 450
474). Stanford, CA: Stanford University Press. Berger, J. (1992). Expectations, theory, and group processes: The Cooley-Mead award presentation address. Social Psychology Quarterly, 55, 3
11. Berger, J. (2007). The standardized experimental situation in expectation states research: Notes on history, uses and special features. In M. Webster, Jr. & J. Sell (Eds.), Laboratory experiments in the social sciences (1st ed., pp. 353
378). New York, NY: Elsevier. Berger, J., Cohen, B. P., & Zelditch, M., Jr. (1966). Status characteristics and expectation states. In J. Berger, M. Zelditch Jr., & B. Anderson (Eds.), Sociological theories in progress (Vol. 1, pp. 29
46). Boston, MA: Houghton Mifflin. Berger, J., Cohen, B. P., & Zelditch, M., Jr. (1972). Status characteristics and social interaction. American Sociological Review, 37, 241
255. Berger, J., & Conner, T. L. (1966). Performance expectations and behavior in small groups. Technical Report no. 18. Laboratory for Social Research, Stanford University. Berger, J., & Conner, T. L. (1969). Performance expectations and behavior in small groups. Acta Sociologica, 12, 186
198.

46

JOSEPH BERGER ET AL.

Berger, J., & Conner, T. L. (1974). Performance expectations and behavior in small groups: A revised formulation. In J. Berger, T. L. Conner, & M. H. Fisek (Eds.), Expectation  states theory: A theoretical research program (pp. 85
109). Cambridge, MA: Winthrop. Berger, J., Conner, T. L., & Fisek, M. H. (1974). Expectation states theory: A theoretical  research program. Cambridge, MA: Winthrop. Berger, J., Conner, T. L., & McKeown, W. L. (1969). Evaluations and the formation and maintenance of performance expectations. Human Relations, 22, 481
502. Berger, J., & Fisek, M. H. (1974). A generalization of the theory of status characteristics and  states. In J. Berger, T. L. Conner, & M. H. Fisek (Eds.), Expectation states expectation  theory: A theoretical research program (pp. 163
205). Cambridge, MA: Winthrop. Berger, J., & Fisek, M. H. (2006). Diffuse status characteristics and the spread of status value: A formal theory. The American Journal of Sociology, 111, 1038
1079. Berger, J., & Fisek, M. H. (2008). Immigration groups and the emergence of status inequality:  An application of spread of status value theory. Presented at the ISA World Forum of Sociology in Barcelona, September 5
9. Berger, J., & Fisek, M. H. (2013). The spread of status value: A theoretical extension. In  & E. J. Lawler (Eds.), Advances in group processes (Vol. 30, pp. 77
107). S. R. Thye Bingley, UK: Emerald Group. Berger, J., Fisek, M. H., & Norman, R. Z. (1989). The evolution of status expectations:  A theoretical extension. In J. Berger, M. Zelditch Jr., & B. Anderson (Eds.), Sociological theories in progress: New formulations (pp. 100
130). Newbury Park, CA: Sage. Berger, J., Fisek, M. H., Norman, R. Z., & Wagner, D. G. (1985). The formation of reward  expectations in status situations. In J. Berger & M. Zelditch Jr. (Eds.), Status, rewards, and influence: How expectations organize behavior (pp. 215
261). San Francisco, CA: Jossey-Bass. Berger, J., Fisek, M. H., Norman, R. Z., & Zelditch, M., Jr. (1977). Status characteristics and social interaction. New York, NY: Elsevier. Berger, J., Norman, R. Z., Balkwell, J., & Smith, R. F. (1992). Status inconsistency in task situations: A test of four status processing principles. American Sociological Review, 57, 843
855. Berger, J., Ridgeway, C. L., Fisek, M. H., & Norman, R. Z. (1998). The legitimation and delegitimation of power and prestige orders. American Sociological Review, 63, 379
405. Berger, J., Rosenholtz, S. J., & Zelditch, M., Jr. (1980). Status-organizing processes. Annual Review of Sociology, 6, 479
508. Berger, J., & Webster, M., Jr. (2006). Expectations, status, and behavior. In P. J. Burke (Ed.), Contemporary social psychological theories (pp. 268
300). Stanford, CA: Stanford University Press. Berger, J., Webster, M., Jr., Ridgeway, C. L., & Rosenholtz, S. J. (1986). Status cues, expectations, and behavior. In E. J. Lawler (Ed.), Advances in group processes (Vol. 3, pp. 1
22). Greenwich, CT: JAI Press. Berger, J., & Zelditch, M., Jr. (1993). Orienting strategies and theory growth. In J. Berger & M. Zelditch Jr. (Eds.), Theoretical research programs: Studies in the growth of theories (pp. 3
19). Stanford, CA: Stanford University Press. Berger, J., Zelditch, M., Jr., Anderson, B., & Cohen, B. P. (1972). Structural aspects of distributive justice: A status value formulation. In J. Berger, M. Zelditch Jr., & B. Anderson (Eds.), Sociological theories in progress (Vol. 2, pp. 119
146). Boston, MA: Houghton Mifflin.

Expectation States Theory: Growth, Opportunities and Challenges

47

Bianchi, A. (2004). Rejecting others’ influence: Negative sentiment and status in task groups. Sociological Perspectives, 47, 339
355. Bianchi, A. J., Kang, S. M., & Stewart, D. G. (2012). The organizational selection of status characteristics: Status evaluations in an open source community. Organization Science, 23, 341
354. Bianchi, A. J., & Lancianese, D. A. (2007). Accentuate the positive: Positive sentiments and status in task groups. Social Psychology Quarterly, 70, 7
26. Bienenstock, E. J., & Bianchi, A. J. (2004). Activating performance expectations and status differences through gift exchange: Experimental results. Social Psychology Quarterly, 67, 310
318. Bierhoff, H. W., Buck, E., & Klein, R. (1986). Social context and perceived justice. In H. W. Bierhoff, R. L. Cohen, & J. Greenberg (Eds.), Justice in social relations (pp. 165
185). New York, NY: Plenum Press. Bunderson, J. S. (2003). Recognizing and utilizing expertise in work groups: A status characteristics perspective. Administrative Science Quarterly, 48, 557
591. Cohen, B. P., Kiker, J. E., & Kruse, R. J. (1969). The use of closed circuit television in expectation experiments. Technical Report no. 29. Laboratory for Social Research, Stanford, University. Cohen, B. P., & Zhou, X. (1991). Status processes in enduring work groups. American Sociological Review, 56, 179
188. Cohen, E. G. (1970). Interracial interaction disability. In A new approach to applied research: Race and education (pp. 98
117). Columbus, OH: Charles E. Merrill. Cohen, E. G. (1982). Expectation states and interracial interaction in school settings. Annual Review of Sociology, 8, 209
235. Cohen, E. G. (1993). From theory to practice: The development of an applied research program. In J. Berger & M. Zelditch, Jr. (Eds.), Theoretical research programs: Studies in the growth of theories (pp. 385
415). Stanford, CA: Stanford University Press. Cohen, E. G., & Lotan, R. A. (Eds.). (1997). Working for equity in heterogeneous classrooms. New York, NY: Columbia University Teachers College Press. Conner, T. L. (1964). Three tasks for use in laboratory small-group experiments. Technical Report no. 11. Laboratory for Social Research, Stanford University. Conner, T. L. (1977). Performance expectations and the initiation of problem solving attempts. Journal of Mathematical Sociology, 5, 187
198. Conner, T. L. (1985). Response latencies, performance expectations, and interaction patterns. In J. Berger & M. Zelditch, Jr. (Eds.), Status, rewards and influence. San Francisco, CA: Jossey-Bass. Cook, K. S. (1975). Expectations, evaluations, and equity. American Sociological Review, 40, 372
388. Correll, S. J. (2001). Gender and the career choice process: The role of biased self-assessments. The American Journal of Sociology, 106, 1691
1730. Correll, S. J. (2004). Constraints into preferences: Gender, status, and emerging career aspirations. American Sociological Review, 69, 93
113. Correll, S. J., Benard, S., & Paik, I. (2007). Getting a job: Is there a motherhood penalty? The American Journal of Sociology, 112, 1297
1338. Dippong, J. (2012). The effects of scope condition based participant exclusion on experimental outcomes in expectation states research: A meta-analysis. Social Science Research, 41, 359
371.

48

JOSEPH BERGER ET AL.

Dovidio, J. F., Brown, C. E., Heltmann, K., Ellyson, S. L., & Keating, C. F. (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., & Mullen, B. (1990). Status, expectations, and behavior: A metaanalytic review and test of the theory. Personality and Social Psychology Bulletin, 16, 541
553. Driskell, J. E., Olmstead, B., & Salas, E. (1993). Task cues, dominance cues, and influence in task groups. Journal of Applied Psychology, 78, 51
60. Driskell, J. E., & Webster, M., Jr. (1997). Status and sentiment in task groups. In J. Szmatka, J. Skvoretz, & J. Berger (Eds.), Status, network, and organization (pp. 179
200). Stanford, CA: Stanford University Press. Entwisle, D. R., & Webster, M., Jr. (1974). Raising children’s expectations for their own performance: A classroom application. In J. Berger, T. L. Conner, & M. H. Fisek (Eds.),  Expectation states theory: A theoretical research program (pp. 211
243). Cambridge, MA: Winthrop. Fararo, T. J., & Skvoretz, J. (1988). Dynamics of the formation of stable dominance structures. In M. Webster, Jr. & M. Foschi (Eds.), Status generalization: New theory and research (pp. 327
350). Stanford, CA: Stanford University Press. Fisek, M. H. (1974). A model for the evolution of status structures in task-oriented discussion  groups. In J. Berger, T. L. Conner, & M. H. Fisek (Eds.), Expectation states theory:  A theoretical research program (pp. 53
84). Cambridge, MA: Winthrop. Fisek, M. H. (1991). Complex task structures and power and prestige orders. In E. J. Lawler,  B. Markovsky, C. Ridgeway, & H. A. Walker (Eds.), Advances in group processes (Vol. 8, pp. 115
134). Greenwich, CT: JAI Press. Fisek, M. H. (2009). Status cues, standards for competence, and graded characteristics. Paper  presented at the annual meeting of the American Sociological Association, August, San Francisco, CA. Fisek, M. H., & Berger, J. (1998). Sentiment and task performance expectations. In  J. Skvoretz & J. Szmatka (Eds.), Advances in group processes (Vol. 15, pp. 23
40). Greenwich, CT: JAI Press. Fisek, M. H., Berger, J., & Moore, J. C., Jr. (2002). Evaluations, enactment, and expectations.  Social Psychology Quarterly, 65, 329
345. Fisek, M. H., Berger, J., & Norman, R. Z. (1991). Participation in heterogeneous and homoge neous groups: A theoretical integration. The American Journal of Sociology, 97, 114
142. Fisek, M. H., Berger, J., & Norman, R. Z. (1995). Evaluations and the formation of expecta tions. The American Journal of Sociology, 101, 721
746. Fisek, M. H., Berger, J., & Norman, R. Z. (1997). Two issues in the assessment of the  adequacy of formal sociological models of human behavior. Social Science Research, 26, 153
169. Fisek, M. H., Berger, J., & Norman, R. Z. (2005). Status cues and the formation of expecta tions. Social Science Research, 34, 80
102. Fisek, M. H., & Hysom, S. J. (2008). Status characteristics and reward expectations: A test of  a theory of justice in two cultures. Social Science Research, 37, 769
786. Fisek, M. H., Norman, R. Z., & Nelson-Kilger, M. (1992). Status characteristics and expecta tion states theory: A priori model parameters and test. Journal of Mathematical Sociology, 16, 285
303.

Expectation States Theory: Growth, Opportunities and Challenges

49

Fisek, M. H., & Wagner, D. G. (2003). Reward expectations and allocative behaviors: A math ematical model. In S. R. Thye & J. Skvoretz (Eds.), Advances in group processes (Vol. 20, pp. 133
148). New York, NY: Elsevier/JAI Press. Foddy, M., & Riches, P. (2000). The impact of task and categorical cues on social influence: Fluency and ethnic accent as cues to competence in task groups. In S. R. Thye, E. J. Lawler, M. W. Macy, & H. A. Walker (Eds.), Advances in group processes (Vol. 17, pp. 103
130). Stamford, CT: JAI Press. Foschi, M. (1989). Status characteristics, standards, and attributions. In J. Berger, M. Zelditch, Jr., & B. Anderson (Eds.), Sociological theories in progress: New formulations (pp. 58
72). Newbury Park, CA: Sage. Foschi, M. (1996). Double standards in the evaluation of men and women. Social Psychology Quarterly, 59, 237
254. Foschi, M. (2000). Double standards for competence: Theory and research. Annual Review of Sociology, 26, 21
42. Foschi, M. (2009). Gender, performance level, and competence standards in task groups. Social Science Research, 38, 447
457. Foschi, M., Lai, L., & Sigerson, K. (1994). Gender and double standards in the assessment of job applicants. Social Psychology Quarterly, 57, 326
339. Foschi, M., Sigerson, K., Lai, L., & Foschi, R. (1990). A computerized setting for expectation states research. Annual West Coast Conference for Small Group Research, Portland, OR. Foschi, M., & Valenzuela, J. (2012). Who is the better applicant? Effects from gender, academic record, and type of decision. Social Science Research, 41, 949
964. Foschi, M., Warriner, G. K., & Hart, S. D. (1985). Standards, expectations, and interpersonal influence. Social Psychology Quarterly, 48, 108
117. Fox, J., & Moore, J. C., Jr. (1979). Status characteristics and expectation states: Fitting and testing a recent model. Social Psychology Quarterly, 42, 126
134. Freese, L. (1974). Conditions for status equality. Sociometry, 37, 147
188. Gallagher, T. J., Gregory, S. W., Bianchi, A. J., Hartung, P. J., & Harkness, S. (2005). An examination of the asymmetry in medical interviews using the expectation states approach. Social Psychology Quarterly, 68, 187
203. Gerber, G. L. (2001). Women and men police officers: Status, gender, and personality. Westport, CT: Praeger. Goar, C., & Sell, J. (2005). Using task definition to modify racial inequality within task groups. Sociological Quarterly, 46, 525
543. Gorman, E. H. (2006). Work uncertainty and the promotion of professional women: The case of law firm partnership. Social Forces, 85, 865
890. Gregory, S. W., Jr., & Kalkhoff, W. (2013). Comprehending the neurological substratum of paraverbal communications: The invention of splitspec systems technology. In D. D. Franks & J. H. Turner (Eds.), Handbook of neurosociology (pp. 369
384). New York, NY: Springer. Hogue, M., & Yoder, J. D. (2003). The role of status in producing depressed entitlement in women’s and men’s pay allocations. Psychology of Women Quarterly, 27, 197
208. Homans, G. C. (1961). Distributive justice. In Social behavior: Its elementary forms (pp. 241
268). New York, NY: Harcourt, Brace and World. Humphreys, P., & Berger, J. (1981). Theoretical consequences of the status characteristics formulation. The American Journal of Sociology, 86, 953
983.

50

JOSEPH BERGER ET AL.

Hysom, S. J. (2009). Status valued goal objects and performance expectations. Social Forces, 87, 1623
1648. Hysom, S. J., & Fisek, M. H. (2011). Situational determinants of reward allocation:  The equity-equality equilibrium model. Social Science Research, 40, 1263
1285. Jackson, L. A., Hunger, J. E., & Hodge, C. N. (1995). Physical attractiveness and intellectual competence: A meta-analytic review. Social Psychology Quarterly, 58, 108
122. Johansson, A. C., & Sell, J. (2004). Sources of legitimation and their effects of group routines: A theoretical analysis. In C. Johnson (Ed.), Research in the sociology of organizations (pp. 89
116). New York, NY: Elsevier. Kalkhoff, W. (2005). Collective validation in multi-actor task groups: The effects of status differentiation. Social Psychology Quarterly, 68, 57
74. Kalkhoff, W., & Barnum, C. (2000). The effects of status-organizing and social identity processes on patterns of social influence. Social Psychology Quarterly, 63, 95
115. Kalkhoff, W., Friedkin, N. E., & Johnsen, E. C. (2010). Status, networks, and opinions: A modular integration of two theories. In S. R. Thye & E. J. Lawler (Eds.), Advances in group processes (Vol. 27, pp. 1
38). New York, NY: JAI Press. Kalkhoff, W., & Thye, S. R. (2006). Expectation states theory and research: New observations from meta-analysis. Sociological Methods and Research, 35, 219
249. Kalkhoff, W., Younts, C. W., & Troyer, L. (2008). Facts and artifacts in research: The case of communication medium, gender, and influence. Social Science Research, 37, 1008
1021. Kalkhoff, W., Younts, C. W., & Troyer, L. (2011). Do second-order expectations transfer to new groups and tasks? An expectation states approach. Social Psychology Quarterly, 74, 267
290. Kuokkanen, M. (1993). An axiomatizaton of status characteristics theory. Social Science Research, 22, 391
414. Leahey, E. (2004). The role of status in evaluating research: The case of data editing. Social Science Research, 33, 521
537. Lockheed, M. E., & Hall, K. P. (1976). Conceptualizing sex as a status characteristic: Applications to leadership training strategies. Journal of Social Issues, 32, 111
124. Lovaglia, M., & Houser, J. A. (1996). Emotional reactions and status in groups. American Sociological Review, 61, 867
883. Lovaglia, M. J., Lucas, J. W., Houser, J. A., Thye, S. R., & Markovsky, B. (1998). Status processes and mental ability test scores. The American Journal of Sociology, 104, 195
228. Lucas, J. W. (2003). Status processes and the institutionalization of women as leaders. American Sociological Review, 68, 464
480. Lucas, J. W., & Phelan, J. C. (2012). Stigma and status: The interrelation of two theoretical perspectives. Social Psychology Quarterly, 25, 310
333. Markovsky, B., Smith, R. F., & Berger, J. (1984). Do status interventions persist? American Sociological Review, 49, 373
382. Melamed, D. (2011). Graded status characteristics and expectation states. In S. R. Thye & E. J. Lawler (Eds.), Advances in group processes (pp. 1
31). Cambridge, MA: Emerald. Melamed, D. (2012). Deriving equity from expectations: A cross-cultural evaluation. Social Science Research, 41, 170
181. Melamed, D. (2013a). Do magnitudes of difference on status characteristics matter for small group inequalities? Social Science Research, 42, 217
229.

Expectation States Theory: Growth, Opportunities and Challenges

51

Melamed, D. (2013b). Yes, magnitudes of difference on status characteristics do matter for small group inequalities. Social Science Research, 42, 496
498. Melamed, D., & Abromaviciute, J. (2013). The implications of social neuroscience for expectation states theories. Sociology Compass, 7, 255
264. Melamed, D., & Savage, S. V. (2013). Status, numbers, and influence. Social Forces, 91, 1085
1104. Mohr, P. B. (1986). Demeanor, status cue or performance? Social Psychology Quarterly, 49, 228
236. Moore, J. C., Jr. (1965). Development of the spatial judgment experimental task. Technical Report no. 15. Laboratory for Social Research, Stanford University. Moore, J. C., Jr. (1968). Status and influence in small group interactions. Sociometry, 31, 47
63. Moore, J. C., Jr. (1985). Role-enactment and self-identity. In J. Berger & M. Zelditch Jr. (Eds.), Status, rewards, and influence: How expectations organize behavior (pp. 262
316). San Francisco, CA: Jossey-Bass. Munroe, P. (2001). Creating a legitimated power and prestige order: The impact of statusconsistency and performance evaluations on expectations for competence and status. Unpublished Ph.D. Dissertation, Stanford University. Norman, R. Z., & Roberts, F. S. (1972). A measure of relative balance for social structures. In J. Berger, M. Zelditch, Jr., & B. Anderson (Eds.), Sociological theories in progress (Vol. 2, pp. 358
391). Boston, MA: Houghton Mifflin. Norman, R. Z., Smith, R. F., & Berger, J. (1988). Processing inconsistent status information. In M. Webster & M. Foschi (Eds.), Status generalization: New theory and research (pp. 169
187). Stanford, CA: Stanford University Press. Oldmeadow, J. (2007). Status generalization in context: The moderating role of groups. Journal of Experimental Social Psychology, 43, 273
279. Oldmeadow, J. A., Platow, M. J., & Foddy, M. (2005). Task-groups as self-categories: A social identity perspective on status generalization. Current Research in Social Psychology, 10, 265
282. Retrieved from http://www.uiowa.edu/∼grpproc/crisp/crisp.10.18.html Oldmeadow, J. A., Platow, M. J., Foddy, M., & Anderson, D. (2003). Self-categorization, status, and social influence. Social Psychology Quarterly, 66, 138
152. Prescott, W. S. (1986). Expectations states theory: When do interventions persist? Unpublished manuscript, Dartmouth College. Pugh, M. D., & Wahrman, R. (1983). Neutralizing sexism in mixed-sex groups: Do woman have to be better than men? The American Journal of Sociology, 88, 736
762. Randel, A. E., Chay-Hoon, L. E., & Earley, P. C. (2005). It’s not just about differences: An integration of role identity theory and status characteristics theory. In M. C. Thomas-Hunt (Ed.), Status and groups, research on managing groups and teams (Vol. 7, pp. 23
42). New York, NY: JAI Press. Rashotte, L. S. (2006). Controlling the status effects of gender. Paper presented at the International Society of Political Psychology Annual Meeting, Barcelona. Rashotte, L. S., & Webster, M., Jr. (2005). Gender status beliefs. Social Science Research, 34, 618
633. Ridgeway, C. L. (1991). The social construction of status value: Gender and other nominal characteristics. Social Forces, 70, 367
386. Ridgeway, C. L. (2000). The formation of status beliefs: Improving status construction theory. In E. J. Lawler, M. Macy, S. R. Thye, & H. A. Walker (Eds.), Advances in group processes (Vol. 17, pp. 77
102). Greenwich, CT: JAI Press.

52

JOSEPH BERGER ET AL.

Ridgeway, C. L. (2011). Framed by gender: How gender inequality persists in the modern world. New York, NY: Oxford University Press. Ridgeway, C. L. (2014). Why status matters for inequality. American Sociological Review, 79, 1
16. Ridgeway, C. L., Backor, K., Li, Y. E., Tinkler, J. E., & Erickson, K. G. (2009). How easily does a social difference become a status distinction: Gender matters. American Sociological Review, 74, 44
62. Ridgeway, C. L., & Balkwell, J. (1997). Group processes and the diffusion of status-value beliefs. Social Psychology Quarterly, 60, 14
31. Ridgeway, C. L., & Berger, J. (1986). Expectations, legitimation, and dominance behavior in task groups. American Sociological Review, 51, 603
617. Ridgeway, C. L., Berger, J., & Smith, R. F. (1985). Nonverbal cues and status: An expectation states approach. The American Journal of Sociology, 90, 955
978. Ridgeway, C. L., Boyle, E. H., Kuipers, K., & Robinson, D. (1998). How do status beliefs develop? The role of resources and interaction. American Sociological Review, 63, 331
350. Ridgeway, C. L., & Correll, S. J. (2004a). Unpacking the gender system: A theoretical perspective on cultural beliefs in social relations. Gender and Society, 18, 510
531. Ridgeway, C. L., & Correll, S. J. (2004b). Motherhood as a status characteristic. Journal of Social Issues, 60, 683
700. Ridgeway, C. L., & Correll, S. J. (2006). Consensus and the creation of status beliefs. Social Forces, 85, 431
454. Ridgeway, C. L., & Diekema, D. (1994). Are gender differences status differences? In C. L. Ridgeway (Ed.), Gender, interaction, and inequality (pp. 157
180). New York, NY: Springer-Verlag. Ridgeway, C. L., & Erickson, K. G. (2000). Creating and spreading status beliefs. The American Journal of Sociology, 106, 579
615. Ridgeway, C. L., & Johnson, C. (1990). What is the relationship between social emotional behavior and status in task groups? The American Journal of Sociology, 90, 1189
1212. Ridgeway, C. L., Johnson, C., & Diekema, D. (1994). External status, legitimacy, and compliance in male and female groups. Social Forces, 72, 1051
1077. Robinson, D. T., & Balkwell, J. W. (1995). Density, transitivity and diffuse status in taskoriented groups. Social Psychology Quarterly, 58, 241
254. Sauer, S. J., Thomas-Hunt, M. C., & Morris, P. A. (2010). Too good to be true? The unintended signaling effects of educational prestige on external expectations of team performance. Organization Science, 21, 1108
1120. Savage, I. R., & Webster, M., Jr. (1972). Source of evaluations reformulated and analyzed. In L. M. LeCam, J. Neyman, & E. L. Scott (Eds.), The sixth Berkeley symposium on mathematical statistics and probability (Vol. 4, pp. 317
327). Berkeley, CA: The University of California Press. Shelly, R. K. (1993). How sentiments organize interaction. In E. J. Lawler et al. (Eds.), Advances in group processes (Vol. 10, pp. 113
132). Greenwich, CT: JAI Press. Shelly, R. K. (1998). Some developments in expectation states theory: Graduated expectations? In E. J. Lawler (Ed.), Advances in group process (Vol. 15, pp. 41
57). Greenwich, CT: JAI Press.

Expectation States Theory: Growth, Opportunities and Challenges

53

Shelly, R. K. (2001). How performance expectations arise from sentiments. Social Psychology Quarterly, 64, 72
87. Shelly, R. K., & Munroe, P. (1999). Do women engage in less task behavior than men? Sociological Perspectives, 42, 49
67. Shelly, R. K., & Webster, M., Jr. (1997). How formal status, liking, and ability status structure interaction: Three theoretical principles and a test. Sociological Perspectives, 40, 81
107. Simpson, B., Willer, R., & Ridgeway, C. L. (2012). Status hierarchies and the organization of collective action. Sociological Theory, 30, 149
166. Skvoretz, J., & Fararo, T. J. (1996). Status and participation in task groups: A dynamic model. The American Journal of Sociology, 101, 1366
1414. Skvoretz, J., Webster, M., Jr., & Whitmeyer, J. M. (1999). Status orders in task discussion groups. In E. J. Lawler (Ed.), Advances in group processes (Vol. 16, pp. 199
218). Stanford, CT: JAI Press. Sobieszek, B. I. (1970). Multiple sources and the formation of performance expectations. Unpublished Ph.D. Dissertation, Stanford University. Stewart, P., & Moore, J. C., Jr. (1992). Wage disparities and performance expectations. Social Psychology Quarterly, 55, 78
85. Thomas-Hunt, M. C., & Phillips, K. W. (2004). When what you know is not enough: The effects of gender on expert’s influence within work groups. Personality and Social Psychology Bulletin, 30, 1585
1598. Thye, S. R. (2000). A status value theory of power in exchange networks. American Sociological Review, 65, 407
432. Thye, S. R. (2010). Status, power and mechanisms of micro-stratification: Theory and new experimental evidence. Unpublished manuscript. Thye, S. R., & Kalkhoff, W. (2009). Seeing the forest through the trees: An updated metaanalysis of expectation states research. Current Research in Social Psychology, 15, 1
14. Retrieved from http://www.uiowa.edu/∼grpproc/crisp/crisp15_1.pdf Thye, S., Willer, D., & Markovsky, B. (2006). From status to power: New models at the intersection of two theories. Social Forces, 84, 1471
1495. Troyer, L. (2000). MacSES v. 5.0. Unpublished software manual. Troyer, L. (2001). Effects of protocol differences on the study of status and social influence. Current Research in Social Psychology. Retrieved from http://www.uiowa.edu/∼grpproc Troyer, L., & Younts, C. W. (1997). Whose expectations matter? The relative power of firstorder and second-order expectations in determining social influence. The American Journal of Sociology, 103, 692
732. Troyer, L., Younts, C. W., & Kalkhoff, W. (2001). Clarifying the theory of second-order expectations: The correspondence between motives for interaction and actor orientation toward group interaction. Social Psychology Quarterly, 64, 128
145. Tuzlak, A., & Moore, J. C., Jr. (1984). Status, demeanor, and influence: An empirical assessment. Social Psychology Quarterly, 47, 178
183. Wagner, D. G. (1988). Status violations: Toward an expectation states theory of the social control of status deviance. In M. Webster, Jr. & M. Foschi (Eds.), Status generalization: New theory and research (pp. 110
122). Stanford, CA: Stanford University Press. Wagner, D. G. (1995). Gender differences in reward preferences: A status based account. Small Group Research, 26, 353
371.

54

JOSEPH BERGER ET AL.

Wagner, D. G. (2007). Symbolic interactionism and expectation states theory: Similarities and differences. Sociological Focus, 40, 121
137. Wagner, D. G., & Berger, J. (1985). Do sociological theories grow? The American Journal of Sociology, 90, 697
728. Wagner, D. G., & Berger, J. (1998). Gender and interpersonal task behaviors: Status expectation accounts. In J. Berger & M. Zelditch Jr. (Eds.), Status, power and legitimacy: Strategies and theories (pp. 229
261). New Brunswick, NJ: Transaction. Wagner, D. G., & Berger, J. (2002). Expectation states theory: An evolving research program. In J. Berger & M. Zelditch, Jr. (Eds.), New directions in contemporary sociological theory (pp. 41
76). Boulder, CO: Rowman and Littlefield Publishers. Wagner, D. G., Ford, R. S., & Ford, T. W. (1986). Can gender inequalities be reduced? American Sociological Review, 51, 47
61. Wagner, D. G., & Harris, R. O. (1995). A handbook for expectation states research. Unpublished manuscript. University at Albany (SUNY), Group Processes Laboratory, Department of Sociology. Wagner, D. G., Larshus, J., & Vogel, M. (2013). Instrumental and expressive attributions in task groups: An expectation states approach. Unpublished manuscript. University at Albany, SUNY. Walker, H. A., & Cohen, B. P. (1985). Scope statements: Imperatives for evaluating theory. American Sociological Review, 50, 288
301. Walker, L. S., Webster, M., Jr., & Bianchi, A. J. (2011). Testing the spread of status value theory. Social Science Research, 40, 1652
1663. Webster, M., Jr. (1969). Source of evaluations and expectations for performance. Sociometry, 32, 243
258. Webster, M., Jr. (1984). Social structures and the sense of justice. In E. J. Lawler & S. Bachrach (Eds.), Research in the sociology of organizations (Vol. 3, pp. 59
94). Greenwich, CT: JAI Press. Webster, M., Jr., & Driskell, J. E., Jr. (1978). Status generalization: A review and some new data. American Sociological Review, 43, 220
236. Webster, M., Jr., & Driskell, J. E., Jr. (1983). Beauty as status. The American Journal of Sociology, 89, 140
165. Webster, M., Jr., & Hysom, S. J. (1998). Creating status characteristics. American Sociological Review, 63, 351
378. Webster, M., Jr., Hysom, S. J., & Fullmer, E. M. (1998). Sexual orientation and occupation as status. In J. Skvoretz & J. Szmatka (Eds.), Advances in group processes (Vol. 15, pp. 1
21). New York, NY: JAI Press. Webster, M., Jr., & Rashotte, L. S. (2010). Behavior, competence, and status. Social Forces, 88, 1021
1050. Webster, M., Jr., Rashotte, L. S., & Whitmeyer, J. M. (2008). Theoretical and intuitive models. Social Science Research, 37, 4
17. Webster, M., Jr., & Smith, R. F. (1978). Justice and revolutionary coalitions: A test of two theories. The American Journal of Sociology, 84, 267
292. Webster, M., Jr., & Sobieszek, B. I. (1973). Conflicting sources of evaluations. Sociometry, 36, 550
560. Webster, M., Jr., & Sobieszek, B. I. (1974). Sources of self-evaluation: A formal theory of significant others and social influence. New York, NY: Wiley-Interscience.

Expectation States Theory: Growth, Opportunities and Challenges

55

Webster, M., Jr., & Whitmeyer, J. M. (1999). A theory of second-order expectations and behavior. Social Psychology Quarterly, 62, 17
31. Webster, M., Jr., & Whitmeyer, J. M. (2001). Applications of theories of group processes. Sociological Theory, 19, 250
270. Webster, M., Jr., & Whitmeyer, J. M. (2002). Modeling second-order expectations. Sociological Theory, 20, 306
327. Webster, M., Jr., Whitmeyer, J. M., & Rashotte, L. S. (2004). Status claims, performance expectations, and inequality in groups. Social Science Research, 33, 724
745. Whitmeyer, J. M. (2003). The mathematics of expectation states theory. Social Psychology Quarterly, 66, 238
253. Whitmeyer, J. M., Webster, M., Jr., & Rashotte, L. S. (2005). When status equals make status claims. Social Psychology Quarterly, 68, 179
186. Willard, D., & Strodtbeck, F. (1972). Latency of verbal response and participation in small groups. Sociometry, 35, 161
175. Willer, D., Lovaglia, M. J., & Markovsky, B. (1999). Part I: Power and influence: A theoretical bridge. In D. Willer (Ed.), Network exchange theory (pp. 229
247). Westport, CT: Praeger. Wood, W., & Karten, S. J. (1986). Sex differences in interaction style as a product of perceived differences in competence. Journal of Personality and Social Psychology, 50, 341
347. Yuchtman-Yaar, E., & Semyonov, M. (1979). Ethnic inequality in Israeli schools and sports: An expectation states approach. The American Journal of Sociology, 85, 576
590. Zelditch, M., Jr. (1980). How are inconsistencies in status and ability resolved? Social Forces, 58, 1025
1043. Zelditch, M., Jr. (1985). Three questions about status. In J. Berger & M. Zelditch, Jr. (Eds.), Status, rewards, and influence: How expectations organize behavior (pp. 73
107). San Francisco, CA: Jossey-Bass. Zelditch, M., Jr. (2013). Thirty years of advances in group processes: A review essay. In S. R. Thye & E. J. Lawler (Eds.), Advances in group processes (Vol. 30, pp. 1
19). Bingley, UK: Emerald Publishing Group Publishing.

THE DEVELOPMENT OF IDENTITY THEORY Jan E. Stets and Peter J. Burke ABSTRACT Purpose
The purpose of this chapter is to review the historical development of identity theory from 1988 to the present, and then outline some thoughts about future directions for the theory. Methodology/approach
The chapter discusses major advances in identity theory over the past 25 years such as the incorporation of the perceptual control system into the theory, the introduction of “resources” in which symbolic and sign meanings are important, new views of the social structure, the relevance of the situation in influencing the identity process, the idea of different bases of identities, broadening our understanding of multiple identities, studying identity change, and bringing in emotions into the theory. Findings
Throughout the review, empirical work is identified and briefly discussed that supports the major advances of the theory. Research limitations
The chapter suggests a number of ways that identity theory may be developed in the future such as examining negative or stigmatized identities. Additionally, there is a discussion as to ways in which the theory may be tied to other theoretical traditions such as affect control theory, exchange theory, and social identity theory.

Advances in Group Processes, Volume 31, 57
97 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0882-6145/doi:10.1108/S0882-614520140000031002

57

58

JAN E. STETS AND PETER J. BURKE

Social implications
Identity theory has had a number of applications to various areas in society, including understanding crime, education, race/ethnicity, gender, the family, and the environment. Originality/value of chapter
This is the most recent overview of identity theory over the past 25 years. It becomes clear to the reader that the theory offers a way of understanding the person as a cognitive, emotional, and behavioral agent who influences the structure of society but who is also influenced by the social structure. Keywords: Emotions; identity; resources; self; verification

IDENTITY THEORY IN 1988 By 1988, we had just left Indiana University (IU), but what remained with us was the intellectual influence of IU theorists Sheldon Stryker and David Heise. We arrived at Washington State University (WSU), where we became influenced by other scholars such as Viktor Gecas and Lee Freese. At WSU, identity theory began to really flourish. Stryker had developed many of the central ideas of structural symbolic interaction (SSI) as well as the main framework of identity theory (Stryker, [1980] 2002). So, we begin with early central ideas. In Stryker’s work, the core idea, taken from Mead, was that “society shapes self shapes social behavior.” This SSI idea gave causal priority to society on the grounds that individuals were enmeshed in networks in society from birth and could not survive outside of preexisting organized social relationships. Society was characterized as social structures comprising patterned behavior and interactions. Two levels of social structure were apparent. The first level included the networks in which people and their identities were embedded, for example, people in their families, classrooms, and work groups in which the parent, student, and worker identities emerged, respectively. Second was the larger bounding social structure of organizations and institutions, which influenced the probabilities that people with their identities entered into various networks. For example, persons with certain class backgrounds would not find easy access to a country club, and children in the inner city would not find access to better schools. Social behavior was conceptualized as role choice behavior, and the primary question was why persons chose one particular course of action among those open to them. For example, why would a person spend

The Development of Identity Theory

59

a Saturday with his children rather than on the golf course? The answer was that individuals were viewed as having multiple identities that might influence behavior, and some identities were more likely to be invoked than others. In the example above of the parent and golfer identities, the parent identity was invoked over the golfer identity. Drawing from Mead and James, the self is made up of many identities. People internalize the meanings that they apply to themselves when they are occupants of positions in the social structure such as father, student, carpenter, or golfer. Thus, people may have a father identity, student identity, carpenter identity, or golfer identity. The positions and the meanings and expectations attached to them come from a common culture that is shared with others. In this common culture, people understand what it means to be a father, student, carpenter, and golfer. Identities thus link persons to the social structure and to culture. Further, with many identities, there are many links: father to family, golfer to the country club, and student to the classroom. Thus, each person is tied in multiple ways to the social structure. In converse fashion, different parts of the social structure are linked through the nexus of identities held by a person, thus providing ties between disparate parts of the social structure. For example, Ralph, being a bus driver for the transit authority and a Parent Teacher’s Association (PTA) member, links the school system to the transit authority. The main thesis of the identity research program at this time was that role choices were a function of identities so conceptualized, and the many identities within the self were organized in a salience hierarchy reflecting the importance of hierarchy as a societal organizational principle (Stryker & Burke, 2000, p. 286). The salience of an identity is the probability that a particular identity will be activated across a variety of situations and thus influence the role choices made by the person. Identities that are more salient are more likely to be enacted or activated across situations. In our example of the parent/golfer above, a more salient parent identity would be more likely to be invoked ahead of the golfer identity. Salience, in turn, is a function of the commitment to the identity, where commitment is understood as the degree to which the person is tied to others in the social structure on the basis of the identity in question, considering both the number and strength of those ties. The greater the number of ties to others and the greater the strength of these ties to others, the stronger is the identity commitment. The parent identity in the above example may have had higher commitment than the golfer identity thus leading to its higher placement in the salience hierarchy than the golfer identity. Thus, commitment (structure) shapes identity salience (self) shapes

60

JAN E. STETS AND PETER J. BURKE

role choice (social behavior). This basic model was subject to numerous tests and applications that both strengthened and refined these ideas (Serpe, 1987; Stryker & Serpe, 1982, 1983). While this early work was being developed, another research line began to emerge that looked at the nature of identities and how they operate to produce behaviors expressing those identities. While more salient identities are likely to be invoked in any situation, at issue was the kind of behavior that individuals invoke given the identity they choose. The solution to understanding the link between identity and behavior was based on the traditional symbolic interaction idea that identities are self-meanings, and that self-meanings develop in the context of meanings of roles and counter roles (Burke, 1980; Burke & Tully, 1977). In the SSI view, behaviors, like identities, were characterized by meaning, and the link between identities and role behaviors existed in the meanings they shared (Burke & Reitzes, 1981). People who have identities with certain meanings choose behaviors that convey these same meanings. The large breakthrough came when it was discovered that it was possible to empirically and reliably measure meaning, including self-meanings (identities) and role behavior meanings using the semantic differential scale (Osgood, Suci, & Tannenbaum, 1957). The methodological innovation made possible a strong program of empirical research to continue to build the theory. Following the work of Osgood et al. (1957) Osgood, May, and Miron (1975), meaning was defined as an internal response to a stimulus that mediated between the stimulus and a behavioral response to that stimulus. This internal response interpreted the stimulus, giving it meaning on which the behavioral response was predicated. To measure this internal mediation response, Osgood and his colleagues devised the semantic differential, which captured meaning as a set of responses to a word or concept stimulus. The responses were captured by the respondents marking their responses on a series of bipolar adjective pairs such as strong
weak. Testing a large number of bipolar response scales and a large number of concept stimuli, Osgood found that the underlying structure of responses could be understood, in part, as consisting of three fundamental dimensions, which were labeled as evaluation (good
bad), potency (strong
weak), and activity (active
passive) (EPA), plus a number of additional dimensions. The fundamental dimensions (EPA) accounted for about 50% of the mediational response, while the remaining 50% of meaning took a large number of additional dimensions to be portrayed. In terms of matching identity meaning to behavior, given the variety of identities and behaviors across society and the vast array of meanings that

The Development of Identity Theory

61

might represent them, it was deemed necessary to discover the relevant dimensions for each identity. Rather than relying on the fundamental EPA dimensions, identity theory sought to distinguish the meaning of any one identity from the meanings of possible counter identities associated with the identity. This procedure, as suggested by Burke and Tully (Burke & Tully, 1977), used a discriminant function to find the relevant adjective pairs that best accomplished this discovery of meanings that distinguished identities from counter identities.1 In this way, a smaller number of relevant adjective pairs could be identified and used to measure the meaning of any particular identity. As Burke and Reitzes (1980) showed, this same procedure could be used to measure the meanings of behaviors that might be linked by common meanings to the identity in question. The semantic differential then began to be used to measure a large number of different identity meanings: gender identity (Burke, 1989b; Burke, Stets, & Pirog-Good, 1988; Burke & Tully, 1977), student identity (Burke & Reitzes, 1980), ethnic identity (White & Burke, 1987), academic identity (Burke & Hoelter, 1988), old age identity (Mutran & Burke, 1979a, 1979b), and body image identity (Stager & Burke, 1982). In all of these cases, the meanings contained in the identities were consistent with the meanings of the behavioral choices made by the individuals. Thus, by 1988, when the first Group Processes Conference was held, we had a good understanding of the nature of identities, the nature of commitment in the way individuals with these identities fit into the social structure, and how that influenced identity activation or enactment through its effect of the salience on the identities. We also knew the mechanism that linked identities and the behavior choices that were made: maintaining consistency in meaning. We turn now to examine how the theory has grown and changed over the last 25 years.

IDENTITY THEORY SINCE 1988 Perceptual Control System Since 1988, there have been a large number of important extensions to identity theory. Perhaps the most significant was the introduction of the perceptual control system into identity theory drawing on the work of Powers (1973). This was first presented in work that examined the origins of distress in identity disruption (Burke, 1991). The idea was that people

62

JAN E. STETS AND PETER J. BURKE

did not just act in ways that were consistent with their identities. Rather they used feedback from others (reflected appraisals) as well as their own direct appraisals to understand the meanings of the behaviors they were enacting. On the basis of this, they altered their behavior in order to make their perceptions of the meaning match the meanings in their identity standard. This was a very subversive idea, which ran counter to much thinking in the social and behavioral sciences. The tradition was that people controlled their behavior. The idea that people controlled their perceptions by engaging in whatever behavior worked to match perceptions of meaning to identity standard meanings was not commonly understood.2 Again, the perceptual control perspective emphasizes the idea that it is the meaning that is important not the behavior itself, and meaning is subject to social confirmation. By looking at others’ responses, one can confirm that one is adjusting the meaning of the situation in the desired manner. This led to the important idea that achieving a match between perceived meanings and the meanings held in the identity standard was identity verification, while failure to achieve this match was identity nonverification. The degree of mismatch or discrepancy between perceptions and the identity standard led to a similar degree of distress (an emotional reaction) and to behavior to correct the situation. From this idea followed a series of studies in which the meanings of the identities as well as the meanings of the reflected appraisals using the same scales were measured. This allowed the calculation of a discrepancy or difference between the reflected appraisals and the identity meanings reflecting the degree to which the identity was not verified. We then showed that the magnitude of this difference or discrepancy predicted changes in behavior over time as people worked to reduce the discrepancy and increase the level of identity verification. For example, Stets and Burke (2005a) showed that in marital interaction when the spousal identity is not verified (for either the husband or the wife), their level of efficacy slips, and they increase their control over their spouse in an attempt to regain the verification that was lost. A later study (Burke, 2006b) examined the leadership identity and leadership behavior in small task-oriented groups. Here it was shown that as the level of leadership behavior displayed in the group by a participant slipped below that implied by their leadership identity, the participant increased his or her leadership behavior in the next discussion. Similarly, if the behavior showed too much leadership for the level implied by their identity, the participant reduced the level of such behavior in the next discussion. In these cases, the change in behavior

The Development of Identity Theory

63

following a discrepancy between perceptions and the identity standard was an inverse linear function of the magnitude of the discrepancy. A positive discrepancy led to a reduction in the behavior meanings, while a negative discrepancy led to an increase in the behavior meanings. There was an unfortunate naming problem that occurred after the identity control system appeared in the literature. We and others began to talk about identity control theory. This was unfortunate because it began to drive a wedge between what Stryker and Burke (2000) called two strands of a single identity theory. Research and discussion began to appear that wanted to “test” the differences between the two strands. There are no incompatible differences. The perceptual control strand incorporates all aspects of the other strand and, in that sense, is more encompassing, but not different. In later work, we returned to using the language of “identity theory” to describe all of the work in this whole theoretical and research program because, in our minds, it is one unified theory, but with researchers testing different ideas in the theory (Burke & Stets, 2009; Stets & Serpe, 2013).

The Introduction of Resources Another big extension to identity theory was not only to deal with symbolic meanings (the traditional province of symbolic interactionism) but also to introduce sign meanings to understand the full range of resources and their control (Freese & Burke, 1994). For Mead (1934), the meaning or response to symbols is shared with others through convention or social agreement. When one uses a symbol, others respond to the symbol in the same way as the person responds to the symbol. In this sense, the symbol has the same meanings to the person who uses it as to the person to whom it is directed. Fundamental here is the notion of shared meanings. However, signs have meaning that is gained by direct and immediate experience rather than social agreement (Freese & Burke, 1994; Lindesmith & Strauss, 1956). For example, what it means to be crammed into an airline seat is not something that is understood through words or symbols. Its meaning is directly experienced. Moving one’s chair into the table at dinner is accomplished by direct experience
not too close, and not too distant, and at the proper orientation. Indeed, most of our interaction with objects in the environment is governed by signs. Certain feelings and perceptions guide our reactions to objects to bring them to be the way that they “should” be, without language or symbols. Keeping one’s car in

64

JAN E. STETS AND PETER J. BURKE

the middle of the lane, stacking up the fireplace wood, scraping off dishes after dinner are all examples of responding to nonlinguistic signs to maintain order. Including sign meanings in the set of meanings that form the identity standards and the meanings that are controlled in the situation greatly broadens the scope of identity theory. It recognizes that people are biosocial beings who exist and are maintained in the world. People are more than symbols or carriers of symbols; people also use and are sustained by signs. The introduction of sign meanings allows identity theory to deal with more than talk and ideas. It allows identity theory to deal with resources that had been the domain of exchange theory. In identity theory, resources are defined by their function: they are anything that sustains and supports individuals, groups, or interaction. The verification process of controlling perceived meanings in the situation to be in line with the meanings held in the identity standard could now include such things as students using money to purchase books as part of the student identity verification process. Resources and resource interactions could now take a central role alongside of symbolic interactions in understanding human behavior and creating and maintaining social structures. The verification process is understood in the same way that it always has been, but now includes both sign and symbolic meanings in the situation being brought into alignment with those meanings in the identity standard. Stets and Cast (2007), for example, showed that identity verification was facilitated by personal, interpersonal, and structural resources. Among the personal resources are beliefs about the individual including one’s worth and efficacy. They noted that positive and efficacious self-feelings help people be more persistent in the face of difficulties in verifying the self. Among the interpersonal resources, which arise out of relationships, are role-taking, trust, and liking, which have been shown to facilitate building and maintaining interpersonal relationships. Finally, structural resources include one’s education, occupational status, and income. Stets and Cast (2007) examined two identities of newly married couples who were studied at three points in time over two years. These identities were the person identity of sociability or friendliness and the role identity of spouse.3 The results showed that all three categories of resources facilitated the verification of both the person identity and the role identity for both men and women across all waves of the study. Thus, resources are important in the verification process that helps to sustain individuals (the person identity) and relationships (the spouse identity).

The Development of Identity Theory

65

Additionally, Stets and Cast found that identity verification facilitated the acquisition of more personal, interpersonal, and structural resources over time, perhaps because identities involve the acquisition and control of resources, and the verification of identities facilitates this acquisition and control. Thus, those whose identities were more strongly verified had more resources available to them in the future: a case of the rich getting richer, or the resource poor losing both the identity verification and future resources. In general, incorporating resources and signs into identity theory has broadened the scope of what it can cover and has opened up new ways to understand the embedding of identities in the social structure.

New Views of the Social Structure During the past 25 years, identity theory had added a new view of social structure in terms of resources. Resources can be divided into actual resources
those resources that are currently functioning to support persons, groups, or interaction (e.g., a table around which people interact, the chairs on which they sit, and the room in which the interaction takes place)
and potential resources
those resources that are not currently functioning as resources, but may function at a future time or after some transformation (e.g., crude oil transformed into gasoline that can be used to move people in an automobile to work). Recall that sign meanings are not necessarily shared but are understood by individuals experientially within any situation; they control actual resources in the situation. Symbolic meanings can be thought of as controlling the potential resources to be used at some time in the future (Freese & Burke, 1994). By controlling sign and symbolic meanings in the verification process, identities control actual and potential resources. Identity theory takes the view that social structure, including the stratification system and all of the institutional processes, involves the allocation of rights and responsibilities for controlling various actual and potential resources, which is the consequence of all the identities working to maintain verification. In this way, identities are intimately tied to the social structure because the operation of identities maintains the flow of resources that, in turn, maintains all of the groups, organizations, and individuals in society. People act to verify their identities. In doing so, in the face of distractions and disruptions, they enact the processes that define and maintain the social system.

66

JAN E. STETS AND PETER J. BURKE

Identity theory also has developed the conceptualization of the social structure by distinguishing large-scale (macro), intermediate (meso), and proximate (micro) structures that both contain and influence identities (Stryker, Serpe, & Hunt, 2005). Large-scale structures are features of the stratification system such as race/ethnicity, class, gender, and socioeconomic status. These structures serve as social boundaries that have consequences for individual life chances including the probability of entering particular networks of social relationships and having access to particular resources. They provide persons with a social identity through which they can identify with others based on sharing both the social location and the meanings associated with a given stratification characteristic. They can also be identified by others as having various rights, responsibilities, and access to resources. Intermediate social structures are more localized networks, for example, neighborhoods, associations, and organizations. They create social boundaries that increase or decrease the probability of particular kinds of social relationships forming. Proximate structures are those closest to interpersonal interactions such as families, departments within larger corporate or educational structures, or social clubs within schools (Serpe & Stryker, 2011; Stryker et al., 2005). Proximate structures provide persons with social relationships directly related to a specific role identity, and enactment of the role identity supports their participation within these structures. Additionally, proximate social structures provide access to others who have counter identities necessary for role enactment (Merolla, Serpe, Stryker, & Schultz, 2012). Social structures affect the likelihood that individuals within them will evolve particular kinds of identities, and this division of social structures into large-scale, intermediate, and proximate levels distinguishes structures on the basis of the way in which they influence identities. For example, social identities are more likely to develop in large-scale social structures, while role identities are more likely to develop in proximate social structures. Person identities, because they are always with people, should emerge across all the social structures: large scale, intermediate, and proximate. The effect of social structures on individual identities is underscored. While individuals develop their own meanings for identities, these identity meanings are influenced by the cultural expectations tied to the social structures within which they are embedded. Also important in understanding the functioning of identities is Serpe’s (1987) distinction between “open” and “closed” structures, wherein the behavioral choices of identities are facilitated or obstructed, respectively,

The Development of Identity Theory

67

foreshadowing Thoits’ (1992) distinction between voluntary and obligatory identities where there is more or less choice involved with the identity. While much interaction reproduces the existing social structures, individuals do have agency to change social structures. At issue is identifying the conditions under which there is pressure to conform to expectations and thus produce a stable social structure compared to modify expectations and thus change the social structure (Serpe & Stryker, 1987). In addition to understanding these different ways in which identities are embedded in the social structure and the consequences of that embedding, identity theory has also looked to understand the ways that identities are always embedded in situations when they are activated.

The Influence of the Situation In identity theory, we have begun to recognize the importance of the situation within which all interaction takes place, and we have brought the situation into the theory showing the relationship between situational meanings and identity meanings (Burke & Franzoi, 1988), the importance of the cognitive situational meanings (framing rules) and affective situational meanings (feeling rules) for identity processes (Stets & Carter, 2012), and the relevance of prior emotions in the situation for the functioning of identities (Stets & Osborn, 2008). Burke and Franzoi (1988) used experiential sampling methods to capture identity meanings and meanings of the situation in which the respondents found themselves at every one-and-a-half-hour intervals (on average) over a two-day period. Two identities (friend and student) occurred often enough in the data that the meanings of these identities and the meanings of the situations in which these identities were activated could be analyzed. Meanings were measured using semantic differential scales to capture the degree of evaluation, potency, and activity for both the identities and the situations. The results showed a strong effect of the evaluation of the situation on the evaluation of the identity, of the potency of the situation on the potency of the identity, and of the activity of the situation on the activity of the identity. Stets and Carter (2012) studied the moral identities of students in eight different situations, paying particular attention to the identity relevant meanings in the situation. To the extent that relevant dimensions of meaning in a situation correspond to dimensions of meaning in one’s identity standard, then we can refer to the meanings in the situation as strong or

68

JAN E. STETS AND PETER J. BURKE

potent for an identity, and the identity is relevant for the situation. For example, situations involving behaviors such as not allowing another student to copy one’s answers during an exam had a relatively low level of moral meaning and thus had lower moral potency than other situations pertaining to not allowing a friend to drive home drunk, which had higher moral potency (Stets & Carter, 2012). Consequently, the moral identity would be less likely to be activated in the former than the latter situation. Stets and Carter argued that potency is a function of cognitive and affective aspects of situations. The cognitive aspect is the framing rules or the interpretation made on a situation; the affective aspect is the feeling rules or how individuals should feel in a situation given the interpretation made by the framing rules. The researchers found that the moral identity predicted the choice to behave morally more strongly when the moral potency of the situation was high than when it was low. They also examined the degree to which respondents felt guilt and shame when they thought others did not see them as acting in the situation as their moral identity would indicate (i.e., there was a discrepancy between the reflected appraisals and their moral identity standard). The reactions of guilt and shame to the discrepancy were much stronger for situations that were morally potent than less morally potent situations. Thus, some of the variation in moral behavior for persons who have the same moral identity is due to situational influences, in this case, the degree to which the situation contained moral meanings relevant to the identity. Another way in which the situation may influence the operation of identities has to do with the fact that when we move from one situation to the next, the feelings generated in one situation may influence identity processes in the new situation. Stets and Osborn (2008) examined the role of people’s feelings across encounters by studying their reactions to feedback across three separate, but distinct, tasks in a laboratory study that simulated a work situation and the worker identity. After the participants (workers) performed each task, they received feedback that exceeded or fell short of what they expected to receive. Across the three tasks, the feedback oscillated from more (an “over-reward”) or less (an “under-reward”) than what they expected (or vice versa). What they found was that positive feelings associated with an initial over-reward (on the first task) persisted beyond the point of their initial arousal. The positive feelings continued to be experienced following feedback on the second and third task, even when the participants received an under-reward for their performance. The positive feelings tempered the negative feelings associated with subsequent

The Development of Identity Theory

69

under-rewards. However, negative feelings did not show the same persistence effects as positive feelings. The negative feelings did not continue beyond the point of their initial arousal unless individuals continued to receive an under-reward on subsequent tasks. In general, the findings showed the carryover effects of positive feelings within an interaction, but not negative feelings. Interestingly, the maintenance of positive feelings appears to act as a buffer, serving to soften the blow of later unexpected outcomes. Thus, emotions may influence interactions beyond their initial encounter to influence feelings in subsequent encounters.

New Bases of Identities While SSI initially thought of identities in terms of role positions in the social structure, such as mother, spouse, or teacher, thus identifying role identities, there is another class of structural positions in society that is important. There are groups and categories to which one belongs such as membership in a local church or the PTA on the one hand, as well as membership in broad social categories such as race, class, and gender on the other hand. These are social identities, and we have come to see the importance of extending identity theory to include them (Stets & Burke, 2000). We have demonstrated that the theory applies to these identities as well, though there are differences between role and group identities that are important. Having a particular social identity means being at one with the members of a particular group or category, being like the others and seeing things from the group’s perspective. In contrast, having a particular role identity means fulfilling the expectations of the role, coordinating with role partners, and manipulating the environment to control resources for which the role has responsibility. In group-based identities, the actor need not interact with other group members, but in role-based identities, some form of interaction and negotiation is usually involved. For group-based identities, similar actions and perceptions create a bond and a group forms. For rolebased identities, relations are reciprocal rather than parallel, there are differences in perspectives, and interaction and negotiation creates microsocial structures within groups. Finally, social identities based on major social structural divisions, such as race, class, and gender, are always with the person across situations, and society makes them almost always relevant. The result is that these social identities have higher salience than

70

JAN E. STETS AND PETER J. BURKE

other social identities and even many role identities, which are not always with the person across situations. Verification is at the heart of both role-based and group or categorybased identities, and while the verification process for each is the same, that is, matching perceptions of self-relevant meanings in the situation to the meanings of identity standard, the process and consequences of each are different. Verifying a role-based identity means engaging in the behavioral requirements of the role, that is, enacting behavior consistent with the role, and others responding appropriately to the behavioral enactment. Verifying a group-based identity means acting like others in the group and gaining acceptance by other group members that one is like them. While verifying a role identity makes one different from others in counter roles, verifying a social identity makes one similar to others in the group or category. The social identity of gender was one of the first identities that was measured and studied (Burke, 2006a; Burke & Cast, 1997; Burke et al., 1988; Burke & Tully, 1977; Stets & Burke, 1996).4 We found that people produced behavior with the same meanings as their identity standard, and they responded with negative emotions and attempts to correct the situation when the identity was not verified. In addition, we have studied the ethnic identity (White & Burke, 1987) and age-based identity (Mutran & Burke, 1979a, 1979b). Recently, we have extended the bases of identities and have shown that identity theory applies to person identities, that is, identities based on the person as a unique biosocial individual (Stets, 1995; Stets & Biga, 2003; Stets & Burke, 1994; Stets & Carter, 2011, 2012). Person identity standards include the meanings that set the person apart from others as a unique individual. These meanings are not attached to roles or groups, but are part of how individuals define themselves. They are always with the person and are relevant in most situations across groups and roles. Because of this, person identities are thought to have higher salience and commitment than other identities, and in some ways act as master identities influencing other role or social identities that persons take on. Since there are many dimensions of meaning that may be relevant to individuals that set them apart from others, it is easier to study person identities by focusing on just one or two dimensions at a time. For example, some people are more controlling than others and this is a characteristic that people want to maintain at the level they feel is appropriate for them (Stets & Burke, 1994, 1996). More recently, attention has been paid to the moral identity, a person identity having to do with the level of morality that one holds for oneself

The Development of Identity Theory

71

(Stets & Carter, 2011, 2012). This research measured the moral identity along the combined dimensions of “caring” and “fairness” and showed that persons with a higher moral identity were less likely to cheat in a laboratory study. Reflected appraisals with respect to the moral identity were also measured using the same scales, and, as mentioned earlier, greater discrepancy between the identity standard and the reflected appraisals (whether the appraisals were too high or too low) led to a higher level of the negative moral emotions of guilt and shame. Thus, like role and social identities, person identities, too, are verified by reflected appraisals matching the meanings held in the identity standard, and their lack of verification leads to negative emotions. While identity theory looks at the verification process of all identities in the same way, that is, that reflected appraisals are brought to match the meanings of the identity standard, we have suggested that the consequences of verifying identities with different bases are different. The argument builds upon the self-esteem theory put forth and tested by Cast and Burke (2002) that self-esteem is a function of the verification of identities. We have argued that there are three dimensions of self-esteem that have been recognized in the literature, self-worth, self-efficacy, and authenticity, and that each is a function of the verification of identities formed on each of the identity bases discussed earlier (Burke & Stets, 2009). Because of its emphasis on performance, the verification of role identities leads to an increase in self-efficacy. Because the verification of social identities is associated with being part of a group of similar others, and therefore accepted by them as a member, verification of social identities leads to an increase in self-worth. Finally, because person identities are a core part of who one is as a person, the verification of person identities leads to an increase in feelings of self-authenticity. We examined the consequences of verification of the gender social identity, the student role identity, and the moral person identity and confirmed this pattern (Stets & Burke, 2013a). Thus, while the verification process is the same across all identities and identity bases, the self-esteem outcomes of that verification are different for identities of different bases. Finally, we point out that in any situation, several identities from different bases may all be active at the same time. People with their person identities may be in a role, which is part of a group. People may find themselves like their role partner as members of the same group. But, they are different because of the different internalized role requirement each plays out as well as each being uniquely different because of the different person identities each has.

72

JAN E. STETS AND PETER J. BURKE

A New Conceptualization of Multiple Identities Prior to 1988, multiple identities were understood in role identity terms, and they were conceptualized as rank ordered within the self, given all the identities that individuals might claim (Stryker, [1980] 2002). If an identity was ranked higher in relative salience, it was more likely to be activated across situations than an identity ranked lower in relative salience. Additionally, there was an analysis of how individuals could hold multiple role identities, but rather than them causing conflict and distress, they could provide meaning and direction in people’s lives (Thoits, 1983, 1986). After 1988, we began to broaden our understanding of multiple identities by considering all the person, role, and social identities: (1) within the person and (2) across persons within a situation. Within the perceptual control system and borrowing from Powers (1973), identities could be understood as forming a hierarchical control system composed of an interlocking set of individual control systems at multiple levels (Burke, 1997; Stets & Harrod, 2004; Tsushima & Burke, 1999). Multiple identities could be at the same level in which the identity standards of each would be set or controlled by a higher level within the system. Multiple identities also could exist at different levels, where one identity was higher than the other within the system, and where the output of the higher identity was the standard for the lower identity. Each of these arrangements has different implications. Further, when we discuss these arrangements across persons, one issue is how individuals coordinate their own identities so that all identities can be verified in the situation. Multiple Identities within Persons Here we discuss identities at the same level rather than identities at different levels because there is empirical research on the former but not the latter. Thus, for identities at the same level, each of the identities has its own perceptual input, standard, and output, though the outputs of the two identities must be combined in some manner as there is only one person acting. If the two identity standards have no overlap in meanings, the two identities can operate independently of each other. The output (behavior) meanings of one identity have no implication in the situation for the output meanings of the other identity. For example, claiming the athlete identity may have no meanings in common with claiming the identity of singer. However, the meanings in the two identity standards may have some overlap or commonality such that the output meanings of one identity may overlap with the output meanings of the other identity, for example,

The Development of Identity Theory

73

claiming the identities of assertive and masculine. In this case, each identity supports the other because the meanings of one align with the meanings of the other. Still yet, the output meanings of one identity could conflict with the output meanings of the other identity, for example, claiming the identities of assertive and feminine. The behavior that verifies one identity may not verify the other identity. The resolution to this conflict would be for the standards of both identities to change so that the meanings of each come into agreement. The standard of one identity may change more than the other, and it is likely that the identity with the higher salience and/or higher commitment will change less because of the stronger ties to others implied by having higher salience or commitment. To maintain such ties, the identity cannot change much without disturbing the whole network. A study of leadership within a group examined the relationship between two identities (the task identity and socioemotional identity) for individuals in task-oriented groups (Burke, 2003). This study showed that the two leadership identities within persons had nothing significant in common. Each operated independently of the other to influence task and socioemotional behavior, and each was shown to independently bring about changes in leadership behavior to increase identity verification. However, it was interesting that disturbances to the performance of one identity were positively correlated with the disturbances to the performance of the other. If people were pushed by situational exigencies to do too much (or too little) task leadership (relative to their identity standard), then those same exigencies pushed their socioemotional leadership to be too high (or too low). Another study on multiple identities within the person examined married people’s gender identity and mastery identity and assessed how their meanings on each identity influenced attempts to control their spouse (Stets, 1995). The findings revealed an overlap in the meanings of the gender identity and the mastery identity with masculine meanings imparting higher levels of mastery, but there was no influence of mastery identity meanings on gender identity meanings. Both gender identity (but not gender) and the mastery identity influenced attempts to control the spouse. Lower levels of the mastery identity and higher levels of the masculine gender identity both increased attempts to control the spouse. While masculinity is consistent with controlling others, it was suggested that those with low mastery might enact controlling behavior to compensate for their perceived lack of control over their environment. In still another study, the researchers examined how social status influenced the verification of three identities (worker, friend, and academic identity) that were held by a sample of people (Stets & Harrod, 2004).

74

JAN E. STETS AND PETER J. BURKE

The meanings in the three identities had little in common, so each of the identities operated quite independently. Yet the verification of each was in part a function of the location of the individual in the status structure of society: persons with higher status (as indicated by age and education) were better able to verify all of the different identities they held. Multiple Identities across Persons How do people coordinate with others in a situation so that everyone’s relevant identities can be verified? While it is possible that the meanings being controlled by the identities of one person are independent of other’s identities in a situation (e.g., what one does has little influence on what another does and vice versa), this is unlikely. More likely, people are activating identities relevant to the situation (and relevant to the identities of others also involved in the situation), and the concern is to control relevant meanings in either a cooperative or a competitive fashion. With multiple people in the situation, and with each trying to verify their identities, the actions of one person may be a disturbance to others and make verification more or less difficult. Several studies have examined the relationships among multiple identities within a situation as individuals attempt to both manifest and verify their identities. In one study, the researchers studied the leadership identities of each of four participants in small task-oriented groups who were working together to accomplish a series of four discussion tasks (Riley & Burke, 1995). The level of the leadership identity of each of the participants as well as the level of leadership behavior that was enacted by each participant was measured. The findings revealed that the leadership identity, in general, predicted the level of leadership behavior, but there were still instances when some participants engaged in more than the expected level of leadership (given the level of their leadership identity), while others engaged in less than their expected level. The findings also revealed that those who engaged in too much or too little leadership behavior for their identities in any of the four tasks were less satisfied with their role in the group, and they acted to increase or decrease their leadership behavior in the next task (if it was less or more than expected given their leadership identities). Participants thus were managing their leadership behavior to both match the meanings in their identity standard and match the coordinated behaviors of others in the task-oriented group to accomplish the task. Perhaps amazingly, they were generally able to do this. In another study, the researchers examined the relationship between the spousal identities of newly married individuals over a period of three years

The Development of Identity Theory

75

(Burke & Stets, 1999). They found that verification of the spousal identity of each partner had the consequence of increasing (1) positive feelings of love for the spouse, (2) trust in the spouse, and (3) a sense of “we-ness” or a strong bond between the spouses. Nonverification of the spousal identity moved the relationship in the opposite direction. While the degree of verification of the spousal identity, like all identities, is subject to exogenous disturbances, the general movement is increasingly stronger bonds between spouses in what we called “mutual verification” contexts, that is, a context in which each spouse verifies the identity of the other spouse at the same time verifying their own spousal identity. Alternatively, lack of verification impels movement toward reduced love, trust, and weaker bonds, leading to marital dissolution (Cast & Burke, 2002). In yet another study examining the relationship between the spousal identities of newly married individuals over a period of three years, the researchers looked at the impact of the reflected appraisals of each spouse on the spousal identity of the other partner as well as the self over time (Cast, Stets, & Burke, 1999). When spouses interact within the family, each person’s spousal identity must complement the spousal identity of the other in order to avoid conflict. In this way, each person’s identity will influence the other’s identity in terms of taking on the meanings of the spousal identity attributed by the partner. The researchers hypothesized that the degree to which each identity would influence the other would be a function of the relative status of the husband and wife, with the spousal identity of the higher-status partner having a stronger influence on the meanings of the spousal identity of the lower-status partner. Indeed, this is what they found when status was measured as the level of education and income of each of the partners. Further, it did not make a difference if the higherstatus partner was the husband or the wife. If the status of each partner was relatively equal, the influence on the spousal identity was equal in both directions
husband to wife and wife to husband. Looking at the way in which multiple identities in a situation influence each other has brought us to consider the question of identity change more generally.

Identity Change Identity change occurs when the meanings in one’s identity shift over time. Since 1988, more theorizing and empirical work has emerged on identity change. One of the things we know is that identity change is ongoing but generally very gradual. Because the perceptual control system is a dynamic

76

JAN E. STETS AND PETER J. BURKE

model, everything, including the identity standard, is always changing, but the changes are generally small and slow, so that one is mostly aware of the stability. Individuals may not find their identity as different from yesterday, last week, or last month. It is only when considering a longer period of time ranging from months to years that they may see a difference. For example, in the study above on the relationship between status and the degree of influence on the meanings of the partner’s spousal identity, the spouse identity of the lower-status spouse changed in the direction defined by the higher-status spouse through reflected appraisals. However, this change occurred slowly over a three-year period, but was cumulative and significant (Cast et al., 1999). There are three ways in which identities may change (Burke, 2006a). Two of these changes we have already discussed. The first occurs when identity standard meanings and behavior meanings conflict, causing a change in both. When behavior meanings do not match identity standard meanings, reflected appraisals (the meanings that others are attributing to the self in the situation) may either exceed or fall short of one’s identity standard meanings. Research has studied whether identity change is more likely to occur when reflected appraisals meanings are higher or lower than identity standard meanings (Cast & Cantwell, 2007). They found that when the reflected appraisals do not match the identity standard, peoples’ identities slowly change in the direction of the discrepant reflected appraisals, but individuals also successfully acted to change the reflected appraisals over time. Both processes occurred simultaneously. However, it did not matter if the reflected appraisals were too high relative to the identity standard or if they were too low; the amount of change in the identity was the same. If situational changes persist and people’s meanings of themselves in those situations are unable to adjust to match their identity meanings, their identity meanings may slowly change. For example, one study tracked the changes in the gender identity meanings of newly married couples over the course of a year upon the birth of their first child (Burke & Cast, 1997). The birth of a child is a change in the situation that is generally irreversible. Becoming parents tends to move individuals to more traditional “gendered” meanings of parenting, and this was clearly the case in this study. The researchers found that in order to successfully accommodate the situational change of a newborn, the gender identity of husbands became somewhat more masculine and the gender identity of the wives became somewhat more feminine. Essentially, changes in the meanings in the situation that cannot be counteracted through the verification process will produce changes in the identity standard meanings in the direction of the

The Development of Identity Theory

77

situation meanings so that the identity can be verified. However, this change is very slow, occurring over weeks, months, or even years. In addition to changes in the situation bringing about changes in identities, the existence of multiple identities that conflict is another source of change. For example, a person who is a minister and has an identity that stipulates meanings of gentleness may also have an identity of masculinity that encourages meanings of toughness. To work together toward the same goal, the meanings in one or both of these identities will change. As mentioned earlier, the less salient and committed identities may be more likely to change or to change more than identities that have higher salience or commitment. Another form of identity change not involving changes in the meaning of the standard centers on changes in the salience of identities. For example, changes in a situation can cause a shift in one’s identity as when individuals enter environments that provide few possibilities for choosing which identities to enact. In other words, the social structure is “closed” rather than “open” (Serpe, 1987; Serpe & Stryker, 1987, 1993). An open structure is one that allows more choice in how to behave compared to a “closed” structure in which individuals have little choice in the way they enact an identity. An example of empirical research on this was a study of five identities held by newly enrolled college students who were followed over the first four months at school (fall term) (Serpe, 1987). The relative salience (probability of enacting the identity across situations) of the coursework, extracurricular, athletic/recreational, personal involvement, and dating identities were measured along with their commitment (ties to others in the social structure based on the identity) at three points in time. While there was stability in these identities over time, there was change in the level of salience of the different identities as the students came into a new (college) environment. The degree of change in salience was, in part, a function of the degree of “openness” of the structure in which the identity was embedded. Thus, the coursework identity, which allowed the least choice, showed the least amount of change in the level of salience as a function of commitment. The athletic/recreational and dating identities were seen as embedded in open structures and thus allowed the most individual choice. Here, the most change occurred in the level of salience as a function of the level of commitment to the identity. In general, identities can and do change both in the meanings that define the identity and in the level of salience of the identity. If a person has difficulty in verifying an identity over time, the identity is likely to change slowly over time, while if the reflected appraisals affirm the identity, it is

78

JAN E. STETS AND PETER J. BURKE

likely to remain stable over time. Similarly, if the social structure in which the identity is embedded allows more freedom of choice, the salience of the identity can change more easily as a function of commitment or the number of people one is connected to given the identity.

The Introduction of Emotions Since 1988, more serious attention has been given to the role of emotion in identity theory beginning with the idea that negative arousal (distress) was experienced with individuals’ identities were not verified in a situation (Burke, 1991). This idea provided an important insight into how emotions emerged within the self and jump-started research on emotions in identity theory over the next 25 years. Essentially, when individuals get support for their identity (Stryker, 2004), or when others in a situation see them in the same way that they see themselves given their identity claim (Burke & Stets, 2009), they will feel positive emotions. In turn, the identity may increase in salience and commitment. Alternatively, the lack of support or shared view as to who one is in the situation generates negative emotions. Correspondingly, the identity may decrease in salience and commitment. The emotional outcomes of the identity verification process have been examined in a longitudinal survey study that followed newly married couples during the first two years of marriage (Burke & Harrod, 2005; Burke & Stets, 1999), and in a series of studies simulating the worker identity in the laboratory (Stets, 2003, 2004, 2005; Stets & Asencio, 2008; Stets & Osborn, 2008). In all these studies, researchers found that when individuals thought that others saw them as failing to meet their identity standard, they experienced negative emotions. However, when they thought that others saw them as exceeding their identity standard, the longitudinal survey found that individuals reported negative feelings even though others’ evaluations were more positive than their own evaluations (Burke & Harrod, 2005). The laboratory studies found individuals reported positive feelings when others saw them as exceeding their identity standards. Identity theory predicts a cognitive consistency process to individuals’ emotional reactions, and the longitudinal survey supported this: people seek evaluations that match their self-views and avoid evaluations that do not match their self-views. However, the laboratory findings were suggestive of a self-enhancement process: people seek positive evaluations and avoid negative evaluations.

The Development of Identity Theory

79

Recently, it has been argued that the cognitive consistency effect may not have emerged in the laboratory studies because at least two important factors were not measured: the reflected appraisal process (how people think that others see them in the situation) and the relevance of the situational meanings for the identity (Stets & Burke, 2014). Researchers used a large data set derived from seven studies that included both a survey and laboratory component to address the emotional responses that occur when identities are not verified (Stets & Burke, 2014). They examined whether individuals showed an enhancement response (they feel good) or consistency response (they feel bad) to identity nonverification in a positive direction (the meanings in the reflected appraisals are more positive than the meanings of the identity standard). They included a measure of both the reflected appraisals and the relevance of the situational meanings for the identity. The results showed that when reflected appraisals and situational meanings were taken into account, there was more evidence for a consistency effect (negative emotions) than an enhancement effect (positive emotions). This helps put prior research into perspective and identifies some of the measurement issues that can make it difficult to distinguish between consistency and enhancement effects. Early on, it was hypothesized that more frequent nonverifying feedback would result in more intense negative emotions (Burke, 1991). The more the individuals receive feedback that others see them differently than how they see themselves, the more they will be unable to initiate or sustain whatever they are doing, and the more distressful their emotional reaction. Contrary to this, findings from the worker identity laboratory studies discussed earlier revealed that negative emotions become less rather than more intense (Stets, 2003, 2005). While a stronger negative response to repeated identity nonverification would indicate that individuals were resisting how others saw them, a weaker negative response would indicate that individuals were modifying their self-views in the direction of others’ views. This is identity change. It was also hypothesized that nonverification from significant others (family and friends) compared to nonsignificant others (strangers and acquaintances) would bring about more intense negative emotions (Burke, 1991). In interaction with a close other, each is likely to verify the identity of the other as they verify their own identity, resulting in a “mutually verifying” relationship. When such a relationship is disrupted, it could be experienced as particularly distressful and intense. This was tested in two ways. Using data from the General Social Survey (GSS), researchers examined whether interaction in the family, comprising significant others,

80

JAN E. STETS AND PETER J. BURKE

resulted in more negative emotions than interaction at work, comprising nonsignificant others (Stets & Tsushima, 2001). Though identity nonverification was not directly tested either at home or at work, the analysis revealed that more intense anger was reported in the family than at work. In a study that extended the worker identity studies discussed earlier, some participants had an opportunity to get to know their coworker for 10 minutes before the study began (this was the “familiar” condition and was a proxy for significant others) compared to not being given this opportunity (the “unfamiliar” condition and a proxy for nonsignificant others) (Stets, 2005). The results showed that familiarity did result in more negative emotions from identity nonverification, although it occurred only when the nonverification occurred once compared to more than once during the study. Because of the limitations in the above two studies (either there was no direct test of the verification process or there was no direct measure of significant others), more empirical work is needed. More recently, an analysis on the source of nonverifying feedback has been expanded to consider the different emotions that may emerge depending on who is responsible for the nonverification: either the person or the other in the situation (Stets & Burke, 2005b). Thus, the attribution process is brought into the theory. For example, when individuals are responsible for their own identity nonverification, they may experience feelings such as embarrassment or shame. Alternatively, when others are responsible for the nonverification, individuals may experience feelings such as annoyance or hostility. While embarrassment or shame are negative feelings directed inward, annoyance or hostility are negative feelings directed outward. Another expansion on the source of nonverifying feedback adds a consideration of the status and power of the nonverifying other relative to the person seeking identity verification (Stets & Burke, 2005b). Here, one’s position in the social structure is brought into the theory. For example, when the self rather than others is responsible for nonverification in a situation, the person may feel shame when others in the situation have higher status than the person, embarrassment when others are of equal status to the person, and discomfort when others have lower status. When others rather than the person are responsible for the person’s nonverification in a situation, the person may feel fear when others in the situation have higher power than the person, anger when others are of equal power, and rage when others have lower power. We do know that those with higher status will be more likely to experience identity verification than those with lower status because they are

The Development of Identity Theory

81

more influential in getting others to confirm their self-views (Cast et al., 1999). Because identity verification produces positive feelings, higher-status people will be more likely to enjoy positive feelings and less likely to experience negative feelings than lower-status people. Two studies support this idea. Using GSS data, the relative status of identities in the home and at work was studied (Stets & Tsushima, 2001). In the home, the parent identity has the highest status, the child identity has the lowest status, and spouses, interacting with each other, have equal status. At work, the employer identity has the highest status, the employee identity has the lowest status, and coworkers, interacting with each other, have equal status. Consistent with the above, those with lower-status identities either at home or at work were more likely to report more intense anger. Further, those with lower-status identities were more likely to report their anger lasting a long time. In another study of the newly married couples, the higher-status person (higher education, occupation, and race) was more likely to have his or her spousal identity verified (compared to the lower-status person), and was less likely to report anger, depression, and distress for identity nonverification than lower-status persons (Burke, 2008). Researchers in identity theory have begun to study specific emotions. Early research studied jealousy and anger in the home and at work (Ellestad & Stets, 1998; Stets & Tsushima, 2001). More recently, moral emotions such as anger, empathy, guilt, and shame have been examined (Stets, 2011; Stets & Carter, 2011, 2012; Stets, Carter, Harrod, Cerven, & Abrutyn, 2008). Like other emotions, moral emotions emerge from the nonverification process. Emotions have been examined not only as an outcome of the identity process but also as a resource to be used for identity verification. Positive emotions can be a resource, regulating the negative feelings that emerge when people experience identity nonverification. This was shown in the Stets and Osborn (2008) research discussed earlier. The positive feelings continued to be experienced following feedback on the second and third task, even when the participants received feedback that fell short of their expectations on those tasks. The positive emotions associated with positive feedback appeared to temper the negative feelings associated with subsequent negative feedback. Negative emotions did not show the same persistence effects as positive emotions. The negative emotions did not continue beyond the point of their initial arousal unless individuals continued to receive negative feedback on subsequent tasks. Thus, emotions do more than signal verifying or nonverifying outcomes. Emotions influence

82

JAN E. STETS AND PETER J. BURKE

interactions beyond their initial encounter to influence feelings in subsequent encounters as well as achieve verification.

Identity Theory Applications While identity theory has developed as a theory over the past 25 years, simultaneously, the theory has found application in a variety of substantive areas. We review some of these areas including crime and law, education, race/ethnicity, gender, the family, and the environment. Crime and Law Interpersonal violence is a serious problem in our society. It has been suggested that identity theory can help us understand domestic violence because aggressive behavior is rooted in issues of self and identity (Stets & Osborn, 2007). An important goal in interaction is the verification of people’s identities. If people experience identity nonverification, they may resort to aggression in an attempt to restore verification. This was investigated in a study of newly married couples (Stets & Burke, 2005a). It was shown that when an individual’s spousal identity was not verified, that individual tended to increase their control over their spouse, which control included acts of aggression. Using aggression in one year significantly reduced identity verification of the spousal identity in the following year, resulting in even more aggression in later years: a spiraling down of the relationship. Recent work has applied identity theory to the criminal identity of incarcerated offenders (Asencio & Burke, 2011). This study showed that the incarcerated offenders, rather than countering nonverifying feedback from peers and significant others as might be expected under identity theory, began to incorporate the noncriminal reflected appraisals into their identity (that they were not a criminal), changing their criminal identity toward a noncriminal identity. This occurred even controlling for the amount of time the individuals were incarcerated. This finding, like the results in the earlier study by Cast et al. (1999), revealed how identities can be changed when the person does not have the power or resources to resist discrepant reflected appraisals as is the case for incarcerated people. Robertson (2009) used identity theory to explain the lapse of judgment that sometimes occurs among lawyers when they defend corporate and government scandals without recognizing their biases. In defending scandals, she argued that lawyers may have two relevant, but competing,

The Development of Identity Theory

83

identities: that of attorney who works to protect and interpret the law, providing sound legal advice to his or her agency, and that of agency employee who works to protect the agency and advocate its goals. Robertson argued that when attorneys allow the employee identity to guide their behavior, they have a stake in a favorable outcome for the corporation, because in winning the case, it facilitates verification of their employee identity. In this way, attorneys acting in the employee identity are subject to the same distortions and biases as the corporation, thus the lawyers place themselves in a position that is unable to offer independent counsel. Partisan bias clouds their judgment. Robertson argues that to prevent this, agencies need to put procedures in place to maintain and strengthen the salience of the lawyer identity and reduce the salience of the employee identity. Education In science, technology, engineering, and mathematics (STEM) disciplines, men greatly outnumber women. In a series of studies, identity theory was used to explain women’s underrepresentation in STEM disciplines and to explore how the underrepresentation might be overcome (Lee, 1998, 2002, 2005). The research revealed that gender meanings of being feminine were held by females for themselves. In contrast, more masculine gender meanings were held for science students, including occupants of the STEM disciplines. For females to enter the STEM disciplines, they would be taking on meanings that were contrary to the meanings of their gender identity. To avoid this gender identity nonverification, they chose not to enter the STEM disciplines. Women who participated in a STEM summer program experienced emotionally satisfying relationships with others involved in science activities, which fostered a scientist identity (Lee, 2002). These good relationships were indicative of affective commitment. In turn, the bonds had the effect of increasing the salience of the scientist identity and engaging in STEM activities. Further, these bonds helped women to maintain science activities after they left the summer program, showing a more prominent or important scientist identity over time (Lee, 2005). Other researchers have examined the educational aspirations of high school students by including students’ academic identity in the standard Wisconsin status attainment model (Burke & Hoelter, 1988). While the original Wisconsin model worked well for White males, only when academic identity was included did the modified model work to predict educational expectations for groups for which it had not worked well previously, namely, White females and Black females. However, the model still did not

84

JAN E. STETS AND PETER J. BURKE

work well for Black males. Since the measure of the meaning of academic identity was invariant across all four groups (White males and females and Black males and females), the researchers reasoned that it must be the meaning of the educational expectations that set Black males apart. A follow-up study confirmed this reasoning by showing that there were two sets of meanings about going to college that were held among the four groups in different proportions (Burke, 1989a). On the one hand, there were a set of social meanings of going to college (being with friends, participating in college social life), while on the other hand there were a set of work meanings (learning more about careers, getting a better job). While people in all groups had varying degrees of each of these views about the meaning of going to college, Black males had predominantly more of the social meanings of going to college, while White males had predominantly more of the work meanings of going to college. When the analysis of the Wisconsin model including academic identity was estimated for only persons who had predominantly more work meanings of college, the model worked well in predicting educational expectations for both White and Black males. When the model was estimated for persons with social meanings of college, the model failed for both White and Black males. This made clear the underlying mechanism of matching identity meanings with situational behavior meanings as important for the verification of identities, as well as the importance of measuring meanings relevant to the identity. Race/Ethnicity Researchers have studied how identities operate across different race/ethnicities as well as how race/ethnicity as a social identity manifests itself. In one study, it was discovered that the salience of the family identity for Blacks, Whites, and Latinos had different sources across the different groups (Owens & Serpe, 2003). The salience of the family identity was a function of self-esteem for Whites and Blacks. However, for Latinos, a salient family identity was a function of commitment both in terms of the number of ties and depth of the ties to the family. This revealed how significant the family was for Latinos compared to Blacks and Whites. This familial emphasis may have other unintended effects for Latinos. For example, other research revealed that compared to Whites and Blacks, Latinos were less likely to have their friend identity verified (Stets & Harrod, 2004). The researchers suggested that because of Latinos devotion to the family, they may be more likely to rely on and confide in kin members than friends. If friendships are not encouraged and developed, it may

The Development of Identity Theory

85

become difficult to acquire the resources such as care, trust, and loyalty, which are necessary to facilitate verification of the friend identity. Other analyses of Whites, Blacks, and Latinos found that increased commitment (including both more and deeper ties) to work and to voluntary association identities for all three ethnic groups was a function of interacting with the same individuals at work and in the voluntary associations (Stryker et al., 2005). Viewed in another way, this supports the identity theory assumption that losing a position and identity in one network threatens relationships and an identity in another network to the degree that the networks have common members. Gender Research has examined the effect of gender identity on school performance on a large sample of 1,688 sixth-, seventh-, and eighth-grade children (Burke, 1989b). The results revealed that gender identity had separate and somewhat different effects on school grades across a range of subjects. Across all subjects, boys and girls with a more feminine gender identity performed better even when controlling for sex, race, grade, IQ, and sex of teacher. While overall girls outperformed boys in these grades, a good portion of the difference appeared to be due not to sex but to the impact of what it means to be a boy or girl, that is, gender identity. Controlling for gender identity reduced the degree to which girls outperformed boys, but did not eliminate it (except in math and science in the sixth grade). Others studied gender identity in adulthood by examining the problemsolving interactions of newly married couples (Stets & Burke, 1996). The focus was on negative and positive behaviors in conversations as coded in the interaction of husbands and wives working to solve one or more issues they agreed were problematic. Based on the dominance and competitive meanings of masculinity, the researchers expected that husbands and wives with a more masculine gender identity would be more likely to use the negative behaviors (e.g., criticisms, defensive talk, and putdowns) and less likely to use positive behaviors (e.g., agreeing and using humor) in the problem-solving interactions. They also expected that, based on expectation states theory, males/husbands (who have a higher status in structure) would use more negative behaviors than females/wives (who have a lower status in the social structure). While the data supported the identity theory predictions about the effects of gender identity (masculinity predicting more negative and less positive behavior than femininity), the results for status were the opposite of the expectations: females/wives engaged in more negative and less

86

JAN E. STETS AND PETER J. BURKE

positive behavior than males/husbands. To understand this reversal with respect to sex differences, Stets and Burke discussed how women, because of their low status, suffer from a stricter standard regarding competence. Women have to work harder than men to be viewed as capable. Thus, acting more “masculine” by using more negative behaviors compared to men is a way to be seen as credible and capable in interaction. Interestingly, rather than reversing the power structure within the marriage, wives’ more negative behavior simply helps equalize power between themselves and their spouses. This same pattern of women having to work harder appeared in another study that examined the effects of leadership identity, legitimation of being a leader, and gender status in small task-oriented discussion groups (Burke, Stets, & Cerven, 2007). The investigators hypothesized that resources that come with both high status and legitimation through authorization of being in the leader position (support for a person’s leader position comes from people in higher positions in the situation) would increase the ability of individuals to verify their leadership identity. The findings showed an asymmetric effect of these factors. Women were underevaluated for their leadership compared to males. Women benefited from legitimation to engage in more leadership behavior than their nonlegitimated counterparts. This benefit of legitimation brought their evaluation up to the level suggested by their leadership identity, thus verifying their identities. Men, on the other hand, were evaluated at the level of their leadership identity without the benefit of having their leadership position legitimated. With legitimation, males were evaluated above their level of leadership identity. This overevaluation resulted in their leadership identities not being verified. Thus, the principle that legitimation and status provide resources that help in the verification of identities must also consider the expectations that come with those resources: women without leadership legitimation suffer from a deficit of expectations, while men with leadership legitimation suffer from expectations that are too high. Family As discussed earlier in this chapter, there is evidence that verification of the spousal identity results in positive outcomes such as increased trust and commitment in the marriage; nonverification of the spousal identity produces the opposite effects (Burke & Stets, 1999). Also discussed earlier in this chapter, research has revealed that the views of the higher-status spouse in the marriage will be more likely to influence (1) the spousal

The Development of Identity Theory

87

identity views of the lower-status spouse and (2) the lower-status spouse’s view of the spousal identity of the higher-status spouse (Cast et al., 1999). In addition, research shows the effects of both structural power (occupation and education) and relationship power (the person who is in love the least has the most power) on spousal identity meanings and spousal behavior meanings (Cast, 2003). At issue is whether married persons behave in ways that are consistent with their own spousal identity or with how their spouse sees them. The results showed that those with the greater structural power and/or relationship power had more influence on their spouse’s role performance than the reverse. Further, the more powerful persons were better able to resist the influence of their spouse on their own role performance. There was no difference in the way the process worked for husbands or wives. In general, the findings revealed that structural and relationship power influence how the spousal identity gets played out in a marriage. Finally, research has investigated the transition to parenthood among newly married couples (Cast, 2004). The results indicated that parenthood per se did not hurt marital well-being when prior levels of well-being were controlled. However, the failure to verify the parent identity had a large impact on both individual and marital well-being. When the parent identity was not verified, parents were more depressed, anxious, and had lower feelings of esteem and efficacy. They also reported less happiness, love, liking, and trust in the marriage. The Environment One other area in which identity theory has been applied is to the environment, particularly environmentally responsive behavior (Stets & Biga, 2003). Historically, researchers have predicted environmental behavior from one’s attitudes about the environment, and only a modest relationship has been found (Tarrant & Cordell, 1997). When one’s environment identity is included in the analysis, that is, seeing oneself as environmentally friendly and supportive of the environment versus environmentally unfriendly and exploitative, environmentally responsive behavior emerges out of the identity process rather than the attitude process. Thus, more research is needed that includes one’s environment identity in studies on one’s behavior toward the environment. The above applications indicate the viability of identity theory in explaining different aspects of social life. Because identity theory focuses on behaviors as a means to control perceptions of sign and symbolic meanings relevant to one’s identities, and because identities are relevant in any

88

JAN E. STETS AND PETER J. BURKE

situation in which people find themselves, identity theory can be applied to a wide range of social and institutional settings. Any theory of criminal behavior or family behavior, for example, would benefit from considering the identity relevant meanings that are controlled by that behavior. In this way, identity theory can help us understand the full range of social behavior with which sociology is concerned. The general principles of identity theory should apply and be useful in a host of social contexts.

IDENTITY THEORY FOR THE FUTURE As we look to the future of identity theory, there are two primary concerns that manifest themselves: the need for future development of the theory and the need to integrate it more strongly into the overall framework of social psychological theories.

Theory Development We suggest three areas of future research development. These involve resources, stigmatized identities, and identity change. Resources More research is needed on the role of resources in the identity process, including understanding both actual resources (resources that are currently supporting identities, interactions, and groups in the situation), and potential resources (resources that are not currently being used, but may be used in the future). Additionally, while there has been some research on resources being used to facilitate verification, as discussed earlier, the focus has been on symbolic meanings to the exclusion of sign meanings. Both sets of meanings are contained in the identity standard. However, it is the sign meanings that are attached to resources in the situation. When these meanings are controlled (especially those contained in role identity standards), they provide the resource infrastructure that allows groups, organizations, and institutions to exist (Burke, 2004). For example, the family, as a group, needs to be sustained. Role identities, such as the spousal identity and the parent identity, need to be verified to help facilitate maintenance of the family unit. In the parent identity, sign meanings will be controlled in order to verify the parent identity such as

The Development of Identity Theory

89

having a crib for the infant, food to feed the child, clothes to keep the child warm, a car to transport the child to the doctor, money, a home, and all other material and nonmaterial resources such as love, support, and care. To the extent that the control of sign meanings verifies identities within the family such as the spousal identity, the parent identity, the child identity, the sibling identity and so forth, then by extension, the family unit is sustained. And, when this is repeated across the thousands of families in the United States, for example, the implications for the flow of material resources across the country are immense. To understand all this, more research on the role of sign meanings in the identity verification process is needed. In addition to the sign meanings that are part of the identity standard for all identities, there are also emotion meanings, that is, emotional responses to signs and symbols. Some identities carry stronger emotion meanings than others. For example, the clown identity likely has happiness as a meaning in the identity standard, the identity of rape victim may have meanings of anger and fear, and the identity of widower may have meanings of grief and sorrow. We hypothesize that the failure to verify identities that contain both cognitive and emotion meaning (compared to those that contain only cognitive meanings) will result in stronger negative reactions because both the cognitive and affective meanings are not verified. If the identities carry weaker emotion meanings, the negative emotional reaction to nonverification might not be as strong. This needs to be investigated. Further, whether the strength of the emotional response is a function of the number of meanings not being verified (more nonverified meanings leading to a stronger emotional response) or the nature of the meanings not being verified (cognitive and emotional meanings, or alternatively, cognitive or emotional meanings) also needs empirical investigation. Stigmatized Identities Most of the identities that have been researched are normative or positive identities. Very little research has examined negative, stigmatized, or counternormative identities. What are the outcomes of verifying negative, stigmatized, or counternormative identities? Two consequences are possible, and research needs to sort through these and perhaps other possibilities. One possibility is that while society, in general, views such identities as negative or stigmatized, the people who hold these identities may not see them as negative in the same way. Because it is the meanings held in the identity standard of the individual that are involved in the verification, if the identity standard does not contain negative meanings for the individual,

90

JAN E. STETS AND PETER J. BURKE

then verification should proceed normally, with positive feelings generated for verification, and negative feelings generated for nonverification. A second possibility is that people who hold the stigmatized identity see these identities as negative. Verification of a negative identity may remind the person of the negative valence attached to who they are, but the verification of who they are may reduce the negative feelings. When nonverification occurs, these negative feelings may become stronger both because of the negative valence of the identity and also because of the failure to verify one’s identity. Verifying a negative or stigmatized identity may be the best of a bad situation
nonverification would be worse. Empirical research needs to test these expectations.

Identity Change The sources of identity change are another area in which research is needed. We have already conducted studies that show identity change results from nonverification and from holding identities with slightly different meanings. Both of these are endogenous sources. We also have shown how the meanings of the identity of one person can influence the meanings of an identity of another person. This is an exogenous source. However, we have not examined exogenous sources beyond this one. For example, the placement of an identity in the social structure may be more or less likely to lead to identity change. This might be the case because of the differential distribution of resources across the social structure with the result that identities higher in the social structure have the necessary resources of power and status for verification compared to identities lower in the social structure. Another source of exogenous identity change that has yet to be investigated is the impact of what have been called “open” versus “closed” structures (Serpe, 1987). In more open systems where choice is possible, people can find positions in the social structure that reinforce the meanings in their identities. In more closed structures where choice is less possible, identities may have to change in order to fit into the rules and responsibilities associated with different positions. Also, the overall balance between endogenous and exogenous sources of identity change would be important to learn. Finally, we need to examine changes in identities involving increasing or decreasing salience, prominence, or commitment. Also unexplored is how and why individuals take on specific identities when they do, and what encourages them to exit or abandon them when they do.

The Development of Identity Theory

91

Links to Other Social Psychological Theories While the development of identity theory per se is important, the relationship between identity theory and other social psychological theories within the overall structure of social psychology also needs to be better understood. Over time, identity theory has been linked to a number of other social psychological theories to show common approaches as well as ways in which each theory can augment the other to the benefit of both. For example, researchers have made associations between identity theory and affect control theory (Smith-Lovin & Robinson, 2006), expectation states theory (Burke, 2008; Cast et al., 1999; Stets & Harrod, 2004), network exchange theory (Burke, 1997), justice theories (Stets, 2003; Stets & Osborn, 2008), legitimation theory (Burke et al., 2007), social comparison theory (Stets & Burke, 2013b), social identity theory (Stets & Burke, 2000), and social movements theory (Stryker, Owens, & White, 2000). But, there is still more work to be done. For example, affect control theory reminds us that we act to maintain not only our own identities in a situation but also the identities of others. At issue for identity theorists is to consider whether, in situations, individuals maintain the identities of others only when it simultaneously sustains their own identities, or whether individuals are motivated to maintain the identities of others as distinct from maintaining their own identities. As another example, and looked at from a different perspective, theorists in expectation states and exchange theories might examine how identity processes operate within their theoretical frameworks. For instance, are performance expectations associated with one’s status in a group more likely to be challenged if one’s identity meanings conflict with those performance expectations? If a high-status male is not assertive in the group, is it because he views himself as having a timid identity? Under what conditions will status meanings or identity meanings have more influence in guiding one’s behavior? In exchange theory, how might behavior choices be different when we take into account not only the structural conditions of the exchange network (such as negatively connected networks or power-imbalanced networks) but also the agency of the individuals’ identities involved in the exchanges. If people are exchanging in different power network structures, how might they behave when, for example, the identity of themselves as “fair” or “unfair” (their fairness identity) is considered? Will we see more equitable exchanges or will the position in the network predict one’s behavior in the exchange? Further, when the fairness identity is not verified in

92

JAN E. STETS AND PETER J. BURKE

an exchange, will individuals modify their exchange behavior by sacrificing more or less depending on whether they are underverified or oververified for being fair and whether they are in a power advantaged or power disadvantaged position? Finally, identity theory historically has focused on role identities, although it is now theorizing and studying social/group and person identities, while social identity theory has consistently emphasized social identities. Within groups people play out various roles, and individuals enact these various roles in different ways, given the unique person identities they bring to their roles. Thus, in situations, group, role, and person identities may not be easily separated, and we may need to examine their simultaneous occurrence. This poses a challenge as to how we might examine their separate effects, even when these effects might not be independent of each other. In this way, the relationship of social identity theory and identity theory can be better understood. Independent of their simultaneous occurrence within a group, we need more research on the relationships among person, role, and social/group identities. Do multiple identities within a person that cross the different bases relate differently than, for example, multiple person identities, multiple role identities, or multiple social/group identities? Are the consequences of the verification of person, role, and social identities different? For example, does the verification of person, role, and social identities form the basis of three dimensions of esteem: authenticity, self-efficacy, and self-worth, respectively? Does the verification of person identities increase feelings of authenticity, the verification of role identities increase self-efficacy, and the verification of social identities increase self-worth? Further, does the verification of one identity, such as a social/group identity, and the self-worth that it generates encourage individuals to pursue the verification of other identities, such as role and person identities, especially when they may be difficult to verify? While social identity theorists generally do not focus on person identities given the depersonalization process that social/group identities activate (persons identify with groups rather than conceive of themselves as individuals/actors), and while role identities have been more the focus in identity theory, person identities also guide behavior in situations. While we are beginning to study person identities such as the moral identity (Stets & Carter, 2006, 2011, 2012; Stets et al., 2008), the control identity (Stets, 1997; Stets & Burke, 1994, 1996), and the environment identity (Stets & Biga, 2003), there are many more person identities that need to be examined as we come to better understand all of human social behavior.

93

The Development of Identity Theory

CONCLUSION Overall, we find there have been many additions as well as elaborations and clarifications of identity theory since 1988. Among the more significant developments are the perceptual control system, the link of symbols and signs to resources, an extension of the bases of identities beyond role identities to group/category and person identities, a clearer recognition of the role of the situation, and the introduction of emotions. There is much more to learn and more theoretical clarity and expansion that is needed. However, the future is bright and the theory remains a promising approach toward understanding how, in the words of Stryker, “society shapes self shapes social behavior.”

NOTES 1. Affect control theory restricted the number of dimensions of meaning to only the EPA dimensions in order to be able to use the same semantic space across a wide variety of concepts and identities. 2. Obviously there are restrictions on the behavior chosen to at least not contradict other identities, including those having to do with morality and propriety. 3. Later, we discuss the extension of identity theory to include two additional bases for identities beyond that of roles: person identities and social identities. 4. Early on, we conceptualized gender identity as a role identity, not having a clear theory about the different bases of identities that we have today. Nevertheless, the general principles held.

ACKNOWLEDGMENT We would like to thank members of the UCR Social Psychology Research Seminar for their comments on an earlier draft of this chapter.

REFERENCES Asencio, E. K., & Burke, P. J. (2011). Does incarceration change the criminal identity? A synthesis of labeling and identity theory perspectives on identity change. Sociological Perspectives, 54(2), 163
182. doi:10.1525/sop.2011.54.2.163 Burke, P. J. (1980). The self: Measurement implications from a symbolic interactionist perspective. Social Psychology Quarterly, 43, 18
29.

94

JAN E. STETS AND PETER J. BURKE

Burke, P. J. (1989a). Academic identity and race differences in educational aspirations. Social Science Research, 18(2), 136
150. Burke, P. J. (1989b). Gender identity, sex, and school performance. Social Psychology Quarterly, 52(2), 159
169. Burke, P. J. (1991). Identity processes and social stress. American Sociological Review, 56(6), 836
849. Burke, P. J. (1997). An identity model for network exchange. American Sociological Review, 62(1), 134
150. Burke, P. J. (2003). Relationships among multiple identities. In P. J. Burke, T. J. Owens, R. T. Serpe, & P. A. Thoits (Eds.), Advances in identity theory and research (pp. 195
214). New York, NY: Kluwer Academic/Plenum. Burke, P. J. (2004). Identities and social structure: The 2003 Cooley-Mead award address. Social Psychology Quarterly, 67, 5
15. Burke, P. J. (2006a). Identity change. Social Psychology Quarterly, 69, 81
96. Burke, P. J. (2006b). Perceptions of leadership in groups: An empirical test of identity control theory. In K. McClelland & T. J. Fararo (Eds.), Purpose, meaning, and action: Control systems theories in sociology (pp. 267
291). New York, NY: Palgrave Macmillan. Burke, P. J. (2008). Identity, social status, and emotion. In D. T. Robinson & J. Clay-Warner (Eds.), Social structure and emotion (pp. 75
93). Burlington, MA: Elsevier. Burke, P. J., & Cast, A. D. (1997). Stability and change in the gender identities of newly married couples. Social Psychology Quarterly, 60(4), 277
290. Burke, P. J., & Franzoi, S. L. (1988). Studying situations and identities using experiential sampling methodology. American Sociological Review, 53(4), 559
568. Burke, P. J., & Harrod, M. M. (2005). Too much of a good thing? Social Psychology Quarterly, 68, 359
374. Burke, P. J., & Hoelter, J. W. (1988). Identity and sex-race differences in educational and occupational aspirations formation. Social Science Research, 17(1), 29
47. doi:10.1016/ 0049-089x(88)90019-1 Burke, P. J., & Reitzes, D. C. (1980). College student identity: Measurement and implications. Pacific Sociological Review, 23, 46
66. Burke, P. J., & Reitzes, D. C. (1981). The link between identity and role performance. Social Psychology Quarterly, 44, 83
92. Burke, P. J., & Stets, J. E. (1999). Trust and commitment through self-verification. Social Psychology Quarterly, 62, 347
366. Burke, P. J., & Stets, J. E. (2009). Identity theory. New York, NY: Oxford University Press. Burke, P. J., Stets, J. E., & Cerven, C. (2007). Gender, legitimation, and identity verification in groups. Social Psychology Quarterly, 70(1), 27
42. doi:10.1177/019027250707000105 Burke, P. J., Stets, J. E., & Pirog-Good, M. (1988). Gender identity, self-esteem, and physical and sexual abuse in dating relationships. Social Psychology Quarterly, 51, 272
285. Burke, P. J., & Tully, J. C. (1977). The measurement of role identity. Social Forces, 55(4), 881
897. Cast, A. D. (2003). Power and the ability to define the situation. Social Psychology Quarterly, 66(3), 185
201. Cast, A. D. (2004). Well-being and the transition to parenthood: An identity theory approach. Sociological Perspectives, 47(1), 55
78. Cast, A. D., & Burke, P. J. (2002). A theory of self-esteem. Social Forces, 80(3), 1041
1068.

The Development of Identity Theory

95

Cast, A. D., & Cantwell, A. M. (2007). Identity change in newly married couples: Effects of positive and negative feedback. Social Psychology Quarterly, 70, 172
185. Cast, A. D., Stets, J. E., & Burke, P. J. (1999). Does the self conform to the views of others? Social Psychology Quarterly, 62(1), 68
82. Ellestad, J., & Stets, J. E. (1998). Jealousy and parenting: Predicting emotions from identity theory. Sociological Perspectives, 41, 639
668. Freese, L., & Burke, P. J. (1994). Persons, identities, and social interaction. Advances in Group Processes, 11, 1
24. Lee, J. D. (1998). Which kids can “become” scientists? Effects of gender, self-concepts, and perceptions of scientists. Social Psychology Quarterly, 61, 199
219. Lee, J. D. (2002). More than ability: Gender and personal relationships influence science and technology involvement. Sociology of Education, 75, 349
373. Lee, J. D. (2005). Do girls change more than boys: Gender differences and similarities in the impact of new relationships on identities and behaviors. Self and Identity, 4, 131
147. Lindesmith, A. R., & Strauss, A. L. (1956). Social psychology. New York, NY: Holt Rinehart and Winston. Mead, G. H. (1934). Mind, self, and society. Chicago, IL: University of Chicago Press. Merolla, D. M., Serpe, R. T., Stryker, S., & Schultz, P. W. (2012). Structural precursors to identity processes: The role of proximate social structures. Social Psychology Quarterly, 75, 149
172. Mutran, E., & Burke, P. J. (1979a). Feeling ‘useless’: A common component of young and old adult identities. Research on Aging, 1, 188
212. Mutran, E., & Burke, P. J. (1979b). Personalism as a component of old age identity. Research on Aging, 1, 37
64. Osgood, C. E., May, W. H., & Miron, M. S. (1975). Cross-cultural universals of affective meaning. Urbana, IL: University of Illinois Press. Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement of meaning. Urbana, IL: University of Illinois Press. Owens, T. J., & Serpe, R. T. (2003). The role of self-esteem in family identity salience and commitment among Blacks, Latinos, and Whites. In P. J. Burke, T. J. Owens, R. T. Serpe, & P. A. Thoits (Eds.), Advances in identity theory and research (pp. 85
102). New York, NY: Kluwer Academic/Plenum. Powers, W. T. (1973). Behavior: The control of perception. Chicago, IL: Aldine. Riley, A., & Burke, P. J. (1995). Identities and self-verification in the small group. Social Psychology Quarterly, 58(2), 61
73. Robertson, C. B. (2009). Judgment, identity, and independence. Connecticut Law Review, 42, 1
48. Serpe, R. T. (1987). Stability and change in self: A structural symbolic interactionist explanation. Social Psychology Quarterly, 50, 44
55. Serpe, R. T., & Stryker, S. (1987). The construction of self and reconstruction of social relationships. In E. Lawler & B. Markovsky (Eds.), Advances in group processes (Vol. 4, pp. 41
66). Greenwich, CT: JAI. Serpe, R. T., & Stryker, S. (1993). Prior social ties and movement into new social relationships. Advances in Group Processes, 10, 283
304. Serpe, R. T., & Stryker, S. (2011). The symbolic interactionist perspective and identity theory. In S. Schwartz, K. Luyckx, & V. Vignoles (Eds.), Handbook of identity theory and research (pp. 225
248). New York, NY: Springer.

96

JAN E. STETS AND PETER J. BURKE

Smith-Lovin, L., & Robinson, D. T. (2006). Control theories of identity, action, and emotion: In search of testable differences between affect control theory and identity control theory. In K. McClelland, & T. J. Fararo (Eds.), Purpose, meaning, and action: Control systems theories in sociology (pp. 163
188). New York, NY: Palgrave Macmillan. Stager, S., & Burke, P. J. (1982). A reexamination of body build stereotypes. Journal of Research in Personality, 16, 435
446. Stets, J. E. (1995). Role identities and person identities: Gender identity, mastery identity, and controlling one’s partner. Sociological Perspectives, 38(2), 129
150. Stets, J. E. (1997). Status and identity in marital interaction. Social Psychology Quarterly, 60(3), 185
217. Stets, J. E. (2003). Justice, emotion, and identity theory. In P. J. Burke, T. J. Owens, P. A. Thoits, & R. Serpe (Eds.), Advances in identity theory and research (pp. 105
122). New York, NY: Kluwer Academic/Plenum. Stets, J. E. (2004). Emotions in identity theory: The effects of status. Advances in Group Processes, 21, 51
76. Stets, J. E. (2005). Examining emotions in identity theory. Social Psychology Quarterly, 68(1), 39
56. Stets, J. E. (2011). Applying identity theory to moral acts of commission and omission. Advances in Group Processes, 28, 97
124. Stets, J. E., & Asencio, E. K. (2008). Consistency and enhancement processes in understanding emotions. Social Forces, 86, 1055
1058. Stets, J. E., & Biga, C. F. (2003). Bringing identity theory into environmental sociology. Sociological Theory, 21, 398
423. Stets, J. E., & Burke, P. J. (1994). Inconsistent self-views in the control identity model. Social Science Research, 23(3), 236
262. Stets, J. E., & Burke, P. J. (1996). Gender, control, and interaction. Social Psychology Quarterly, 59(3), 193
220. Stets, J. E., & Burke, P. J. (2000). Identity theory and social identity theory. Social Psychology Quarterly, 63(3), 224
237. Stets, J. E., & Burke, P. J. (2005a). Identity verification, control, and aggression in marriage. Social Psychology Quarterly, 68(2), 160
178. Stets, J. E., & Burke, P. J. (2005b). New directions in identity control theory. Advances in Group Processes, 22, 43
64. Stets, J. E., & Burke, P. J. (2013a). Self-esteem and identities. Unpublished manuscript. Stets, J. E., & Burke, P. J. (2013b). Social comparison processes in identity theory. In Z. Krizan & F. X. Gibbons (Eds.), Communal functions of social comparison (pp. 39
59). New York, NY: Cambridge University Press. Stets, J. E., & Burke, P. J. (2014, forthcoming). Emotions and identity non-verification. Social Psychology Quarterly, 77. Stets, J. E., & Carter, M. J. (2006). The moral identity: A principle level identity. In K. McClelland & T. J. Fararo (Eds.), Purpose, meaning, and action: Control systems theories in sociology (pp. 293
316). New York, NY: Palgrave MacMillan. Stets, J. E., & Carter, M. J. (2011). The moral self: Applying identity theory. Social Psychology Quarterly. 74, 192
215. Stets, J. E., & Carter, M. J. (2012). A theory of the self for the sociology of morality. American Sociological Review, 77, 120
140.

The Development of Identity Theory

97

Stets, J. E., Carter, M. J., Harrod, M. M., Cerven, C., & Abrutyn, S. (2008). The moral identity, status, moral emotions, and the normative order. In D. T. Robinson & J. ClayWarner (Eds.), Social structure and emotion (pp. 227
251). San Diego, CA: Elsevier. Stets, J. E., & Cast, A. D. (2007). Resources and identity verification from an identity theory perspective. Sociological Perspectives, 50, 517
543. Stets, J. E., & Harrod, M. M. (2004). Verification across multiple identities: The role of status. Social Psychology Quarterly, 67, 155
171. Stets, J. E., & Osborn, S. N. (2007). Identity theory and domestic violence. In N. A. Jackson (Ed.), Encyclopedia of domestic violence (pp. 375
381). New York, NY: Routledge. Stets, J. E., & Osborn, S. N. (2008). Injustice and emotions using identity theory. Advances in Group Processes, 25, 151
179. Stets, J. E., & Serpe, R. T. (2013). Identity theory. In J. DeLamater & A. Ward (Eds.), Handbook of social psychology (pp. 31
60). New York, NY: Springer. Stets, J. E., & Tsushima, T. (2001). Negative emotion and coping responses within identity control theory. Social Psychology Quarterly, 64, 283
295. Stryker, S. ([1980] 2002). Symbolic interactionism: A social structural version. Caldwell, NJ: Blackburn Press. Stryker, S. (2004). Integrating emotion into identity theory. Advances in Group Processes, 21, 1
23. Stryker, & Burke, P. J. (2000). The past, present, and future of an identity theory. Social Psychology Quarterly, 63, 284
297. Stryker, S., Owens, T. J., & White, R. W. (Eds.). (2000). Self, identity, and social movements. Minneapolis, MN: University of Minnesota Press. Stryker, S., & Serpe, R. T. (1982). Commitment, identity salience, and role behavior: A theory and research example. In W. Ickes & E. S. Knowles (Eds.), Personality, roles, and social behavior (pp. 199
218). New York, NY: Springer-Verlag. Stryker, S., & Serpe, R. T. (1983). Toward the theory of family influence in the socialization of children. In A. Kerchoff (Ed.), Research in the sociology of education and socialization (Vol. 4, pp. 47
74). Greenwich, CT: JAI. Stryker, S., Serpe, R. T., & Hunt, M. O. (2005). Making good on a promise: The impact of larger social structures on commitments. Advances in Group Processes, 22, 93
123. Tarrant, M. A., & Cordell, H. K. (1997). The effect of respondent characteristics on general attitude-behavior correspondence. Environment and Behavior, 29(5), 618
637. doi:10.1177/0013916597295002 Thoits, P. A. (1983). Multiple identities and psychological well-being: A reformulation and test of the social isolation hypothesis. American Sociological Review, 49, 174
187. Thoits, P. A. (1986). Multiple identities: Examining gender and marital status differences in distress. American Sociological Review, 51, 259
272. Thoits, P. A. (1992). Identity structures and psychological well-being: Gender and marital status comparisons. Social Psychology Quarterly, 55(3), 236
256. Tsushima, T., & Burke, P. J. (1999). Levels, agency, and control in the parent identity. Social Psychology Quarterly, 62(2), 173
189. White, C. L., & Burke, P. J. (1987). Ethnic role identity among Black and White college students: An interactionist approach. Sociological Perspectives, 30(3), 310
331.

RELATIONAL COHESION, SOCIAL COMMITMENTS, AND PERSON-TOGROUP TIES: TWENTY-FIVE YEARS OF A THEORETICAL RESEARCH PROGRAM Shane R. Thye, Aaron Vincent, Edward J. Lawler and Jeongkoo Yoon ABSTRACT Purpose
This chapter analyzes the ways that individuals develop person-to-group ties. The chapter reviews the development and evidentiary basis of the theory of relational cohesion, the affect theory of social exchange, and the theory of social commitments. Methodology/approach
We survey twenty-five years of published literature on these theories, and review unpublished theoretical tests and extensions that are currently in progress. Findings
The research program has grown substantially over the past twenty-five years to encompass more varied and diverse phenomena. The findings indicate that structural interdependencies, repeated exchanges,

Advances in Group Processes, Volume 31, 99
138 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0882-6145/doi:10.1108/S0882-614520140000031008

99

100

SHANE R. THYE ET AL.

and a sense of shared responsibility are key conditions for people to develop affective ties to groups, organizations, and even nation-states. Research limitations/implications
The research implies that if people are engaged in joint tasks, they attribute positive or negative feelings from those tasks to their local groups (teams, departments) and/or to larger organizations (companies, communities). To date, empirical tests have focused on microlevel processes. Practical implications
Our work has practical implications for how managers or supervisors organize tasks and work routines in a way to maximize group or organizational commitment. Social implications
This research helps to understand problems of fragmentation that are faced by decentralized organizations and also how these can be overcome. Originality/value of the chapter
The chapter represents the most complete and comprehensive review of the theory of relational cohesion, the affect theory of social exchange, and the theory of social commitments to date. Keywords: Affect; cohesion; solidarity; cooperation; group formation; social exchange

This chapter reviews a program of research that has been constantly evolving over the past twenty-five years and is the product of a long-term collaborative effort. The original work, anchored in the theory of relational cohesion, began in the early 1990s and addressed at that time a relatively unexplored phenomenon
how power dynamics produce emotional reactions that foster social bonds, cohesion, and commitment (Lawler & Yoon, 1993, 1996). Since then, the theory has been tested in a diverse array of settings, evolved to address new questions, spawned new but intellectually related theories, and been generally transformed into a broader, more encompassing account of how human emotions create person-to-group bonds evidenced by feelings of cohesion and acts of commitment. We last published a review of our work more than a decade ago (Thye, Yoon, & Lawler, 2002); since that time much new theory and evidence has emerged on the problem of person-to-group ties. The overall goal of this chapter is

Person-to-Group Ties

101

to provide the most comprehensive review of our accomplishments to date, assess the current state of our work, and look forward to the future of the program. We first identify the broader theoretical foundations of the work, then review the primary branches of our work and the intellectual questions that we have addressed. Our focus is concentrated squarely on person-to-group ties. Importantly, we ground these ties in person-to-person interactions. In an ever increasingly individualized world, how do we form and maintain ties to larger groups such as communities, work organizations, or even nations? We theorize that such ties are essentially formed from the “bottom up” (see Lawler, Thye, & Yoon, 2009) and that these ties are primarily affective and emotional (see Lawler & Thye, 2006). Microinteractions generate personto-group ties due to the emotional experiences of people when they interact with others. We show how, through social interaction, relations and groups can become social objects that are valued in and of themselves. Even within interactional settings that begin as purely instrumental and transactional, positive emotions generated in interaction sometimes foster resilient ties to salient relations, groups, or even organizations.

THEORETICAL FOUNDATIONS Our theoretical research program is intellectually anchored in the ideas and principles found in the works of Durkheim (1915), Homans (1950), and Emerson (1972a, 1972b). Durkheim focused on the consequences of joint activities in preliterate societies. According to Durkheim, joint activities have the capacity to produce social order due to the emotional consequences and feelings these create. His primary assertion is that collective activity (e.g., religious rituals) serves as an integrating force due to the emotional by-products (collective effervescence) of such activities. Collective effervescence affirms and reasserts social ties by making salient the overarching group and the important ties that individuals have to that group. These ties are strengthened by such rituals and therefore are easier to maintain when threats to social order are present. In terms of our research program, Durkheim’s primary contribution is to suggest that emotions generated in group contexts facilitate the formation of emotional ties to larger, macro-units. Homan’s emphasis is decidedly more micro. Homans argued that there are three dimensions that determine the emergence of stable

102

SHANE R. THYE ET AL.

relationships: the nature of the activities that people are engaged in, how often people interact with each other, and the sentiments of affection or sympathy that people develop toward one another. While Durkheim was focused on the emergence of person-to-group ties, Homans’ analysis centers more on person-to-person ties. Despite the difference, both imply that emotion from repeated social interaction at the microlevel creates and maintains social ties to more macro-entities. We emphasize the point that social interaction can produce emotions and attachments to others or larger social units. Richard Emerson’s work in the early 1970s also provides an important contextual backdrop. Emerson’s focus was on exchange, power, and dependence processes. In his power-dependence theory, the power (P) of an actor A over another actor B is determined by the dependence (D) of B on A, thus PAB = DBA. Dependence, in turn, is determined by the availability of exchange partners and the value of the goods they hold. In short, powerful actors are those who (i) possess highly valued goods and (ii) are the sole supplier of those goods. This is what makes drug dealers and popular teenagers powerful. Note that part of what determines power for Emerson (1972a, 1972b) is the availability of other exchange partners; prior to Emerson, most work on power focused on dyads. Emerson’s insight is that dyads are embedded in webs of other dyadic relations. Early work in our own research program picks up on this structural theme, and uses Emerson’s notion of power and dependence. The difference is that whereas Emerson’s focus was primarily on power, resource distributions, and divisiveness, our work focuses on the emotional and cohesive consequences of power use. Emerson did note that cohesion is likely to occur within highly dependent relationships; however, cohesion was not a central focus. Thus, our work shows how mutual dependence can facilitate cohesion by way of emotional and affective processes. Emotions generated at the microlevel, even in relations initially formed for instrumental reasons, facilitate the transformation of relationships and networks into group entities (Lawler & Thye, 1999). Emotions serve as a fundamental connection between micro and macro contexts.

THEORETICAL DEFINITIONS AND SCOPE Two terms, emotion and commitment, require some discussion to make clear the context and focus of the research. An emotion is defined as a

Person-to-Group Ties

103

relatively short-lived positive or negative evaluative state that involves neurophysiological, neuromuscular, and sometimes cognitive elements (Damasio, 1999; Izard, 1977; Kemper, 1978). Emotions can be triggered by any number of sources and are unique in that they are often not entirely under our control. The emotions of concern are everyday feelings such as mild forms of pleasure and excitement. Social interactions almost invariably generate emotional states along pleasure and excitement dimensions. Our work has focused on the way that social interaction, of various kinds, produces positive emotions and the conditions under which an attribution process leads to stronger or weaker affective attachments to groups. Such attachments entail commitments that can take various forms. Kanter (1968) identifies three general forms of commitment: continuance, affective, and normative. Continuous commitment refers to the tendency for actors to remain in a group or relationship due to the benefits it provides or the losses that would result by exiting. The classic example is continuing to work for an organization when one does not have a more attractive employment offers. Continuance commitments are wholly instrumental and imply an overarching market framework in which relations are about transactions or exchanges (Emerson, 1972b). Affective and normative commitments differ from continuance in that the group itself plays a more central role as an important and salient object. Affective commitment refers to emotional feelings or sentiments that tie individuals to a group or organization (Lawler, 1992). Here the group or organization takes on intrinsic, affective value itself, beyond any instrumental benefit. Normative commitment is defined as a sense of obligation to a group or organization based on its formal and informal rules. These obligations may be instrumental and/or affective (see Lawler et al., 2009, chapter 2). The primary intent of our research program, over the years, has been to understand how instrumental motives in social relations (continuous forms of commitment) come to spark both affective and normative forms of commitment over time (Lawler, Thye, & Yoon, 2008; Lawler et al., 2009). We define social commitments as person-to-group ties that are affective or are normative with an affective basis. Social commitments are said to exist if these affective and normative elements are part of the person-to-group bond. The context of our theory and research is elaborated further by making explicit the scope conditions the theory operates within. Scope conditions specify the domain to which a theory applies (Cohen, 1989; Walker & Cohen, 1985). Most of the scope conditions specified in our research are commonly found in social exchange theories, with some modification. There are specifically six primary scope conditions assumed: (1) A network

104

SHANE R. THYE ET AL.

of three or more people who seek individual profit or gain. (2) The social structure provides incentives for the actors to exchange with at least one other actor within the network. (3) Actors initially choose partners to exchange with solely on the basis of who will provide the greatest individual gain or benefit. (4) Interactions occur in the context of an ongoing social unit, such as a network, group, organization, or community. (5) The immediate or local unit is salient to the people in it, that is, people recognize that they are interacting within a social unit. (6) There are larger, more distant units within which local units are nested. At least one of these is salient in that actors are aware that they are interacting within it. Together these scope conditions define a broad area of applicability. In the following, we first review early work that links interaction and emotion to cohesion processes and commitment. This section illustrates how instrumental exchange-based ties are transformed into expressive ties. Relational cohesion theory (Lawler & Yoon, 1996) is the focus in this discussion. The second section examines the next phase of the work, analyzing how structural and cognitive factors relate to the attribution of emotions to relevant social units, in particular groups based on different forms of exchange and networks of differing size. Social unit attribution of emotion is the mechanism by which individual emotions lead to affective group commitments. The affect theory of social exchange (Lawler, 2001) is the focus here. Finally, the emphasis is broadened to consider more generally how structural and cognitive dimensions of shared responsibility moderate emotional attributions to social groups, both locally and to more distant social entities. The theory of social commitments (Lawler et al., 2009) is the focus here.

RELATIONAL COHESION THEORY Relational cohesion theory (Lawler, Thye, & Yoon, 2000; Lawler & Yoon, 1996) (i) proposes structural conditions under which an interaction-toemotion process will occur and (ii) explains how and why this process leads to the relational unit becoming an object of awareness and commitment (Lawler & Yoon, 1996). Relational cohesion theory begins with a simple unit of social exchange, that is, a dyadic structure in which each person has an alternative (i.e., another person to exchange with or some standing payoff). The theory presumes that actors are likely to seek exchanges that will maximize their individual gain. Elaborating, the theory proposes that

Person-to-Group Ties

105

repeated exchanges are likely to generate positive emotional outcomes. Over time, the theory presumes that, in part, the emotions will be attributed to the relation itself. Thus, a sense of coming together or “relational cohesion” emerges from the positive emotions produced by exchange. In short, this fosters an expressive relationship between the actors that moves beyond the initial instrumental foundation for exchange. These expressive elements emerge to the degree that the emotions generated from repeated exchange facilitate the motivational and cognitive effects mentioned above. The theory does not claim that the expressive component fully overrides the instrumental aspect. Even with the expressive elements of a relationship in place, instrumental aspects may still motivate behavior. The theory connects exogenous structural conditions of interdependence to an endogenous affective process that ultimately leads to commitment. The theory is displayed in Fig. 1. The two structural conditions capture both the zero-sum and nonzero-sum dimensions of Emerson’s (1972b) approach to power dependence. The three measures of commitment involve both instrumental and expressive elements: (i) staying in a relation despite better alternatives, which is the standard instrumental measure of commitment, (ii) gift-giving, which entails the unilateral giving of token gifts to represent an expressive aspect of commitment, and (iii) contributing to a joint venture that carries an inherent risk such as a prisoner’s dilemma. The theory asserts that frequent exchange produces positive emotions that in turn facilitate perceptions of the relation as coming together or relational cohesion. The endogenous process is the core of the theory as portrayed in Fig. 1. The theory predicts a sequence of indirect steps
that is, the structural

Fig. 1.

Theoretical Model for the Theory of Relational Cohesion. Source: Adapted from Lawler and Yoon (1996).

106

SHANE R. THYE ET AL.

conditions of power and dependence should not have direct effects on cohesion or commitment. Said differently, controlling for interaction frequency and positive emotion there should be no direct effects of structural interdependence (power) on commitment. As detailed below, this sequence of steps has been verified in a number of empirical settings. The emotional mechanism stands in contrast to the classic answer to the commitment problem in social exchange theory
that being uncertainty reduction (Emerson, 1972a; Kollock, 1998). The classic explanation for commitment is that as actors exchange, they become more familiar with the actions and preferences of others and maintain that relationship to avoid the uncertainty associated with new exchange partners or markets. Thus, repeated exchange, via uncertainty reduction, produces commitment to a given partner. Relational cohesion theory provides an alternative and complementary causal path by which cohesion is generated through positive affect. We next discuss some of the empirical research and tests of the theory.

Tests of the Theory in Dyads and Triads The theory has been tested across a significant number of studies, both inside (Lawler et al., 2000; Lawler & Yoon, 1993, 1996, 1998) and outside of the laboratory (Price & Collett, 2012; Taylor & Pillemer, 2009; Yoon & Thye, 2002). Here we discuss the most important tests and extensions. Relational cohesion theory was first tested in 1996 by Lawler and Yoon. Three distinct experiments were conducted, each examining a separate form of commitment behavior (recall from the prior discussion that the specific measures of commitment are stay behavior, gift-giving, and contributing to a joint venture). Each experiment examined the exchanges of subjects (college students) who represented companies negotiating the price of a product (e.g., iron ore); one was buying and the other was selling. The actual negotiations between subjects took place over a computer. Subjects had no face-to-face contact and no reason to expect future interaction beyond the study itself. Importantly, these conditions should militate against cohesion and commitment. In each condition, there were 12 episodes of negotiation. Subjects were told to think of each episode as a distinct year. Further, episodes were independent of each other, meaning each outcome had no relation to the prior outcome. There were three to five rounds within each episode. The negotiation for each year would end when subjects agreed on the terms of the exchange, or if no agreement was reached by the end of round 5. The only communications between subjects

Person-to-Group Ties

107

were a series of numbers that represented individual offers and points of agreement within a given round of negotiation. The initial test employed a 2 × 2 fully-crossed experimental design; unequal dependence (relative power) crossed by mutual dependence (total power). These manipulations were implemented by giving subjects an alternative partner to exchange with in the event that they failed to reach agreement in their focal dyad. Subjects were given a probability distribution of agreements at different levels of profit for the subject, and the expected value of the alternative manipulated equal
unequal and low
high total power. Unequal dependence was manipulated by giving one subject a larger alternative profit (expected value) relative to their partner. High mutual dependence was achieved by giving both subjects worse alternative exchanges relative to exchanging with each other, thus making them more dependent on one another. The expected value of the alternative was always lower than the expected value of the exchange, but was varied depending on condition. Because of the availability of the alternative, exchange between partners was somewhat problematic (grand mean = 0.62), a condition important to studying the variability of the frequencies of exchange. All variables from the theoretical model were measured during the experiment (see Fig. 1). Measures of the endogenous process were taken after the eighth round of negotiation. Exchange frequency was measured as the proportion of rounds during which subjects reached agreement. Emotions were measured with a series of questionnaire items that examined pleasure/satisfaction (e.g., contented/discontented, happy/unhappy, pleased/displeased, satisfied/unsatisfied, and joyful/not joyful) and interest/ excitement (e.g., excited/bored, motivated/unmotivated, enthusiastic/ unenthusiastic, energetic/bored, interested/not interested). Relational cohesion was also measured via an index of questionnaire items regarding how subjects perceived the relationship (e.g., close/distant, divisive/cohesive, converging/diverging, cooperative/conflictual, integrating/fragmenting, solid/fragile, team oriented/self-oriented). The three measures of commitment were administered across the three experiments. These included stay behavior, gift-giving, and contribution to a joint venture that were administered at the end of each episode from episodes 9 to 12. To measure stay behavior the alternative payoff for each subject was increased such that they were close to the expected profits from an agreement in the focal dyad. Thus, the total number of agreements across the last four episodes is the specific indicator of stay behavior. Giftgiving was examined by giving subjects the option to give a small token gift after each of the last four episodes. To measure contribution to a joint

108

SHANE R. THYE ET AL.

venture, a prisoner’s dilemma was presented to the subjects after each of the last four episodes. In both the gift-giving and contribution conditions, subjects had no knowledge of their partner’s actions until after the study was complete. This was to ensure that the commitment behaviors themselves did not become a part of the negotiation or exchange. Results from the three 1996 experiments provide strong support for relational cohesion theory. Equal power relations generated more frequent exchange relative to unequal power relations. High total power relations also generated more frequent exchange relative to low total power relations. The endogenous frequency-to-emotion-to-cohesion links also were strongly supported. Exchange frequency was positively related to positive emotions, and importantly, the effect of the power-dependence structure on positive emotions was indirect through the frequency of interaction. Also, theoretically consistent, exchange frequency had an indirect effect on perceived cohesion through positive emotions. These results were predicted by the theory. To examine the final theoretical link, each of the three commitment behaviors was regressed on all important variables within the theory. Again, as expected per the causal model (see Fig. 1), relational cohesion was the primary cause of commitment behavior. The only other variable with a significant effect was a positive effect of exchange frequency on stay behavior. The results, taken together, provide exceptionally strong support for the theory. All predicted effects were found and they were found in ways that were theoretically predicted. One unexpected significant relationship was the direct positive effect of exchange frequency on relational cohesion. This most likely represents an uncertainty-reduction process that is complementary to the affective process predicted by the theory of relational cohesion.

Pockets of Cohesion Shortly after the initial test, Lawler and Yoon (1998) examined the relational cohesion process in a network context, the “stem” network (Fig. 2). In this four-person network, each person may exchange with, at most, one other person on a given episode. This creates a condition wherein Anna (A) can exchange with up to three others while Beth (B) and Gina (G) may exchange with only two. The person with the least amount of power, Diane (D), has only one person (A) with whom to exchange. As predicted by network exchange theories, this study found that the stem network tends

109

Person-to-Group Ties

Fig. 2.

Stem Network. Source: Adapted from Lawler and Yoon (1998).

to “break” into component dyads such that B and G (the equal power dyad) tend to exchange at rates comparable to that of A and D (the unequal power dyad). Yet, despite the lack of difference in exchange rates, pleasure/satisfaction and perceptions of cohesion were much higher in the B
G relation relative to the A
D relation. This coheres with a key prediction from relational cohesion theory, that unequal power weakens the endogenous interaction-to-emotion-to-cohesion process. Said differently, the relational cohesion process tends to operate more strongly in the B
G equal power dyad. This essentially demonstrates that dyadic “pockets of cohesion” can emerge in larger networks if the conditions specified by relational cohesion theory are present. The study suggests that relational cohesion processes operate and are sensitive to processes beyond the immediate dyad.

Extension to Productive Exchange Subsequent studies extended the theory to three-person groups that interact via other forms of exchange (Lawler et al., 2000). Two primary motivations drove this work. We wondered how the endogenous process might operate (i) in a three-person productive exchange context, and (ii) how this process might be related to the uncertainty-reduction process at the center of social exchange explanations of commitment. Productive exchange involves three or more actors who contribute to a joint good that no actor can produce

110

SHANE R. THYE ET AL.

solely or with a single partner. An example is three colleagues who produce a collaborative paper that requires the unique talents of each. In more conceptual terms, the exchange occurs between each actor and the group entity. Coordination and trust are important factors here as there is significant potential for free riding. In game-theoretical terms, a productive exchange is framed best as an assurance game as opposed to a social dilemma. Two experiments were run to gain insights into these issues (Lawler et al., 2000). The research procedures were similar to those of the 1996 tests, with a few differences owing to the different exchange context. First, there were 16 episodes (years) in which subjects could decide to contribute or not. These decisions were made simultaneously, making salient issues of coordination and trust. Second, the design again was a 2 × 2 relative power by total power, but in this study the alternative was a group instead of an individual. Finally, in order to examine uncertainty reduction, a questionnaire measure assessing perceived predictability (uncertainty) of others’ behavior was administered along with the emotion measures (items included). The first experiment examined gift-giving as the measure of commitment whereas the second examined contributions to a social dilemma. Results provide strong support for the emergence of relational cohesion in a productive exchange context. As predicted, productive exchange rates were positively related to dependence on the group, and were higher when individuals were equally dependent on the group relative to conditions of unequal dependency. The frequency of productive exchange generated both positive emotions (pleasure/satisfaction and interest/ excitement) and, importantly, the perceived predictability of others’ behavior. This is direct evidence that both positive emotion and uncertainty reduction operate as complimentary independent pathways between power dependence and commitment behavior. However, the mediating processes are somewhat different. Predictability had no effect on cohesion whereas positive emotion did have an effect, indicating a stronger role for emotion in the generation of cohesion. Further, predictability had a stronger effect on contributions to a social dilemma, while positive emotion had a stronger effect on gift-giving. This suggests that uncertainty reduction is a source of instrumental behaviors (where free riding may occur), whereas emotions are a source of gift-giving (a more symbolic gesture). In sum, this study affirmed the relational cohesion process in a productive form of exchange and verified that positive emotions and uncertainty reduction operate through distinct pathways to produce commitment behaviors.

111

Person-to-Group Ties

Extension to Organizations Two years later, Yoon and Thye (2002) asked if aspects of the relational cohesion process might help explain commitment in an organizational setting. Extending the logic of the prior relational cohesion studies, they offer a dual process model that suggests features of the work environment (autonomy, pay, tenure, and so on) impact job satisfaction (a form of positive emotion) and perceptions of organizational support (global beliefs that the organization supports and cares about the employee). Specifically, the model presumes that job satisfaction and overall organizational support represent emotional and cognitive processes that mediate the impact of work conditions on organizational commitment. They tested the model using survey data from a sample of employees drawn from two large Korean organizations (N = 2,443). One advantage of this data is that it provided an opportunity to see if the cultural and normative processes unique to Korea would affect the emotional precursors of commitment and if the relational principles of commitment developed in the United States would generalize to Eastern settings. Overall, the results provided clear and consistent support for the model. The main findings are that job satisfaction and perceived organizational support operate through independent channels to mediate the impact of work conditions on commitment. At a theoretical level, the results are consistent with relational cohesion theory; employee interactions produce emotions and cognitions that independently yield commitment. By the early 2000s, relational cohesion theory and subsequent tests had focused on how structural conditions facilitate interactions between actors that, when successful, unleash an emotion to cohesion process. At that time, much research in the exchange tradition had focused on negotiated exchange among actors connected in larger exchange networks (see Cook, Emerson, Gillmore, & Yamagishi, 1983; Markovsky, Skvoretz, Willer, Lovaglia, & Erger, 1993; Markovsky, Willer, & Patton, 1988; Willer, 1999). Negotiated exchange contexts entail both cooperative and competitive elements. As the scope of relational cohesion theory expanded from explaining dyads (and those embedded in networks) to three-person productive exchange groups, we asked how the theory might deal with even larger networks
those similar to the networks that received so much attention in the mid-1980s to early 1990s. Such exchange networks were conceived as “pure” exchange networks that provide nothing more than opportunities for and constraints on interaction. We suspected that there might be conditions under which actors in competitive exchange networks come to view

112

SHANE R. THYE ET AL.

the overarching network as a group entity. Thus, the next phase of our work extended the theory to an even broader, more complex unit
multiactor exchange networks where other network dynamics may unfold.

Being Able to Choose Exchange Partners Moving toward this more network-oriented view, Lawler, Thye, and Yoon (2006) examined how cohesion and commitment might be altered in structurally enabled versus structurally induced relations. In this study, actors exchanged in a box network (Fig. 3), wherein one relation (A
B) provides more value than any of the others. This creates a condition where A and B exchange voluntarily (we call this relation structurally enabled), while G and D exchange because there is no better alternative (we call this relation structurally induced). The overall idea was that exchange partners might perceive greater control and develop stronger affective relations in the enabled A
B relation. There are good theoretical reasons to believe so; a greater sense of control or autonomy has been shown to produce a variety of positive emotional reactions (Bandura, 1997; Izard, 1991; Westcott, 1988). By extension, we reasoned that actors in structurally enabled relations would develop higher levels of emotion and cohesion relative to actors in induced relations. The study confirmed these predictions. Actors in the A
B enabled relation experienced greater exchange frequency, higher levels of pleasure/satisfaction and interest/excitement, greater perceptions of cohesion, and greater perceptions of control. The differences between the enabled and induced relations were consistent at virtually every moment in the endogenous process. Overall, the theory nicely extends to explain cohesion in this broader network context.

1,200 A

B

600

600

G

D 400

Fig. 3.

Box Network. Source: Adapted from Lawler et al. (2006).

113

Person-to-Group Ties

Can Networks Become Groups? The following question motivated this next line of inquiry: Can networks of agents who compete against one another to acquire scarce resources, under some circumstances, come to see the overarching network as a group in its own right? Theoretically, we ask how the dyadic relational cohesion process can be extended and transformed to produce network-wide cohesion (Thye, Lawler, & Yoon, 2011). This project developed a new conception of structural cohesion to address this issue (cf. Moody & White, 2003; Wasserman & Faust, 1994). Structural cohesion is essentially a networkbased measure that takes two properties of the network into consideration: the prevailing power in the network and its density
both of which are suggested by relational cohesion theory. The basic idea is that the relational cohesion process, at the dyadic level, is most likely to occur in complex networks that (i) have a large proportion of equal power relations and (ii) are configured such that most people in the network can interact with most others. In network terms, we developed a new measure of structural cohesion comprising a numerical index that considers the prevailing power conditions in the network (strong power, weak power, equal power) and the level of network density. Extrapolating from the work on power in dyads, we reasoned that equal power networks (e.g., isolated dyads and triangles) would display high levels of relational cohesion for the same reason that equal power dyads do. At the same time, for those dyad-generated emotions and perceptions of cohesion to have effects at the group level, individuals must be able to interact with many other group members. The second element of our structural cohesion measure captures this aspect of the interaction, that is, network density. Thus, while prior measures of network cohesion tend to focus on graph-theoretic ideas (i.e., the number of cut points, cliques, or structural isomorphism), ours focuses on the network properties that produce interactions and enable network members to see themselves as members of a common group. The broad sweeping implication is that actors will come to perceive the network itself as a social unit, and direct their behavior at least in part toward the larger group or entity. At issue are the fundamental processes of group formation. Five exchange networks were examined in order to test for the networkto-group phenomenon (Thye et al., 2011). Each network differed in terms of structural cohesion (see Table 1). In all networks, subjects engaged in a series of negotiated exchanges, similar to that of prior exchange network research. Negotiations involved dividing 32 units of profit during each of .

114

Table 1.

SHANE R. THYE ET AL.

Predicted Group Formation across Five Competitive Networks.

Structural Cohesion and Predicted Group Formation

Name and Network

Highest

4-Full

Equal power

1.00

Triangle

Equal power

0.67

4-line

Weak power

0.87

3-branch

Strong power

0.67

Strong power

0.50

Moderate Lowest

4-branch

Type of Network Power

Likelihood of Inclusion L(i)

Source: Adapted from Thye et al. (2011).

the 20 episodes of exchange. Actors could exchange with only one partner during each exchange episode, so the networks were “negatively” or “exclusively” connected. After episode 16, questionnaires were administrated examining positive emotion, uncertainty, and dyadic cohesion. Cognitive perceptions of the group were measured after the 20th episode. This index consisted of items that asked to what degree actors were mutually dependent, had a common bond, and were in a similar situation with others in the network. We also measured a behavioral dimension of group formation. This consisted of a dictator game that was administered following episodes 17
20. Specifically, actors were given 100 profit points to distribute as they saw fit across all network members at the end of the last four episodes. Subjects did not find out the results of the dictator game till after episode 20 in order to prevent the use of the profit points in a manner related to the episodic exchanges. The results support the predicted effects of structural cohesion on perceived group affiliation and resource sharing at the network level (i.e., group formation). Structural cohesion increases rates of exchange, and consistent with the logic of relational cohesion theory, this leads to increased positive emotion and cohesion in the dyadic relations. Increased cohesion in the dyadic relations is important as this leads to (i) cognitive perceptions of the network itself as a group and (ii) an increased likelihood of resource sharing among members of the perceived group. Thus, the same relational cohesion process that unfolds in dyads can be used to understand when networks of self-interested agents, each vying for scarce resources, come to see themselves as members of a common group.

115

Person-to-Group Ties

The importance of this phenomenon is the bridging of both micro- and macrolevel processes. This research shows that the relational cohesion process can have macrolevel effects. Cohesion, generated at the dyadic level by way of interaction and positive emotion, leads to perceptions of a group affiliation to the network within which the dyad is nested. Cohesion in dyads essentially leads people to infer that others in their network, even ones with whom they do not interact, are collectively oriented and members of a common entity. In this way, individuals form “group” ties by way of the dyadic interaction that occurs in structurally cohesive networks. Importantly, this occurs even when the “network” provides no benefits, is not a source of common identity or affiliation, and comprises self-interested actors who seek to maximize their own profit. The theory identifies an emotional process through which this occurs (see Thye et al., 2011). In conclusion, interaction at the micro- or immediate level is the focus of relational cohesion theory (Lawler & Yoon, 1996; Thye et al., 2002). The theory does not make predictions about group or macrolevel cohesion processes. However, as seen above, Thye et al. (2011) provide evidence that structurally cohesive networks facilitate a collective sense of shared experience throughout actors in a network, even when actors are only able to interact with limited number of others. The network may come to constitute a common focus for an actor (Collins, 1975), actors may infer that others located in similar contexts experience similar emotions (Lawler, Thye, & Yoon, 2013), and positive emotions can spread across different relations to make the network salient as a group entity (Barsade, 2002).

Comparing Dyads and Triads More recently, we revisited a classical sociological question posed, but not entirely answered, by Simmel (Yoon, Thye, & Lawler, 2013). Do dyads or triads have a greater capacity to generate group formation? Simmel argues that triads tend to generate greater cohesion and solidarity than dyads. Social interaction in dyads is more personal, involves more affect or emotion, and reveals greater variability because individual personalities have freer reign. Triads, on the other hand, are the smallest structure in which the group is an entity distinct from any given individual. Simmel asserts that triads tend to constrain emotions, reduce individuality, and generate behavioral convergences or uniformity. To examine Simmel’s proposition, we compared a dyad in which neither actor has an alternative partner with a triad in which people are competing

116

SHANE R. THYE ET AL.

with one another to secure an exchange partner. Exchange in the triad creates a condition of tertius gaudens (“rejoicing third”) in Simmel’s terms. One might expect that such exchanges attenuate the development of the group ties rather than reinforce them. In order to develop a cohesive relationship, individuals in triads must somehow resolve the problem of potential exclusion. In dyads, exclusion can also occur, as when the two people cannot resolve conflict due to a misalignment of personal interests. The results support Simmel’s classic prediction regarding the stabilizing effects of a third party in triads. Triads generated more uniformity and convergence in exchange behavior (frequency of agreement, equality of profit divisions) and greater sense of cohesion. This study also supports the idea that positive affect and uncertainty reduction are dual processes that lead to cohesion and commitment in exchange relations and networks (Lawler et al., 2000). The results clearly show that these mediating processes differentiate dyads and triads. Cohesion in dyads is based solely on positive affect net of the effect of frequent or repetitive exchange, whereas cohesion in triads is based primarily on an uncertainty-reduction process. The importance of positive affect in dyads could be due to the sense of relief or the pleasure in finding shared interests with another on whom one is highly dependent. Uncertainty is not clearly situational here but mainly concerns the orientation or approach of the other to the negotiations. In a triad, uncertainty is more situational in that each has alternative partners available and they are competing for partners. The relative importance of uncertainty reduction in triads and affect in dyads is generally consistent with Simmel’s ideas about impact of a third actor and the significant role of emotion in the dyad.

THE AFFECT THEORY OF EXCHANGE The affect theory of social exchange (Lawler, 2001; Lawler et al., 2008, 2009) builds and expands upon the theoretical contributions of relational cohesion theory. In particular the affect theory explicates the broader conditions by which emotions are attributed to relational or group entities, focusing in particular on the tasks people undertake in their social interactions. The theory specifies both cognitive and structural factors under which actors are likely to attribute positive emotions to the group as a salient social object. Unlike Thye et al. (2011), the affect theory of social exchange does not explicitly apply to an exchange network context.

117

Person-to-Group Ties

The affect theory focuses on social interaction in a more abstract, broader sense. Joint tasks and shared responsibility are central theoretical concepts in the affect theory. These concepts and the overall theoretical framework are discussed next.

Joint Tasks and Shared Responsibility Tasks organize instrumental behaviors and typically include methods or procedures to successfully complete the task and reach a goal. All tasks include both structural (objective) and cognitive (subjective) dimensions that shape and constrain how actors define and approach the task and how they interpret success or failure (Lawler, 2001; Lawler et al., 2008, 2009). Tasks may also vary along an individual to collective continuum. Getting dressed is usually an individual task while playing basketball constitutes a collective or joint task. The affect theory of social exchange focuses on joint (collective) tasks, and how the jointness of a task relates to perceptions of shared responsibility. A joint task involves two or more actors who cannot accomplish the task on their own. The actors must coordinate their actions with each other or face potential costs associated with task failure. Importantly, joint tasks as we refer to them here are not just in reference to the jointness of an outcome. Instead, the affect theory focuses on the activities and behaviors within the context of a joint task, and how distinguishable individual contributions are. The question is, how indistinguishable are individual contributions to task success? Where joint tasks fall along this dimension has important implications for the development of person-group affective ties.

Structural and Cognitive Task Dimensions Joint tasks can vary in a variety of ways that differ on structural and cognitive dimensions. Both dimensions can be positively, but also negatively, related to one another in any given joint task context. For example, a work team composed of four members may be assigned a specific task within an organization. The structural dimension of jointness is high because it is mandated to be a team product. However, presume that within the process of accomplishing the task one individual may have undertaken a majority of the work. From that individual’s point of view, and possibly the other group members as well, the subjective dimension of jointness will be low.

118

SHANE R. THYE ET AL.

This example demonstrates the dual nature of jointness and points to the two fundamental conditions the theory specifies for social unit attributions. Structural jointness refers to the degree to which each individual’s efforts or contributions to a joint task activity are separable (distinguishable) or nonseparable (indistinguishable) (Lawler et al., 2008). The focus here is on how individual contributions become blended or combined into a team or group outcome, an idea traced to Oliver Williamson’s analysis of governance structures (Williamson, 1975). Structural jointness is the precursor to cognitive jointness. Given a structurally joint task, individuals are likely to make subjective inferences regarding how well individual contributions were actually combined within a group task. This constitutes the cognitive dimension of jointness. The cognitive condition is captured by the degree to which the joint task or activity generates a sense of shared responsibility. In the affect theory, shared responsibility is the key moderator determining when emotions are attributed to group units. A sense of shared responsibility increases the salience of the group, therefore increasing the likelihood that positive emotions will be attributed to the group in the context of task success. The subjective dimension is closely related to the objective dimension but can vary by a number of other factors. One factor of particular importance is the framing of the task by leaders or authorities. When leaders define tasks in collective terms, members of the group are more likely to attribute individual feelings to the social unit as a result of higher perceived shared responsibility. Given this reasoning, there are two core propositions within the affect theory of social exchange adapted from Lawler (2001): Proposition 1. The greater the nonseparability of task activities and outcomes, the greater the perception of shared responsibility. Proposition 2. The greater the perception of shared responsibility for success or failure at a joint task, the more inclined actors are to attribute the resulting global emotions to social units (relations, networks, or groups). Taken together, these two propositions posit a causal chain by which task properties generate shared responsibility, which in turn fosters social unit attributions of emotion. A theoretical link can be drawn from the affect theory to relational cohesion theory. Relational cohesion theory predicts that conditions of high total power and equal relative power lead to greater exchange rates, and in turn, triggers the interaction-to-emotion-tocohesion process. High total power and equal relative power entail greater

119

Person-to-Group Ties

outcome interdependence (Lawler et al., 2008). From the perspective of the affect theory, outcome interdependence may be cast as a kind of joint interaction. The two theories are different but identify complimentary processes. The affect theory of social exchange explicates how and why social unit attributions are possible within social relations. This stands in contrast to the attribution processes emphasized by psychologists, where the focus is on internal (to the person) versus external (situational) attributions (see Graham, 1991; Kelley, 1967; Weiner, 1985). Our work simply points to how the salience of the social unit, and the emotional by-products of interaction, can in part overcome these otherwise common attributional tendencies. Importantly, the outcome of this process may extend above and beyond material benefits. In other words, through this process, groups can become valued in and of themselves, take on normative and moral properties, and even in some cases, become an important source of identity.

Predictions for Forms of Exchange In social exchange theorizing, there are four basic forms of exchange: negotiated, reciprocal, generalized, and productive (Emerson, 1981; Lawler, 2001; Molm, 1994) (Fig. 4). To test the affect theory of exchange, we conducted a study that compared the impact of all four forms of exchange on the development of group ties. In negotiated exchange, subjects negotiated how to divide a resource pool with one or both of their respective partners. In the reciprocal exchange condition, resources were given unilaterally to

Fig. 4.

Form of Exchange. Source: Adapted from Lawler et al. (2008).

120

SHANE R. THYE ET AL.

one or both of their partners (and can be received from either partner as well). The generalized exchange network allowed subject A to give to B, B can give to C, and C can give to A. Finally, in productive exchange, subjects contributed resources to a common effort from which individual benefit could be derived. Negotiated and reciprocal exchanges entail “direct” exchange, meaning that resources are given and received directly from one actor to another and vice versa. Productive and generalized exchanges entail “indirect” exchange where another entity (in the case of productive exchange, the group, and for generalized exchange, another person) mediates the benefits that each individual can receive. To our knowledge, this study represents the first time all four forms of exchange were studied under the same experimental protocol. The theory makes three main predictions. First, of the four forms of exchange, productive exchange will generate the highest levels of group cohesion and the strongest person-to-group ties. Productive exchange entails the greatest degree of jointness, and therefore, by the logic of the theory, should produce higher levels of perceived shared responsibility relative to the other forms of exchange. Second, generalized exchange will produce the lowest levels of group cohesion and person-to-group ties, due to relatively low structural jointness and resulting perceptions of shared responsibility. Third, negotiated and reciprocal exchange, that is, the direct forms of exchange, will fall in-between productive and generalized exchange, but negotiated exchange involves greater task jointness compared to reciprocal exchange. Thus, it is predicted that negotiated exchange has the greater capacity for person-to-group ties. The development of group cohesion and person-to-group ties is predicted to be as follows: Productive > ½Negotiated > Reciprocal > Generalized Productive exchange is predicted to facilitate person-to-group affective ties owing to the task structure. In productive exchange, three or more actors work together to jointly produce an event or good, from which each individual actor benefits. This joint good cannot be produced by any single individual or dyadic pair. In generalized exchange, givers and receivers are not matched in pairs. Because each individual act of giving or receiving involves only two actors, there is lower task jointness. Thus, based on the theoretical principles above, productive exchange should generate stronger social commitment than generalized exchange because productive exchanges are more likely to produce positive emotions and social unit attributions of the produced emotions.

121

Person-to-Group Ties

The theory predicts that negotiated exchange will generate stronger affective ties than reciprocal exchange. In negotiated exchange, agreements between actors or parties are required for benefits to be received. The task is inherently joint as benefits cannot be gained unless both parties agree to the terms of the (typically binding) contract. An example would be a professor with another job offer negotiating a raise with a Dean. The task of negotiating a new contract is inherently joint as both parties share in the responsibility for the outcome. In comparison, reciprocal exchange involves unilateral acts of giving with no specific expectation regarding reciprocity. As such, there is less jointness in reciprocal exchange and the theory predicts a weaker sense of shared responsibility in reciprocal exchange (see Lawler, 2001; Lawler et al., 2008).

The Experimental Test The predictions were tested via an experiment in which subjects had to decide whether or not to give resources to one or both available partners during each of 48 opportunities to interact. Subjects were told that they represented small computer companies, and that each company possessed resources of value to the companies represented by their potential exchange partners. Subjects were told that their pay was contingent upon their success at the task, with success being operationalized as maximizing profits for the company they represented. In the event of no agreement, subjects would receive a default payoff from another company beyond the focal triad. This alternative option is important because it decreases the baseline incentive to exchange, meaning that company profits were not entirely based on their exchanges with the two available exchange partners. Questionnaires were administered both midway through the experiment and at the conclusion to measure theoretical constructs at the appropriate point in time. Specifically, we measured perceptions of shared responsibility, perceived cohesion, positive emotions, and rates of exchange behavior (giving) over the course of the interaction. The post questionnaire included measures of affective commitment to the social unit, which in this case was the network, as well as other measures of network cohesion. The predictions were strongly supported by the experimental results. First, as predicted, productive exchange generated the strongest effects on all theoretically relevant dependent variables. Productive exchange produced the highest rates of exchange behavior, generated more positive

122

SHANE R. THYE ET AL.

feelings, had higher levels of perceived cohesion, and had the strongest reported affective attachments to the social unit. Second, productive exchange was the only condition in which perceptions of network cohesion actually increased between the midpoint questionnaire and the final questionnaire. Third, logically consistent with the theory, productive exchange generated the highest perceptions of shared responsibility. Overall, the results strongly indicate that productive exchange is an important form of interaction at the microlevel that can have important effects on the macrolevel person-to-unit ties. The key prediction was confirmed. Also, consistent with our predictions, generalized exchange produced the weakest effects on theoretically relevant variables. Lower rates of exchange behavior (frequency), weaker positive emotions, lower perceived network cohesion, and weaker ties to the network were all present in generalized exchange condition relative to the three other forms of exchange. Also of note, generalized exchange was the only form of exchange in which perceptions of network cohesion actually deteriorated over the course of the interaction. This suggests that generalized exchange has limited internal potential to promote or sustain emergent group ties (Lawler et al., 2008). As will be addressed shortly, these results present a counterpoint to previous research suggesting the important role of generalized exchange in the formation of group solidarity and the facilitation of pro-social behavior (see Bearman, 1997; Ekeh, 1974; Gilmore, 1987; Molm, Collett, & Schaefer, 2007a; Uehara, 1990). We turn next to our most recent theorizing and look to the future of our research program.

THE THEORY OF SOCIAL COMMITMENTS The theory of social commitments, posited in a recent book (see Lawler et al., 2009), combines theoretical contributions from relational cohesion theory and the affect theory of social exchange. The crux of the theory focuses on how individuals form and maintain relational ties to groups as a function of joint tasks. Three ideas form the overarching framework of the theory. First, when people work together on a task, they are likely to feel good when their work is successful and feel bad if not. Second, if joint tasks and experiences are repeated over time, individuals may come to attribute their feelings at least in part to the context they share with others. Third and last, given a joint task, it is plausible that individuals will attribute their

Person-to-Group Ties

123

feelings to the social unit, be it a group, sports team, community, or organization. While these ideas are in our prior theorizing, the theory of social commitments broadens the implications to examine forms of interaction other than exchange. In essence, the theory explicates the broader implications of developing ties to dyads, networks, groups, organizations, communities, or even nation-states. The central theoretical question addressed is, Under what conditions are people likely to attribute their feelings to group level social units within the context of repeated social interaction? Given that there are multiple levels of social units, from the dyad to the local group to even larger units such as communities and organizations, a second concern is, Under what conditions are emotions and feelings attributed to local groups versus broader more distant social units? To answer these questions, we use insights of an earlier theory of nested-group commitments (Lawler, 1992; Mueller & Lawler, 1999). The theory of social commitments incorporates theoretical ideas about ties to local and larger social units from that prior theory. The theory of nested-group commitments (Lawler, 1992) distinguishes proximal (local, immediate) from distal (removed, overarching) groups. The theory asserts that positive events and feelings are more likely to be attributed to the proximal group, whereas negative events and feelings are more likely to be attributed to more distal groups. Therefore, within a nested-group context, positive feelings generated from group interaction are likely to facilitate the development of solidarity and cohesion within the local group but not necessarily to any overarching, larger group. The theory predicts that attributions of positive events to the proximal group are likely to happen when tasks are designed and controlled locally. Whereas, when tasks are designed and controlled by a more distant group, positive events are relatively more likely to be attributed to the distal group. The theory does not necessarily specify an inverse relationship between local and distant groups. That is to say, crediting the local group for positive events does not necessarily imply failing to credit a more distal social unit. The broader implication of this line of thought is that grouplevel mechanisms of responsibility inform people about the degree to which tasks are joint, the degree of shared responsibility within the group, and where (local vs. distal) that responsibility lies. The theory of social commitments indicates how the effects of control, responsibility, and accountability bear on the strength of affective ties people develop to local groups and larger ones in which they are nested (Lawler et al., 2009).

124

SHANE R. THYE ET AL.

Propositions The broader, integrative theory of social commitments proposes that joint tasks generate a sense of shared responsibility and that local, more immediate social units are typically given greater responsibility for positive events and feelings than larger, more removed social units (see Lawler, 1992; Lawler et al., 2009). The theoretical argument can be expressed as a series of four broad propositions: (1) The greater the nonseparability of individual contributions to task success or failure, the stronger the perceptions of shared responsibility. (2) The stronger the perceptions of shared responsibility for success or failure at a joint task, the greater the likelihood that resulting feelings and emotions will be attributed to the group or social unit. (3) Social unit attributions of positive emotion produce stronger affective attachments to the group, whereas attributions of negative emotion weaken affective attachments to the group. (4) Stronger affective commitments lead to more group-oriented behavior, such as an increased likelihood to collaborate, increased willingness to compromise when individual and group goals conflict, and to, in general, give more effort on the group’s behalf.

RELATION TO OTHER WORK Generalized Exchange While we predict that generalized exchange should generate fewer and weaker person-to-group ties, there is a good deal of research suggesting that generalized exchange is an important form of interaction that gives rise to cohesion and solidarity (Bearman, Moody, & Stovel, 2004; Ekeh, 1974; Molm, 1994; Molm et al., 2007a). In contrast with our predictions, Molm et al. (2007a) found higher levels of cohesion and solidarity in generalized exchange, relative to reciprocal and negotiated exchange. However, there are significant differences between the programs of research that may have accounted for the differing results. Most important, from our standpoint, is that our study included an opportunity cost for giving in generalized exchange, whereas there is no such cost in the typical research of Molm and colleagues. In other words, in our study, actors could reap some benefit (i.e., retain profit) by not giving resources to another, and thus, they would forego this benefit if they chose to give. In Molm’s standard condition, there is no such opportunity cost for giving; sending resources to

Person-to-Group Ties

125

another subject is essentially “free” for all actors. This difference alone could account for the fact that we found much lower rates of giving in generalized exchange than did Molm. When Molm’s research included such a cost (see Molm et al., 2007a, p. 236), the rates of giving were more similar to the rates in our study. We suspect that one reason for the difference in solidarity and cohesion across the two studies is the differential rates of exchange behavior (giving), which are due to these different cost conditions. Recall that greater rates of exchange facilitate greater positive emotions in the context of interaction success, ultimately leading to higher cohesion. Simple but important protocol differences across these studies likely produce variance in the levels of interaction or exchange, which in turn account for the differing levels of cohesion. We hypothesize that generalized exchange will engender cohesion if there are strong incentives to give resources (or a lack of barriers to giving as there are in Molm et al., 2007a). Second, generalized exchange should lead to cohesion if the interactants already share a group affiliation or identity (see related discussion in Lawler et al., 2008). Generalized exchange has been shown to generate strong solidarity and cohesion when it increases the salience or reinforces a preexisting group identity or affiliation (see Ekeh, 1974, for an early example in the exchange tradition). Our theory and research protocol examine exchange under sparse conditions, and in doing so, focuses on the underlying exchange processes themselves. We purposefully eliminate the confounding effects of common or shared group identities. The experimental procedures were designed to examine the four different forms of social exchange and their potential to endogenously generate micro orders, net of the impact of preexisting social categories (Lawler et al., 2008). The logic of our theorizing suggests that generalized exchange will increase cohesion and solidarity when (i) the exchange reflects or is symbolic on an already established group identity or affiliation or (ii) there are strong structural incentives or corresponding lack of barriers to initiate exchange. Absent these two conditions our research indicates that generalized exchange has a relatively weak potential to create and maintain cohesion and solidarity. A more recent study that bears on the generalized exchange issue is that by Willer, Flynn, and Zak (2012), who compared two different Internet exchange websites. They argued that generalized exchange would generate a stronger group identity relative to more direct forms of social exchange. The two Internet sites studied were Freecycle and Craigslist. Freecycle is a website that allows users to give away items to other Freecycle users, with no strings attached (a form of group generalized exchange), whereas

126

SHANE R. THYE ET AL.

Craigslist entails formal direct exchange of goods between parties. As predicted, Willer et al. (2012) found that the Freecycle site had a stronger effect on solidarity in that it strengthened the identification with the group (i.e., other Freecycle users) compared to Craigslist users. This finding seems to conflict with our theorizing and results. We note that unlike Craigslist where the exchange of good is guided by market forces, Freecycle users share a preexisting group affiliation (i.e., a community of people dedicated to the unilateral giving of items to others in that community). Without this preexisting group affiliation, the development of a stronger group identity relative to Craigslist may not have occurred. In fact, the most comprehensive study to date directly comparing these forms of exchange in the purest sense indicates that it would not (Lawler et al., 2008). It seems more probable that in the presence of an exogenous shared identity, generalized exchange fosters greater cohesion and solidarity due to a strengthening of that shared group identity. In this respect, the users of Craigslist and Freecycle were undoubtedly different in many respects prior to the study of either website. Our research shows, ceteris paribus, it is unlikely that generalized exchange alone has much capacity to produce solidarity and cohesion. The relationship between exogenous social identities and how the forms of exchange combine to yield solidarity and cohesion deserves more attention in future research.

Negotiated and Reciprocal Exchange Another point of contrast between our theoretical research program and that of others involves the comparison of negotiated and reciprocal exchanges (Lawler et al., 2008; Molm, Schaeffer, & Collett, 2007b). The primary theoretical mechanism of concern for Molm and colleagues is the salience of conflict in negotiated exchange. Specifically, they assert that reciprocal exchange generates greater solidarity relative to negotiated exchange due to the increased salience of conflict in negotiated exchange. Our research program has asserted that negotiated exchange will generate stronger levels of solidarity and exchange due to stronger perceptions of shared responsibility relative to the reciprocal exchange setting. Molm and colleagues have found clear support for their hypothesis under conditions of unequal power, but have not found much support under conditions of equal power (see Molm et al., 2007b). Similarly, we find no difference in cohesion generated between negotiated and reciprocal exchange within an equal power context (Lawler et al., 2008). It is important to remember that the two research programs focus on two distinct theoretical mechanisms: perceptions of shared responsibility and

Person-to-Group Ties

127

the salience of conflict. Understanding when reciprocal and negotiated exchanges are likely to generate solidarity and cohesion may be dependent on establishing which mechanism is stronger in particular contexts, especially given that the experimental settings across the two programs differ in important ways. Kuwabara (2011) empirically took up the question in two experiments that examine how these distinct mechanisms affect the exchange process across different contexts. The first study compared two types of negotiated exchange: distributive and integrative. Distributive exchange involves dividing a fixed amount of profit, whereas integrative exchange provides the opportunity to expand the joint profit through joint problem solving. Integrative exchange is a more joint task and should therefore elicit stronger perceptions of shared responsibility and greater perceptions of cooperation. Study 2 compared one-way and two-way trust games, arguing that the two-way trust game constituted a more explicitly joint task. In study 1, integrative exchange generated the highest levels of cohesion, whereas distributive exchange produced the lowest levels of cohesion. In study 2, a repeated two-way trust game (in which the subjects rotated in the roles of truster and trustee) produced more cohesion than the standard one-way trust game (subjects played only one role). These results support both the role of joint tasks and shared responsibility (Lawler et al., 2008) and perceptions of a cooperative or nonconflictual structure (Molm et al., 2007b). Again, the nature of the task and the interaction context seem to trigger one or the other mechanism. Kuwabara’s (2011) research coheres well with our theoretical predictions, demonstrating that tasks with greater jointness promote attributions of positive emotion to the social unit. It also sheds light on the processes central to Molm et al.’s (2007a, 2007b) predictions; exchange settings that generate lower perceptions of conflict generate higher levels of solidarity and cohesion. It appears that negotiated exchange can have differing effects on cohesion, depending on which theoretical mechanism (shared responsibility vs. perceptions of conflict) is stronger or more salient in the setting. This does not imply a mutually exclusive relationship between the mechanisms. Rather, the nature of the task and the manner in which the interaction is conducted must be taken into consideration.

Joint Tasks and Group Ties Beyond the Lab Our theorizing has been tested recently in studies of commitment among nursing home staff and teachers. Taylor and Pillemer (2009) examined the

128

SHANE R. THYE ET AL.

effects that joint tasks and shared responsibility have on staff turnover in nursing homes. They conceptualized “caregiving” as a joint task in which contributions from different employees were nonseparable. In line with this theoretical logic, they predicted that success at caregiving would generate positive emotions and emotions directed toward the organization, in turn, making employees less likely to leave as reflected in lower turnover rates. Two longitudinal surveys (two waves, six months apart) examined 20 different nursing homes across New York State. Results indicate that the perceived success of caregiving has an indirect effect on turnover rates, mediated by positive feelings about the nursing home. There was not a direct effect, providing support for the central mediating role of positive feelings and emotions. Overall, the study demonstrates that joint tasks in which contributions are nonseparable generate commitment behaviors (staying, in this case) so long as such tasks engender positive feelings toward the organization. In a different context, Price and Collett (2012) examined cohesion and commitment (again turnover rate) among elementary teachers using the logic of the affect theory of social exchange. Using a nationally representative sample, they measured levels of task interdependence (shared control over school policy in different contexts), frequency of cooperative action, enthusiasm and satisfaction, perceptions of cohesion, and commitment to the school (intent to stay). Consistent with the theory, perceptions of shared control and responsibility as well as perceptions of cooperative interaction where positively related to positive emotions (measured as enthusiasm about teaching and satisfaction with the school). These positive emotions were related to stronger perceptions of school cohesion, and cohesion was related to an increased likelihood (intent) of remaining at the school. Both of the above studies provide supporting evidence for the theory beyond an exchange context and outside of the experimental laboratory.

Shared Responsibility and Collective Emotions The theory of social commitments applies to a broad array of interaction, even those in which actors cannot read each other’s emotional cues. That is to say, the theory applies even when individuals are interacting over some impersonal, non-face-to-face medium of communication. In fact, the first test of relational cohesion theory illustrated that common emotions and

129

Person-to-Group Ties

cohesion could emerge even when actors were physically apart. In such settings, collective emotions may not occur or be weaker due to the reduction in the potential for emotional contagion (Barsade, 2002). Nevertheless, it is possible that the commitment process still unfolds. Recent work by Lawler et al. (2013) aims to specify conditions under which “collective emotions” are likely to emerge, despite physical distance between individuals. Collective emotions are common feelings by members of a social unit as a result of shared experiences (Bar-Tal, Halperin, & De Rivera, 2007). Collective emotions involve mutual inferences or awareness of one another’s emotions even without direct evidence (e.g., facial expressions, communications, and bodily signals). Such inferences may generate collective emotions and thereby strengthen social unit attributions. The relevant question is as follows: When will actors who are not face-to-face infer that others have similar feelings? If actors infer common emotional states with others, this essentially “collectivizes” their feelings, transforming their individual feelings to collective feelings perceived to be shared by others in their group. Our solution to the question of collective emotions in the formation of group ties is expressed in two main points. First, as the relational cohesion process emerges, that is, repeated exchange fosters positive emotions; actors are likely to infer that their partners are experiencing the same emotional states. We reason that this is due to a “burden of proof” principle. Stated differently, people will infer that others involved in the same joint task will experience the same feelings and emotions, absent any indication otherwise. Second, social unit attributions serve as a mechanism that “collectivizes” individual feelings through emotional inferences. Collective emotions should increase the salience and awareness of a shared affiliation and its role in the group environment. In this way, social unit attributions and collective emotions work to strengthen each other, as both shed light on commonalities in the setting and emotions felt among group members.

NEW QUESTIONS As our research program continues to grow, it is currently expanding along a number of theoretical frontiers. Several questions are being addressed by projects at various stages of development. Three are highlighted and summarized below.

130

SHANE R. THYE ET AL.

Does the Theory Apply to Cooperative Groups? The development of group ties from social exchange processes has been the primary emphasis of our theoretical paradigm. That said, social commitments theory takes the principles and theoretical contributions from the exchange-based work and applies them to broader contexts, focusing on ties formed to larger-scale social units such as organizations. Social commitments theory applies to cooperative task groups in which individual and collective goals are not necessarily at odds. The theory predicts that task groups with joint tasks should generate stronger and more affective group ties than task groups with individualized tasks (see Lawler et al., 2009). A project in progress tests this idea (Thye & Lawler, 2010). Specifically, this project examines interaction in four-person groups who have 10 minutes to solve a collective problem. That problem is to rank order 15 items that will help you survive if you are lost at sea. The “lost at sea” task asks you to imagine that you and three others have been stranded in a rubber life raft that is afloat in the middle of the Atlantic Ocean. Your goal is to rank order a variety of items (shaving mirror, diesel fuel, chocolate, rum, mosquito netting, and so on) in the order of their importance for your survival. The task is desirable for a number of reasons: (i) it is engaging, (ii) has a known correct solution provided by the US Coast Guard, and (iii) is not intuitive (e.g., shaving mirror is very important for signaling, while mosquito netting is practically useless because there are no mosquitoes in the middle of the ocean). Subjects complete the task and along the way we measure emotions, cohesion, and commitment to others in the group. Specifically, this project asks the following question: What other kinds of jointness will promote person-to-group ties? Using the aforementioned procedures, we recently completed an experiment that manipulates (i) if interaction is joint or individual, (ii) if incentives (i.e., pay structure) are primarily based on individual efforts of the group effort, and (iii) if members share a common identity or possess distinct identities (manipulated as a preference for a common artist or for distinct artists). Social commitments theory contends that social interaction within joint tasks is necessary for the transformation of ties from instrumental to expressive. Shared interests and shared identities also may be necessary for developing enduring group ties, and in fact, much of micro-sociology emphasizes these factors. The current project will determine if the effects of these factors are additive or multiplicative.

Person-to-Group Ties

131

Do Status Hierarchies Weaken Group Ties? Another new project incorporates ideas from the theory of social commitments with those from status characteristics theory. Status characteristics theory (SCT) has a rich tradition of research in sociology (see Berger, Fisek, Norman, & Zelditch, 1977; Berger & Webster, 2006). The theory’s primary focus is to explain power and prestige gradients in collectively oriented task groups. The theory specifies how, in the absence of other information, actors form expectations of competence from characteristics such as age, gender, race, and physical attractiveness. In turn, these expectations impact task contributions, actor evaluations, social influence, and other elements of an observable power and prestige order. The key insight, and link to our current work, is that the preceding sentence implies that the proportion of work within a task group is unequal when status hierarchies are present. That is, SCT suggests that those perceived as more competent do a greater share of the work. Logically, we infer that this should have a bearing on perceptions of shared responsibility. We reason that status stratification in a task group should weaken the formation of group ties by weakening perceptions of shared responsibility. The broader implication is that status stratification weakens the objective dimension of task jointness, and consequently, the emotions (see Lovaglia & Houser, 1996) and perceptions that we have shown trigger person-to-group ties. These assertions are currently being tested (Vincent, 2013).

Does Relational Cohesion Have a Neurological Basis? The most recent branch of our work seeks to examine the neurological bases of relational cohesion in dyads. A phenomenon known as interbrain synchronization (see Dumas, Lachat, Martinere, Nadel, & George, 2011) is currently being examined in an empirical project being conducted at Kent State University by Kalkhoff and his colleagues (see Kalkhoff, Thye, & Lawler, 2011). Interbrain synchronization occurs when multiple individuals experience covariations in their neural activity (i.e., hyperconnectivity) during social interaction. At a very basic level, this phenomenon can sometimes be detected visually when raw electroencephalogram (EEG) signals for electrode pairs for two individuals begin to move in harmony, as if being generated by a single brain instead of two (Condon & Ogston, 1966).

132

SHANE R. THYE ET AL.

While interbrain synchronization is a fairly recent neurological discovery, synchronization in general is a fundamental aspect of human interaction (Collins, 1981), occurs across a variety of social contexts (Kalkhoff, Dippong, & Gregory, 2011), and can even be seen in the earliest moments of life between newborns and mothers (Condon & Sander, 1974). Of significant interest to us are the studies that have demonstrated a relationship between certain kinds of interaction and interbrain synchronization. Covariation in EEG signals across electrode pairs proximate to structures such as the anterior cingulate cortex (potentially related to joint attention) and the medial orbitofrontal cortex (potentially linked with cooperation), and those occurring within the beta band frequencies (related to attentional focus) and the gamma band frequencies (related to emotions) are of particular theoretical interest. Neuroscientists have found evidence of interbrain synchronization of this sort when two individuals play cooperative card games or guitar duets, imitate each other’s hand movements, and so on. The question that has not been asked, and of interest to us is, How does the brain represent and support the social structural conditions that give rise to relational cohesion? We suspect that the exchange structures we have studied for the past twenty years may foster synchronization. Task jointness, conditions of power (high total power, equal relative power), and network properties (i.e., structural cohesion) may potentially be structural factors that relate to patterns of interbrain synchronization. As a first step in this research, we are currently seeking to replicate the original findings of the 1996 relational cohesion study conducted by Lawler and Yoon (1996), but now with the addition of event-related potential (ERP) measurement. In other words, EEG activity across a pair of negotiating actors is being measured as subjects negotiate under the original Lawler and Yoon (1996) protocol. The project promises to shed light on how neurological activity relates to the development of solidarity across pairs of actors in an exchange context. At least since Durkheim (1915) sociologists have theorized that collective interaction is a precursor to solidarity. Our newest research promises to reveal the neurosociological processes that undergird and produce this phenomenon.

CONCLUSIONS Our social commitments theory provides a theoretical understanding of how social interaction facilitates affective ties to overarching social units.

Person-to-Group Ties

133

Affective ties to social units differ from ties to specific individual others, and have important order generating effects on both the micro- and macrolevel. With ever changing interaction dynamics, primarily due to advances in social media, person-to-unit affective ties provide a means of stability and order in an increasingly global world (see also Hechter & Horne, 2009, for a related discussion). Person-to-unit ties help resolve problems of social order for two primary reasons. First, even though affective ties to larger units require social interaction as the foundation, these interactions only need to occur among a very small proportion of actors within a population. Second, social unit attributions make it possible for local, individualized emotions to have macrolevel effects. Macro orders can be strengthened by positive microlevel emotions and, in turn, weakened by negative emotions (see Turner, 2000, 2007). The theory of social commitments (Lawler et al., 2009) abstracts and generalizes from three prior theories related to commitment in exchange relations: nested-commitment theory (Lawler, 1992), relational cohesion theory (Lawler & Yoon, 1996), and the affect theory of social exchange (Lawler, 2001; Lawler et al., 2008). Joint tasks serve as the central structural condition within which the theory of social commitments specifies affective attribution to social units. Perceptions of shared responsibility (a function of task jointness) foster the increased salience of the social relation, and increase the likelihood that positive emotions and feelings generated within that context are attributed to the group. Actors are more likely to attribute positive emotions and feelings to local or immediate groups, and more likely to attribute negative emotions and feelings to larger, more distal units. This nested-commitments principle implies that local units are credited for success, and distant units are given more blame for negative experiences. As a result, larger social units are more susceptible to problems of social order. Macro-units can overcome this problem by generating responsibility for the joint task structure and emotions at the microlevel, or by reducing the distance between micro- and macrolevels, that is, tightly embedding smaller units in larger ones. The theory of social commitments takes a “bottom-up” approach to the problem of social order. The theory specifies how social order generated at the microlevel can impact order at the macrolevel by way of task structure and positive affect. This process can also operate in the reverse, that is, to produce social order from the top down. For example, task jointness may be a function of the macrostructures that combine to influence interactions at a microlevel. Different cultures may have different rules and norms pertaining to perceptions of shared responsibility, and different organizations

134

SHANE R. THYE ET AL.

may develop different procedural methods, ranging from individual autonomy to an overall collective effort. Overall, the theory of social commitments contends that microlevel interactions are an important component in the understanding of macrolevel structures and culture. Both “top-down” and “bottom-up” approaches affect each other, and joint tasks and shared responsibility are important at both micro- and macrolevels. In closing, our theory and research broadly speaks to the contemporary debates regarding the decline of close personal ties and social capital (cf. Stolle & Hooghe, 2004). It seems clear that the nature and range of social ties are changing in significant ways, in part, because of new technological means of connecting with one another, the globalization of economic markets and, in particular, work. Some scholars assert that our social interactions increasingly emulate economic “spot markets” and are increasingly disjointed or spread out, resulting in weaker ties to traditional units such as work organizations and local communities (Cappelli, 1999). Putnam (2000) points to increases in social isolation, a decline in trust, and lower participation in local communities as the culprit. Others find that networks of close ties have shrunk from 1985 to 2004 (McPherson, SmithLovin, & Brashears, 2006); yet still others indicate little or no changes in family and friendship ties over recent decades (see Fischer, 2011). Our research speaks to these issues in two ways. First, it suggests that the sole emphasis on person-to-person ties is misplaced. Person-to-group ties, even those shared among casually or weakly connected individuals, may be sufficient to generate and sustain a sense of order and community among strangers in large (even massive) impersonal networks (e.g., the extensive YouTube communities that exist; see Wesch, 2009, for an excellent discussion). Second, analyses of social ties tend to neglect the emotional dynamics of the social ties. Our research suggests that even the most individualistic, instrumentally based ties can generate affective sentiments about specific others and the larger encompassing community. The confluence of affectively laden personal ties and a salient group affiliation is a powerful source of order and stability, even in the spot market oriented and highly transactional world of today.

ACKNOWLEDGMENT This material is based on work supported by the National Science Foundation under Collaborative Grant Numbers SBR-9817706 and SBR-9816259 to the University of South Carolina and Cornell University.

135

Person-to-Group Ties

REFERENCES Bandura, A. (1997). Social learning theory. Englewood Cliffs, NJ: Prentice Hall. Barsade, S. (2002). The ripple effect: Emotional contagion and its influence on group behavior. Administrative Science Quarterly, 47, 644
675. Bar-Tal, D., Halperin, E., & De Rivera, J. (2007). Collective emotions in conflict situations: Societal implications. Journal of Social Issues, 63, 441
460. Bearman, P. (1997). Generalized exchange. American Journal of Sociology, 102, 1383
1415. Bearman, P. S., Moody, J., & Stovel, K. (2004). Chains of affection: The structure of adolescent romantic and sexual networks. American Journal of Sociology, 110, 44
91. Berger, J., Fisek, M. H., Norman, R. Z., & Zelditch, M., Jr. (1977). Status characteristics in social interaction: An expectation-states approach. New York, NY: Elsevier. Berger, J., & Webster, M., Jr. (2006). Expectations, status, and behavior. In P. J. Burke (Ed.), Contemporary social psychological theories. Stanford, CA: Stanford University Press. Cappelli, P. (1999). The new deal at work: Managing the market-driven workforce. Boston, MA: Harvard Business School Press. Cohen, B. P. (1989). Developing sociological knowledge: Theory and method (2nd ed.). Chicago, IL: Nelson-Hall. Collins, R. (1975). Conflict sociology: Toward an explanatory science. New York, NY: Academic Press. Collins, R. (1981). On the microfoundations of macrosociology. American Journal of Sociology, 86, 984
1014. Condon, W. S., & Ogston, W. D. (1966). Sound film analysis of normal and pathological behavior patterns. Journal of Nervous and Mental Diseases, 143, 338
457. Condon, W. S., & Sander, L. W. (1974). Synchrony demonstrated between movements of the neonate and adult speech. Child Development, 45, 456
462. Cook, K. S., Emerson, R., Gillmore, M. R., & Yamagishi, T. (1983). The distribution of power in exchange networks: Theory and experimental evidence. American Journal of Sociology, 89, 275
305. Damasio, A. R. (1999). The feeling of what happens: Body and emotion in the making of consciousness. New York, NY: Harcourt Brace. Dumas, G., Lachat, F., Martinere, J., Nadel, J., & George, N. (2011). From social behavior to brain synchronization: Review and perspectives in hyperscanning. IRBM, 32, 48
53. Durkheim, E. (1915). The elementary forms of religious life. New York, NY: Free Press. Ekeh, P. (1974). Social exchange theory. Cambridge, MA: Harvard University Press. Emerson, R. M. (1972a). Exchange theory. Part I: A psychological basis for social exchange. In J. Berger, M. Zelditch Jr., & B. Anderson (Eds.), Sociological theories in progress (Vol. 2, pp. 38
57). Boston, MA: Houghton Mifflin. Emerson, R. M. (1972b). Exchange theory. Part II: Exchange rules and networks. In J. Berger, M. Zelditch Jr., & B. Anderson (Eds.), Sociological theories in progress (Vol. 2, pp. 58
87). Boston, MA: Houghton Mifflin. Emerson, R. M. (1981). Social exchange theory. In M. Rosenberg & R. H. Turner (Eds.), Social psychology: Sociological perspectives (pp. 30
65). New York, NY: Basic Books, Inc. Fischer, C. S. (2011). Family and friends in America since 1970. New York, NY: Russell Sage Foundation.

136

SHANE R. THYE ET AL.

Gilmore, D. D. (1987). Honor, honesty, shame: Male status in contemporary Andalusia. In D. D. Gilmore (Ed.), Honor and shame and the unity of the Mediterranean (pp. 90
103). Washington, DC: American Anthropological Association. Graham, S. (1991). A review of attribution theory in achievement contexts. Educational Psychology Review, 3, 5
39. Hechter, M., & Horne, C. (2009). Theories of social order: A reader. Stanford, CA: Stanford University Press. Homans, G. C. (1950). The human group. New York, NY: Harcourt Brace Jovanovich. Izard, C. E. (1977). Human emotions. New York, NY: Plenum. Izard, C. E. (1991). The psychology of emotions. New York, NY: Plenum Press. Kalkhoff, W., Dippong, J., & Gregory, S. W. (2011). The biosociology of solidarity. Sociology Compass, 5, 936
948. Kalkhoff, W., Thye, S., & Lawler, E. J. (2011, August). Relational cohesion and commitment in exchange relations: The role of inter-brain synchronization. Presented at the conference for Theory and Research on Group Processes, Las Vegas, NV. Kanter, R. M. (1968). Commitment and social organization: A study of commitment mechanism in utopian communities. American Sociological Review, 33, 499
517. Kelley, H. H. (1967). Attribution theory in social psychology. In D. Levine & D. E. Berlyne (Eds.), Nebraska symposium on motivation (pp. 220
266). Lincoln, NE: University of Nebraska Press. Kemper, T. D. (1978). A sociological interactional theory of emotions. New York, NY: Wiley. Kollock, P. (1998). The emergence of exchange structures: An experimental study of uncertainty, commitment, and trust. American Journal of Sociology, 100, 183
214. Kuwabara, K. (2011). Cohesion, cooperation, and the value of doing things together: How economic exchange creates relational bonds. American Sociological Review, 76, 560
580. Lawler, E. J. (1992). Power processes in bargaining. Sociological Quarterly, 33, 17
34. Lawler, E. J. (2001). An affect theory of social exchange. American Journal of Sociology, 107, 321
352. Lawler, E. J., & Thye, S. R. (1999). Bringing emotions into social exchange theory. Annual Review of Sociology, 25, 217
244. Lawler, E. J., & Thye, S. R. (2006). Social exchange theory of emotions. In J. E. Stets & J. H. Turner (Eds.), Handbook of the sociology of emotions (pp. 295
320). New York, NY: Springer. Lawler, E. J., Thye, S. R., & Yoon, J. (2000). Emotion and group cohesion in productive exchange. American Journal of Sociology, 106, 616
657. Lawler, E. J., Thye, S. R., & Yoon, J. (2006). Commitment in structurally enabled and induced exchange relations. Social Psychology Quarterly, 69(2), 183
200. Lawler, E. J., Thye, S. R., & Yoon, J. (2008). Social exchange and micro social order. American Sociological Review, 73, 519
542. Lawler, E. J., Thye, S. R., & Yoon, J. (2009). Social commitments in a depersonalized world. New York, NY: Russell Sage Foundation. Lawler, E. J., Thye, S. R., & Yoon, J. (2013). The emergence of collective emotions in social exchange. In C. Scheve & M. Salmela (Eds.), Collective emotions. Oxford: Oxford University Press. Lawler, E. J., & Yoon, J. (1993). Power and the emergence of commitment behavior in negotiated exchange. American Sociological Review, 58, 465
481.

Person-to-Group Ties

137

Lawler, E. J., & Yoon, J. (1996). Commitment in exchange relations: Test of a theory of relational cohesion. American Sociological Review, 61, 89
108. Lawler, E. J., & Yoon, J. (1998). Network structure and emotion in exchange relations. American Sociological Review, 63, 871
894. Lovaglia, M. J., & Houser, J. A. (1996). Emotional reactions and status in groups. American Sociological Review, 61, 864
880. Markovsky, B., Skvoretz, J., Willer, D., Lovaglia, M. J., & Erger, J. (1993). The seeds of weak power: An extension of network exchange theory. American Sociological Review, 58(2), 197
209. Markovsky, B., Willer, D., & Patton, T. (1988). Power relations in exchange networks. American Sociological Review, 53, 220
236. McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2006). Social isolation in America: Changes in core discussion networks over two decades. American Sociological Review, 71, 353
375. Molm, L. D. (1994). Dependence and risk: Transforming the structure of social exchange. Social Psychology Quarterly, 57, 163
176. Molm, L. D., Collett, J. L., & Schaefer, D. R. (2007a). Building solidarity through generalized exchange: A theory of reciprocity. American Journal of Sociology, 113, 205
242. Molm, L. D., Schaefer, D. R., & Collett, J. L. (2007b). The value of reciprocity. Social Psychology Quarterly, 70, 199
217. Moody, J., & White, D. R. (2003). Structural cohesion and embeddedness: A hierarchical concept of groups. American Sociological Review, 68, 103
107. Mueller, C. W., & Lawler, E. J. (1999). Commitment to nested organizational units: Some basic principles and preliminary findings. Social Psychology Quarterly, 62, 325
346. Price, H. E., & Collett, J. L. (2012). The role of exchange and emotion on commitment: A study of teachers. Social Science Research, 41, 1469
1479. Putnam, R. D. (2000). Bowling alone: The collapse and revival of American community. New York, NY: Simon and Schuster. Stolle, D., & Hooghe, M. (2004). Inaccurate, exceptional, one-sided, or irrelevant? The debate about the alleged decline of social capital and civic engagement in western societies. British Journal of Political Science, 35(1), 149
167. Taylor, C., & Pillemer, K. (2009). Using affect to understand employee turnover: A context specific application of a theory of social exchange. Sociological Perspectives, 52, 481
504. Thye, S., & Lawler, E. J. (2010). Collaborative research: The emergence of social order in groups. National Science Foundation Grant; SES-0956796 to USC & SES-0957982 to Cornell; 5/15/2010
5/14/2012. Thye, S. R., Lawler, E. J., & Yoon, J. (2011). The emergence of embedded relations and group formation in networks of competition. Social Psychology Quarterly, 74, 387
413. Thye, S. R., Yoon, J., & Lawler, E. J. (2002). The theory of relational cohesion: Review of a research program. In S. R. Thye & E. J. Lawler (Eds.), Advances in group process (pp. 89
102). Oxford, UK: Elsevier. Turner, J. H. (2000). On the origins of human emotions: A sociological inquiry into the evolution of human affect. Stanford, CA: Stanford University Press. Turner, J. H. (2007). Human emotions: A sociological theory. New York, NY: Routledge. Uehara, E. (1990). Dual exchange: Theory, social networks, and informational social support. American Journal of Sociology, 96, 521
557.

138

SHANE R. THYE ET AL.

Vincent, A. (2013, August). Status and the development of cohesion. Presented at The conference for Theory and Research on Group Processes, Denver, CO. Walker, H. A., & Cohen, B. P. (1985). Scope statements: Imperatives for evaluating theory. American Sociological Review, 50(3), 288
301. Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications. Cambridge, UK: Cambridge University Press. Weiner, B. (1985). ‘Spontaneous’ casual thinking. Psychological Bulletin, 97, 74
84. Wesch, M. (2009). YouTube and you: Experiences of self-awareness in the context collapse of the recording webcam. Explorations in Media Ecology, 8(2), 19
34. Westcott, M. R. (1988). The psychology of human freedom: A human science perspective and critique. New York, NY: Springer. Willer, D. (1999). Network exchange theory. Westport, CT: Praeger. Willer, R., Flynn, F. J., & Zak, S. (2012). Structure, identity, and solidarity: A comparative field study of generalized and direct exchange. Administrative Science Quarterly, 57, 119
155. Williamson, O. (1975). Markets and hierarchies. New York, NY: Free Press. Yoon, J., & Thye, S. R. (2002). A dual process model of organizational commitment. Work and Occupations, 29(1), 97
124. Yoon, J., Thye, S. R., & Lawler, E. J. (2013). Exchange and cohesion in dyads and triads: A test of Simmel’s hypothesis. Social Science Research, 42, 1457
1466.

BACK TO THE FUTURE: 25 YEARS OF RESEARCH IN AFFECT CONTROL THEORY Neil J. MacKinnon and Dawn T. Robinson ABSTRACT Purpose
To provide a comprehensive review of theoretical and research advances in affect control theory from 1988 to 2013 for academic and student researchers in social psychology. Design/methodology/approach
Against the background of a concise history of affect control theory from its beginnings in the 1960s to its maturation in the late 1980s, a comprehensive review of research and publications in the last 25 years is reported in five sections: Theoretical Advances (e.g., self and institutions, nonverbal behavior, neuroscience, artificial intelligence); Technological Advances (e.g., electronic data collection, computer simulations, cultural surveys, equation refinement, small groups analysis); Cross-Cultural Research (archived data and published analyses); Empirical Tests of the Theory; and Substantive Applications (e.g., emotions, social and cultural change, occupations/ work, politics, gender/ideology/subcultures, deviance, criminology, stereotyping, physiological behavior).

Advances in Group Processes, Volume 31, 139
173 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0882-6145/doi:10.1108/S0882-614520140000031003

139

140

NEIL J. MACKINNON AND DAWN T. ROBINSON

Findings
Reveals an impressive number of publications in this area, including over 120 articles and chapters and four major books, and a great deal of cross-cultural research, including European, Asian, and Middle-Asian cultures. Research limitation/implications (if applicable)
Because of limitations of space, the review does not cover the large number of theses, dissertations, and research reports. Originality/value
No other review of affect control theory with this scope and detail exists. Keywords: Affect control; emotions; identities; self; social interaction; social institutions

Affect control theory (ACT) (Heise, 1979, 2007; MacKinnon, 1994; SmithLovin & Heise, 1988) proposes that people construct and interpret social interaction in the direction of confirming established cultural sentiments for the operative identities and actions, and that when events strain cultural sentiments, people initiate restorative actions or cognitive revisions to bring affective disturbances back into line with established cultural sentiments. The formalization of the theory comprises a propositional statement of the theory (MacKinnon, 1994; MacKinnon & Heise, 1993) and elegant mathematical models and computer programs for predicting and simulating initial and restorative actions, recognition and cognitive revisions of events (labeling and attribution processes), and emotions. Affect control theory developed against the backdrop of raging intellectual debates between structural functionalist scholars, who gravitated toward the power and parsimony of top-down models describing how social structure constrains individual outcomes, and the symbolic interactionist scholars who argued for attention to agency and creativity of individuals who jointly and continually re-create social structures. The theory derived from a synthesis of symbolic interactionist theory with three theories from psychological social psychology that relied on cutting-edge tools and technologies. The method for measuring affective meaning in affect control theory drew from the EPA (evaluation, potency, and activity) structure of semantic differential scales established by Osgood and associates (Osgood, 1962; Osgood, May, & Miron, 1975; Osgood, Suci, & Tannenbaum, 1957); the empirical equations of affect control theory for

Affect Control Theory

141

predicting the affective reaction to events is an extension of Gollob’s work on impression change (Gollob, 1968; Gollob & Rossman, 1973); and the development of the principle of affect control drew from the perceptual control systems theory of Powers (1973), albeit the general principle of affect control was stated by Heise some years earlier (1969a, p. 212; see Heise, 2007, p. 129). The theory’s origin coincided with the roots of mathematical sociology in the 1960s and 1970s.1 In the first couple of decades of its development, affect control theory was on the vanguard with respect to methodology and technology in sociological theory
offering one of the earliest examples of a mathematically well-specified social theory programmed into software that enabled computer simulations to explore its implications. In the last 25 years, affect control theory has developed into a major contemporary theory in sociological social psychology, focusing largely on emotions and identity processes in social interaction and, as discussed below, has recently advanced a cybernetic theory of self and a theory of social institutions based on identities and identity processes (MacKinnon & Heise, 2010). Although this chapter is devoted to a review of affect control theory research since 1988, we establish a baseline by briefly summarizing its beginnings in the late 1960s/early 1970s and its maturation into a coalescent theory and program of research during the late 1970s/early 1980s. We then review theoretical and technological advances of affect control theory from 1988 to 2013, cultural and cross-cultural work, experimental tests of the theory, and applications of the theory to numerous substantive areas. Under the heading of theoretical developments, we discuss important expositions of affect control theory and expansions of the theory to deal with self, social institutions, interaction in small groups, nonverbal behavior, neuroscience, artificial intelligence, and interaction with the physical environment and technology. Under the rubric of technological advances, we include the development of a computer program (Surveyor) for collecting data online and a program of research (Magellan) devoted to collecting data from cultures around the world using this technology, the methodology of surveying cultures, refinement of impression change equations and expansion to other cultures, the Group Simulator Program, and a version of affect control theory employing a Bayesian statistics approach. Applications to substantive areas include research on emotions; social structure, culture, and social and cultural change; occupations and work; politics; gender, ideology, and subcultures; deviance and criminology; and stereotyping.

142

NEIL J. MACKINNON AND DAWN T. ROBINSON

Recently, research in this tradition has begun to revisit core assumptions and working procedures of the theory in light of new methodologies, new technologies, and new knowledge. Affect control theory research continues to engage with newly emerging technologies and methodologies, and both the form and progress of the theory have been
and continue to be
heavily shaped by these advances. We will trace the way that such technological and methodological advances have paved the way for new accumulations of empirical knowledge and even afforded us novel opportunities to revisit fundamental assumptions in the theory
shedding new light on old questions and taking us “back to the future.”

FROM NASCENCE TO MATURATION The beginning of affect control theory can be traced to Heise’s early publications on semantic differential scales for measuring affective meaning (1965, 1969b), stemming from the work of Osgood and associates (see above references), and on impression change in simple events (1969a, 1970) stemming from the work of Gollob and associates (see above references). Heise extended Gollob’s work on impression change for the actor of an event and the evaluation dimension of meaning to all three components of simple events (Actor, Behavior, and Object-Person) and to all three dimensions of the semantic differential (evaluation, potency, and activity). In 1971, at the University of North Carolina, Heise developed the mathematical model for the minimization or control of deflection
the disjunction between cultural sentiments and current feelings resulting from social interaction. In 1972, he wrote an early version of Interact, a computer program for simulating social interaction based on the affect control model for minimizing deflection. Following the collection of a new dictionary of more than 1,000 EPA measures of social identities and behaviors in 1974 and the reestimation of the impression change equations with new data in 1975, the first detailed description of affect control theory appeared in a 1977 article in Behavioral Science. A research monograph on the program Interact (Heise, 1978) was followed by a paradigmatic statement of the theory in a major book (Heise, 1979), which presented the mathematical model; illustrated its power with computer simulations; and connected the theory to attitude research, cybernetic theory, and symbolic interactionism.

Affect Control Theory

143

During the same period (1978
1979), an National Institute of Mental Health (NIMH) grant enabled the collection of new EPA data for over 2,000 identities, behaviors, identity modifiers, and settings, and the development of more complex impression change equations based on analysis of more than 500 event sentences. This project supported the research of a number of graduate students at University of North Carolina (UNC), including Lynn Smith-Lovin who completed her dissertation in 1979. Neil MacKinnon joined this research program while a visiting scholar at the university from 1978 to 1979, and upon his return to Canada conducted large-scale studies of affect control theory funded by multiple grants from the Social Sciences and Humanities Research Council of Canada (SSHRC) beginning in 1981 and extending through the 1980s, 1990s, and into the early 2000s. During this period, there were major new specifications of the impression change equations that formed the basis of the theory (Heise, 1985b; Heise & Smith-Lovin, 1981; Smith-Lovin & Heise, 1982), as well as key extensions to deal with settings (Smith-Lovin, 1979), emotions, and attributions (Averett & Heise, 1987; Heise, 1985a). The first major cross-cultural investigation compared the impression change equations for U.S. college students, Egyptians and Lebanese living in the United States, and Catholic high school students in Northern Ireland (Smith-Lovin, 1987a). In 1987, Heise developed Attitude, a program for collecting EPA data with computers. A documentation of the programs Attitude and Interact was published a year later (Heise & Lewis, 1988). By this time the basic structure of the theory was well-developed, and the theory was ready to go into a period of testing, refinement, and elaboration. In 1987, a special issue of the Journal of Mathematical Sociology was devoted to advances in ACT and summarized the state of theory and empirical research to date. Edited by Smith-Lovin and Heise, this included articles on an overview of the theory (Heise, 1987); impression change (SmithLovin, 1987a); the effect of settings on impressions (Smith-Lovin, 1987b); and modifying identities: amalgamations, attributions, and emotions (Averett & Heise, 1987). This collection also included the first two empirical examinations of ACT predictions
one examining the affective bases of likelihood judgments (Heise & MacKinnon, 1987) and the other experimentally examining behavioral responses to deflection (Wiggins & Heise, 1987). The publication of this special issue as a book a year later (Smith-Lovin & Heise, 1988) coincides with the boundary between the nascence to maturation period summarized in this section and advances in affect control theory in the past 25 years to which the remainder of this chapter is devoted.

144

NEIL J. MACKINNON AND DAWN T. ROBINSON

THEORETICAL ADVANCES SINCE 1988 Twenty-five years ago, Linda Molm organized the first of what would become an annual conference bringing together group processes scholars from a variety of traditions. Lynn Smith-Lovin attended that conference and the development of affect control theory research since that time was greatly influenced by interactions between affect control theory scholars and researchers within other group processes traditions. The rest of this chapter outlines the development of the theory since that conference. One of the major influences that interaction with other, largely experimental, group processes researchers had on the trajectory of affect control theory research was to inspire a wave of empirical tests of the theory’s implications
much of it in laboratory settings. As described below, there have been dozens of tests of the theory since that time, including laboratory experiments, surveys, and field observations. In addition, much of the early part of the last 25 years was spent capitalizing on new methods and procedures to transform affect control theory. Moving to the use of GraecoLatin square condition assignments in the survey design for estimating the impression change equations enabled drastically more efficient designs. This change, combined with turning to a personal computer-facilitated data collection allowed affect control researchers to collect new equation data in an array of new cultures. As we will see later, this led us down the path of heightened attention to differences in the representations of event processing in various cultures.

Expositions of the Theory Affect control theory developed in more or less discrete steps as its initial focus on impression change and management in interpersonal dyadic events was expanded to incorporate ancillary theories of emotions, attributions, labeling, and other social psychological phenomena. These various strands of theory were integrated into a formal statement of affect control theory in the early 1990s (MacKinnon, 1994; MacKinnon & Heise, 1993). This formalization comprises 24 propositions dealing with the theory’s fundamental assumptions on symbols, language, and affective meaning; its core principles of affective reaction, affect control, and reconstruction of events; the application of these fundamental assumptions and principles to the recognition, production, and reconceptualization of events; and the production, expression, and function of emotions in social interaction.

Affect Control Theory

145

The exposition of affect control theory by MacKinnon (1994) greatly expanded the earlier connections between the theory and symbolic interactionism made by Heise (1979), with an entire chapter on affect control theory and the social psychology of George Herbert Mead; and another chapter on identities and roles, within which affect control theory is compared with the identity theories of Stryker (1968, 1980), McCall and Simmons (1978), Burke (Burke & Reitzes, 1981, 1991; Burke & Tully, 1977), and Alexander and Wiley (1981). In addition, this exposition of affect control theory dealt extensively with more fundamental issues such as the relation between cognition and affect and the nature of emotion and motivation, and how affect control theory addresses these larger issues. Other expositions of the theory include mathematical accounts of the theory by Heise (2000a, 2000b, 1989a, 1990a), the introduction of the scope conditions of the theory (Robinson, 2007; also in Robinson & Smith-Lovin, 2006), a comparison between affect control theory and the control system ideas in identity theory (Smith-Lovin & Robinson, 2006), and a nontechnical and highly accessible presentation of the theory by Heise (2007). The first book-length exposition of the theory since MacKinnon (1994), Heise’s book Expressive Order consists of three parts, the first of which provides a verbal, reader-friendly, and nontechnical introduction to ACT. Besides covering the usual bases, Expressive Order introduces new discussions of self and social institutions, drawing from the later book by MacKinnon and Heise (2010), then in progress. The second part of Expressive Order provides a detailed presentation of the underlying mathematics of ACT, and the third part provides a description and history of the research program, the application of affect control theory to general topics in social psychology and substantive areas of research, and instructions on conducting simulations with affect control theory software. MacKinnon and Heise’s (2010) book on identities, self, and social institutions develops a new affect control theory of self and establishes the boundaries distinguishing affect control theory as an identity theory from its extension as a theory of self. MacKinnon and Heise also compare the affect control theory of identities and self with other structural symbolic interactionist theories (e.g., Burke, 1991; McCall & Simmons, 1978; Stryker, 1980), processual symbolic interactionist theories (e.g., Perinbanayagam, 2000; Wiley, 1994), as well as two identity theories from psychology
social identity theory (e.g., Tajfel, 1969, 1970, 1981) and self-categorization theory (Turner, 1985; Turner, Hogg, Oakes, Reicher, & Wetherell, 1987; Turner, Oakes, Haslam, & McGarty, 1994). Each theory is summarized and compared with the affect control theory of identities and self in terms of

146

NEIL J. MACKINNON AND DAWN T. ROBINSON

the following criteria: whether the theory takes affect as well as cognition into account, how it explains human motivation, whether it incorporates cybernetic and semiotic perspectives, and how it explains the integration and stability of self. While not an exposition of affect control theory per se, Scholl (2013) contributed to our appreciation of the scope of the theory by demonstrating that the EPA dimensions of meaning in the form of affiliation, power, and activation can be identified in five diverse areas of psychological research
the perceptions of emotions, verbal communication, nonverbal communication, interpersonal behavior, and personality. Identifying empirical connections between the ten possible pairs of these five research domains, he concludes that “humans construct their social world along these three dimensions of socio-emotional perception and action” (2012, p. 5), that these dimensions correspond to an evolutionary need for coordination among individuals, and that affect control theory captures the socioemotional consistency in this dimensional space that makes cooperative social life possible.

New Theoretical Domains The past 25 years has seen affect control theory expand into a variety of new domains. In the following sections, we describe elaborations of theory that have enabled it be applied to new processes and phenomena. Self and Institutions Although affect control theory has often invoked the concept of self in various ruminations about the theory (e.g., MacKinnon, 1994; Moore & Robinson, 2006; Robinson & Smith-Lovin, 1992), strictly speaking affect control theory is a theory about identities, not self. And while the concept of institutions has also been invoked in establishing context for defining a situation (e.g., Heise, 1979), affect control theory is certainly not a theory of institutions. MacKinnon and Heise (2010), however, extended the principle of affect control from the level of identities to the higher control level of self and situated this theory of self explicitly in an empirically grounded, complementary theory of social institutions based on the cultural meanings of identities. Specifically, the book advances a cybernetic model of the selfprocess wherein individuals reaffirm their self-sentiments after disconfirming institutional experiences by selecting and enacting compensatory or redeeming identities. In addition, it details how every society possesses an

Affect Control Theory

147

implicit cultural theory of people based on the semantics of identity-nouns and how individuals employ this implicit cultural knowledge to construct their selves. The book advances a symbolic interaction, semantic approach to institutions by extracting semantic networks of identities from online and standard dictionary definitions, then employing these semantic networks to empirically enumerate and identify social institutions. Heise is currently working on additional analyses with expanded dictionary definitions, free association data, and collocate2 analyses of texts from the last two centuries, expanding the number of institutions identified in the MacKinnon and Heise (2010) book. Nonverbal Behavior Heise (1989b) showed how the equations of ACT could be used to predict the effects of emotion displays on the reidentification of participants in social events. Rashotte (2001, 2002a, 2000b, 2003) showed how nonverbal, visual cues such as demeanor and emotional expression affect the impressions we create during social interaction. Neuroscience Schro¨der, Stewart, and Thagard (2013) recently proposed a neurological model of intentions applied to belief, planning, and action that draws upon the EPA structure of affective meaning and ideas from affect control theory and the affect control theory of self discussed in the preceding section. Schro¨der and Thagard (2013) used affect control theory to help explain the underlying mechanisms of automatic human behavior resulting from the temporary activation of concepts in an individual’s mind exemplified by experimental priming effects. Specifically, they used affect control theory to explain how experimental priming procedures constrain automatic behavior by biasing the conceptual framing of social situations to correspond to the affective meanings of a culture (See also Thagard & Schro¨der, Forthcoming). Controlling Affect of and by Inanimate Objects Recent extensions of affect control theory open up the possibility of affect control theory as the basis for social intelligence in nonhuman actors and as the basis for understanding social interactions with nonsocial objects. Heise (2004) and Troyer (2004) proposed affect control theory as a basis for incorporating emotion in the design of socially intelligent computer agents. The Bayesian version of Interact (Hoey, Schro¨der, & Alhothali, 2013) described below also allows for modeling interaction between

148

NEIL J. MACKINNON AND DAWN T. ROBINSON

humans and computer systems. Lulham (2004) applied affect control theory to the influence of architectural design of the physical environment on inmate feelings in juvenile justice centers. Shank (2010) expanded the scope of affect control theory from human
human to human
technology interaction. MacKinnon (2003) discussed the latter kind of interaction in a presentation to professionals involved in knowledge presentation.

TECHNOLOGICAL ADVANCES SINCE 1988 By separating the theoretical advances from the technological advances in affect control theory during the last 25 years, we make somewhat of a faulty distinction. Affect control theory is a partially grounded theory, meaning that some of the core knowledge of the theory is captured in the empirically estimated and measured equations and sentiment dictionaries. In this way, while there exists a theoretical structure that is pan-cultural, the theory can only be implemented (implications derived) in the context of cultural models. Some of the theory’s core assumptions are represented in the procedures that it uses to generate those sentiment dictionaries and profiles. Moreover, new techniques for estimating equations and for manipulating/interpreting those equations are not just technological advances, but because ACT is a mathematical theory they are also theoretical advances. Nonetheless, in the following sections, we highlight several technological and methodological advances in the last few decades that have major implication for the structure, content, and future of the theory.

Project Magellan and Program Surveyor Important updates were made to the simulation program Interact and the data collection program Attitude after 1988. Interact was reprogrammed to run online as a Java applet and additional documentation of the program was published by Schneider and Heise (1995). This applet was housed on the Indiana University website and ran on the user’s browser. Recently, the Java applet version of the program has been replaced by a freestanding full application. The Attitude program collecting sentiments data with EPA scales (Heise & Lewis, 1988) was also completely revised as an online data collection tool, called Surveyor. For years, this program resided on the Indiana University server, allowing affect control theory researchers to

149

Affect Control Theory

collect data over the Internet, with the data being stored at Indiana University. This facilitated new data collections in the United States and Canada, as well as new cultures
including Japan, China, and Germany. Magellan is the name assigned to an ongoing research project of collecting data sets on sentiments in different cultures around the world, while Surveyor is the name of the program for accomplishing this objective (Heise, 2001, 2010). Surveyor collects data on sentiments and returns the ratings of respondents via the Internet. Besides English, revisions of the program to accommodate indigenous languages are available in Japanese, Bengali, German, Spanish, simplified Chinese, and Arabic.3 Surveying Cultures Two publications (Heise, 2010; Smith, Ike, & Li, 2002) offer wisdom acquired in the trenches about how best to acquire data about cultures. While traditional survey research focuses on individual differences among respondents, surveys of cultural norms focus on consensus among respondents. A new book, Surveying Cultures (Heise, 2010), not only assembles and codifies the methods employed in affect control theory for measuring cultural sentiments from a culture-as-consensus model, but also presents the methodology in a way that can be adopted by anyone engaged in the study of cultural norms. Among other things, this work details the use of bipolar scales and methods of data collection using the Internet, describes archival repositories of EPA cultural sentiments data, and explains differences between traditional survey research and the culture-as-consensus model in conceiving measurement error and reliability. Drawing from his experience collecting ACT sentiment dictionary and event level data in Japan and China, Herman Smith (Smith et al., 2002) published a paper reflecting on methodological concerns especially relevant to this activity
identifying problems and solutions involved in collecting cross-cultural data on affective sentiments with the Internet. Impression Change Equations: Refinement and Expansion to Other Cultures Impression change equations model the affective reaction of individuals to the various components of ABO (Actor-Behavior-Object) events such as “The teacher scolded the student.” As discussed earlier, impression change equations had been developed prior to 1988 for ABOS (Actor-BehaviorObject-Setting) events and for MI (Modifier-Identity) combinations,

150

NEIL J. MACKINNON AND DAWN T. ROBINSON

combining an identity with a trait or emotion. Since 1988, impression change equations for self-directed action (AB events) have been created and studied (Britt & Heise, 1992; Smith & Francis, 2005). In addition, the basic impression change equations for all of the cultures were reestimated using stepwise regression (Heise, 1991). Theoretical discussions of impression change published after 1988 can be found in Heise (2000a, 2007) as well as expositions of the theory cited earlier. Impression change equations are fundamental to affect control theory because they are the empirical basis for derivation of the theory’s impression-management equations that embody the principle of affect control and predict restorative actions in the wake of events producing large affective deflections. However, affect control theory also proposes that deflections that cannot be resolved by restorative actions stimulate the reidentification of the interactants and actions involved in the event. Discussions of reidentification processes can be found in the various expository references to affect control theory cited earlier. However, there have been a number of important developments since 1988. For example, Heise (1989b) incorporated the effect of emotion displays into models for the reidentification of actors and object-persons in events, and using eventvignettes, Nelson (2006) explored which components of improbable or bizarre events are more likely to be reidentified. Prior to 1988, full impression change equations had been developed for only the United States and Canada. Since then, equations have been developed for Japanese (Smith, Matsuno, & Ike, 2001; Smith, Matsuno, & Umino, 1994) and German cultures (Schro¨der, 2011). In addition, the late Herman Smith began to develop impression change equations in a study of mainland Chinese culture in 1999, but this work has not been published to date, and Robinson and Smith-Lovin are currently engaged in developing impression change equations for several Arabic-speaking cultures. Early in the affect control theory research program, affect control theory researchers assumed that while there are subcultural and crosscultural differences in cultural sentiments (EPA ratings of identities, actions, settings, traits, and emotions), impression change equations may reflect culturally universal ways of affective processing. Indeed, the stepwise methodology used to estimate the impression change equations rested on this assumption to allay the conventional critique of such methods as being prone to producing highly sample-dependent results that capitalize on both random and systematic error. Estimations of impression change equations using these techniques resulted in a number of two-way and three-way interaction terms. While some of these made intuitive and theoretical sense,

Affect Control Theory

151

the interpretation of others was more problematic. Recently, Heise (2012b) circulated a white paper proposing one solution to this problem
namely, that the construction of impression change models should be made in two phases: model specification using analysis of variance and model estimation using regression analysis. This proposed approach capitalizes on the experimental design used to collect event data for affect control theory equations by employing ANOVA models in the specification phase. This dual procedure eliminated many interaction terms, resulting in much simpler and interpretable equations. While substantial structural similarities in equations have been found across sex and culture, notable differences have also appeared in studies to date (see references in this section and in the above section on the early development of affect control theory). Therefore, over the last few decades affect control theory researchers have moved away from the assumption that impression change processes are universal and have proceeded on the assumption that new equations should be developed for different cultures or distinctive subcultures within a given culture (Heise, 2007, p. 134; 2010, p. 107). Another novel analysis circulated by Heise (2012c) provided the inspiration of a reexamination of the idea of cultural consensus
even in impression change dynamics. In this paper, Heise rated all of the 128 vignettes necessary to estimate his own personal set of impression change equations, finding key ways in which his impressions differ substantially from those of mainstream culture as represented in the equations contained within Interact. This leads us to the question of whether we need (for maximum precision) separate equations for each individual, and even whether personalized impression equations are stable over time. To begin to address this question, Smith-Lovin and colleagues (SmithLovin, Heise, Freeland, & Mageed, 2010; Smith-Lovin, Heise, & Rogers, 2010) (see URL address in note 4) collected innovative data in which they asked every respondent participate in 10 data collection session to rate all of the stimuli necessary to estimate a separate equation for each individual respondent. One of these data collection efforts was in Arabic and surveyed a highly diverse population of individuals from cultures as different as Iraq, Egypt, Syria, and Lebanon. This study was a pilot study for a planned multi-country study of Arabic linguistic culture. These respondents, who were surveyed in Durham or Charlotte, NC, varied in culture of origin, religion, gender, and length of time in the United States. By no means did this sample of respondents resemble a culture in our conventional understanding of the concept. The respondents shared a linguistic

152

NEIL J. MACKINNON AND DAWN T. ROBINSON

culture alone. A hierarchical linear modeling approach to modeling the impression change equations in this sample, however, revealed that the grand mean equations explained nearly all of the variance (Kriegel, 2013). In the equations predicting evaluation, or goodness, there was not enough unexplained variance to proceed with any lower-level analysis. In the potency and activity equations, there was barely enough unexplained variance. Even so, none of the available demographic (gender, religion, years in the United States) explained the observed variation. The upshot of this is that individuals
even those who originate from what we would typically consider highly variable cultures
respond similarly to social events. Moreover, and importantly, Kriegel found that models estimated using a hierarchical linear modeling approach were much simpler than the models previously estimated in the affect control theory tradition. Current work is using these same techniques to analyze the degree of shared culture in the U.S. equations. Capitalizing on the recent development of methods for analyzing nested data may offer a new path toward reinvestigating old assumptions about cultural consensus at the level of both concepts and responses to social events.

Agent-Based Modeling Affect control theory has traditionally employed a system-based approach for conducting computer simulations at the level of the social situation. Over the last few decades, there has been increasing use of agent-based approaches to simulating social phenomena. The first agent-based modeling analyses making use of affect control theory appeared as a trio of papers at the first ACT conference in 2002 (Friedkin & Johnsen, 2002; Wang, 2002; Youngreen, 2002). Friedkin and Johnsen (later published in 2003) incorporated the idea and measurement of fundamental sentiments into their social influence network theory. Wang (2002) examined the stability of cross-cultural friendships by simulating emergent patterns of interactions in which agents were guided by sentiments and equations from differing cultural models within the theory. Youngreen (2002) investigated the emergence of friendships in a classroom setting. More recently, Heise (2012a) developed a version of the program Interact for simulating social interaction in small groups. The Group Simulator Program uses an agent-based modeling approach and can simulate interaction in many kinds of small groups. Heise (2013) illustrated the power of the program by comparing computer simulations of jury deliberations

Affect Control Theory

153

with empirical findings from Strodtbeck, James, and Hawkins’ (1957) classic research on mock jury trials using Bales’ IPA (Interaction Process Analysis).

Bayesian Version of Affect Control Theory As described earlier, the traditional version of affect control theory is a theory of situated identities. And, while the theory has occasionally been applied to selves and relationships, the traditional formulation of the theory does not lend itself well to those endeavors. MacKinnon and Heise’s (2010) foray into new affect control theory based theories of selves and institutions offer a partial solution to this limitation. An additional remedy is offered by a novel recent reframing of affect control theory using a Bayesian formulation. Hoey et al. (2013) proposed a version of affect control theory based on Bayesian statistics, allowing researchers to entertain multiple hypotheses about behaviors and identities simultaneously as a probability distribution and generating affectively believable interactions among people. This new specification will allow simulations to represent identity labels, emotions, or behaviors, as probability distributions that alter in response to observed ongoing interactions. Thus, the theory begins with hypotheses based on identity meanings, but can “learn” about the persona of an individual over time. This approach may be the key for testing new ideas postulated in MacKinnon and Heise (2010), but also opens up a whole host of more rigorous and more realistic hypothesis testing avenues for the theory.

CROSS-CULTURAL RESEARCH Among the most important contributions affect control theory researchers have made in the past quarter century is the development of a host of new sentiment dictionaries and impression change equations for different cultures. These afford more opportunity than ever before for examining crosscultural dynamics of emotions and behaviors in situated interactions. Below is a listing of data sets available for cross-cultural analysis.4 U.S.A.: Georgia and North Carolina, 2013
2014: Ratings of roughly 2,400 Behaviors, Identities, and Modifiers are being collected at Duke University and the University of Georgia at the time of this writing. Each stimulus is

154

NEIL J. MACKINNON AND DAWN T. ROBINSON

being rated by 50 men and 50 women, roughly balanced between the two research sites (Smith-Lovin and Robinson). U.S.A.: North Carolina, 2010: Ratings of roughly 1,000 stimuli, including Identities, Behaviors, Modifiers, and Events were collected at Duke University using a freestanding (not via the Internet) applet version of Surveyor (Smith-Lovin, Heise, and Rogers). U.S.A. (Arabic): North Carolina, 2010: Ratings of 40 Identities, 32 Behaviors, 18 Modifiers, and 256 events were collected at Duke University using a freestanding (not via the Internet) applet version of Surveyor (Smith-Lovin, Heise, Freeland, and Mageed). U.S.A.: Indiana, 2003: Ratings of 500 Identities, 500 Behaviors, 300 Modifiers, and 200 Settings were collected at Indiana University using the Internet and the Surveyor program (Francis and Heise). U.S.A.: Texas, 1998: Ratings of 443 Identities, 278 Behaviors, 65 Modifiers, and 1 Setting were collected at Texas Tech University with program Attitude (Schneider). U.S.A.: North Carolina, 1978: Ratings of 721 Identities, 600 Behaviors, 440 Modifiers, and 345 Settings were obtained with paper questionnaires from 1,225 North Carolina undergraduates (Smith-Lovin and Heise). Canada: Ontario, 1980
1986: Data on 843 Identities and 593 Behaviors were obtained from 5,534 Guelph, Ontario, undergraduates with paper questionnaires in 1980
1983, and 495 Modifiers rated by 1,260 Guelph undergraduates were added in 1985
1986 (MacKinnon). Canada: Ontario, 2001
2003: Data on 993 Identities, 601 Behaviors, 500 Modifiers, and 200 Settings were gathered with the Attitude program from Guelph, Ontario, undergraduates in 2001
2002. Data on settings were gathered with the Surveyor program at Guelph in 2003 (MacKinnon). Japan, 1989
2002: Data on 713 Identities, 455 Behaviors, 426 Modifiers, and 300 Settings were collected over several surveys using the Attitude or Surveyor programs (Smith, Matsuno, Ike, and Umino). Mainland China, 1991: Ratings of 449 Identities, 300 Behaviors, 98 Emotions, 150 Traits, and 149 Settings were obtained with the Attitude program from about 380 undergraduate students at Fudan University in Shanghai, Peoples Republic of China (Smith and Yi Cai).

155

Affect Control Theory

Germany, 1989: Ratings of 442 Identities, 295 Behaviors, and 67 Modifiers were obtained with the Attitude program from 520 Mannheim students (Schneider). Germany, 2008: Mean affective ratings of 1,100 concepts (Schro¨der). Northern Ireland, 1977: Ratings of 528 Identities and 498 Behaviors were obtained with paper questionnaires from 319 Belfast teenagers in Catholic high schools in 1977 (Willigan and Heise). India, 2013: Ratings of 2,000 concepts on Bengali rating forms by over 40 Calcutta respondents, equally male and female, with each respondent rating all 2,000 concepts (Mukherjee and Heise). Egypt, 2013
2014: Ratings of roughly 1,500 stimuli, including Behaviors, Identities, Modifiers, and Events by 2,000 Cairo respondents, with each stimulus being rated by 50 men and 50 women (Smith-Lovin, Robinson, and Latif).

Cross-Cultural Analysis Early in the history of ACT, researchers operated on the assumption that the impression change equations captured basic, perhaps universal, processes involved with responding to social events (Smith-Lovin, 1987a), and the sentiment dictionaries (EPA ratings of social elements, averaged within a culture) captured the more interesting cultural variation. Consequently, much of the major cross-cultural analysis in ACT has focused on comparing sentiments. MacKinnon and Keating (1989) conducted a cross-cultural comparison of the United States and Canada on the structure of emotions in EPA space and found that while there were cultural differences in both affective range and intensity of emotion word ratings, the structure of emotions was the same. Langford and MacKinnon (2000) conducted a cross-cultural comparison of Canada and the United States on the gendering of traits, again finding highly similar patterns in the two cultures with respect to the gender stereotyping of traits. Schneider found considerably more differences between sentiments in the United States and Germany in cross-cultural comparisons of the association among sexual eroticism, emotions, and violence (1996, 1999a, 2005), the affective meaning of role identities (1999b), and neo-conservatism (1999c). From this research, Schneider (2002a) proposed an intracultural standard based on gender with which to analyze cross-cultural differences in situations where researchers are familiar with only one of the cultures being

156

NEIL J. MACKINNON AND DAWN T. ROBINSON

compared. Continuing his research on cross-cultural comparisons between the United States and Germany, Schneider (2004) compared the ideal type of authority in the two countries; and Schneider and Schro¨der (2012) located leadership styles based on Weber’s ideal types in EPA space and employed this frame for a cross-cultural comparison of business managers in the United States and Germany at different points in time. In a comparison of the EPA ratings of 56 stereotyped groups across the United States, Germany, and Japan, Schro¨der, Rogers, Ike, Mell, and Scholl (2013) found greater within-culture than across-culture consensus and a greater similarity between the United States and Germany and were able to relate these cross-cultural differences to the relative standings of countries on general cultural dimensions. Finally, Smith and Yap (2006) compared American and Japanese cultures in terms of guilt and shame, finding cultural differences in the meanings of the emotions as well as related behaviors. A series of studies by Herman Smith led ACT researchers to question the early assumption about the cultural universality of the impression change equations. Smith et al. (1994, 2001); Smith (2002); and Smith and Francis (2005) compared American and Japanese cultures with respect to impression change and attribution processes, and the role of settings, emotions, moods, and personality traits in these processes, and found several differences that they interpreted as culturally meaningful.

EMPIRICAL TESTS OF THE THEORY There have been many explorations and applications of affect control theory employing survey, ethnographic, and computer simulation modes of research both prior to and after 1988. However, up to 1988, there were only two empirical tests of the theory (Heise & MacKinnon, 1987; Wiggins & Heise, 1987).5 In the last 25 years, however, there have been a number of tests of the theory’s predictions by affect control theory researchers. An experiment by Robinson and Smith-Lovin (1992) based on self-verification theory verified a key assumption of affect control theory that people try to confirm current impressions of themselves, in this case by selectively interacting with others; and an additional experiment by Robinson and Smith-Lovin (1999) confirmed that emotion display is an effective strategy for identity maintenance. Smith-Lovin and Douglass validated affect control theory predictions by getting members of two religious subcultures to rate the likelihood of events

157

Affect Control Theory

in a survey design where the selected events were predicted by affect control theory to be deflecting in one subculture and confirming in another. Robinson, Smith-Lovin, and Tsoudis (1994) conducted an experimental study on the ameliorating effects of remorse on the responses of subjects to mock criminal confessions. An experimental study of identity maintenance and cognitive test performance confirmed affect control theory predictions revealing that cognitive test performance varied by its connection to a salient, valued identity (Youngreen, Conlan, Robinson, & Lovaglia, 2009). Employing a German dictionary of EPA profiles (Schro¨der, 2008) and German impression change equations (Schro¨der, 2011), Schro¨der and Scholl (2009) confirmed the affect control principle experimentally, demonstrating that subjects in a leadership task were more likely to choose behaviors causing less affective deflection and to label their emotional experiences with emotion terms having low Euclidean distances to those predicted by ACT. Schro¨der, Netzel, Schermuly, and Scholl (2013) found that the affect control theory model successfully predicted nonverbal observations in actual dyadic interaction for both static baseline deflections and dynamic sequences of interpersonal behavior. In a vignette study of affect control theory’s predictions about consequent emotions (Heise & Calhan, 1995), students were randomly assigned to take the role of the actor or object of the behavior while imagining themselves in 128 situations and reported the emotion they felt. The study supported predictions from the theory. A later reanalysis (Heise & Weir, 1999) examining scatter plots of the distances from affect control predictions and the frequencies of emotion choice found that the distributions of these errors provided even more detailed support of the theory’s predictions. The positions of the reported emotions were usually very close (in EPA space) to the theoretical emotion predicted by affect control theory, and emotions positioned far (in EPA space) from the theoretical emotion predicted by affect control theory were rare.

SUBSTANTIVE APPLICATIONS Emotions The last few decades saw a considerable amount of attention to the affect control theory model of emotions (See Robinson, Smith-Lovin, & Wisecup, 2006 for a comprehensive review). In affect control theory, emotions serve

158

NEIL J. MACKINNON AND DAWN T. ROBINSON

the role of signaling the extent of confirmation and disconfirmation of social identities by experienced events (Smith-Lovin, 1990). Consequently, they can be used by social observers to inform new identity impressions of others (Heise, 1989b; Robinson & Smith-Lovin, 1992, 1999). The interaction between the evaluation of identities and emotions amplifies the negativity of emotions for people with stigmatized identities (Heise & Thomas, 1989); the expression of emotion in groups (Heise & O’Brien, 1993); and transitional paths connecting emotions in EPA space (Lively & Heise, 2004). ACT researchers have examined the structure of emotions in the United States (Morgan & Heise, 1988) and in Canada (MacKinnon & Keating, 1989) and applied the affect control theory model of emotions to empathic solidarity (Heise, 1998). Smith-Lovin (2002, 2003) developed new affect control theory arguments reconciling emotion and rationality (1993) and describing how parallel processing of multiple social definitions of events can lead to mixed emotions. MacKinnon and Goulbourne (2006) developed an application of the affect control theory of emotions to the case of depression. Creative substantive analyses of emotions include Schneider’s (1996) cross-cultural comparison of the United States and Germany with respect to sexual-erotic emotions; Francis’ (1997a, 1997b) articles on therapeutic ideologies and emotion management in support groups; Lee and Shafer’s (2002) application of affect control theory to the emotional experience of trail users; Lively and Powell’s (2006) article on emotional expression at work and at home; Smith and Yap’s (2006) comparison of American and Japanese cultures in terms of guilt and shame mentioned earlier; Schro¨der and Scholl’s (2009) experimental study on emotions and behavior in a leadership task; and Smith and Schneider’s (2009) critical analysis of emotions models. Additional affect control research on emotion is reviewed in the section on empirical tests of the theory earlier. There are also a number of recent extensions of the affect control theory of emotions. Using ideas from affect control theory in conjunction with other theories of identity and self, Doan (2012) developed and tested a model identifying the factors and mechanisms through which ephemeral emotions become persistent moods. Employing Interact computer simulations to explore empirical connections between affect control theory and expectations states theory, Dippong (2013) found that attempts to reduce deflection through behavior may result in changes in expectations over the course of an interaction. And Rogers, Schro¨der, & von Scheve (2014) use affect control theory as a basis for developing a multilevel model of emotion construction connecting cultural, interactional, individual, and neural levels of analysis.

159

Affect Control Theory

Social Structure, Culture, and Social and Cultural Change A simulation analysis (Robinson, 1996) of affective dynamics and network formation revealed that affect control theory predicts that the sentiments associated with some identities make some relationships more comfortable than others, influencing the formation of friendship affiliations and cliques. In particular, affect control theory predicts homophily along the evaluation dimension of identities, but structural equivalence along the potency dimension. Later analyses (Robinson, Smith-Lovin, Doan, & Moore, 2014) show how these patterns are qualified when social setting is taken into account. Lovaglia, Youngreen, and Robinson (2005) connected performance in an identity with the sentiment associated with the identity. MacKinnon and Luke (2002) analyzed changes in EPA profiles of identities over a 15-year period as reflections of social and cultural change; and Dunphy and MacKinnon (2002) applied affect control theory and Interact simulations to folklore research. In a creative integration of their dynamic models of influence networks with affect control theory, Friedkin and Johnsen (2003) showed how fundamental sentiments associated with identities mediate the effects of status characteristics on interpersonal influence in task groups.

Occupations and Work Heise (1990b) analyzed aging and institutional careers in terms of sentiments about sequenced identities. MacKinnon and Langford (1994) analyzed the meaning of occupational prestige scores, showing how the effects of income and education are mediated by the EPA profiles of occupational identities. Smith (1995) showed how differences in cultural sentiments can cause stress in business relations between Americans and Japanese; and Schneider (2002b) showed how differences in cultural sentiments can cause stress when local offices have to obey top-down edicts of multinational corporations. Moore and Robinson (2006) connected the selection of occupational identities to existing self-views.

Politics Schneider (1999c) applied affect control theory to analyze neo-conservatism in the United States and Germany. Britt and Heise (2000) explained how

160

NEIL J. MACKINNON AND DAWN T. ROBINSON

a social movement converts the negative emotions of shame and loneliness into feelings of pride and solidarity by traveling through intermediate emotional states of fear and then anger. Heise and Lerner (2006) applied affect control theory to international relations among countries in the Middle East, and Heise (2006) extended this study to show how social sentiments develop and change in sequences of social interaction between nations. Troyer and Robinson (2006) demonstrated the relevance of affect control theory to political science by applying the affect control theory of emotion to political identity and action.

Gender, Ideology, Subcultures A number of affect control theory researchers have used ACT to theorize about gender in interpersonal interaction. Smith-Lovin and Robinson (1992) applied affect control theory to gender and conversational dynamics. In a series of publications, Kroska (1997, 2001, 2002, 2003) connected gender ideology with the meaning of role identities and the division of labor in the home. Lively, Steelman, and Powell (2010) also addressed issues of equity, emotion, and household division of labor. Smith, Umino, and Matsuno (1998) studied gender-differentiated sentiments in Japan; Lee (1998) used affect control theory in conjunction with identity theory (e.g., Burke, 1991) to measure identities and to determine how gender, selfconcept, and perceptions of scientific others affect students’ interests in science, mathematics, and engineering; and Langford and MacKinnon (2000) compared the affective basis for the gendering of traits between Canada and the United States. Affect control theory ideas and measurement procedures have also been fruitfully applied to ideology and subcultures. Smith-Lovin and Douglass (1992) demonstrated that affect control theory successfully predicts behavior and emotions in Unitarian and gay fundamentalist religious subcultures. Francis (1997a, 1997b) examined the therapeutic ideologies of different support groups, finding support for the affect control principle in strategies used by therapy group leaders. Berbrier (1998) employed affect control theory as a framework to bring emotion into an interpretation of the rhetoric of contemporary white supremacist subculture. Thomassen (2002) studied sentiment change in Alcoholic Anonymous; King (2001, 2008) connected Internet cultures with social sentiments for Internet identities and behaviors. Hunt (2008, 2010) studied involvement, identity-roles, conformity, and deviance in Jamband subcultures. And, Sewell and Heise

161

Affect Control Theory

(2010) explored the copresence of Black and White cultures in the United States using 1970s data on cultural sentiments.

Deviance and Criminology There are an appreciable number of papers applying affect control theory to deviance or criminology, one of the earliest of which was the article by Scher and Heise (1993) on affect and the perception of justice. Kalkhoff (2002) examined delinquency and violence as affect control in support of the revival of the subcultural approach in criminology. A series of experimental studies (Robinson et al., 1994; Tsoudis, 2000; Tsoudis & Smith-Lovin, 1998, 2001) investigated how the expression of emotions by perpetrators and victims affects sentencing outcomes in mock criminal trials. In another series of articles, Kroska and Harkness (2006, 2008, 2011) applied affect control theory to the stigma of mental illness. Schneider (1996, 1999a, 2005) showed a greater propensity for sexual eroticism to be connected to violence in the United States compared to Germany and proposed a model of sexual constraint and emancipation based on differences between the two cultures in the association between shame and violence. Schneider (2009) also applied affect control theory to show how the willful exposure to pain in religious and sexual contexts can be rendered into a positive affective experience. Thomassen’s (2002) study of Alcoholics Anonymous and Hunt’s (2010, 2012) study of Jamband subculture also represent studies in deviance.

Stereotyping MacKinnon and Bowlby (2000) identified three deficits in the social cognition approach to stereotyping and intergroup relations in psychology that can be corrected by an affect control theory approach, and conducted three studies based on intergroup attitudes for Canadian regional identities to substantiate this claim. Schneider (2007) showed how politically correct stereotyping in Texas is largely unfounded. Rogers, Schro¨der, and Scholl (2013) compared the dimensional structure and predictions of the stereotype content model of Cuddy et al. (2009) with the EPA structure and predictions of ACT. The comparison of the EPA ratings of 56 stereotyped groups across the United States, Germany, and Japan by Schro¨der et al. (2013) found greater within-culture than across-culture consensus and

162

NEIL J. MACKINNON AND DAWN T. ROBINSON

greater similarity between the United States and Germany and explained these cross-cultural differences in terms of the relative standings of countries on general cultural dimensions.

Physiological Behavior Robinson, Rogalin, and Smith-Lovin (2004) summarized the state of the physiological measurement literature, connecting various measurable responses to constructs in sociological theories. Drawing from extant empirical evidence, they offered predictions about which physiological measures might most closely map onto which theoretical constructs. In an experiment crossing status with competitive outcomes, Civettini (2012) found support for one of these predictions when she reported that individuals in high deflection conditions (high-status losers, low-status winners), released more cortisol while both low- and high-status winners experienced more positive emotion than losers (consistent with affect control theory predictions). Robinson et al. (2012) found support for more of these predictions
showing that warming in and around the eyes (as measured by infrared thermography) corresponded closely to deflection, while warming in the brow and cheeks corresponded more closely to positive and negative emotion.

SUMMARY AND CONCLUSION In summary, we have reviewed progress in affect control theory and research in the last 25 years prefaced by a capsule summary of its beginnings in the late 1960s/early 1970s to its coalescence and maturation in the 1970s/1980s. We structured our review in terms of theoretical and technological advances, cultural and cross-cultural work, empirical tests of the theory, and applications of the theory to numerous substantive areas. Under the rubric of theoretical developments, we included expositions of affect control theory and expansions of the theory to deal with self, social institutions, interaction in small groups, nonverbal behavior, neuroscience, artificial intelligence, and interaction with the physical environment and technology. Under technological advances, we included methods and computer programs for collecting cultural data, the refinement of impression change equations and their expansion to other cultures, the

Affect Control Theory

163

development of a computer program for simulating interaction in small groups, and the development of a Bayesian statistics version of ACT. Under applications to substantive areas we included studies of emotions; social structure, culture, and social and cultural change; occupations and work; politics; gender, ideology, and subcultures; deviance and criminology; and stereotyping. The number of publications on affect control theory in the last 25 years has been impressive. In addition to four major books since 1988 (Heise, 2007, 2010; MacKinnon, 1994; MacKinnon & Heise, 2010), there have been over 120 refereed articles and chapters, with more on the horizon, as well as a sizable number of theses, dissertations, presentations, and research reports not included in this review. The numerous and diverse substantive areas to which affect control theory has been applied attest to the scope of the theory. The expansion of affect control theory research from North American to European, Asian, and Middle-Eastern cultures in the last 25 years attests to its cross-cultural generalization. Finally, the application of the theory to emerging areas of interdisciplinary research such as artificial intelligence and neuroscience suggests a productive future. One of the ways to characterize the current work in the field is that we are coming around to revisit old assumptions and procedures with new methods and new data. The history and development of affect control theory has always been shaped by the available tools and technologies. In the last few years, the use of newer methods and procedures are shifting our attention again. Small handheld devices that allow field collection of affect control theory data, services such as Amazon’s Mechanical Turk, and resources such as Time-sharing Experiments for the Social Sciences (TESS) are all allowing more distal collection of affect control theory data, affording researchers opportunities to access more variable populations than ever before. This is allowing us to reexamine questions about cultural consensus of meaning across demographic divides. Two recent dissertations (Rogers, 2013; Wisecup, 2011) and a new research paper (Ambrasat, Von Scheve, Schauenburg, & Schroder, 2014) address these questions squarely and offer not only substantial support for the consensus model but also (Ambrasat et al., 2014) evidence for systemic relationships between social structure and the degree of consensus. Use of nested equation data and hierarchical linear modeling techniques have opened up the possibility of working with personal equations for individuals to yield ever more precise predictions. This approach is also allowing us to test hypotheses about the variability of internalized culture between people, the degree of consensus within samples of respondents, and the

164

NEIL J. MACKINNON AND DAWN T. ROBINSON

identification of subcultures. The development of a new Bayesian affect control theory model (Hoey et al., 2013) is allowing us to incorporate some entirely new processes, not captured by earlier specifications of the theory. New measurement techniques are affording us more opportunities to test affect control theory predictions in the context of unfolding interactions, rather than retrospectively through the use of questionnaires (Civettini, 2012; Robinson et al., 2012). These technological and methodological advances are allowing us to reexamine questions from the past and find new ways forward
taking us back to the future.

NOTES 1. The first book using that title was, Introduction to Mathematical Sociology, published in 1964 by James Coleman and the Journal of Mathematical Sociology launched in 1971. 2. The term collocate refers to words frequently occurring with other words. 3. At present, Surveyor no longer exists as an online tool. The applet version has since been replaced by a freestanding application (Morgan & Morgan, 2014). Compared to the applet version, this new application has the advantage of being user-adaptable for various research designs, of being able to collect and store data offline on the user’s storage medium, and being friendlier to non-Roman characters for conducting international research. At time of this writing, there are plans in the place to develop a new online version of this application. Two websites http://www. indiana.edu/∼socpsy/ACT/ and http://act.uga.edu/ house these programs, along with a host of other resources for affect control researchers, including bibliographies and data archives. 4. Archived data sets, metadata, and citation guidance can be accessed at: http:// act.uga.edu/data/ 5. Wiggins and Heise (1987) experimentally verified the principle of evaluative balance in social interaction (the BeOe interaction term in the impression formation equation for actor evaluation in ACT), which specifies that actors treat positively evaluated objects with kindness and negatively evaluated objects with disdain in order to create positive impressions of themselves. Heise and Mackinnon (1987) investigated whether deflection predicts perceived likelihood of events by examining likelihood judgments in response to 515 social events.

REFERENCES Alexander, C. N., Jr., & Wiley, M. G. (1981). Situated activity and identity formation. In M. Rosenburg & R. H. Turner (Eds.), Social psychology: Sociological perspectives (pp. 269
282). New York, NY: Basic Books.

Affect Control Theory

165

Ambrasat, J., Von Scheve, C., Schauenburg, G., & Schroder, T. (2014). Consensus and stratification in affective meanings of human sociality. Unpublished manuscript. Averett, C., & Heise, D. R. (1987). Modified social identities: Amalgamations, attributions, and emotions. The Journal of Mathematical Sociology, 13(1
2), 103
132. Berbrier, M. (1998). Half the battle: Cultural resonance, framing processes, and ethnic affectations in contemporary white separatist rhetoric. Social Problems, 45, 431
450. Britt, L., & Heise, D. R. (1992). Impressions of self-directed action. Social Psychology Quarterly, 55(4), 335
350. Britt, L., & Heise, D. R. (2000). From shame to pride in identity politics. In S. Stryker, T. J. Owens, & R. W. White (Eds.), Self, identity and social movements (pp. 252
268). Minneapolis, MN: University of Minnesota Press. Burke, P. J. (1991). Identity processes and social stress. American Sociological Review, 56(6), 836
849. Burke, P. J., & Reitzes, D. C. (1981). The link between identity and role performance. Social Psychology Quarterly, 54(3), 83
92. Burke, P. J., & Reitzes, D. C. (1991). An identity theory approach to commitment. Social Psychology Quarterly, 54(3), 239
251. Burke, P. J., & Tully, J. C. (1977). The measurement of role identity. Social Forces, 55(4), 881
897. Civettini, N. H. (2012). Self-enhancement versus self-verification: Physiological and self-report responses to status dissonance. Advances in Group Processes, 29, 201
223. Coleman, J. S. (1964). Introduction to mathematical sociology. New York, NY: Free Press Glencoe. Cuddy, A. J., Fiske, S. T., Kwan, V. S., Glick, P., Demoulin, S., Leyens, J. P., et al. (2009). Stereotype content model across cultures: Towards universal similarities and some differences. British Journal of Social Psychology, 48(1), 1
33. Dippong, J. (2013). Using simulated interactions to explore emotional processes and status organizing processes: A joint application of expectation states theory and affect control theory. Advances in Group Processes, 30, 195
229. Doan, L. (2012). A social model of persistent mood states. Social Psychology Quarterly, 75(3), 198
218. Dunphy, T., & MacKinnon, N. J. (2002). A proposal for integrating folklore and affect control theory. Electronic Journal of Sociology, 6, 3. Francis, L. E. (1997a). Emotion, coping, and therapeutic ideologies. Social Perspectives on Emotion, 4, 71
102. Francis, L. E. (1997b). Ideology and interpersonal emotion management: Redefining identity in two support groups. Social Psychology Quarterly, 153
171. Friedkin, N. E., & Johnsen, E. C. (2003). Attitude change, affect control, and expectation states in the formation of influence networks. Advances in Group Processes, 20, 1
29. Gollob, H. F. (1968). Impression formation and word combination in sentences. Journal of Personality and Social Psychology, 10, 341
353. Gollob, H. F., & Rossman, B. B. (1973). Judgments of an actor’s power and ability to influence others. Journal of Experimental Social Psychology, 9, 391
406. Heise, D. R. (1965). Semantic differential profiles for 1,000 most frequent English words. Psychological Monographs: General and Applied, 79(8), 1. Heise, D. R. (1969a). Affectual dynamics in simple sentences. Journal of Personality and Social Psychology, 11(3), 204.

166

NEIL J. MACKINNON AND DAWN T. ROBINSON

Heise, D. R. (1969b). Some methodological issues in semantic differential research. Psychological Bulletin, 72(6), 406. Heise, D. R. (1970). Potency dynamics in simple sentences. Journal of Personality and Social Psychology, 16(1), 48. Heise, D. R. (1978). Computer-assisted analysis of social action: Use of program INTERACT and SURVEY. UNC75. Chapel Hill, NC: University of North Carolina, Institute for Research in Sociology. Heise, D. R. (1979). Understanding events: Affect and the construction of social action. Cambridge: Cambridge University Press. Heise, D. R. (1985a). Facial expression of emotion as a means of socialization. Electronic Journal of Sociology, 1, 1. Heise, D. R. (1985b). Affect control theory: Respecification, estimation, and tests of the formal model. Journal of Mathematical Sociology, 11(3), 191
222. Heise, D. R. (1987). Affect control theory: Concepts and model. Journal of Mathematical Sociology, 13, 1
33. Heise, D. R. (1989a). Modeling event structures. Journal of Mathematical Sociology, 14(2
3), 139
169. Heise, D. R. (1989b). Effects of emotion displays on social identification. Social Psychology Quarterly, 52(1), 10
21. Heise, D. R. (1990a). Affect control model technical appendix. In T. D. Kemper (Ed.), Research agendas in the sociology of emotion (pp. 271
280). Albany, NY: SUNY Press. Heise, D. R. (1990b). Careers, career trajectories, and the self. In J. Rodin, C. Schooler, & K. W. Schaie (Eds.), Self-directedness: Cause and effects throughout the life course (pp. 59
84). New York, NY: Lawrence Erlbaum Associates. Heise, D. R. (1991). OLS Equation Estimations for Interact. Technical Report. Retrieved from http://www.indiana.edu/∼socpsy/papers/EQ_Estimations.pdf Heise, D. R. (1998). Conditions for empathic solidarity. In P. Doreian & T. J. Fararo (Eds.), The problem of solidarity: Theories and models (pp. 197
211). Amsterdam: Gordon & Breach. Heise, D. R. (2000a). Affect control theory and impression formation. In E. F. Borgatta & R. J. Montgomery (Eds.), Encyclopedia of sociology (Vol. 1, pp. 41
47). New York, NY: MacMillan. Heise, D. R. (2000b). Thinking sociologically with mathematics. Sociological Theory, 18(3), 498
504. Heise, D. R. (2001). Project Magellan: Collecting cross-cultural affective meanings via the internet. Electronic Journal of Sociology, 5, 3. Heise, D. R. (2004). Enculturating agents with expressive role behavior. In S. Payer & R. Trapp (Eds.), Agent culture: Human
agent interaction in a multicultural world (pp. 127
142). Mahwah, NJ: Lawrence Erlbaum Associates. Heise, D. R. (2006). Sentiment formation in social interaction. In K. McClelland & T. J. Fararo (Eds.), Purpose, meaning, and action: Control systems theories in sociology (pp. 189
211). New York, NY: Palgrave. Heise, D. R. (2007). Expressive order: Confirming sentiments in social actions. New York, NY: Springer. Heise, D. R. (2010). Surveying cultures. Hoboken, NJ: Wiley. Heise, D. R. (2012a). Group Simulator. Version Retrieved from http://www.indiana.edu/∼ socpsy/ACT/SmallGroups/GroupSimulator.html. Accessed on September 12, 2012.

Affect Control Theory

167

Heise, D. R. (2012b). Methodology in impression change research. Unpublished white paper. Indiana University. Heise, D. R. (2012c). State of mind: Self-surveying an impression formation system. Unpublished manuscript. Indiana University. Heise, D. R. (2013). Modeling interactions in small groups. Social Psychology Quarterly, 76(1), 52
72. Heise, D. R., & Calhan, C. (1995). Emotion norms in interpersonal events. Social Psychology Quarterly, 58(4), 223
240. Heise, D. R., & Lerner, S. J. (2006). Affect control theory in international interactions. Social Forces, 85(2), 993
1010. Heise, D. R., & Lewis, E. (1988). Programs interact and attitude: Software and documentation. Dubuque, IA: Wm. C. Brown Publishers, Software. Heise, D. R., & Mackinnon, N. J. (1987). Affective bases of likelihood judgments. The Journal of Mathematical Sociology, 13(1
2), 133
151. Heise, D. R., & O’Brien, J. (1993). Emotion expression in groups. In M. Lewis & J. M. Haviland (Eds.), The handbook of emotions (pp. 489
497). New York, NY: Guilford Press. Heise, D. R., & Smith-Lovin, L. (1981). Impressions of goodness, powerfulness, and liveliness from discerned social events. Social Psychology Quarterly, 44(2), 93
106. Heise, D. R., & Thomas, L. (1989). Predicting impressions created by combinations of emotion and social identity. Social Psychology Quarterly, 52(2), 141
148. Heise, D. R., & Weir, B. (1999). A test of symbolic interactionist predictions about emotions in imagined situations. Symbolic Interaction, 22(2), 139
161. Hoey, J., Schro¨der, T., & Alhothali, A. (2013). Affect control processes: Probabilistic and decision theoretic affective control in human-computer interaction. arXiv preprint arXiv:1306.5279. Hunt, P. M. (2008). From festies to tourrats: Examining the relationship between jamband subculture involvement and role meanings. Social Psychology Quarterly, 71(4), 356
378. Hunt, P. M. (2010). Are you kynd? Conformity and deviance within the jamband subculture. Deviant Behavior, 31(6), 521
551. Hunt, P. M. (2012). Examining the affective meanings of interaction settings in the jamband music subculture. The Journal of Public and Professional Sociology, 4(1), 5. Kalkhoff, W. (2002). Delinquency and violence as affect-control: Reviving the subcultural approach in criminology. Electronic Journal of Sociology, 6(3). King, A. B. (2001). Affective dimensions of internet culture. Social Science Computer Review, 19, 414
430. King, A. B. (2008). Finding online subcultures in shared meanings. Social Science Computer Review, 26, 137
151. Kriegel, D. J. (2013). Arabic impression change. Unpublished Masters Thesis. University of Georgia, Athens, GA. Kroska, A. (1997). The division of labor in the home: A review and reconceptualization. Social Psychology Quarterly, 60, 304
322. Kroska, A. (2001). Do we have consensus? Examining the relationship between gender ideology and role meanings. Social Psychology Quarterly, 64, 18
40. Kroska, A. (2002). Does gender ideology matter? Examining the relationship between gender ideology and self-and partner-meanings. Social Psychology Quarterly, 65, 248
265.

168

NEIL J. MACKINNON AND DAWN T. ROBINSON

Kroska, A. (2003). Investigating gender differences in the meaning of household chores and child care. Journal of Marriage and Family, 65(2), 456
473. Kroska, A., & Harkness, S. K. (2006). Stigma sentiments and self-meanings: Exploring the modified labeling theory of mental illness. Social Psychology Quarterly, 69(4), 325
348. Kroska, A., & Harkness, S. K. (2008). Exploring the role of diagnosis in the modified labeling theory of mental illness. Social Psychology Quarterly, 71(2), 193
208. Kroska, A., & Harkness, S. K. (2011). Coping with the stigma of mental illness: Empiricallygrounded hypotheses from computer simulations. Social Forces, 89(4), 1315
1339. Langford, T., & MacKinnon, N. J. (2000). The affective bases for the gendering of traits: Comparing the United States and Canada. Social Psychology Quarterly, 35, 34
48. Lee, B., & Shafer, C. S. (2002). The dynamic nature of leisure experience: An application of affect control theory. Journal of Leisure Research, 34(3), 290
310. Retrieved from http://js.sagamorepub.com/jlr/article/view/622 Lee, J. D. (1998). Which kids can “become” scientists? Effects of gender, self-concepts, and perceptions of scientists. Social Psychology Quarterly, 61, 199
219. Lively, K. J., & Heise, D. R. (2004). Sociological realms of emotional experience. American Journal of Sociology, 109(5), 1109
1136. Lively, K. J., & Powell, B. (2006). Emotional expression at work and at home: Domain, status, or individual characteristics? Social Psychology Quarterly, 69(1), 17
38. Lively, K. J., Steelman, L. C., & Powell, B. (2010). Equity, emotion, and household division of labor. Social Psychology Quarterly, 73(4), 358
379. Lovaglia, M. J., Youngreen, R., & Robinson, D. T. (2005). Identity maintenance, affect control, and cognitive performance. Advances in Group Processes, 22, 65
91. Lulham, R. (2004). Controlling your feelings in juvenile justice centres: Applying affect control theory to an investigation of the influence of the physical environment. Proceedings of the environmental design research association conference, Albuquerque, NM. MacKinnon, N. J. (1985/1988/1998). Final reports to Social Sciences and Humanities Research Council of Canada on Projects 410-81-0089, 410-86-0794, and 410-94-0087. Department of Sociology and Anthropology, University of Guelph. Guelph, ON. MacKinnon, N. J. (1994). Symbolic interactionism as affect control. Albany, NY: State University of New York Press. MacKinnon, N. J. (2003). Keynote address
Symbolic interaction and knowledge presentation: From cognitive to affective models. Preparing for the future of knowledge presentation. Chicago, IL: International Institute for Information Design, Institute of Design, Illinois Institute of Technology. MacKinnon, N. J., & Bowlby, J. W. (2000). The affective dynamics of stereotyping and intergroup relations. Advances in Group Processes, 17, 37
76. MacKinnon, N. J., & Goulbourne, M. M. (2006). The affect control theory of emotions: The case of depression. In K. A. McClelland & T. J. Fararo (Eds.), Purpose, meaning, and action: Control systems theories in sociology (pp. 237
266). New York, NY: Palgrave Macmillan. MacKinnon, N. J., & Heise, D. R. (1993). Affect control theory: Delineation and development. In J. Berger & M. Zelditch, Jr. (Eds.), Theoretical research programs: Studies in the growth of theory (pp. 64
103). Stanford, CA: Stanford University Press. MacKinnon, N. J., & Heise, D. R. (2010). Self, identity, and social institutions. New York, NY: Palgrave Macmillan.

Affect Control Theory

169

MacKinnon, N. J., & Keating, L. J. (1989). The structure of emotions: Canada-United States comparisons. Social Psychology Quarterly, 52, 70
83. MacKinnon, N. J., & Langford, T. (1994). The meaning of occupational prestige scores. The Sociological Quarterly, 35(2), 215
245. MacKinnon, N. J., & Luke, A. (2002). Changes in identity attitudes as reflections of social and cultural change. Canadian Journal of Sociology/Cahiers canadiens de sociologie, 27, 299
338. McCall, G. J., & Simmons, J. L. (1978). Identities and interactions (2nd ed.). New York, NY: Free Press. Moore, C. D., & Robinson, D. T. (2006). Selective identity preferences: Choosing from among alternative occupational identities. Advances in Group Processes, 23, 253
281. Morgan, G. P., & Morgan, J. H. (2014). Surveyor 3.0: A light-weight data driven approach for capturing affective meanings. Durham, NC: Department of Sociology, Duke University. Morgan, R. L., & Heise, D. (1988). Structure of emotions. Social Psychology Quarterly, 51(1), 19
31. Nelson, S. M. (2006). Redefining a bizarre situation: Relative concept stability in affect control theory. Social Psychology Quarterly, 69, 215
234. Osgood, C. E. (1962). Studies on the generality of affective meaning systems. The American Psychologist, 17(1), 10. Osgood, C. E., May, W. H., & Miron, M. S. (1975). Cross-cultural universals of affective meaning. Urbana, IL: University of Illinois Press. Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement of meaning. Urbana, IL: University of Illinois Press. Perinbanayagam, R. S. (2000). Presence of self. Lanham, MD: Rowman and Littlefield. Powers, W. T. (1973). Behavior: The control of perception. New York, NY: Hawthorne. Rashotte, L. S. (2001). Some effects of demeanor on the meaning of behaviors in context. Current Research in Social Psychology, 6(2002), 92
102. Rashotte, L. S. (2002a). Incorporating nonverbal behaviors into affect control theory. Electronic Journal of Sociology, 6, 3. Rashotte, L. S. (2002b). What does that smile mean? The meaning of nonverbal behaviors in social interaction. Social Psychology Quarterly, 65, 92
102. Rashotte, L. S. (2003). Written versus visual stimuli in the study of impression formation. Social Science Research, 32(2), 278
293. Robinson, D. T. (1996). Identity and friendship: Affective dynamics and network formation. Advances in Group Processes, 13, 91
111. Robinson, D. T. (2007). Affect control theory. In G. Ritzer (Ed.), Blackwell encyclopedia of sociology. Oxford, UK: Blackwell Publishing. Robinson, D. T., Clay-Warner, J., Moore, C. D., Everett, T., Watts, A., Tucker, T. N., & Thai, C. (2012). Toward an unobtrusive measure of emotion during interaction: Thermal imaging techniques. Advances in Group Processes, 29, 225
266. Robinson, D. T., Rogalin, C., & Smith-Lovin, L. (2004). Physiological measures of theoretical concepts: Some ideas for linking deflection and emotion to physical responses during interaction. Advances in Group Processes, 21, 77
115. Robinson, D. T., & Smith-Lovin, L. (1992). Selective interaction as a strategy for identity maintenance: An affect control model. Social Psychology Quarterly, 55(1), 12
28. Robinson, D. T., & Smith-Lovin, L. (1999). Emotion display as a strategy for identity negotiation. Motivation and Emotion, 23(2), 73
104.

170

NEIL J. MACKINNON AND DAWN T. ROBINSON

Robinson, D. T., & Smith-Lovin, L. (2006). Affect control theory. In P. J. Burke (Ed.), Contemporary social psychological theories (pp. 137
164). Stanford, CA: Stanford University Press. Robinson, D. T., Smith-Lovin, L., Doan, L., & Moore, C. D. (2014). How setting constrains friendship: An affect control analysis. Paper presented at SUNBELT XXXIV: The annual meetings of the International Network for Social Network Analysis, St. Pete Beach, FL. Robinson, D. T., Smith-Lovin, L., & Tsoudis, O. (1994). Heinous crime or unfortunate accident? The effects of remorse on responses to mock criminal confessions. Social Forces, 73(1), 175
190. Robinson, D. T., Smith-Lovin, L., & Wisecup, A. K. (2006). Affect control theory. In J. Turner & J. Stets (Eds.), Handbook of the sociology of emotions (pp. 179
202). New York, NY: Springer. Rogers, K. B. (2013). Mapping the social ecology of culture: Social position, connectedness, and influence as predictors of systematic variation in affective meaning. Unpublished Dissertation. Duke University. Durham, NC. Rogers, K. B., Schro¨der, T., & Scholl, W. (2013). The affective structure of stereotype content behavior and emotion in intergroup context. Social Psychology Quarterly, 76(2), 125
150. Rogers, K. B., Schro¨der, T., & von Scheve, C. (2014). Dissecting the sociality of emotion: A multilevel approach. Emotion Review, 6, 124
133. Scher, S. J., & Heise, D. R. (1993). Affect and the perception of injustice (Vol. 10, pp. 223
252). Advances in Group Processes. Schneider, A. (1996). Sexual-erotic emotions in the US in cross-cultural comparison. International Journal of Sociology and Social Policy, 16(9
10), 123
143. Schneider, A. (1999a). The violent character of sexual-eroticism in cross-cultural comparison. International Journal of Sociology and Social Policy, 18, 81
100. Schneider, A. (1999b). Emergent clusters of denotative meaning. Electronic Journal of Sociology, 4(2). Schneider, A. (1999c). US neo-conservatism: Cohort and cross-cultural perspective. International Journal of Sociology and Social Policy, 19(12), 56
86. Schneider, A. (2002a). Probing unknown cultures. Electronic Journal of Sociology, 6(3). Schneider, A. (2002b). Computer simulation of behavior prescriptions in multi-cultural corporations. Organization Studies, 23, 105
131. Schneider, A. (2004). The ideal type of authority in the United States and Germany. Sociological Perspectives, 47(3), 313
327. Schneider, A. (2005). A model of sexual constraint and sexual emancipation. Sociological Perspectives, 48(2), 255
270. Schneider, A. (2007). Politically correct stereotyping: The case of Texans. International Journal of Contemporary Sociology, 44(1), 87. Schneider, A. (2009). The rhythm of the whip. Social Psychology Quarterly, 72(4), 285
289. Schneider, A., & Heise, D. R. (1995). Simulating symbolic interaction. Journal of Mathematical Sociology, 20, 271
287. Schneider, A., & Schro¨der, T. (2012). Ideal types of leadership as patterns of affective meaning a cross-cultural and over-time perspective. Social Psychology Quarterly, 75(3), 268
287.

Affect Control Theory

171

Scholl, W. (2013). The socio-emotional basis of human interaction and communication: How we construct our social world. Social Science Information, 52(1), 3
33. Schro¨der, T. (2008). Mean affective ratings of 1,100 concepts by a diverse sample of Germans in 2007 [computer file]. Distributed at Affect Control Theory Website. Retrieved from http://www.indiana.edu/∼socpsy/ACT/Interact/JavaInteract.html Schro¨der, T. (2011). A model of language-based impression formation and attribution among Germans. Journal of Language and Social Psychology, 30, 82
102. Schro¨der, T., Netzel, J., Schermuly, C. C., & Scholl, W. (2013). Culture-constrained affective consistency of interpersonal behavior. Social Psychology, 44(1), 47
58. Schro¨der, T., Rogers, K. B., Ike, S., Mell, J. N., & Scholl, W. (2013). Affective meanings of stereotyped social groups in cross-cultural comparison. Group Processes & Intergroup Relations, 16(6), 717
733. Schro¨der, T., & Scholl, W. (2009). Affective dynamics of leadership: An experimental test of affect control theory. Social Psychology Quarterly, 72(2), 180
197. Schro¨der, T., Stewart, T. C., & Thagard, P. (2013). Intention, emotion, and action: A neural theory based on semantic pointers. Cognitive Science. Published online before print. doi:10.1111/cogs.12100 Schro¨der, T., & Thagard, P. (2013). The affective meanings of automatic social behaviors: Three mechanisms that explain priming. Psychological Review, 120(1), 255
280. Sewell, A. A., & Heise, D. R. (2010). Racial differences in sentiments: Exploring variant cultures. International Journal of Intercultural Relations, 34(4), 400
412. Shank, D. B. (2010). An affect control theory of technology. Current Research in Social Psychology, 15(10), 1
13. Smith, H., & Schneider, A. (2009). Critiquing models of emotions. Sociological Methods & Research, 37(4), 560
589. Smith, H. W. (1995). Predicting stress in American-Japanese business relations. Journal of Asian Business, 12, 79
89. Smith, H. W. (2002). The dynamics of Japanese and American interpersonal events: Behavioral settings versus personality traits. Journal of Mathematical Sociology, 26(1
2), 71
92. Smith, H. W., & Francis, L. E. (2005). Social vs. self-directed events among Japanese and Americans. Social Forces, 84(2), 821
830. Smith, H. W., Ike, S., & Li, Y. (2002). The pre-surveyor experience. Electronic Journal of Sociology. Retrieved from http://sociology.org/content/vol006.003/smith_ etal.html Smith, H. W., Matsuno, T., & Ike, S. (2001). The affective basis of attributional processes among Japanese and Americans. Social Psychology Quarterly, 64, 180
194. Smith, H. W., Matsuno, T., & Umino, M. (1994). How similar are impression change processes among Japanese and Americans? Social Psychology Quarterly, 57, 124
139. Smith, H. W., Umino, M., & Matsuno, T. (1998). The formation of gender-differentiated sentiments in Japan. Journal of Mathematical Sociology, 22(4), 373
395. Smith, H. W., & Yap, M. (2006). Guilty Americans and shameful Japanese? An affect control test of Benedict’s thesis. In Purpose, meaning and action: Control systems theories in sociology (pp. 213
236). New York, NY: Palgrave Macmillan. Smith-Lovin, L. (1979). Behavior settings and impressions formed from social scenarios. Social Psychology Quarterly, 42(1), 31
43.

172

NEIL J. MACKINNON AND DAWN T. ROBINSON

Smith-Lovin, L. (1987a). Impressions from events. Journal of Mathematical Sociology, 13(1
2), 35
70. Smith-Lovin, L. (1987b). The affective control of events within settings. Journal of Mathematical Sociology, 13(1
2), 71
101. Smith-Lovin, L. (1990). Emotion as the confirmation and disconfirmation of identity: An affect control model. In T. D. Kemper (Ed.), Research agendas in the sociology of emotions (pp. 238
270). New York, NY: SUNY Press. Smith-Lovin, L. (1993). Can emotionality and rationality be reconciled? A comment on Collins, Frank, Hirshleifer, and Jasso. Rationality and Society, 5(2), 283
293. Smith-Lovin, L. (2002). Roles, identities, and emotions: Parallel processing and the production of mixed emotions. In Y. Kashima, M. Foddy, & M. Platow (Eds.), Self and identity: Personal, social, symbolic (pp. 125
143). New York, NY: Lawrence Erlbaum Associates. Smith-Lovin, L. (2003). Self, identity, and interaction in an ecology of identities. In P. J. Burke, T. J. Owens, P. A. Thoits, & R. T. Serpe (Eds.), Advances in identity theory and research (pp. 167
178). New York, NY: Plenum. Smith-Lovin, L., & Douglass, W. (1992). An affect control analysis of two religious subcultures. In V. Gecas & D. Franks (Eds.), Social perspectives on emotions (Vol. 1, pp. 217
248). Greenwich, CT: JAI Press. Smith-Lovin, L., & Heise, D. R. (1982). A structural equation model of impression formation. In N. Hirschberg & L. Humphreys (Eds.), Multivariate applications in the social sciences (pp. 195
222). Hillsdale, NJ: Lawrence Erlbaum. Smith-Lovin, L., & Heise, D. R. (Eds.). (1988). Analyzing social events: Advances in affect control theory. New York, NY: Gordon and Breach Scientific. Smith-Lovin, L., & Robinson, D. T. (1992). Gender and conversational dynamics. In C. Ridgeway (Ed.), Gender, interaction, and inequality (pp. 122
156). New York, NY: Springer-Verlag. Smith-Lovin, L., & Robinson, D. T. (2006). Control theories of identity, action and emotion: In search of testable differences between affect control theory and identity control theory. In K. McClelland & T. Fararo (Eds.), Purpose, meaning and action: Control systems theories in sociology (pp. 163
188). New York, NY: MacMillan. Strodtbeck, F. L., James, R. M., & Hawkins, C. (1957). Social status in jury deliberations. American Sociological Review, 22, 713
719. 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. (1980/2000). Symbolic interactionism: A social structural version. Menlo Park, CA: Benjamin-Cummings/Blackburn Press. Tajfel, H. (1969). Cognitive aspects of prejudice. Journal of Social Issues, 25(4), 79
97. Tajfel, H. (1970). Experiments in intergroup discrimination. Scientific American, 223(5), 96
102. Tajfel, H. (1981). Human groups and social categories: Studies in social psychology. New York, NY: Cambridge University Press. Thagard, P., & Schro¨der, T. (Forthcoming). Emotions as semantic pointers: Constructive neural mechanisms. In L. F. Barrett & J. A. Russell (Eds.), The psychological construction of emotions. New York, NY: Guilford. Thomassen, L. (2002). An alcoholic is good and sober: Sentiment change in AA. Deviant Behavior: An Interdisciplinary Journal, 23, 177
200.

Affect Control Theory

173

Troyer, L. (2004). Affect control theory as a foundation for the design of socially intelligent systems. Paper presented at the Proceedings of 2004 American Association for Artificial Intelligence Symposium on Architectures for Modeling Emotion: Cross Disciplinary Foundations. Palo Alto, CA. Troyer, L., & Robinson, D. T. (2006). Contributions of a sociological perspective on affect to the study of political action. In D. P. Redlawsk (Ed.), Feeling politics: Emotion in political information processing (pp. 47
56). New York: Palgrave Macmillan. Tsoudis, O. (2000). The likelihood of victim restitution in mock cases: Are the ‘rules of the game’ different from prison and probation? Social Behavior and Personality, 28(5), 481
498. Tsoudis, O., & Smith-Lovin, L. (1998). How bad was it? The effects of victim and perpetrator emotion on responses to criminal court vignettes. Social Forces, 77(2), 695
722. Tsoudis, O., & Smith-Lovin, L. (2001). Criminal identity: The key to situational construals in mock criminal court cases. Sociological Spectrum, 21(1), 3
31. Turner, J. C. (1985). Social categorization and the self-concept: A social cognitive theory of group behavior. Advances in Group Processes: Theory and Research, 2, 77
122. Turner, J. C., Hogg, M. A., Oakes, P. J., Reicher, S. D., & Wetherell, M. S. (1987). Rediscovering the social group: A self-categorization theory. New York: Basil Blackwell. Turner, J. C., Oakes, P. J., Haslam, S. A., & McGarty, C. (1994). Self and collective: Cognition and social context. Personality and Social Psychology Bulletin, 20, 454
463. Wang, J. (2002). Cross-cultural friendships. Paper presented at the Research Agendas in Affect Control Theory, Highland Beach, FL. Wiggins, B., & Heise, D. R. (1987). Expectations, intentions, and behavior: Some tests of affect control theory. The Journal of Mathematical Sociology, 13(1
2), 153
169. Wiley, N. (1994). The semiotic self. Chicago, IL: University of Chicago Press. Wisecup, A. K. (2011). Do we have consensus? Examining the sources of systematic variation in cultural identity meanings. Unpublished Dissertation. Duke University, Durham, NC. Youngreen, R. (2002). Emergent friendship networks in the classroom: A simulation employing affect control theory. Paper presented at the Research Agendas in Affect Control Theory, Highland, FL. Youngreen, R., Conlan, B., Robinson, D. T., & Lovaglia, M. J. (2009). Identity maintenance and cognitive test performance. Social Science Research, 38, 438
446.

ELEMENTARY THEORY: 25 YEARS OF EXPANDING SCOPE AND INCREASING PRECISION$ David Willer, Pamela Emanuelson, Michael J. Lovaglia, Brent Simpson, Shane R. Thye, Henry Walker, Mamadi Corra, Steven Gilham, Danielle Lewis, Travis Patton, Yamilette Chacon and Richard Chacon ABSTRACT Purpose
This exposition explains how Elementary Theory works and how it has been developed over the last two-and-a-half decades. Both increased scope and heightened precision are covered. Methodology/approach
Theoretic methodology is explained. Using that method formal models are constructed analogous to empirical events. Those models predict events, design experiments, and guide applications in the field. $

Authors are ordered first by number of sections contributed and then alphabetically. The first author is solely responsible for errors.

Advances in Group Processes, Volume 31, 175
217 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0882-6145/doi:10.1108/S0882-614520140000031005

175

176

DAVID WILLER ET AL.

Findings
There is a widely held belief in sociology that theory becomes more vague and imprecise as its scope broadens. Whereas broader generalizations are more vague than narrower ones, this exposition shows that abstract theory becomes more precise as its scope broadens. Research limitations/implications
Here implications and limitations are closely connected. Regarding implications, this exposition shows that scientific explanations and predictions are viable today in sociology but only when exact theory is employed. Regarding limitations, the theory and research included in this exposition make clear why the empiricist search for regularities that dominates sociological research is so very limited in its results. Originality/value of chapter
This exposition demonstrates that theory is the method of all the sciences and in particular the science of sociology. Keywords: Social theory; social structure; power; influence; coercion; exchange

In 1984, four years before the first Group Process Meeting, the first author of this chapter asked whether social theory could take a methodological form like the physical theories of Galileo, Newton, and Einstein. Could a social theory begin with a few irreducible elements, as Einstein put it, which are “as simple and few in number as possible”? (Einstein, [1933] 1954, p. 272). Could a social theory develop procedures to combine those elements into more and more complex models for social structures? Could social actors be placed in those modeled structures such that they would be, not static, but dynamic and thus predictive? If all of the foregoing is possible, could the models of that theory be used as blueprints to design experiments to test the theory? (D. Willer, 1984).1 Those who have followed the development of Elementary Theory know that the answer to all the foregoing questions is “yes.” Yes, a few elementary parts can be built into dynamic models for social structures that are blueprints for experiments. Thirty years of building and running those experiments has provided unprecedented support for Elementary Theory. The support is unprecedented in three regards: (1) the broad scope and wide variety of phenomenon covered by a single theory, (2) the number of contrasting models subjected to experimental test, and (3) the exactness of

Elementary Theory: 25 Years of Expanding Scope

177

prediction across that broad range. Yes, a theory of this methodological form does work as well in sociology as it works in physics. This form of theory also undergirds chemistry and biology. Any who harbor doubts regarding the latter need only reflect that Watson and Crick’s breakthrough came, not through generalization from research, but upon building a theoretic model for the double helix (Watson, 1968). Conventional wisdom suggests, to the contrary, that not the physical and biological sciences, but economics should be emulated, a suggestion supported by the plethora of economists who have commented on sociology (e.g., Fehr & Gintis, 2007). There are good reasons to reject economics as a model for our science. Economics theorizes only one social structure, the market, and only in one form, perfect competition (Samuelson, 1983). In the eighth chapter of Network Exchange Theory, more than 25 structures were modeled and predictions for them were experimentally supported (D. Willer, 1999). Economists experimentally study a wide variety of games, but in none of their studies have they shown that payoff matrices like those of their games actually exist outside their laboratories (Kagel & Roth, 1995). Later in this chapter, Strategic Analysis, Elementary Theory’s procedure for finding games in structures will be explained. Microeconomics and macroeconomics are wholly distinct: their micro
macro problem has been unresolved for well over a century. As also shown later, Elementary Theory has no micro
macro problem. Its experimental social structures are small, but its most important applications outside the lab are to very large social structures (D. Willer & Walker, 2007). Nevertheless, Elementary Theory shares this with economics: both use a rational actor model. But the rational actor models of the two are quite different. Regardless of context, economic man is a commodity maximizer, nothing more (Olson, 1965; Samuelson, 1983). By contrast, the rational actor of Elementary Theory acquires its values from the social structures in which it acts. In Elementary Theory, the social actor who is proself at the office will be prosocial in the household when those are the social values of the two social structures (D. Willer & Anderson, 1981). Furthermore, Elementary Theory’s actor is driven, not by one interest, but by two. Like economic man, it seeks its best outcome, but, in addition, avoids its worse outcome (D. Willer, 1984). The two interests are not the same and, as will be clear in the next section, it is the latter that is central to explaining power in structures. The focus of Elementary Theory is on explaining power in social structures, a focus it shares with the classics of Marx, Weber, and Simmel. But Marx was a nineteenth century thinker while Weber and Simmel

178

DAVID WILLER ET AL.

bridged the nineteenth and twentieth centuries. Most of their work is at least 100 years old. Isn’t their focus outmoded today? Perhaps even slightly quaint? After all, economics does without any conception of power and without formulations for influence, exploitation, domination, and legitimacy as well. Were Marx, Weber, and Simmel wrong to focus on structures of power? To the contrary, here is a great irony of the intellectual history of our field. As sociology in the United States has come to have a leading place among the sociologies of the world, theories of social power have almost disappeared. The irony is that, during that same time period, the United States rose to be the most powerful nation the world has ever seen. Though it is inside the paramount national power in the world, mainstream sociology today understands less about contemporary power structures than did Marx, Weber, and Simmel about the structures in which they lived. Perhaps it is well that the direction of development of Elementary Theory is directly opposed to that of mainstream sociology. The purpose of this chapter is to explain the development of Elementary Theory, and, in so doing, offer to the reader a toolbox of derivations to use in historical and contemporary explanation. As space allows, it will show how that theory offers insights into the working of both macro- and microstructures now and in the past. The chapter is organized in the following way. The section immediately to follow explains Elementary Theory’s core insight of how social structures produce power differences. The reader will find that, as anticipated by Popper, theory explaining how social structures produce power is quite simple. (See note 1.) The section after that briefly reviews the basic formulations of Elementary Theory as they were developed and tested just prior to the 1988 Group Process Meeting. Unlike some theories of exchange, the basic conceptions of Elementary Theory remain unchanged up to today. The body of the chapter is organized temporally, tracing the scope extensions and applications of the theory as well as the bridges built and tested between Elementary Theory and other formal theories. This work will be covered in the order of its development.

THE SECRET OF POWER IN STRUCTURES Here is the secret of how power is exercised in social structures. Power structures make those high in power obstinate and those low in power

179

Elementary Theory: 25 Years of Expanding Scope

obedient. How are those attitudes produced? Power structures connect social relations in such a way that high-power actors face little or no loss upon disagreeing with those low in power, while, in the same process, the opportunities for better agreements for those low in power shrink to nothing. The pairing of obstinacy with obedience in exchange structures results in exchange ratios favoring the obstinate high-power actor over the obedient low-power actor. Similarly, the same pairing in coercive structures results in heightened levels of coercive exploitation favoring obstinate coercer over obedient coercee.

Power in Exchange Structures Exclusionary power in exchange structures is explained by reference to Fig. 1(a) where social actors labeled A and B are connected by paired positive sanctions representing exchange. To simplify this explanation, the payoffs at agreement of the two actors in each of the two exchange relations sum to 10. For example, when A agrees with B1 gaining 6, B1 gains 4; or when A agrees with B2 gaining 3, B2 gains 7, and similarly for all agreements. In the absence of agreement
when in “confrontation”
the payoff is zero. When A can exchange with B1 or with B2, but not both, Fig. 1(a) network is an exclusionary power structure, so called because one B is always excluded from exchanging. It is the simplest exclusionary structure. Larger more complicated structures are also exclusionary power if those high in power can exclude those low in power from exchange and some are actually excluded. Those structures are called strong power because their agreements occur at the extreme favoring high-power actor(s).2 The power process works through each actor’s two interests. Each has an interest in gaining the best payoff, and an interest in avoiding the worst payoff. In Fig. 1(a), initially A’s worst payoff is zero. As soon as an exchange is negotiated with either B1 or B2, however, A’s worst payoff in (a)

+

B1 +

Fig. 1.

A

+

+

(b)

B2

+

D1 –

C

+

–

D2

Exchange and Coercive Structures.

180

DAVID WILLER ET AL.

one relation is the payoff from exchanging in the other. If the first agreement is with B1, a rational A will accept no offer worse than or equal to B1’s from B2. Thus is A obstinate. Initially, B1 or B2’s best payoff is nearly all of the value in the relation, leaving to A the minimum that will move A to exchange. As soon as an exchange is negotiated between A and either B, however, the best payoff in the other relation is slightly smaller to that B than the payoff from agreement in the first. For example, if the first agreement is with B2, B1’s best hope declines to just less than B2’s payoff in exchange with A. Thus do the Bs become obedient to A’s demands. Once begun, the power process of all strong power structures continues step-by-step to the end point of the “negotiation set.” At that end point, the payoff to those low in power who exchange is minimal, while the payoff to those low in power who are excluded from exchange is zero. By contrast, the payoff to the high-power actor(s) is the maximum possible for the relations of the structure. This process is produced by the conjunction of exclusion with exchange. When there is no exclusion, when the A can exchange with both B1 and B2, all exchanges are at equal power exactly as they would be in isolated dyads (Brennan, 1981).

Power in Coercive Structures We will now show that structural power in coercion works exactly as it does in exchange
with this one important caveat. When unaffected by structure, exchange relations are equipower, but in unaffected coercive relations, the coercer exercises power over the coercee. Why are coercive relations counted as power relations? The answer lies in this: only then can the term “power” be consistently applied across the two kinds of relations. For example, when unaffected by any power structure, the A
B exchange relations introduced earlier will have payoffs of 5 to A and 5 to B; that payoff distribution is counted as equipower. Now introduce coercion by giving A one negative sanction that, when sent, costs B 10. Then, the payoffs will be 7.5 to A and 2.5 to B, clearly a power exercise. Therefore, the isolated coercive relation is a power relation. Nevertheless, there is also coercive structural power. Its effect is to increase power exercise beyond that of the coercive dyad. Here is how coercive structural power works. In Fig. 1(b), two C
D coercive relations are connected at C: they are represented by negative sanctions from C to the Ds and positive sanctions from Ds to C. Initially,

Elementary Theory: 25 Years of Expanding Scope

181

each D holds 10 resources valuable to both D and C, while C holds one negative sanction worth nothing to C: when sent the negative wipes out the value of resources initially held by D. Let C always send one negative sanction to one of the two Ds. Then C can receive valued resources from the other D. As in exchange, the power process works through the actors’ two interests
the interest in gaining the best outcome and the interest in avoiding the worst outcome. Initially, C’s worst outcome is a payoff of zero. As soon as a possible agreement is reached with either D1 or D2, however, C’s worst outcome in the other relation is the payoff from the agreement in the first. If the first agreement is with D1, C will accept no offer worse than or equal to D1’s from D2. Thus is C obstinate. Initially, D1 or D2’s best payoff is to keep nearly all of its resources, sending to C the minimum to avoid C’s negative sanction. As soon as an agreement between C and either D is reached, however, the best payoff possible in the other relation is just slightly smaller to that D than the payoff from agreement in the first. For example, if the first agreement is with D2, D1’s best hope declines to just less than D2’s payoff in its relation with C. Thus do the Ds become obedient to C’s demands. As in the exchange structure, in the coercive structure, the power process continues step-by-step to the end point of the “negotiation set” where the low-power D gains the minimum, the high-power C the maximum, and the other D, having received the negative sanction, has a payoff of zero. As in exchange, this process is not produced because the network branches from C to two Ds. When C can gain resources from both D1 and D2, both Ds will send five resources to C, just as they would in an isolated coercive relation. In fact, strong coercive structures can have any number of coercers and coercees as long as (1) the coercers jointly send negatives to a “surplus” of coercees, (2) the negatives are received by coercees offering the fewest resources, and (3) coercees cannot escape, or act collectively to oppose the power exercise (D. Willer, 1987).

Power Structures as the Secret of Primitive Accumulation We know how exchange and coercive structures behave because they have been studied under controlled conditions of the laboratory. But does this understanding have explanatory power outside the limits of the lab? Here, theory just discussed is applied to Marx’s Secret of Primitive Accumulation ([1867] 1967), of how capital in England was first accumulated through exploitation of the emerging working class. As Marx explained, there never

182

DAVID WILLER ET AL.

was a labor market free of external regulation. The two kinds of external regulation define the two forms of primitive accumulation. The first form was instituted under the Elizabethan labor laws: they imposed a strong coercive structure on top of the market for labor. The unemployed when first discovered by the state were branded. When second discovered lost an ear. When third detected were either hung in chains or sent indentured to the new world (Marx, [1867] 1967). This coercive system was not expensive to administer for two reasons. Obviating bureaucratic paperwork, its victims carried on their persons the evidence of their past. Second, since England is an island, escape from this hellish system was effectively impossible. The first form of primitive accumulation, having driven wages to subsistence and below, was a great success for the emergent capitalist class. But it had one fault. It killed off the working class almost as rapidly as it later used up coal in the boilers of its steam engines. As a result, the first form of primitive accumulation could continue only as long as the enclosure movement in the countryside produced a large flow of new laborers to the towns and cities. When that flow slowed, the second form was instituted. The second form of primitive accumulation was “welfare.” Those who were unemployed, instead of being branded and transported, were now fed and housed so they would not die too rapidly. Thus did welfare assure the capitalist class a surplus of unemployed labor with which to threaten the employed (Marx, [1867] 1967). This form of primitive accumulation is a strong exchange structure. In the transition to the second form, wages increased to the point that some of the working class could reproduce itself while the now reduced flow of people from countryside to town kept the labor market overfull. At the same time, wages did not rise further due to laws forbidding the organization of the working class either in one factory or class wide. Now the impact of coercion was not directly upon wages, but indirect. Workers seeking to act collectively were coerced. With collective action blocked, there was no opportunity to countervail power exercised through the strong exchange structure. Those acquainted with the economic history of early modern England will recognize that labor laws governing the labor market drove down wages and furthered capital accumulation. How they did so is now covered by experimentally tested derivations from Elementary Theory. As that application shows, experimental research can be used to undergird historical explanation. At issue is not generalization from the lab to history for nothing said here supports such a chancy, even futile, enterprise. To the contrary, the bridge between the laboratory and history is theory, and

183

Elementary Theory: 25 Years of Expanding Scope

always theory that applies in both settings and thus connects the two. Further historical explanations are given later.3

BASIC CONCEPTS Having shown how Elementary Theory works, this section briefly reviews its basic concepts and how they are built into the formulations like those just discussed. Following Einstein, we begin with the theory’s simplest elements and move to the principles and laws that govern actors who, through their decisions and actions, produce the dynamics of the models.4 In Elementary Theory, “sanctions” are the simple elements out of which actors, relations, and structures are composed. A sanction is a social act sent by one actor and received by a second, which has either a positive payoff or negative payoff to the receiver. Only the sender decides whether or not to send the sanction. Both arcs in Fig. 2 are sanctions, while the nodes A and B are actors. All actors have preference systems. For example, in the figure, B prefers to receive the positive sanction to receiving no sanction and prefers to receive no sanction to receiving the negative sanction.5 Sanctions can be given numerical values that may be “objective” payoffs as in $10 or “subjective” payoffs as in 10 utility units. In its calculations, Elementary Theory does not compare payoffs between actors. Therefore, any scales can be used as long as scale values are all commensurate within each actor. Since the theory does not make interpersonal utility comparisons, it matters not whether the payoffs of any pair of actors are expressed on the same scale or different scales. Positive and negative sanctions are not rewards and punishments. A sanction is an elementary act defined only by its direction of preference effect. But a reward could be either of two very different acts as could a punishment. For example, you can reward me by giving me a gift or by ending the beating you are giving me. That is to say, a reward is either starting a positive act or stopping a negative act. Similarly, you can punish + A

B Positive Sanction

Fig. 2.

– A

B Negative Sanction

Types of Sanctions.

184

DAVID WILLER ET AL.

me by drawing a pistol and shooting me or by stopping my pay. That is to say, a punishment is either starting a negative act or ending a positive act. Terms like “reward” and “punishment” are not elementary ideas: they are compounds that group together very different acts and should be avoided. Elementary Theory allows the theorist to model a social structure together with actors’ value systems and beliefs with a single construction. It does so by assuming that actor’s values and beliefs reflect the structure in which they act, an idea borrowed from Marx. This is not orthodoxy: it is parsimony. At the same time, as influenced by Weber ([1918] 1968), when needed, more complex models can be built where actors carry values from one structure to impose them on another, as he asserted rational ascetics did.6 When the values and beliefs of all actors in a modeled structure reflect the structure, the model is substantially simpler than when any actor’s values and beliefs are not reflective. For example, when any actor is misinformed, the actual structure and the believed structure must both be modeled resulting in a model twice as complex as the one where all actors’ beliefs are reflective. A model with three misinformed actors will be at least three times as complex and similarly as reflectivity of beliefs declines. Striving for parsimony, it is best to begin model constructions with actors having complete and accurate beliefs because, once a model for the structure is built, actors’ values and beliefs have also been specified. The first decision rule of Elementary Theory is its first principle: P1. All social actors act to maximize their expected payoffs. P1 says that actors are rational, but obviously they are not like the rational actors of economic theory. Since actors acquire their values from the structure in which they act, P1 does not assert that all actors are selfish profit seekers. To the contrary, it asserts that actors seek to maximize in light of the values given by the structure in which they act. Furthermore, rationality can take two forms. Parametric rationality optimizes among alternatives fixed by the environment (Elster, 1986, p. 7). When prices are given by the market, economic actors do not interact with each other, they select things from the market. The economists use parametric rationality, a use that is justified by appeal to parsimony. Because the actors of Elementary Theory interact, however, they are not parametric unless their information is highly restricted. As defined by Schelling, rationality is strategic when “each player’s best choice of action depends on the action he expects the other to take” (1970, p. 86). Strategic rationality is the default assumption of Elementary Theory and necessarily so because, in deciding to act, actors in social relationships

185

Elementary Theory: 25 Years of Expanding Scope

take into account what they expect the other will do. Strategic rationality is also the default assumption of game theory. In fact, von Neumann and Morgenstern (1944) introduced game theory with the strategic
parametric distinction. But more than 25 years earlier, Weber distinguished the two types of rationality and used the distinction to differentiate economics with its parametric rationality from sociology with its strategic rationality ([1918] 1968, pp. 22
23).

Quantity and Resistance We now show how to find the range of exchange ratios that can occur in a relation by application of P1. Then having introduced the second principle and the resistance equation, we show how to predict the exchange ratio at agreement. Turning to coercion, we show that the rate of coercive exploitation is predicted in exactly the same way with exactly the same tools. Though exchange and coercion differ fundamentally, because both are social relations, exactly the same procedures are applied to both. These calculations are for isolated relations, relations that are not embedded in structures. In the section to follow, new applications of resistance calculate how structures affect power in exchange and coercive structures. For the exchange relation of Fig. 3 or any other mixed motive social relation, the condition of “confrontation” occurs when actors cannot agree.

X

X

A

B

1

9

1

2

8

2

3

7

3

.

.

.

.

.

.

9

1

9

+1

–1

B

A

+10

0

Fig. 3.

An Exchange Relation.

186

DAVID WILLER ET AL.

When exchange is in confrontation, since there is no agreement, no sanctions flow. Therefore, the payoffs to A and B are, respectively, PAcon = 0 and PBcon = 0. If an exchange occurs, B will send its sanction. Therefore, payoffs to A and B will vary only with X, the amount sent by A. Assume that X is lumpy in units of 1: there can be a flow of 7 sanctions from A to B or of 6 sanctions, but not of any value between such as 6.4. To the right is the payoff matrix that maps payoffs from exchange to A and B as determined by the valuations given in the relation. Importantly, the range of those payoffs is limited by Pcon. No actor will agree to an exchange where the payoff is zero as it is at confrontation. Note that PAmax = 9 when A receives B’s sanction and sends only X = 1, and PBmax = 9 when B sends its sanction and A sends X = 9. These are the initial conditions of the relation from which the exchange ratio of the relation will be predicted. Resistance equations are laws of the theory. They take many forms depending on the position of the actor, the relation, and the structure. Resistance takes its simplest form in the dyad where A’s resistance is RA =

PA max − PA PA − PA con

ð1Þ

Given the values for confrontation and agreement, the resistance factor to the right in Eq. (1) gives the strength of A’s two interests: in the numerator is A’s interest in gaining a better payoff and in the denominator is A’s interest in avoiding confrontation. To determine the point of agreement between A and B, compare the working of those two interests between the actors by invoking the second principle. P2. Agreements occur at equal resistance for undifferentiated actors in a full information system. The principle asserts that actors agree when the benefit of payoffs at agreement and the cost of avoiding confrontation are mutually balanced. Therefore, setting RA = RB, RA =

PA max − PA PB max − PB = = RB PA − PA con PB − PB con

Plugging in values for Pmax and Pcon, 9 − P A 9 − PB = PA − 0 P B − 0

ð2Þ

187

Elementary Theory: 25 Years of Expanding Scope

As can be inferred from Fig. 3’s payoff matrix, PA = 10 − X

and PB = X

Plugging in 9 − ð10 − XÞ 9 − X = ð10 − XÞ − 0 X − 0 So, X = 5, PA = 5, and PB = 5. B sends its sanction while A sends X = 5 sanctions in return. That the two actors each gain 5 is an intuitive result
not a bad thing. The result of applying resistance is not always intuitive, however. In the coercive relation of Fig. 4, resistance is applied exactly as it was above. As seen in that figure, C has a negative sanction that is costless to transmit: the effect of the negative sanction is
10 on D. C’s negative sanction is threatened and will be sent only at confrontation. D holds positive sanctions, each of which is a loss of 1 to D and a gain of 1 to C. Unknown is Z, the number of sanctions transmitted by D. These values are displayed on the coercive relation: together with P1 they give the payoff matrix to the right in Fig. 4. That payoff matrix, like the one in Fig. 3, is the negotiation set of the relation. As in the case of X, for simplicity we assume that Z values are lumpy in units of one. When Z = 1, PD = PDmax = −1. When Z = 9, PC = PCmax = 9. Pcon values are also highly unequal: PCcon = 0

0

–10

C

D

Z

Fig. 4.

Z

C

D

1

1

–1

2

2

–2

3

3

–3

.

.

.

.

.

.

9

9

–9

A Coercive Relation.

188

DAVID WILLER ET AL.

while PDcon = −10. From the payoffs of the negotiation set, we can infer that PC = Z and PD = − Z Plugging these values into Eq. (2), 9−Z − 1 − ð − ZÞ = Z −0 − Z − ð − 10Þ Simplifying, 9−Z Z −1 = Z 10 − Z So, Z = 5. Thus, PC = 5 and PD = −5. In coercive relations, only positive sanctions flow at agreement: the negative sanction is transmitted only at confrontation. As already mentioned, all coercive relations are power relations, even ones not embedded in structures. At agreement, C, the coercer, is exercising power over D, the coercee. The power events are the positive sanctions transmitted by D to C. Confrontation is not a power event: it is fulfilling a threat.

The Mathematics of Power in Structures The amount of power exercised in structures is found by an iterative solution that parallels the negotiation processes of the structure. The solution begins with a high-power and a low-power actor negotiating to a potential agreement as if they were in a dyad
as just calculated. With that potential agreement in hand, the high-power actor turns to a second low-power actor and negotiates in light of the outcome of the first negotiation. As will now be seen for exchange, the initial conditions for the first negotiation are the initial conditions of the exchange dyad while the initial conditions of the second negotiation are altered by the outcome of the first negotiation
and similarly through further iterations to the equilibrium solution for the structure. The running example will be the Fig. 1(a) structure as composed of two Fig. 3 exchange relations. Therefore, initially for A and B, as in the dyad, PA max = 9; PA con = 0

and

PB max = 9; PB con = 0

Elementary Theory: 25 Years of Expanding Scope

189

Therefore, RA =

9 − ð10 − XÞ 9 − X = = RB ð10 − XÞ − 0 X − 0

and X = 5, PA = 5, and PB = 5 as above. But now, prior to exchanging, A turns to negotiate with the other B. Initial conditions have changed: PA max = 9; PA con = 5

and

PB max = 4; PB con = 0

Therefore, RA =

X−1 4−X = = RB 10 − X − 5 X

And solving, X = 2.5, PA = 7.5, and PC = 2.5. Why this result? When A turns to the second B, it knows that, as we do, were that negotiation to fail, it can exchange with the first B for PA = 5. Therefore, PAcon = 5, which increases A’s power. To attract A, the second B must make an offer better for A than the offer of the other B. Therefore, PCmax = 4. With those changes in initial conditions, resistance predicts exchange ratio and payoffs that are substantially better for A. This was the first step of the power process. For the second step of the power process, A turns back to the first B, but now the initial conditions have changed to PA max = 9; PA con = 7:5

and

PB max = 2:5; PB con = 0

Plugging in and solving, X = 1.25, PA = 8.75 and PB = 1.25. Now PA values are very close to PAmax. The power dynamics of all strong exclusionary exchange structures are very similar. The running example up to now was the two branch, but exactly the same results will be seen in any strong power structure. That is to say, in any structure in which one set of positions, the i’s are never excluded, and some of the second set of positions, the j’s are always excluded
and in which all i’s are connected to all j’s but no i or j is connected to the other. When t is the current iteration and t − 1 the previous iteration, the general equation for the equal resistance solution in exclusionary structures is Ri =

Pi max − Pti Pjt − 1 − Ptj = = Rj Ptj Pti − Pit − 1

ð3Þ

190

DAVID WILLER ET AL.

Equation (3) introduces two new terms, Pit − 1 and Pjt − 1 . Both refer to payoffs in the previous round. For the resistance of i, the high-power actor, the payoff at confrontation is increasing over rounds. Thus, the denominator Pti − Pit − 1 reads the payoff to i in the current round less the (smaller) payoff in the previous round. For the resistance of the low power j, Pmax is declining. Thus, the numerator Pjt − 1 − Ptj reads the (larger) payoff to j in the previous round less the payoff to j in the current round. Said somewhat differently, as the process goes forward, Pit − 1 increases and Pjt − 1 decreases. Strong power in coercive structures also develops iteratively. As will be remembered, the initial conditions of the coercive dyad are PC max = 9; PC con = 0 and

PD max = − 1; PD con = − 10

As above, resistance predicts PC = 5 and PD = −5. Let those be the values of a tentative agreement between C and D1 in the Fig. 1(b) coercive structure. Now C turns to D2 where initial conditions for that negotiation have become PC max = 9; PC con = 5 and

PD max = − 6; PD con = − 10

Plugging these values into Eq. (2), 9−Z − 6 − ð − ZÞ = Z −5 − Z − ð − 10Þ So, Z = 7.5, PC = 7.5, and PD = −7.5. Clearly, power develops in the strong coercive structure much as it does in the strong exchange structure. As in that structure here, the end point of the process is for the high-power actor’s payoff to approach the maximum possible for the relation PC ≈ PCmax = 9. Certainly, exchange and coercion are not the same, but strong power is strong power whether the structure is composed of exchange relations or coercive relations. As shown elsewhere, strong power is also the same in conflict structures (D. Willer, Simpson, Szmatka, & Mazur, 1996).

INCLUSION AND POWER A position is inclusively connected when more than one of its relations must be completed to benefit from any one. Inclusive connections pervade all human societies. All divisions of labor produce inclusion. For example, if

Elementary Theory: 25 Years of Expanding Scope

191

any one of Adam Smith’s four pinmakers stopped work, unless replaced, no pins would be made and the capitalist, for whom the pins were produced, could not profit. Threshold effects also produce inclusion. If four people are needed to move a rock, the efforts of three or fewer will come to naught. The scope of Elementary Theory was expanded by recognizing “inclusive” connections as a second basis for structural power in exchange structures. In the simplest inclusively connected exchange network, a centrally placed A actor must complete exchanges with all peripheral actors in order to benefit. When NA is the number of relations connected to A and A needs to exchange QA times, then when NA = QA, A is inclusively connected. For inclusion, power is at the periphery. Consider the B
A
C inclusively connected two-branch structure. Let A’s first exchange be with B for a possible payoff of PAb > 0. Being inclusively connected, if A’s second exchange is not completed, PAb, the payoff from the first exchange is lost to A. Therefore, PAccon = −PAb. Hence, the effect of inclusion is to reduce the resistance of the central actor while the resistance of peripheral actors remains unchanged. More generally, when N = Q > 2, the exchange ratio for each subsequent exchange is successively less favorable to A than previous exchanges. Applying resistance in a three-branch structure, the first two exchanges are like the first and second of the two-branch structure, while for the third exchange, PAdcon = −(PAb + PAc). More generally, for an A inclusively connected to N = Q peripherals labeled B1, B2, … Bi, for the ith exchange PAb ðIÞ con = − ðPAb1 þ PAb2 þ ⋯ þ PAbi Þ Then the resistance of an inclusively connected A for exchange with an ith peripheral B is RAbi ðIÞ =

PAmax − PAbi PAbi − PAbicon

The effect of inclusion for a sequence of settlements in the N = Q branch is to successively reduce resistance of the central position while leaving the resistance of peripherals unaffected. As a result, the exchange ratio for the Qth exchange is less favorable to A than the rate of Q
1, the Q − 1 is less favorable than Q − 2, and similarly for all connected relations. When the first inclusion experiments were run, subjects in peripheral positions very quickly came to understand that later exchanges would give them better payoffs. All tried to outwait the others. As a result, dead

192

DAVID WILLER ET AL.

silence prevailed because no peripheral would negotiate until the last seconds of each period of negotiation. Then, at the last seconds, all exchanged at effectively the same time. We were surprised that exchanges became effectively simultaneous. Though surprised, looking back at Elementary Theory we realized that exchanges need not be sequential. When all Bs know, as they came to know in the experiment, that exchange in the last exchange is preferred to any prior exchange, all will attempt to outwait others and all settlements become effectively simultaneous. Simplifying the calculation of resistance for actors in inclusively connected positions, for simultaneous exchanges rates, will not be differentiated by order of exchanging. Instead, PAb Qcon = − ðQ − 1ÞPAb In words, the cost to the central position of not exchanging in any one relation is the loss of benefit from all other exchanges. (For Q exchanges, the number of all other exchanges is Q
1.) Substituting, for an inclusively connected A when exchanging simultaneously with Q Bs, RAbQ ðIÞ =

PAmax − PAb PAmax − PAb = PAb þ ðQ − 1Þ PAb QPAb

It follows that the resistance of the inclusively connected actor in a branch with Q simultaneous exchanges is the resistance of the actor in the prototype relation divided by Q. That is, RAbQ ðIÞ = RAb =Q As the size of an inclusively connected branch increases, the power of the central position decreases such that exchange ratios become less and less favorable. Experiments on two-branch through seven-branch networks support the forgoing predictions for inclusive connection (Patton & D. Willer, 1990; D. Willer & Skvoretz, 1997a).

CONNECTION AND POWER Not configuration alone, but configuration and type of connection determine the distribution of power in social structures. As seen for exchange

Elementary Theory: 25 Years of Expanding Scope

193

networks, the central position is high power when the connection is exclusion and low power when the connection is inclusion. Beyond exclusion and inclusion, there are three further types of connection. All five are defined using Ni, the number of connections at i; Mi, the maximum number of exchanges in which i can benefit; and Qi, the minimum number that i must complete to benefit. Then, Qi is a subset of Mi, which is a subset of Ni and i is inclusively connected if Ni = Mi = Qi > 1 i is exclusively connected if Ni > Mi ≥ Qi = 1 i is null connected if Ni = Mi > Qi = 1 i is inclusive-exclusively connected if Ni > Mi ≥ Qi > 1 i is inclusive-null connected if Ni = Mi > Qi > 1 The effects of these types of connections are quite general and straightforward
at least for networks such as branches where only one node has multiple relations. When inclusively connected the central position is less powerful than peripherals, when exclusively connected the central position is more powerful than peripherals, and when null connected the central position exchanges equally with peripherals. Furthermore, when inclusion is mixed with exclusion and null, the effect of inclusion is completely subverted. Thus, for a given number of peripherals, exchange ratios of the inclusive-exclusively connected branch are identical to those of an exclusive branch, and exchange ratios of the inclusive-null connected branch are identical to those of a null branch (D. Willer & Skvoretz, 1997a). The effect of inclusion in these two cases is subverted, not eliminated because, when peripherals act collectively, inclusion effects reemerge. (See section “Coalitions and Collective Goods”.) An example will explicate the working of an inclusive-exclusive connection. A corporate actor is seeking to hire people to work on an assembly line. Unless all positions on the line are filled, the assembly will be incomplete and the product will not be marketable. Therefore, the corporate actor is inclusively connected. At the same time, more workers are seeking employment than are needed. Therefore, the corporate actor is exclusively connected. Due to the surplus of workers, the corporate actor will exercise power over the workers and that power will not be affected by the inclusive connection. Since inclusive connections pervade all human societies, why has inclusion long remained obscure? The answer seems to be that inclusion is fragile; its effects are subverted when either exclusion or null is admixed. Nevertheless, inclusion could be the key to countervailing the massive

194

DAVID WILLER ET AL.

power differences of modern societies. Just as Smith’s pinmakers could have power over the capitalist hiring them, so could assembly line workers in a modern factory. To do so, however, they would have to control replacement
which is to say, eliminate the effect of exclusion.

WEAK POWER Weak power is a self-limiting form of exclusionary structural power discovered by researchers working on a fundamental problem of network exchange: predicting the relative power of positions in an exchange network of any size and shape. In contrast, structural power differences produced by exclusion such as those described in the original formulation of Elementary Theory produce large resource differences that increase toward the extreme between high-power and low-power actors. Over a series of exchanges, almost all of the available profit eventually migrates to the high-power actor. Compared to such “strong power” effects on resources, weak power effects are so subtle that the structural conditions that produce them were not immediately apparent. Markovsky, D. Willer, and Patton (1988) developed the Graph-theoretic Power Index (GPI) to predict which positions in an exchange network would have strong power over other positions. The GPI works by counting paths radiating out from a position in an exchange network. A path directly connecting to another position adds power. When that other position, however, has a path connecting to a third actor, it detracts from the original actor’s power, now two steps back along the path. In general, odd-length paths add power to a position by providing an exchange partner while evenlength paths subtract power from a position by providing a potential exchange partner with an alternative. For nearly all exchange networks when actors can exchange at most once per round of exchange, the GPI accurately predicts which positions will have a strong power advantage or disadvantage and which positions will have power equal to their immediate neighbors. The GPI also predicts that a de facto break will occur in exchange networks when actors accept the more generous offers of an equal-power actor over those of a high-power actor. Not all seemingly equal power actors are equal. As researchers investigated different exchange networks, some networks predicted to have equal power, in fact, developed small resource differences between positions. Consider a four-actor line network, A1
B1
B2
A2. The B actors might

Elementary Theory: 25 Years of Expanding Scope

195

seem to have a power advantage given that A actors have no alternative to exchange with a B. The GPI, however, scores them all equally. The small resource advantage of B actors found by researchers was confirmed by extending the theory to explain weak power. Markovsky, Skvoretz, D. Willer, Lovaglia, and Erger (1993) showed that when actors in the fouractor line network begin to exchange, B actors will occasionally exchange with each other. When that happens, A actors will be deprived of profitable exchange and thus increase offers to B actors on subsequent rounds. But actor A need offer only a small increase in profit to B, who can expect no better than an equal profit split in exchange with the other B. Markovsky et al. (1993) developed a probability model, based on the likelihood that a position would be included in exchange, to predict the relative weak power advantage or disadvantage of positions in networks rated equipower by the GPI. One additional advance was needed to solve the problem of predicting power in exchange networks. The likelihood of being included model could predict the relative power of positions in an exchange network. The ultimate goal of research programs in network exchange is to predict the exact profit that different positions in a given network will garner after a series of exchanges. Reaching that goal required extending the likelihood of being included model by incorporating the resistance equation from Elementary Theory (Lovaglia, Skvoretz, D. Willer, & Markovsky, 1995). In Elementary Theory, exchange occurs at the point where two actors balance their interests in gaining more resources from exchange with their interests in avoiding confrontation that precludes exchange. The resistance equation relates two fundamental motivations, greed and fear. Lovaglia et al. (1995) used likelihood of inclusion to estimate an actor’s interest in gaining more resources and avoiding conflict. The resulting formula gave accurate predictions for the profit garnered by positions in a variety of exchange networks. Developments in elementary theory continued to improve the prediction of resource distributions. Simpson and D. Willer (1999) introduced new, simpler methods for differentiating strong, equal, and weak power networks and finding network breaks. GPI is no longer used. When li is the likelihood that a position is included in exchanges, strong power networks contain only two types of positions, nonexcludable for which li = 1 and excludable for which li < 1. Excludable positions are connected only to nonexcludable. In equal power networks, li values for all positions are the same. All networks that are not strong or equal are weak power. Finally, network breaks occur only between high-power positions in strong power subnetworks and positions that are never excluded.

196

DAVID WILLER ET AL.

The Simpson and D. Willer procedures were further simplified in two further papers. To avoid the complication of calculating li values, Girard and Borch (2003) substituted “excludable” and nonexcludable and showed that distinction was enough to differentiate the three types of networks and find breaks. Subsequently, in “Testing Ten Theories” (2008), D. Willer and Emanuelson proposed substituting Friedkin’s Expected Value procedure to simplify the calculation of likelihoods.

POWER AND INFLUENCE Bierstedt considered Karl Marx to be highly influential but not powerful. “Stalin, on the other hand, is a man of influence only because he is first a man of power” (1950, p. 732). In recent sociological research, power and influence are seen as distinct social processes (D. Willer, Lovaglia, & Markovsky, 1997). A position of power alters behavior and attitudes through sanctions; failure to comply has a cost. Influence alters behavior and attitudes in the absence of sanctions through advice and persuasion. With status-based influence, for example, a patient might follow the advice of a doctor to undertake a life-threatening medical procedure with the expectation that the doctor is competent and shares an interest in the patient’s successful health outcome. While both power and influence are fundamental to social interaction, they are highly correlated. Individuals and organizations high in power are usually high in influence as well. Showing how power and influence are related is an important research problem. Karl Marx might propose that power underlies all effective influence. Max Weber might disagree, seeing influence as separate and perhaps equally important. As explained in the next section, Thye (2000) demonstrated that individuals high in status-based influence easily gain power through exchange because exchange partners place a higher value on resources associated with those high in status. Gaining influence from a position of power, as Bierstedt proposed, also occurs. Power confers access to resources that can be marshaled to accomplish a variety of goals. And because the exercise of power can occur without visible evidence other than accumulating resources, those in positions of power are given credit for the ability to amass those resources. The resulting expectations of competence for those in positions of power result in influence similar to the influence of those high in status.

Elementary Theory: 25 Years of Expanding Scope

197

Despite firm theoretical underpinning for the proposition that a position of power can produce influence, a laboratory demonstration of the phenomenon has proven elusive. In a series of experiments, participants in high-power and low-power positions exchanged resources. They then worked on a cooperative task. The previously high-power partner was no more influential than was the low-power partner (Lovaglia, 1995). Lowpower participants, however, reported negative emotions and attitudes directed toward their high-power partner. D. Willer et al. (1997) showed how emotional reactions to the exercise of power create resistance that mediates between power and the influence that might result from the accumulation of resources. Work continues in the attempt to explain the high influence attained by people in positions of power, especially those that exercised their power aggressively, sowing negative emotion and resistance along the way. R. Willer, Troyer, and Lovaglia (2005) proposed that while people upon whom power is used react negatively, observers of power used on others might not react as strongly. Perhaps those directly subjected to power would resist the influence of the powerful, while others seeing only the resulting differences in resources would be influenced by the apparent competence of those in power. A person in power, then, could exercise power harshly over relatively small numbers of people while gaining influence over many that are aware of the exercise. The result would be increased influence overall. In an experiment, participants watched a network exchange setting in which one exchange partner had a power advantage over another. In the second phase of the experiment, participants worked on a cooperative task with a partner they had observed exchanging in either the high-power or low-power position in the previous exchange setting. Partners that had previously been observed in a high-power position had significantly more influence over observers than did those previously observed in a low-power position. A second proposal to explain the influence of the powerful holds that after accumulating resources through a position of power, the powerful can then gain status through philanthropy, contributing to the collective good. R. Willer (2009) found that those making substantial contributions to collective action gained influence. R. Willer, Troyer, Youngreen, and Lovaglia (2012) found that after a philanthropic gesture, power users were held in higher esteem by observers. It is likely that the passage of time, separation from the event, and subsequent good works all serve to increase the influence of those that amass resources through the exercise of power.

198

DAVID WILLER ET AL.

STATUS VALUE AND STATUS INFLUENCE Inspired by Lovaglia et al.’s work on power and status, Thye began a line of theorizing in the late 1990s aimed at the converse connection between status and power. The status value theory of power (Thye, 1999, 2000; Thye, D. Willer, & Markovsky, 2006) explains how status characteristics like race or gender affect the perceived status value of resources, and subsequently, the development of power in exchange relations. The initial theory relied on a simple insight
the status value of any item (i.e., a basketball) might be related to the status of its owner (a high school amateur vs. Lebron James). From Elementary Theory and virtually all other theories of power, then, actors who possess valued goods have power over actors with less-valued goods, all else being equal. That status produces power is now axiomatic in tests inside (Berger & Fi ¸sek, 2006, Forthcoming; Thye, 2000; Thye et al., 2006) and outside the laboratory (Ayres, 1991). A series of tests in status-differentiated dyads and triangles showed that participants tried harder to acquire the chips associated with high-status partners, assumed they were generally more important, and were willing to accept less money to get them. As a result higher-status actors exercised power in exchanges with lower-status partners. These studies showed that status characteristics alter the perceived value of resources, and this translates into power. Later work quantified this effect via the resistance equations and allowed for precise ratio level predictions for the status-topower effect (Thye et al., 2006). A second mechanism, status influence links status to power through influence processes (Thye et al., 2006). In summarizing this branch of the research program, the conclusion is clear: status produces power and it does so through a variety of mechanisms.

STRATEGIC ANALYSIS Applications of game theory across the social sciences assume, but fail to show, that games are contained in social structures (Roth, 1995). Strategic Analysis is a theoretic procedure for finding games in social structures. Its application shows that 1. low-power actors in all strong power structures play a dilemma game, often an N-person prisoner’s dilemma game, with each other that must be overcome to cooperate;

Elementary Theory: 25 Years of Expanding Scope

199

2. by contrast, high-power actors face no dilemma blocking their cooperation; 3. the dilemma faced by low-power actors and its absence for those high in power taken together explain why low-power actors free ride and thus how power is exercised; and 4. successful coalition formation by those low in power eliminates the dilemma, allowing power to be countervailed. The procedures of Strategic Analysis are found in D. Willer and Skvoretz (1997b) and Borch and D. Willer (2006). Here is an example using the simplest power structure with multiple high- and low-power positions. Let two A positions be connected to three B positions and all be limited to maximally one exchange. At equal power, two As and two Bs exchange: each would gain half the payoffs and one B would be excluded. Thus, each B’s expected payoff is 1/2 × 2/3 = 1/3. However, if only two Bs make the equal power offer and one B free rides offering to take X when 1/2 > X > 1/3, the free rider always exchanges and the As are exercising power. The motive to free ride is that X > 1/3, and the free rider always exchanges because X < 1/2. It is easily seen that, by contrast, there is no X such that any high-power actor ever gains more by free riding. Therefore, free riding drives power to the extreme. Free riding is controlled and power is countervailed by a coalition that is organized to reverse the X > 1/3 motive by coercion or the 1/2 > X motive by sharing.

COALITIONS AND COLLECTIVE GOODS A coalition is a group of two or more actors acting collectively against one or more others to gain more favorable outcomes than they would gain acting independently. Research on coalitions and collective action has long been central to sociology. But most of this work (e.g., Caplow, 1956; Gamson, 1961) has taken the underlying reason for coalition formation, such as power or resource inequalities, as initial conditions. In contrast, research extending Elementary Theory to collective action takes recognizable power structures as initial conditions. It then predicts whether or not collective action will occur in those structures (Simpson & Macy, 2001; D. Willer & Skvoretz, 1997b) and the size of collective goods that can be gained via collective action (Simpson & D. Willer, 2005). This section briefly reviews extensions of Elementary Theory to find collective goods embedded in power structures.7

200

DAVID WILLER ET AL.

Building on the theory’s typology of network connections, Simpson and D. Willer (2005) used resistance equations to differentiate structures in which collective goods are embedded from those in which they are not, and to predict the sizes of those goods. Consider, for instance, two networks: an inclusively connected branch structure (Br666) and an exclusively connected branch structure (Br641). Simpson and D. Willer predicted that peripherals in the exclusively connected network would benefit from coalition formation, whereas peripherals in the inclusively connected structure would not. This is because the interdependencies that exist along the periphery in inclusively connected networks favor occupants of the peripheral positions, as explained in previous sections on “Inclusion and Power” and “Connection and Power.” In contrast, the interdependencies that exist between peripherals in exclusively connected networks disadvantage them, as bidding wars create exchange inequalities that increasingly favor the central actor. A critical mass of otherwise low-power peripherals acting collectively can mitigate these bidding wars. For instance, a coalition of low-power actors might agree to reduce the number of offers it sends to a central actor by N
M, thereby effectively transforming the structure into a null-connected network that eliminates the power that would have been exercised over the periphery. In this case, the size of the collective good, derivable from resistance equations, is the difference between the payoffs in a null-connected network and the payoffs in an exclusively connected network (Simpson & D. Willer, 2005). That collective goods are latent in power structures is even clearer in networks with mixed connection types. Recall from the earlier section on network connections that, inclusion and exclusion can both be present simultaneously. For example, consider a network in which relations branch from a single central position to four peripheral positions. When the central position must exchange with two or more of the peripherals, it is inclusively connected. When the central position can exchange with at least one fewer than its four peripherals, it is exclusively connected. When both conditions are present at the same time, as in Br432, the network is inclusive-exclusively connected. When peripherals in an inclusive-exclusively connected network act independently, exchange ratios are effectively identical to those in pure exclusively connected networks. This is because exclusive connection subverts the effects of inclusion. But Simpson and D. Willer (2005) argued that peripherals in inclusive-exclusively networks can, via collective action, transform the structure into a virtual inclusively connected network. For instance, if two (N
M) peripherals in a Br644 network withhold offers

Elementary Theory: 25 Years of Expanding Scope

201

from the central actor, the central actor becomes inclusively connected to the remaining four peripherals. Following the predictions for inclusion, the exchanges that occur in this virtual inclusive network are expected to favor the peripherals. Experimental results reported by Simpson and D. Willer are fully consistent with these predictions. Summing up, most research on coalitions and collective action has taken the potential size of a collective good that may be gained via collective action as given. Extensions of Elementary Theory, in contrast, take social structure as an initial condition and derive the size of collective goods that may be gained from successful collective action in that structure. A critical insight of these extensions is that networks that generate identical exchange ratios when actors act independently can lead to radically different outcomes when they act collectively. Thus, collective action can make manifest aspects of social structure that otherwise remain latent.

ORDERING AND GATEKEEPING Ordered exchange is the second “variant” and the seventh power condition studied by Network Exchange Theory (NET) (Walker et al., 2000).8 Ordering occurs when agreements in a series of relations must be completed in a given order. The A
B
C network is ordered when B must reach agreement first with A before negotiating with C. More generally, exchanges are ordered when “n > 1” exchanges must be completed in a given sequence (Corra, 2000, 2008; Corra & D. Willer, 2002). Gatekeeping, a familiar example of ordering, occurs when A controls B’s access to C. Elementary Theory asserts that the “ordering effect” is B’s cost of gaining access to C, a cost that benefits A at B’s expense (Corra, 2000). In exchange networks, gatekeeping takes the form of ordering, a structural power condition.9 Let G be the gatekeeper, C the client, and V the access valued by C. (“V” may be another actor to whom C seeks access or may be something valued by C. For gatekeeping, it only matters that C cannot access V without first exchanging with G, the gatekeeper.) Let the G
C
V relations each have a resource pool. Then, the gatekeeper’s fee is the inequality of the resource division in the G
C relation. The G
C
V structure is thus a gatekeeping structure because C must complete its agreement with G before it can exchange with V. As shown in Corra and D. Willer (2002), as the value of the C
V exchange increases, so does the inequality of the G
C exchange. Said somewhat differently, the fee to

202

DAVID WILLER ET AL.

the gatekeeper varies with the value of the access (Corra, 2000). Application of Elementary Theory’s resistance equations predicts the amount of increase of fee for a given value of access (Corra & D. Willer, 2002). The following are the main results of Corra and D. Willer’s (2002) investigation of gatekeeping: 1. Gatekeeper’s fees are determined by the value of the access granted to clients. 2. To gain fees, gatekeepers must monopolize control of access. 3. Multiple gatekeepers can gain fees only by organizing a shared monopoly, for example, by licensing. Ordering is a “variant” of inclusive connection. A “variant” is a structural power condition that (1) is not a connection type, but (2) has effects that are quantitatively similar to one of the types, and (3) interacts with other power conditions in the same way as that type. For example, research by Corra and D. Willer (2002) showed that the effects of ordering, like those of inclusive connection, are eliminated when combined with exclusive and null connection. This finding leads to the conclusion that, whereas gatekeepers occupy a position of power, to successfully demand fees, they must monopolize the gatekeeping role. Research currently underway seeks to investigate the impact of status differences on gatekeeper
client relations. For example, are greater fees gained when the gatekeeper’s status is higher than the client’s? What if the gatekeeper’s status is lower than the clients? Can a fee yet be gained? Three sets of hypotheses may be drawn. The first set predicts the effects of status influence as it works in the gatekeeper
client relation. The second predicts the effect of inflated or deflated status value of the access on the fees paid by clients. Given those two applications, a third set of hypotheses is drawn. The third set predicts the joint effect of status influence and status value when there are status differences in both gatekeeper
client and client
object relations.10

LEGITIMACY AND POWER Legitimacy finds its origin in structural power differences that generate power exercise through which structurally advantaged actors exploit lowpower actors to gain greater shares of social and material benefits. Highpower actors have an interest in retaining their power advantage, whereas

Elementary Theory: 25 Years of Expanding Scope

203

low-power actors have an interest in reducing or eliminating the advantages held by high-power actors. These interests give rise to legitimacy processes (Walker, Forthcoming). Legitimacy processes operate at both the collective and individual levels. Validity is legitimacy at the collective level. It establishes obligations for group members to conform to rules, beliefs, membership criteria, and standards of conduct. Propriety is individual-level legitimacy wherein individuals evaluate the appropriateness and desirability of constitutive characteristics. Those who attribute propriety to constitutive elements of groups or social structures conform to constitutive standards. Those who disapprove either deviate from rules and standards or conform because they are monitored to ensure that they do not disrupt the social order (Zelditch & Walker, 2003). High-power actors can use legitimacy processes to reinforce power by defining power structures as legitimate and alternative structures as illegitimate. For example, central to power exercise in contemporary United States is its capitalist economy. It is not surprising that more Americans have positive images of capitalism (61% vs. 33%) and more have negative images of socialism (36% vs. 58%) (Gallup, 2010).11 Legitimacy Theory implies that the degree to which individuals attribute propriety (have positive impressions of) to social structures (e.g., economic systems) varies with their authorization by superordinate individuals and institutions (e.g., opinion leaders, government officials, and federal court decisions). Propriety also varies with the degree to which the masses or important reference groups (i.e., “people in general” or members of one’s ethnic or other identity group) endorse or offer strong support. To investigate endorsement, Walker and D. Willer (2007) built two strong power networks and assigned experimental participants to low-power positions in them. In the opening rounds, participants negotiated individually and power exercise approached maximal high power
low power differentials as predicted by Elementary Theory. The participants were then given the option of forming coalitions against the lone high-power actor. In half the networks, participants were told that a majority of their peers endorsed (i.e., supported) coalition formation
a power-reducing strategy
while in the remaining networks they were told that a majority of their peers did not endorse the strategy. Walker and D. Willer (2007) report that networks in which coalition formation was legitimized by endorsement were (a) more likely to form coalitions, (b) less likely to have free riders (i.e., they had more unanimous coalitions), and (c) more likely to countervail power than networks for which coalition formation was not legitimized.

204

DAVID WILLER ET AL.

SOCIAL VALUES AND RATIONAL CHOICE Elementary Theory, like many rational choice theories, has focused research on self-interested actors (Von Neumann & Morgenstern, 1944; D. Willer & Emanuelson, 2008). Yet, there are cases in which actors do not behave in a narrowly self-interested manner, instead giving and cooperating at higher than predicted rates for games. There is a body of literature on Social Value Orientations (hereafter, SVO) that offers not one, but three orientations to rational action that can account for such behaviors (Bogaert, Boone, & Declerk, 2008; Messick & McClintock, 1968; Van Lange, 1999). These orientations are determined by preferences for how outcomes should affect self and other. Only one orientation, the individualist, corresponds to the selfinterested actor. The competitor seeks to maximize differences between self and other, whereas the prosocial seeks to maximize joint outcomes while maintaining equality.12 How does the addition of these orientation types alter power in social relations and structures? For example, would prosocials seek fairness over exploitation when occupying a powerful position? To expand the scope of Elementary Theory, utility functions were derived for each orientation type and then integrated into the resistance equation (D. Willer, Gladstone, & Berigan, 2013). A series of experiments investigated how actors of differing orientations behave in the equal power dyad, weak power A
B
B
A four line, and strong power exchange structure like that shown in Fig. 1. The integration of SVO and resistance created metric predictions that substantially diverged from previous predictions using only the individualist orientation. For instance, prosocials were expected to cooperate with each other and be exploited by proselfs, regardless of structure. A second experiment compared rates of coalition formation between prosocials and proselfs. Here, too, prosocials were expected to exhibit higher rates of cooperation by acting collectively. Surprisingly, none of the predictions were supported. In all experiments, the actions of the prosocials and the proselfs were indistinguishable. Unlike previous Elementary Theory research where participants were asked to assume a proself orientation, in these experiments they were prompted to “Exchange as you see fit.” While SVO research clearly shows that differences in social values strongly affect behavior in many contexts, in economic exchange conditions only behavior consistent with a proself orientation is called forth. In that context, the classic rational actor model remains a reasonable and “useful approximation of behavior” (Roth, 1995, p. 78). These results suggest that some social contexts can call forth one specific value orientation. If exploitative relations like exchange (and perhaps also

Elementary Theory: 25 Years of Expanding Scope

205

coercion) call forth only the proself orientation, then simpler theory using only that one orientation is needed for prediction and explanation in that context.

POWER-AT-A-DISTANCE IN FLOW NETWORKS In the structures discussed earlier, the exercise of power was limited to adjacencies. Power was limited to adjacencies because positions could only benefit at the expense of adjacent others. That is an important scope limit. The most important power structures in ancient and modern societies contain power that is exercised well beyond adjacencies. Power can be exercised over distant others, however, when resources move beyond adjacent positions to reach positions distant in the network (D. Willer, 2003). When resources move from one end of the network to the other, power exercise, both local and distant, is dependent on the series of power conditions linking the two ends of the network. Elementary Theory is the only theory in any social science predicting the exercise of power-at-a-distance. Power-at-a-distance in flow networks may be produced by as many as three structural conditions: ordering, inclusion, and exclusion. The simplest flow network A ↔ B ↔ C is not just two dyads stuck together: it has ordering and inclusion. Let A initially hold resource α that is valued only by C, and let A and C be connected only through B. Initially B and C have a medium of exchange ($) valued by all positions. With the goal of making a profit, B buys α from A using $ and then seeks to sell α to C for more $ than paid. Since A and C are not connected, the exchange sequence is fixed: A
B then B
C. Contrary to speculations by a number of sociologists and economists (e.g., Marsden, 1983), theory and experiments show that the middleman is disadvantaged in both exchanges. In the A
B exchange, B is disadvantaged by ordering because benefit from the exchange with A can be realized only through exchange with C. Therefore, B’s resistance to A equals A’s resistance to B only when A’s payoff is much higher than B’s. Inclusion disadvantages B in the B
C exchange. If B cannot sell α to C, B will neither benefit from α nor recoup the $ paid to A. B’s greater potential for loss lowers B’s resistance relative to C’s. Therefore, B is also less powerful than C. More generally, when the middleman’s first exchange is conditioned only by ordering and the second only by inclusion, the middleman is low in power in both. If the middleman has exclusive alternatives in either or both

206

DAVID WILLER ET AL.

exchanges, however, it will be high power in the exchange. Research prior to the study of flow networks is again supported here. Exclusion masks the effect of either ordering or inclusion or both, reversing power to the benefit of the middleman. Furthermore, when A has multiple Bs, each B multiple Cs and each C multiple Ds, such that exclusion occurs at each step, A exercises power-at-a-distance because it has power over all other positions in the structure and benefits from their exchanges. Two competing models predict the distribution of power in flow networks. First introduced was the Simultaneous Model where resistance equations for all subnetworks are solved simultaneously (D. Willer, 2003). While experiments in a number of networks support that model, it does not predict well for the A
B
C network. The Sequential Model (Emanuelson, 2008) solves resistance equations in the order that exchanges occur. As value is appropriated by actors who complete early exchanges, the value available for negotiation in later exchanges is reduced. Sequential model predictions fit the A
B
C network and are as good or better than simultaneous model predictions for all experimentally tested networks.

TESTING TEN THEORIES In 2008, ten exchange theories were tested against each other: all predicted exchange outcomes from conditions of structure. Previously, four of the ten, Power-Dependence Theory, Core Theory, Elementary Theory, and Expected Value Theory were compared and Elementary Theory’s predictions were found to be the most precise (Lovaglia et al., 1995; Skvoretz & D. Willer, 1993). Since those papers, three of the four theories had been refined and five new contending theories had emerged. With the goal of providing definitive grounds for selecting among competing theories, D. Willer and Emanuelson (2008) evaluated predictive precision using the most comprehensive extant data set. In addition to evaluating precision, theories were evaluated for logical consistency, parsimony, and scope. Seeking parsimony, Expected Value Theory’s method of calculating likelihoods of being included was combined with Elementary Theory’s resistance to create a 10th theory called Expected Value-Resistance. With the exception of Elementary Theory, the predictive mechanisms of the theories were developed specifically for resource pool networks where each node is limited to, at most, a single exchange. Therefore, with that one exception, the tested theories were sharply scope limited to networks that

Elementary Theory: 25 Years of Expanding Scope

207

have been proven to exist only in sociological experimental laboratories. By contrast, Elementary Theory can predict for relations in which commodities are exchanged both immediately and at a distance and for coercive relations as well. Although all theories could generate predictions for each of the experimental networks tested, some showed more theoretical merit and others less. Notable results of the test included Power-Dependence, perhaps the most prominent of the theories, being ranked ninth of ten in precision. Sixth-ranked Substitutable Exchange was found to contain logical inconsistencies that irrevocably undermined its utility. Network Nash was both logically consistent and relatively simple, but ranked seventh in precision. Elementary Theory was found to be logically consistent and ranked first in precision, but it lacked parsimony insofar as calculating likelihood of being included is concerned. Expected Value-Resistance, the new theory offered by D. Willer and Emanuelson, was second in precision and far more parsimonious than any other. While all 10 theories clearly demonstrated that sociology can be and should be a precise science, with one exception, Expected Value-Resistance, all theories are so complex in application that computer programs are needed to apply them. By 2008, however, those programs were either not publically available or could not be run on then modern computers, or both, which sharply limited their utility.

ONE-SHOT EXCHANGE AND THE SHADOW OF THE FUTURE In game theory, strategies and outcomes can differ between one-shot and repeated games. For example, in one-shot P/D games defection dominates but when repeated tit-for-tat gives higher payoffs. In strong and equal power exchange structures, rational actors adopt the same strategies whether or not exchanges are repeated. But, when structures are weak power, the presence or absence of a future affects both actors’ strategies and exchange outcomes. Emanuelson and D. Willer (2009) predict the effect of the “Shadow of the Future” on strategy and outcomes in one-shot and repeated weak power exchange networks. When exchanges are one-shot there is no future, and rational actors will not exchange in suboptimal relations. For example, in A
B
B
A, the B
B is a suboptimal relation. Because the Bs receive only equal

208

DAVID WILLER ET AL.

divisions when exchanging with each other, in the one-shot network, each A need offer its B only one more point to be assured of exchange. By contrast, when exchanges are repeated, suboptimal relations become threats used by higher-power actors like the Bs to increase the size of offers made by lower-power positions. When threats are believed, payoffs to higherpower actors increase beyond those found in one-shot exchanges. When not believed, threats are carried out. Therefore, exchanges in suboptimal relations should occur only in networks with repeated exchanges. To predict payoffs in one-shot networks, application of the “Rationality Principle” (P1) is sufficient, but to predict outcomes in repeated exchange networks, resistance equations must be applied. Theory predicts and experiments found that lower-power positions were never excluded in one-shot networks and their payoffs were only slightly less than payoffs to higher-power positions. In contrast, higher-power actors in repeated exchange networks strategically used the opportunity to exclude lower-power exchange partners by exchanging in suboptimal relations, exchanges that increased their total payoffs across the sequence of exchanges beyond the minimal differences found in one-shot networks.

ANALYZING LARGE-SCALE NETWORKS Resource pools have been widely used in experiments in part because subjects find dividing resources easier to understand than sending and receiving valued resources. Nevertheless, resource pool networks pose uniquely difficult prediction problems that rapidly compound with network size. It is easier, for example, to predict for a labor market containing hundreds of thousands of nodes where labor is exchanged for money than to predict for some networks as small as eight nodes connected by resource pools. One reason why Elementary Theory is not size limited is because it is not restricted to resource pools. To predict outcomes for large resource pool networks, Elementary Theory uses “Domain Analysis,” a procedure for cutting large networks into smaller parts called “domains” (D. Willer, Emanuelson, & Van Assen, 2012). Domains are embedded subnetworks that act as they would were they not part of a larger structure. Using Domain Analysis, Elementary Theory can find embedded strong, equal, and weak power domains, and predict activity in each domain such that all exchange outcomes are known.

209

Elementary Theory: 25 Years of Expanding Scope

Domain Analysis is demonstrated for Fig. 5(a) where positions A through J are labeled I if always included and E if not. As shown in Fig. 5(b), relations not found in strong power networks are removed so that a boundary can be drawn around the revealed G
C
H strong power domain. In Fig. 5(c), the relations are replaced and the network relabeled. When relabeling, relations that cross into the strong power domain like B
C, C
D, and C
I are ignored. Since exchange will never happen between I and a high-power position, the C
D relation is removed. In Fig. 5(d), the A, B, F domain is equal power and the D, E, I, J domain is weak power. Two relations extend between domains: B
C and C
I. When B and I seek to exchange with C, since they are competing with the low-power H and G for C’s single exchange, they will be low power. It follows that B and I will seek exchange

(a)

BE

CI

DI

EE

FE

GE

HE

IE

JE

B

C

D

E

I

J

EE

A

(b)

G

F

(c)

AE

FE

BE

C

DI

G

H

IE

Equal

(d)

A

F

Fig. 5.

H

Strong

JE

Weak

B

C

D

E

G

H

I

J

Applying Domain Analysis.

210

DAVID WILLER ET AL.

with C iff excluded in their equal and weak power networks, respectively. Therefore, likelihoods of exclusion can be calculated for the D, E, I, J domain as if it had no connection to any other. The analysis of Fig. 5 is now complete, but other networks may need one further rule. The rule is that relabeling cannot produce a strong power component. For example, when an analysis shows a five line sequence A
B
C
D
E after strong power domains are revealed, it will not be relabeled E
I
E
I
E. Instead, initial labels apply. To accurately predict exchange outcomes of Fig. 5 and other networks as large or larger, they must be analyzed into their component parts. Since Domain Analysis is the only mode of analysis thus far developed, exchange outcomes in large resource pool networks cannot be predicted without it.

EXPERIMENTAL METHODS A science advances by devising theory (i.e., principles and laws) that organize and explain phenomena within its domain. But constructing theory is only one element of the scientist’s primary mission. Unlike the products of historical and philosophical inquiry, scientific principles must be tested. Elementary Theory has survived more than three decades of challenging tests. Many, but by no means all, of those tests have been conducted by laboratory experiments and that is as it should be. An experiment is “an inquiry for which the investigator plans, builds, or otherwise controls the conditions under which phenomena are observed and measured” (D. Willer & Walker, 2007, p. 2). As explained by D. Willer and Walker (2007), scientific experiments are driven by theory and are only superficially similar to method of difference experiments that are so common in the social sciences. The logic of scientific experimentation is well-known in the physical sciences where theorists build models that are translated into concrete laboratory operations. Laboratory procedures generate observations that support or disconfirm predictions derived from theory. Theories that garner reproducible support are refined by expanding their scope or by improving their precision. Theories that fail to gain support are either reformulated or replaced. Research and development in Elementary Theory has followed the scientific experiment paradigm. Early Elementary Theory experiments including those of Brennan (1981) used counters (e.g., poker chips) for resources and university students as subjects in various exchange and coercion settings. The findings from those

Elementary Theory: 25 Years of Expanding Scope

211

experiments, and subsequent studies in computer-mediated negotiation settings, overwhelmingly support predictions of Elementary Theory. The philosopher of science Toulmin (1953) suggested that one should not go into the lab without good reason to know results beforehand. In 25 years of Elementary Theory research surprises have been the exception.

PREHISTORIC SOCIAL DYNAMICS Research in prehistoric social dynamics aims to understand the social relations and social structures of ancient societies and to explain their social dynamics. That research’s current focus is on the advent of social complexity through the rise of the state. An important challenge in this task is the lack of direct evidence of the social relations within which people acted. Nor is it known how those relations were embedded in structures. While anthropological and archaeological studies provide indirect evidence, the application of experimentally tested theory is central to inferring ancient structures and deriving their dynamics (Chacon, 2009; Emanuelson & D. Willer, 2012; D. Willer, Chacon, Emanuelson, Chacon, & Lewis, 2013). In fact, this research employs a methodology new to prehistorical investigations. It applies experimentally tested theories, including Elementary Theory and Status Characteristics Theory (Berger, Cohen, & Zelditch, 1966), to infer the structures of ancient societies. It is well-known that states evolved out of chiefdoms and that chiefdoms were built on status lineage structures. What is not known is why the chiefdom
status lineage structure developed independently on all inhabited continents. Sociological theory, grounded in extensive experimental research, asserts that status structures, of which status lineage is a very early type, solve collective action problems allowing its members to pursue common goals in a coordinated way (Simpson, R. Willer, & Ridgeway, 2012). It follows that societies with status lineage structures have a significant competitive advantage in warfare, hunting at other collective actions over societies that lack structures capable of solving those problems.

FUTURE DIRECTIONS It makes no sense to summarize a chapter that is, itself, a summary of past work. Far better to look toward the future. Today Elementary Theory

212

DAVID WILLER ET AL.

offers a well-developed toolbox of formulations that include a wide array of experimentally tested models to apply for historical and (as in the previous section) prehistorical explanation as well as general procedures for developing new models as needed. Papers already published and in preparation show that Elementary Theory models tested in the laboratory can be usefully applied to explain historical transitions. In addition to the recent work on the rise of the state, D. Willer et al. (1996) applied a number of Elementary Theory models to explain the fall of the Roman Empire in the west. Elementary Theory should be used because it has important implications for explanation of social structures. For example, studies of organizations take for granted that administration will be through a series of offices arranged in a tree-like structure. Also taken for granted is the system of promotion wherein higher positions are filled by promotion from below. In this promotion system, as Weber put it, every private soldier carries a marshal’s baton in his knapsack. This system of hierarchy with mobility is effectively universal today. It is found in all societies regardless of culture, language, or property system. And where it is found, the hierarchy of positions is also a hierarchy of pay and perquisites of office. It is time to ask why. Here is why. Hierarchy/mobility is a structure of domination. Importantly, it is not sufficient that a structure have pay and perk differentials from bottom to top. When there is no mobility, as in feudal and other traditional structures, the hierarchical system is not a power structure. Allow mobility, however, and officials, motivated to move up, will be obedient (Bell, Walker, & D. Willer, 2000). In fact, hierarchical models with and without mobility have been experimentally tested. Only those with mobility were power structures (D. Willer, 1987). There is certainly room for work in pure theory, work to advance Elementary Theory further. Nevertheless, the time has come to use it. Of course, sociology has a long history of using theory for explanation, but the theory used has been classical theory, especially that of Marx and/or Weber. The idea of using contemporary theory, more particularly experimentally tested theory, is a radical proposal. All the better.

NOTES 1. It is Popper’s position that the subject matter of the social sciences, contrary to the complaints of our colleagues, is simpler than that of the physical sciences

Elementary Theory: 25 Years of Expanding Scope

213

(1957, p. 129). He goes on to assert that developing social science theory is far easier than physical theory because social science has rationality with which to construct its models for which there is no parallel in the physical sciences (pp. 130
131). Only if Popper is right, that our subject matter is simpler and building models for it easier, is it possible for a theory as simple as Elementary Theory to have the broad scope that experiments demonstrate that it does. This chapter covers that scope. Significantly, given Popper’s points, we will report on research showing that the principle of rationality has far broader application than recent attacks on that principle are thought to indicate. 2. Strong power occurs in any exchange network where (1) high-power positions can jointly exchange with fewer than the number of those low in power, (2) those low in power have no better alternative than exchanging with those high in power, and (3) those low in power cannot act collectively to oppose the power exercise. An overfilled labor market, the structure that is central to stratification in all capitalist societies, is an exclusionary structure with multiple high-power capitalists and multiple low-power workers, only some of which are employed. 3. As in this example and more generally, Elementary Theory has no micro
macro problem. 4. Comprehensive introductions to the theory are found in D. Willer (1999) and Walker et al. (2000). 5. As in Fig. 3, normally sanctions are bisigned and may include payoff valuations. 6. With the advent of multiple theories of exchange, some theorists came to believe that there was a need to develop a “theory of value,” and one exchange theorist, Emerson, went so far as to begin such a construction. From the point of view of Elementary Theory, no such construction is needed. Values are in structures. Actors do not create their own values, they internalize them. Even ascetics do not create their own values. According to Weber, they internalize them from their religions and apply them in their economic behavior. 7. The discussion to follow uses a shorthand to designate kinds of structures. The prefix “Br” means that the structure is a tree network with relations to multiple nodes branching from a single central node. Three numbers follow. The first is the number of relations connected at the central node, the second is the maximum number of peripheral nodes that can exchange with the central node, and the third number is the minimum number that must exchange for the central position to benefit. For example, in a Br432, there is a single central node and four relations to four peripheral nodes, at most three of the four can exchange with the central node, and at least two must exchange for the central node to benefit. 8. The five types of connection are five conditions of power: they are exclusion, inclusion, null, exclusion-inclusion, and null inclusion. There are also two variants: hierarchy/mobility that is a variant of exclusion and ordering a variant of inclusion. Five connection types plus two variants give a total of seven power conditions. For hierarchy/mobility, see later. 9. In gatekeeping, exchanges are ordered, but not all ordered relations involve gatekeeping. For example, when the middleman B buys a commodity from A and sells to C, B is not a gatekeeper because, before the sale to C, B owns the commodity.

214

DAVID WILLER ET AL.

10. Extending previous research to status-differentiated networks requires application of two theories: Elementary Theory (ET) to model the gatekeeper
client relation (Corra & D. Willer, 2002) and Status Characteristics Theory (SCT) to model the impact of status on that relation. (See Walker et al., 2000, for an overview of ET and Wagner & Berger, 2002, for an overview of SCT.) 11. Respondents were asked whether they had positive or negative images of the concepts “just off the top of your head.” The surveys did not define the terms. As might be expected, the frequency of positive and negative responses varied with political party orientation and general political outlook (i.e., conservative, moderate, or liberal). 12. Research frequently collapses competitors and individualists into one category, the proself (De Cremer & Van Lange, 2001; Simpson, 2004; Simpson & R. Willer, 2008).

ACKNOWLEDGMENT The authors would like to thank The National Science Foundation for grants supporting the development and testing of Elementary Theory and its relations to other theories.

REFERENCES Ayres, I. (1991). Fair driving: Gender and race discrimination in retail car negotiations. Harvard Law Review, 104, 817
872. Bell, R., Walker, H. A., & Willer, D. (2000). Power, influence, and legitimacy in organizations: Implications of three theoretical research programs. Research in the Sociology of Organizations, 17, 131
177. Berger, J., Cohen, B. P., and Zelditch, M., Jr., (Eds.). (1966). Sociological theories on progress (Vol. 1). Boston, MA: Houghton Mifflin. Berger, J., & Fi ¸sek, H. (Forthcoming). The spread of status value: A theoretical extension. In Advances in group processes (Vol. 30). Bingley, UK: Emerald. Berger, J., & Fi ¸sek, M. H. (2006). Diffuse status characteristics and the spread of status value: A formal theory. American Journal of Sociology, 111(4), 1038
1079. Bierstedt, R. (1950). An analysis of social power. American Sociological Review, 15, 161
184. Bogaert, S., Boone, C., & Declerk, C. (2008). Social value orientation and cooperation in social dilemmas: A review and conceptual model. British Journal of Social Psychology, 47, 453
480. Borch, C., & Willer, D. (2006). Power, embedded games and coalition formation. Journal of Mathematical Sociology, 30, 77
111. Brennan, J. (1981). Some experimental structures. In D. Willer & B. Anderson (Eds.), Networks exchange and coercion: The elementary theory and its applications (pp. 189
206). New York, NY: Elsevier.

Elementary Theory: 25 Years of Expanding Scope

215

Caplow, T. (1956). A theory of coalitions in the Triad. American Sociological Review, 21, 489
493. Chacon, Y. (2009). An inquiry into Inka power structures. Master thesis, Department of Sociology, University of South Carolina, Columbia, SC. Corra, M. (2000). Applying resistance to ordering in exchange networks: A theoretical extension. Current Research in Social Psychology, 5, 84
96. Retrieved from http:// www.uiowa.edu/∼grpproc Corra, M. (2008). Inclusion and ordering: The compounding effects of two distinct but related structural power conditions. Social Behavior and Personality, 36(9), 1161
1178. Corra, M., & Willer, D. (2002). The Gatekeeper. Sociological Theory, 20, 180
205. De Cremer, D., & van Lange, P. A. M. (2001). Why prosocials exhibit greater cooperation than proselfs: The roles of social responsibility and reciprocity. European Journal of Personality, 15, 5
18. Einstein, A. ([1933] 1954). On the method of theoretical physics. In S. Bargmann (Ed.), Ideas and opinions (pp. 270
276). New York, NY: Crown Publishers. Elster, J. (Ed.). (1986). Rational choice. New York, NY: New York University Press. Emanuelson, P. (2008). Extension and refinement of network exchange theory. PhD Dissertation, University of South Carolina. Emanuelson, P., & Willer, D. (2009). One-shot exchange networks and the shadow of the future. Social Networks, 31, 147
154. Emanuelson, P., & Willer, D. (2012). Toward a more general circumscription theory: Two models for the transition from chiefdom to the state. Unpublished manuscript. Department of Sociology, University of South Carolina, Columbia, SC. Fehr, E., & Gintis, H. (2007). Human cooperation and social cooperation: Experimental and analytical foundations. Annual Review of Sociology, 33, 43
64. Gamson, W. A. (1961). A theory of coalition formation. American Sociological Review, 26, 373
382. Girard, D., & Borch, C. (2003). Optimal seek simplified. Current Research in Social Psychology, 8, 225
242. Kagel, J., & Roth, A. (Eds.). (1995). The handbook of experimental economics. Princeton, NJ: Princeton University Press. Lovaglia, M. J. (1995). Status and power: Exchange, attribution and expectation states. Small Group Research, 26, 400
426. Lovaglia, M. J., Skvoretz, J., Willer, D., & Markovsky, B. (1995). Negotiated exchanges in social networks. Social Forces, 74, 123
155. Markovsky, B., Skvoretz, J., Willer, D., Lovaglia, M. J., & Erger, J. (1993). The seeds of weak power: An extension of network exchange theory. American Sociological Review, 58, 197
209. Markovsky, B., Willer, D., & Patton, T. (1988). Power relations in exchange networks. American Sociological Review, 53, 220
236. Marsden, P. (1983). Restricted access in networks and models of power. American Journal of Sociology, 88, 686
717. Marx, K. ([1867] 1967). Capital. New York, NY: International Publishers. Messick, D. M., & McClintock, C. (1968). Motivational basis of choice in experimental games. Journal of Experimental Social Psychology, 4, 1
25. Newport, F. (2010). Socialism viewed positively by 36% of Americans. Gallup politics. Retrieved from http://www.gallup.com/poll/125645/socialism-viewed-positively-americans.aspx. Accessed on June 15, 2013.

216

DAVID WILLER ET AL.

Olson, M. (1965). The logic of collective action. Cambridge, MA: Harvard University Press. Patton, T., & Willer, D. (1990). Connection and power in centralized exchange networks. Journal of Mathematical Sociology, 16, 31
49. Popper, K. (1957). The poverty of historicism. New York, NY: Routledge. Roth, A. E. (1995). Introduction. In J. H. Kagel & A. E. Roth. (Eds.), The handbook of experimental economics (pp. 3
20). Princeton, NJ: Princeton University Press. Samuelson, P. (1983). Foundations of economic analysis. Cambridge, MA: Harvard University Press. Schelling, T. (1970). The strategy of conflict. Cambridge, MA: Harvard University Press. Simpson, B. (2004). Social values, subjective transformations, and cooperation in social dilemmas. Social Psychology Quarterly, 67, 385
395. Simpson, B., & Macy, M. W. (2001). Collective action and power inequality: Coalitions in exchange networks. Social Psychology Quarterly, 64, 88
100. Simpson, B., & Willer, D. (1999). A new method for finding power structures. In D. Willer (Ed.), Network exchange theory (pp. 270
284). Westport, CT: Praeger. Simpson, B., & Willer, D. (2005). The structural embeddedness of collective goods: Connection and coalitions in exchange networks. Sociological Theory, 23, 386
407. Simpson, B., & Willer, R. (2008). Altruism and indirect reciprocity: The interaction of person and situation in prosocial behavior. Social Psychology Quarterly, 71, 37
52. Simpson, B., Willer, R., & Ridgeway, C. (2012). Status hierarchies and the organization of collective action. Sociological Theory, 30(3), 149
166. Skvoretz, J., & Willer, D. (1993). Exclusion and power: A test of four theories of power in exchange networks. American Sociological Review, 58, 801–818. Thye, S. R. (1999). Status influence and status value. In D. Willer (Ed.), Network exchange theory (pp. 248
255). Westport, CT: Praeger. Thye, S. R. (2000). A status value theory of power in exchange networks. American Sociological Review, 65(3), 407
432. Thye, S. R., Willer, D., & Markovsky, B. (2006). From status to power: New models at the intersection of two theories. Social Forces, 84(3), 1471
1495. Toulmin, S. (1953). The philosophy of science. New York, NY: Harper. Van Lange, P. A. M. (1999). The pursuit of joint outcomes and equality in outcomes: An integrative model of social value orientation. Journal of Personality and Social Psychology, 77, 337
349. Von Neumann, J., & Morgenstern, O. (1944). Theory of games and economic behavior. Princeton, NJ: Princeton University Press. Wagner, D., & Berger, J. (2002). Expectations states theory: An evolving research program. In J. Berger & M. Zelditch, Jr., (Eds.), Contemporary sociological theory: New directions (pp. 41
76). New York, NY: Rowman and Littlefield. Walker, H. A. (Forthcoming). Legitimacy and inequality. In J. D. McLeod, E. J. Lawler, & M. Schwalbe (Eds.), Handbook of the Social psychology of inequality. New York, NY: Springer. Walker, H. A., & Willer, D. (2007). Peer endorsement: Legitimizing collective action and countervailing power. Paper presented at the annual meeting of the American Sociological Association, New York, NY. Walker, H., Thye, S., Simpson, B., Lovaglia, M., Willer, D., & Markovsky, B. (2000). Network exchange: Theory recent developments and new directions. Social Psychology Quarterly (Millennial Edition), 63, 324–337.

Elementary Theory: 25 Years of Expanding Scope

217

Watson, J. (1968). The double helix. New York, NY: Atheneum. Weber, M. ([1918] 1968). Economy and society. Berkeley, CA: The University of California Press. Willer, D. (1984). Analysis and composition as theoretic procedures. The Journal of Mathematical Sociology, 10, 241
270. Willer, D. (1987). Theory and the experimental investigation of social structures. New York, NY: Gordon and Breach. Willer, D. (2003). Power-at-a-distance. Social Forces, 81, 1295–1334. Willer, D. (Ed.) (1999). Network exchange theory. Westport, CN: Praeger. Willer, D., & Anderson, B. (Eds.). (1981). Networks exchange and coercion: The elementary theory and its applications. New York, NY: Elsevier. Willer, D., Chacon, Y., Emanuelson, P., Chacon, R. J., & Lewis, D. (2013). From influence to power: The path through chiefdoms to the state. Paper presented at the American Sociological Association meetings, New York, NY. Willer, D., & Emanuelson, P. (2008). Testing ten theories. The Journal of Mathematical Sociology, 32, 165
2003. Willer, D., Emanuelson, P., & Van Assen, M. (2012). Analyzing large scale exchange networks. Social Networks, 34, 171
180. Willer, D., Gladstone, E., & Berigan, N. (2013). Social values and social structure. Journal of Mathematical Sociology, 37, 113
130. Willer, D., Lovaglia, M. J., & Markovsky, B. (1997). Power and influence: A theoretical bridge. Social Forces, 76, 571
603. Willer, D., Simpson, B., Szmatka, J., & Mazur, J. (1996). Social theory and historical explanation. Humboldt Journal of Social Relations, 22, 63
84. Willer, D., & Skvoretz, J. (1997a). Network connection and exchange ratios: Theory, predictions and experimental tests. Advances in Group Process, 14, 199
234. Willer, D., & Skvoretz, J. (1997b). Games and structures. Rationality and Society, 9, 5
35. Willer, D., & Walker, H. (2007). Building experiments: Testing social theory. Stanford, CA: The Stanford University Press. Willer, R. (2009). Groups reward individual sacrifice: The status solution to the collective action problem. American Sociological Review, 74, 23
43. Willer, R., Troyer, L., & Lovaglia, M. J. (2005). Influence over observers of structural power: An experimental investigation. Sociological Quarterly, 46, 263
277. Willer, R., Youngreen, R., Troyer, L., & Lovaglia, M. J. (2012). How do the powerful attain status? The roots of legitimate power inequalities. Managerial and Decision Economics, 33, 355
367. Zelditch, M., & Walker, H. A. (2003). The legitimacy of regimes. In S. R. Thye & J. Skvoretz (Eds.), Advances in group processes (Vol. 20, pp. 217
249). Oxford, UK: Elsevier.

PERCEPTIONS OF ABILITY AND ADHERENCE TO RULES, GUIDELINES, AND TRADITION Jeffrey W. Lucas, Wesley S. Huey, Marek N. Posard and Michael J. Lovaglia ABSTRACT Purpose
This chapter develops and tests a theory on relationships between perceptions of ability and adherence to rules, guidelines, and tradition. Drawing from theory and research on status processes in groups, the theory proposes that adherence to rules can provide an alternative to task ability in demonstrating competence at a group task and that persons who perceive themselves to be low in ability will become especially likely to strictly adhere to rules. Methodology/approach
In an experimental study, participants received feedback that they had high or low ability at a group task that involved making judgments about bonuses in a fictitious organization. Findings
Supporting the theory, participants who perceived themselves to be low in ability gave less money to employees technically ineligible for raises, even when the reason for the ineligibility was arguably trivial.

Advances in Group Processes, Volume 31, 219
240 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0882-6145/doi:10.1108/S0882-614520140000031004

219

220

JEFFREY W. LUCAS ET AL.

Research limitations/implications
The proposed theory and supportive results have a number of theoretical implications for how status processes shape individual behavior in groups. For example, the theory might help explain collective enforcement against free riding, with people low in ability being motivated to enforce norms against free riding to compensate for their perceived lack of ability to contribute. Practical/social implications
It is easy to conjure examples in which persons who are seen as exceptionally competent also seem to be given wide leniency in adhering to rules. The theory and experimental test presented here can help in understanding the extent to which the following of rules may be seen as the domain of the incompetent. Keywords: Group processes; status; performance expectations; rules

INTRODUCTION Any fool can make a rule. And every fool will mind it. — Henry David Thoreau

When Thoreau made his journal entry about every fool obeying rules so long as they are made, he was reflecting on rules for speaking and writing English and how, so long as they exist, even educated persons will obey the rules without reflection. Thoreau’s insight, however, may be only part of the truth. Rule followers may be more strategic and more mindful than blind in following rules, especially if knowledge of rules, adherence to them, and enforcement of them are seen as demonstrations of competence. So are rules the domain of fools? We all follow countless rules every day. At the same time, it is easy to conjure examples of persons who are seen to be exceptionally competent and are also given wide leniency in obeying rules. From artists to athletes to members of elite military units, we tend to see those with the most ability as able to operate in ways that transcend the rules that apply to mere mortals. Perhaps rules are viewed by elites as the domain of the less competent, and perhaps the less competent cling to rules as evidence of their competence. We develop a theory that proposes that the less ability people feel they have in a setting, the more rigidly attached they will become to rules, procedure, and tradition, and also the more likely they will be to impose them on others. We develop our explanation based on status processes in groups and describe an experimental test of one element of the proposed theory.

Ability Perceptions and Rule Adherence

221

THEORETICAL DEVELOPMENT Our interest in the topic of ability perceptions and adherence to rules stemmed from an experience one of the authors had when applying for summer research funding from his university. The university had a program whereby faculty members could receive one month of summer salary support to work on research activities. The author developed a draft of a proposal to spend the summer months writing a grant proposal to the National Science Foundation (NSF). When the draft was complete, he met with the chair of the campus committee that reviewed applications. The committee chair said that the proposal was just the type of work the campus wanted to support and that it would be funded. After submitting his proposal, the author heard back from the committee that Institutional Review Board (IRB) approval was needed for the proposed research. The author then sent a note to the committee explaining that the research had not yet been developed, that the plan for the summer was to begin to design the research, that it would not be possible to write an IRB protocol without spending more time developing the work, that even NSF would not require IRB approval until they were ready to release moneys from a funded proposal, and that any IRB approval obtained at the time would be long expired before the research would actually be done. The committee responded that IRB approval was nevertheless necessary. The author contacted his IRB and was able to obtain an approval in principle for the basic idea of the proposed research. He then forwarded this to the committee albeit after the initial deadline for proposal submission. Sometime later, the author heard back from the committee that although they evaluated the proposal as strong, they ultimately decided that it would not be fair to applicants who had all of their materials in on time to grant an award to someone who had been negligent in obtaining IRB approval. After receiving this news, the author met with the chair of the campus committee to determine what happened. The committee chair informed the author that committee deliberations had been taken over by a group of new assistant professors who made evaluations of proposals nearly entirely on the basis of how well the proposals followed the rules. The author’s reaction to this was to conclude that the assistant professors on the committee did not feel as though they had the ability to competently evaluate the quality of the proposals, so they instead used criteria on which they felt they could demonstrate competence; namely, how well proposals adhered to stated rules and guidelines.

222

JEFFREY W. LUCAS ET AL.

This experience with the campus committee gave rise to our theory. We propose that the less ability people perceive that they have in a setting, the more likely they will become attached to rules, procedure, and tradition. Our explanation is largely grounded in ways in which status and expectations guide group interactions. We develop the explanation below, beginning with the ubiquity of status orders in groups.

The Pervasiveness of Hierarchies Task groups stratify. Bales, Strodtbeck, Mills, and Roseborough (1951) made the rather accidental discovery that members of human task groups initiate acts of communication toward other group members and toward the group as a whole in the same rank order as they are also targets of communicative acts from other group members. More generally, task groups tend toward a pattern of social relations in which some members control the flow of communication within the group relative to other group members. Presumably, Bales and colleagues reasoned, this degree of control experienced by members at the top of the rank order corresponds to some sociologically significant factor worth exploring as a feature of the group dynamic. Bales and colleagues had stumbled upon status as a differentiating characteristic of group members, and status order as a characteristic of the group as a whole (Fisek, Berger, & Norman, 1991). Important theoretical and practical implications flowed from the discovery of Bales and colleagues, giving birth to theoretical traditions in group process research in sociology, most notably expectation states theory (Berger, Cohen, & Zelditch, 1966) and the modeling of group member participation in group discussion (Skvoretz, 1988; Skvoretz & Fararo, 1996). These programs have, in turn, spawned theory and research designed to explain the remarkable regularity, ubiquity, and robustness of status orders in groups (Goetsch & McFarland, 1980), including their legitimacy (Ridgeway & Berger, 1986; Zelditch, 2001), their transmission from setting to setting (Ridgeway, 1991, 2006), and their relationship to power as a distinct property of individuals and group structures. Lovaglia (1995) showed how structural power differences among group members lead to status group members come to admire and esteem (to confer status upon) those in the group who control access to valued resources. As discussed below, Thye (2000) showed the reverse process also holds
group members lower in the status order came to value resources possessed by those higher in the status order.

Ability Perceptions and Rule Adherence

223

What is plainly clear from a half century of sociological research in small groups (see Burke, 2003, for a review) is that status hierarchies in groups are, as Bales and colleagues predicted, fruitful objects of sociological study, and we need not look hard to find them. The entirety of empirical findings in these research programs point to what may be the nearest we come to a sociological law
that task groups stratify. Additionally, as discussed below, group members value higher-status more than lower-status positions in stratified groups.

Status as a Valued Commodity Research on status processes in groups assumes that people tend to value more, rather than less, status in social settings (Sewell, Haller, & Portes, 1969). In the group processes tradition, a growing body of work explains the conditions that lead individuals to value status as a resource (Huberman, Loch, & O¨nc¸u¨ler, 2004; Thye, 2000; Willer, 2009a, 2009b). Thye’s (2000) status value theory of power, for example, proposes that status affects how actors perceive the value of material resources they exchange with their partners. Results from a series of experiments showed that actors perceived material resources they received from high-status partners as more valuable than the same resources from low-status partners. Thus, status differences may distort how actors perceive the value of material resources they exchanged in groups. This distortion, in turn, leads to power advantage in groups for those with higher status. In addition to shaping how individuals perceive material resources, status itself is a valued resource to group members. In an experimental study, Huberman et al. (2004) had participants interact in groups of four to play a two-stage game. In the first stage of the game, each member of the group received a set number of game cards. The members then made independent decisions about how many of their cards they wished to allocate toward winning a lottery to participate in the second stage of the game. The experimenter would take all of the cards allocated by all group members in the first stage and randomly pick one card, with the owner of that card being the winner of the first stage. Then, in the second stage, the experimenter would take all of the cards the first-stage winner “spent” in the first round and replace them with “loser” cards. The experimenter would then randomly draw a card from the stack (the first-stage winner’s remaining original cards and the “loser” cards). If the experimenter drew one of the remaining original game cards, the player received a prize.

224

JEFFREY W. LUCAS ET AL.

The rules of the game were such that in the first stage of the experiment participants had to make a tradeoff between allocating more cards to increase their chances of winning the first stage and allocating fewer of their cards to increase their chances of winning the second stage, if they made it. In other words, the more cards participants allocated in the first stage, the greater their chances of winning that stage, but the lower their subsequent chances of winning the second stage. The experiment had two conditions. In one condition, the winner of the first stage was announced and received public recognition from all other players in the form of applause. In the other condition, the winner of the first stage was not publicly recognized. Huberman et al. (2004) conducted their experiment with undergraduate students as participants in Hong Kong, Turkey, the United States, Germany, and Finland. In the United States, Hong Kong, and Turkey, participants in the public recognition condition invested significantly more resources in the first stage than did participants in the condition without public recognition. Because the only difference between the two conditions was the presence or absence of a status component, the results demonstrated that participants in these cultures valued status independent of monetary consequences. Huberman et al.’s results were noteworthy in indicating the importance of reputational incentives. If people intrinsically value status, then they may sacrifice their material resources to gain status in groups. This is what Willer (2009a, 2009b) proposes in the status value theory of collective action. The theory states that actors want to gain status and will do so by contributing material resources to others during collective tasks. In a series of experiments, Willer (2009a) found that group members who contributed more to the collective good were rewarded with higher status, more influence, more cooperation, and gifts of higher value from fellow group members. Most important for our purposes, actors who received status from fellow group members for their contributions became motivated to give more to the group in future. A good deal of research, then, indicates that status hierarchies are pervasive and that persons tend to value higher- rather than lower-status positions. Thus, we should expect individuals to attempt to identify ways to gain status in groups.

Status in Groups Status in its broadest sense is the standing of an individual relative to others in a group. More narrowly, social status in a work group or organization is

Ability Perceptions and Rule Adherence

225

conceived by researchers as an informal rank based on expectations that group members have for an individual’s contributions to group goals. Weber (1978) proposed that honor, the respect and prestige accorded to a valued member of the community, is an important dimension of social status distinct from power and wealth. Homans (1974) proposed that status is based on an individual’s ability to contribute, in contrast to power that is based on an individual’s ability to extract resources. Most recent work on status in groups relies on status characteristics theory and its formal development explaining how expectations for contributions to group goals create stable status hierarchies (Berger, Fisek, Norman, & Zelditch, 1977; Berger, Rosenholtz, & Zelditch, 1980). Status characteristics theory was developed to explain the common observation about social groups noted above. When people work together, some attain more influence than others over group decisions. They are accorded higher prestige. Not only do such differences emerge among individuals in a group, but those differences soon crystalize into a stable status hierarchy. Status characteristics theory explains how characteristics of people, such as gender and race, produce expectations for performance. These expectations operate outside conscious awareness to alter the expected value and prestige of group members. The expectations associated with individuals’ status characteristics then generate informal yet stable hierarchies commonly observed in task groups (Berger et al., 1977). A status characteristic is a trait or attribute of a person (a) that has at least two states with differential social value, one state expected to be more valuable to group success than at least one other, and (b) associated with each state are expectations for the value of that individual’s contributions to group goals (Berger, Cohen, & Zelditch, 1972; Berger et al., 1977). Status characteristics theory identifies two categories of characteristics that produce performance expectations, specific and diffuse. A specific status characteristic produces initial expectations for ability in a limited sphere of activity such as accreditation in bookkeeping or a recent performance evaluation that then generalize to other settings and tasks. In contrast, a diffuse status characteristic initially produces general expectations. In the United States, for example, gender produces general expectations that men’s contributions will be higher than those of women in many areas of activity (Pugh & Wahrman, 1983). Importantly, these expectations form despite the fact that women in modern organizations perform at least as well as men (Ridgeway, 2001). Four elements characterize a diffuse status characteristic in a given society (Berger & Fisek, 2006): (1) social significance, (2) states of the characteristic (e.g., male
female or young
old)

226

JEFFREY W. LUCAS ET AL.

that can partition the population, (3) differential status evaluations (or levels) for each state in terms of honor, prestige, and general social worth, and (4) high and low conceptions of the general capacities consistent with those status evaluations. Performance expectations then contribute to the relative prestige and expected value of group members, resulting in a prestige hierarchy (Berger et al., 1972, 1977). Status characteristics theory has been refined as a model predicting the combined effects of individuals’ varied status characteristics on expectations for their value to a task group. The expectations produced by both specific and diffuse status characteristics then produce observable behavioral inequalities in groups. Higher-status group members (a) are given more opportunities to perform, (b) perform more, (c) receive higher evaluations for their performances, and (d) are more influential over group decisions than lower-status group members. A stable prestige hierarchy forms with members ranked in terms of their performances, evaluations, and influence. Status characteristics theory shows how the prestige of individuals in a task group results from expectations for their contributions to group goals (Berger & Zelditch, 1985). For example, Anderson, Willer, Kilduff, and Brown (2012) found that some individuals accept lower status because they believe themselves to be less capable of performing in ways that will help the group succeed. Thye (2000) and Podolny (1993) demonstrated the usefulness of conceiving of higher status more broadly in work groups and among organizations as the higher expected value of an individual to the group. Status characteristics such as race and gender affect perceptions of the value of measurably equal contributions. The contributions of higherstatus group members are expected to have more value than similar contributions from lower-status group members. Status characteristics theory has been applied to a variety of settings where the worth of individuals or organizations is evaluated (Cohen & Lotan, 1995; Foschi, Lai, & Sigerson, 1994; Lovaglia, Lucas, Houser, Thye, & Markovsky, 1998; Sauder, Lynn, & Podolny, 2012). In settings where the contributions required for success are uncertain, group members use a variety of characteristics and cues to form expectations for the likely contributions others will make to the group (Berger et al., 1980). Some expectations are based on characteristics, such as race, gender, or even motherhood, perceived to be associated with success or failure in organizations (Correll, Benard, & Paik, 2007). Other cues are related to behavior. Anderson, Brion, Moore, and Kennedy (2012) found that individuals who exuded self-confidence, even overconfidence, attained

Ability Perceptions and Rule Adherence

227

higher status because group members expected them to perform well. Similarly, a record of successful performance in the past can promote expectations for competent contributions in the future. Perhaps the most common way for researchers to generate a specific status characteristic is to report the results of an individual’s performance on a test, even if the test does not measure an ability relevant to group success (Berger et al., 1977; Berger, Norman, Balkwell, & Smith, 1992). Expectations whether derived from specific or diffuse characteristics generalize, contributing to an overall impression of an individual’s value to a work group. Thus, evaluations of performance in prior settings are a source of expectations for the quality of future performances.

Criteria for Determining Competence Status characteristics theory and research models status as a self-sustaining process in which expectations for a group member’s ability to contribute to group tasks alter perceptions of competence on any particular task and subsequent performance evaluations. Performance evaluations then create expectations for ability to contribute. If group members hold low expectations for an individual’s ability to contribute, then that individual’s performances will be devalued, making it difficult for low-status group members to increase their status through competent performance on group tasks. The scope of status characteristics theory is limited to collectively oriented groups working toward a valued goal. We adopt those scope conditions for our explanation, but our explanation (as discussed later) requires additional conditions as well. In particular, the explanation only applies when there are criteria for determining competent contributions to a group other than ability at the task. In a basketball game, for example, the person with the highest level of basketball ability and who makes the most baskets will be seen as contributing to the group goal of winning a game and will thus gain status. However, perhaps an individual who helps the team win by motivating fellow players might also gain status for contributing to the group. In this situation, motivating fellow players serves as an alternative to task ability at demonstrating competence at the group’s task. We propose that adherence to rules and procedures might at times provide an alternative to ability at a group task for demonstrating competence in a task setting. When individuals feel their task performances will not be highly evaluated, whether because of their low status or their inability, they

228

JEFFREY W. LUCAS ET AL.

might be able to demonstrate competence by adherence to rules and enforcement of them. Moreover, by focusing on rule following rather than task success, low-status members can make contributions without directly competing with high-status members less focused on rules.

PROPOSITIONS On the basis of the theoretical development above, we conclude that (1) status hierarchies are pervasive, (2) individuals value high-status positions, (3) high-status positions are based on perceptions of competence, (4) perceptions of high ability at a group’s task are associated with high expectations for task competence, and (5) adherence to rules and guidelines can provide an alternative to task ability for demonstrating competence. From these observations, we develop a set of propositions about relationships between perceptions of ability and adherence to rules, procedure, and tradition. The explanation we present builds on and extends status characteristics theory. Thus, we adopt the scope of SCT: that groups work collectively toward a valued goal. Additionally, however, our explanation requires the following scope conditions. First, there must be ways alternative to task ability for determining the competence of the contributions of group members. Second, there must be rules and procedures that are formally expected to guide the group’s activities. And third, there must be tradition from which members can draw, either previous interactions of the group or a history of processes for similar types of groups. It follows from points (1) through (5) from our theoretical development above that if an individual is low in ability at a group’s task, he or she will likely expect to be in an undesirable low-status position in the status hierarchy that is likely to emerge. Thus, the theory leads us to expect people who perceive they have low ability at a group task to favor the status quo, to resist change: Proposition 1. The less ability individuals perceive that they have in a setting, the more likely they will be to favor the status quo. If persons believe they have low ability in a group setting, they also likely will favor criteria other than task ability in determining competence at the group’s task. Thus, we propose that individuals who perceive they

Ability Perceptions and Rule Adherence

229

have low ability in a setting will more strictly adhere to rules and be more likely to impose rules and guidelines on others: Proposition 2. The less ability individuals perceive that they have in a setting, the more likely they will become to strictly follow specified rules and guidelines for the group task. Proposition 3. The less ability individuals perceive that they have in a setting, the more likely they will be to impose specified rules and guidelines for the task on others in the group. The theory proposes, then, that persons who perceive themselves to be low in ability in a group setting will favor rules because adherence to rules provides an alternative method to task ability in demonstrating task competence and subsequently attaining high-status positions. They will also be likely to impose rules on others because they favor rules to task ability in judging competence. And, we expect that these persons who perceive themselves to be low in ability will favor tradition because any changes to the group structure will be likely to result in relatively low status for them. We describe below an experimental test of one element of the theory.

METHODS We designed an experimental study to test one aspect of the proposed theory, the one most closely related to the original experience with the campus committee that gave rise to the theory: that the less ability people perceive that they have in a setting, the more likely they will be to impose rules on others. In the study, we led participants to believe that they had high or low ability at a group’s task before they worked in a group (in fact, fictitious) in which they imagined they were members of a compensation committee in an organization. Prior to participants arriving for the study, Research Assistants randomly assigned them to be in one of two conditions: high ability or low ability. When participants arrived, Research Assistants sat them alone in rooms with computer terminals. Instructions informed participants that they would be working with participants on computers in other rooms on a compensation committee task. Participants and their partners would act as the compensation committee in a fictitious organization, determining how to divide a raise pool among several employees nominated for raises by their supervisors.

230

JEFFREY W. LUCAS ET AL.

After receiving instructions on the study and filling out a demographic questionnaire, participants learned that they would take a test of personal response style. Instructions said: After reading instructions for the study, the first thing you will do today is take an assessment of what is called personal response style. We are giving the assessment for the purpose of developing diverse work groups of the types found in work organizations. Personal response style is a widely researched concept that reflects ways in which people respond to different situations. After dozens of studies on the concept and instruments used to assess it, researchers have identified two general personal response styles that in the research literature are typically labeled S2 and Q2. All groups in the study today will be comprised of a mix of S2s and Q2s. There are no right or wrong answers in the assessment you will take—it’s simply an instrument designed to assess whether you more closely classify as an S2 or Q2 in personal response style.

The world is in fact not divided into two categories on personal response style, and we used the information above and the instrument we created to assign participants to experimental conditions. We drew heavily from Ridgeway and Erickson (2000) in constructing this manipulation, including in the use of the categories S2 and Q2. Participants then began the personal response style assessment test. The test comprised 10 items that described various scenarios in organizations and asked participants to select among multiple-choice options on how they would respond to the scenarios. For example, one item asked the best course of action for a manager to take who plans to institute several major changes in an organization. After completing the test, participants received a form describing results of their personal response style assessment. By random assignment, they learned (depending on condition) that test results indicated that they were either S2s or Q2s. Participants then began the compensation committee task. Instructions asked participants to imagine that they were members of a compensation committee tasked with determining how to distribute a raise pool of $12,000 among four employees nominated for raises by their supervisors. There were rules specifying which employees were eligible for raises
most notably that supervisor nominations were due by a certain date and that employees must have been at the organization for 12 months before being eligible for a raise. The compensation committee also had rules that no one employee could receive a raise larger than $7,500 and that they must distribute the full $12,000.

Ability Perceptions and Rule Adherence

231

For each of four employees nominated for a raise, participants received detailed information, including time at the company, salary, experience, salary relative to other employees with similar responsibilities, dates and amounts of last raises, and other information. Additionally, each nominee package was accompanied by a letter from a supervisor recommending a raise for the nominee.1 Participants evaluated four nominees for raises in the following order: Nominee 1: A competent employee who was worthy of and eligible for a raise. Nominee 2: A competent employee who was exceptionally deserving of a raise and that the company badly needed to keep in the face of an outside offer. The employee’s supervisor, however, submitted the nomination letter to the compensation committee 11 days late. The lateness of the nomination was indicated by a date on the letterhead for the nomination 11 days past the committee deadline and an opening sentence that read “Although this nomination is being submitted beyond the nomination deadline, I hope that the committee strongly considers a merit raise for….” Nominee 3: An employee who was eligible for a raise but also not reasonably deserving of a raise. In the nomination letter, the employee’s supervisor noted that the employee was generally incompetent but expressed the hope that perhaps a raise would motivate the nominee to work harder. Nominee 4: A competent employee who was technically ineligible for a raise. The employee had only been at the company nine months, and guidelines required that employees had worked at the company for 12 months before being eligible for a raise from the compensation committee. The nominator noted this in the recommendation letter, saying in part “Normally, I would oppose this type of action (as my record shows) if only because it is contrary to organizational policy. However, she has matured in her job so rapidly that she truly deserves a raise.” The nomination goes on to note that relationships with key clients are dependent on the rapport the nominee has with them. Thus, the second and fourth nominees were technically ineligible for raises, the former because of a late letter from a recommender and the latter because of a tenure at the company that was too short. The first

232

JEFFREY W. LUCAS ET AL.

and fourth nominees were women (by first name
no explicit information was given on gender), the second and third men. All nominators had male names. For both nominators and nominees, we attempted to select common names unlikely to communicate information about race and/or social class. After receiving information on the employees nominated for raises, participants determined how much of the $12,000 raise pool they felt each nominee should receive. They then received reports indicating the raise amounts that each of the three other persons in their group recommended. The partners were fictitious, and all participants received the same reports from partners. Participants then evaluated their partners’ selections and made final raise recommendations.

Independent Variable The independent variable in the study was our manipulation of participants’ perceptions of their ability at the group task. After we told participants that they were S2s or Q2s, we gave them information indicating that S2s are particularly good at the task being carried out in the study. These instructions in part said: The compensation committee task you will be completing with your group today is one that involves making complex, collaborative decisions with sometimes ambiguous information. We designed it carefully to be a task that is particularly well-suited to persons in the S2 category on personal response style. It is a type of task on which persons in the S2 category of personal response style do very well. Persons who are Q2s on personal response style, in contrast, tend to do poorly on this type of task. Again, however, neither category of personal response style is better on tasks in general—we simply designed a task to be well-suited to the S2 style.

Thus, we led participants randomly assigned to be S2s to believe they had high ability at the group task and participants randomly assigned to be Q2s to believe they had low ability at the group task.

Dependent Variable Our dependent measure was initial raise determinations by participants. In particular, we were concerned with the amount of raises that participants gave to nominees who were technically ineligible for raises. Consistent with Proposition 3, we predicted that participants who perceived they were low

Ability Perceptions and Rule Adherence

233

in ability would be more likely to impose rules on others. More specifically, we predicted: Hypothesis. Participants in the low-ability condition will give lower raises on average to nominees technically ineligible for raises than participants in the high-ability condition. We describe results of our data collection later.

RESULTS Participants were undergraduate students at a large Mid-Atlantic university. We recruited 106 participants for the study. We rejected data from 23 participants because of excessive suspicion about the purposes of the study. In almost all cases, the suspicion resulted from participants either not believing that S2 and Q2 really represent categories of personal response style and/or that the task was better suited to S2s than to Q2s. Our rejection rate of 21.70% is high, and future studies with a more successful manipulation of ability perceptions would give us more confidence in the results we report here. Among the 83 participants from whom we kept data, 38 were in the high-ability condition and 45 were in the lowability condition; 39 (46.99%) participants were men and 44 (53.01%) were women; 48 (57.83%) participants were identified as White, 35 (42.17%) indicated either other categories of race/ethnicity or multiple categories. Participants overwhelmingly followed the rules for distribution of raises
that no employees could receive raises larger than $7,500 and that the entire raise pool must be distributed. Additionally, participants changed very little in the amounts they recommended between their initial and final sets of recommendations, and we could identify no pattern indicating that participants in one condition were significantly more likely to modify their raise determinations than participants in the other condition. Thus, we only report here results on initial raise determinations, which was our primary dependent variable of interest. We predicted that participants in the low-ability condition would recommend significantly smaller raises to technically ineligible nominees than would participants in the high-ability condition. Table 1 displays mean raise recommendations by participants’ condition and type of nominee, as well as results of t-tests. The first row of the table shows that participants in the high-ability condition on average gave $3,566 (SD = 1,747) of the

234

JEFFREY W. LUCAS ET AL.

Table 1.

Mean Compensation Awarded by Condition.

Condition

Eligible nominee (competent) Eligible (incompetent) Ineligible (materials late) Ineligible (tenure too short) Ineligible nominees combined

High Ability N = 38 Mean (SD)

Low Ability N = 45 Mean (SD)

t

p

$3,566 (1,747) $761 (1,218) $4,934 (2,572) $2,384 (1,910) $7,318 (3,199)

$4,058 (2,197) $1,844 (2,393) $3,780 (3,063) $1,700 (1,851) $5,480 (3,600)

1.11 2.53 1.84 1.65 2.44

0.259 (two-tailed) 0.010* (two-tailed) 0.033* (one-tailed) 0.052 (one-tailed) 0.008** (one-tailed)

* = p < 0.05, ** = p < 0.01. We did not make predictions about differences in raises for employees eligible for raises, and we thus report two-tailed p values for these differences. We report one-tailed values for the predicted differences in raises given to employees technically ineligible for raises.

raise pool to the competent employee eligible for a raise, compared to participants in the low-ability condition who on average gave $4,058 (SD = 2,197). The second row shows results for the employee who was eligible for a raise but who was described by his boss as incompetent. The third and fourth rows of the table are the subject of our predictions
employees technically ineligible for raises. The final row combines the amounts that participants recommended in raises for the two technically ineligible employees. We predicted that participants in the low-ability condition would give lower raises to nominees ineligible for raises than would participants in the high-ability condition. For the employee ineligible for a raise because the nomination letter came in late, participants in the high-ability condition on average recommended raises of $4,934 (SD = 2,572), compared to participants in the low-ability condition who on average recommended $3,780 (SD = 3,063). This difference is consistent with our prediction and significant (t = 1.84, one-tailed p = 0.033). For the employee ineligible for a raise because of a tenure with the organization shorter than the designated minimum period, participants in the high-ability condition on average recommended a raise of $2,384 (SD = 1,910), compared to $1,700 (SD = 1,851) in the low-ability condition. This difference is in the predicted direction but not significant (t = 1.65, one-tailed p = 0.052). Combining average raise amounts given to the two technically ineligible employees produces a difference between high- and low-ability conditions of $7,318 (SD = 3,199) to $5,480 (SD = 3,600), a difference that is significant (t = 2.44, one-tailed p = 0.008).

Ability Perceptions and Rule Adherence

235

We conducted regression analyses that controlled for race and gender of participants in examining relationships between experimental condition and raise amounts given to ineligible nominees. Coefficients for gender and race did not approach significance in any regression analyses. Coefficients for condition were, however, significant: in the model predicting raises to the employee whose nomination came in late, condition was significant at t = 1.78, one-tailed p = 0.039; in the model predicting raises to employees who had too short a tenure, condition was significant at t = 1.764, onetailed p = 0.041; and in the data combining raises to ineligible employees, condition was significant at t = 2.44, one-tailed p = 0.009. In addition to average raise determinations, we compared the proportion of participants in each condition who denied raises altogether (i.e., recommended a raise of $0) to the nominees who were technically ineligible for raises. Six of 38 participants (15.8%) in the high-ability condition denied raises to the employee whose nomination letter came in late, compared to 15 of 45 participants (33.33%) in the low-ability condition. This difference is not significant (chi-square = 3.355, p = 0.067). Eleven of 38 participants (28.9%) in the high-ability condition denied raises to the employee who had too short a tenure in the organization, compared to 20 of 45 (44.44%) in the low-ability condition. This difference is also not significant (chi-square = 2.115, p = 0.146). In all, 12 of 38 participants (31.58%) in the high-ability condition denied raises to either or both of the ineligible employees, compared to 25 of 45 participants (55.56%) in the low-ability condition. This difference is significant (chi-square = 4.794, p = 0.029).

DISCUSSION Results provided generally strong support for our prediction that participants who perceived they had low ability at the group task would be more likely to impose rules on others than participants who perceived they had high ability. In t-test results, participants in the low-ability condition gave significantly lower raises to one of the two nominees technically ineligible for a raise than did participants in the high-ability condition. In regression analyses that controlled for race and gender of participants, participants in the low-ability condition gave significantly lower raises to both ineligible employees than did participants in the high-ability condition. In analyses of the proportion of participants in each condition who denied raises altogether to ineligible nominees, results show that significantly more

236

JEFFREY W. LUCAS ET AL.

participants in the low-ability condition denied a raise to either or both of the ineligible employees than did participants in the high-ability condition. A flaw in our design limits our ability to compare mean raise determinations across the types of nominees. Because we told participants that they were required to distribute the entire $12,000 raise pool, decisions about how much they gave each nominee were dependent on their raise decisions for other nominees. For example, participants in the low-ability condition gave significantly larger raises to the incompetent employee eligible for a raise than did participants in the high-ability condition. This result is difficult to interpret. It may be that participants in the low-ability condition believed that these nominees deserved larger raises than did participants in the high-ability condition. It also might be that participants in the low-ability condition had remaining raise money they had to spend after denying raises to the ineligible employees and giving the maximum to the competent and eligible employee, and so they gave relatively larger raises to the incompetent employees. In retrospect, our design would have been improved had participants made independent raise determinations for each nominee.

CONCLUSION If rules are in fact seen largely as the domain of those low in ability, and if conversely a reputation of high ability earns the “privilege” of operating outside the rules, the consequences can be substantial, even fatal. In 1994, Air Force Commander Bud Holland crashed a B-52 aircraft while engaged in practice maneuvers for an air show. Lt Col Holland had a reputation as an exceptionally skilled B-52 pilot. Comments from his superiors included that he was “as good as a B-52 aviator as I have seen” and “probably the best B-52 pilot that I know in the wing and probably one of the best, if not the best within the command” (Kern, 1995, p. 7). In retrospect, it appears that these perceptions of Lt Col Holland’s competence may have allowed him to operate outside the rules, routinely breaking regulations for the B-52s he flew; the allowances he was accorded even extended to his regularly parking in the center of a red-curbed “No Parking” zone outside the wing headquarters building. As much as Lt Col Holland’s superiors thought of his flying ability, junior crewmembers who were required to fly with him painted a dramatically different picture. Their comments included “I was thinking that he was going to try something again … at this airshow and possibly kill

Ability Perceptions and Rule Adherence

237

thousands of people,” “I’m not going to fly with him … he’s going to kill somebody someday and it’s not going to be me,” and “Lt Col Holland broke the regulations or exceeded the limits … virtually every time he flew” (Kern, 1995, p. 7). Lt Col Holland’s peers and subordinates made a number of efforts to have him grounded for numerous and flagrant violations of air discipline, but their efforts were repeatedly met with deaf ears among those who had the power to ground Lt Col Holland. Ultimately, on June 24, 1994, Lt Col Holland planned and briefed an air show flight profile through the Wing Commander level that grossly exceeded regulations. During practice maneuvers, Lt Col Holland began a tight 360 turn around the control tower at only 250 feet of altitude, the aircraft banked past 90°, stalled, and crashed, killing all four crewmembers. The flight plan was egregious to the point that the nurse manager in emergency services for the air show said in advance of the flight “I just hope he crashes on Friday, not Sunday, so I will not have so many bodies to pick up” (Kern, 1995). We cannot recreate the conditions that led to the crash of Lt Col Holland’s B-52, and so we cannot conclusively determine what factors contributed to the leadership oversight in not grounding Lt Col Holland before the tragic air disaster occurred. In retrospect, however, it seems that Lt Col Holland built such a high reputation of ability that it allowed him to be viewed by many as not subject to the same rules and regulations as others. In a case study of the event, Tony Kern (1995) noted the “breaking the rules that became a pattern in Lt Col Holland’s flying activities” and said “Many aviators report that rules and regulations are ‘bent’ on occasion, and some individuals seem to be ‘Teflon coated’ …” (p. 4). We propose that status processes lead to a relationship between perceptions of ability and adherence to rules, procedure, and tradition. In particular, the theory we present here proposes that persons who perceive themselves to be low in ability will become more likely to strictly adhere to rules and procedures, to impose rules on others, and to favor tradition. Our experimental study provided strong support for one element of the proposed theory: we found that persons who perceived themselves to be low in ability became significantly more likely to impose rules, including an arguably trivial rule (whether the nomination letter arrived on time), on others than did persons who perceived themselves to be high in ability. We see a number of potential theoretical and practical implications of this work. For example, research finds that individuals are motivated to contribute to public goods in order to enhance their individual status (Willer, 2009a). The proposed explanation and the tentative support for it

238

JEFFREY W. LUCAS ET AL.

we provide here indicate that enforcing rules (both by following them or monitoring others’ compliance) might be seen by those with low ability as a contribution to the group. This might help explain collective enforcement against free riding. In other words, perhaps not only are people motivated to contribute to collective action to increase their status, but also people with low ability are motivated to enforce norms against free riding to compensate for their perceived lack of ability to contribute. Although our experiment was not designed to test relationship between outcomes of concern in status characteristics theory, the explanation we propose might also have the ability to extend work in the theory to situations in which means of demonstrating competence other than task ability, such as adherence to rules, are available to group members. If group members favor evaluation criteria that best match their own relative contributions, it could have interesting implications for modeling status processes in groups. Practically, the explanation we present can potentially have implications for work groups in a variety of contexts. In response to the 1919 Boston police strike, the city fired nearly 1,000 police officers and hired about 1,500 replacement workers from a pool of unemployed World War I veterans. In subsequent generations of Boston police, the replacement police came to have a reputation for low competence as well as for strict adherence to tradition and protocol. We cannot know whether this response was driven by concerns about the legitimacy of their positions, responses to others’ questioning their competence, their own concerns about their competence, or something else, but the behavior is consistent with the explanation we present here, and our explanation might help explain the conditions under which individuals become strictly attached to rules and procedures in a variety of contexts. Future research would be valuable in testing additional elements of the proposed theory. For example, research might assign participants to high and low perceived ability conditions and then give them a set of rules to follow when completing a task, some reasonable and some unreasonable. The theory would predict that participants in the low-ability condition would be more likely to follow rules, even when they were unreasonable. Research also might assign participants to low- and high-ability conditions before measuring adherence to tradition or willingness to embrace change. As we noted above, our manipulation of ability perceptions generated relatively high levels of suspicion, and research also would be valuable in introducing alternative methods of manipulating perceptions of ability.

Ability Perceptions and Rule Adherence

239

NOTE 1. We thank Cameron Anderson for making available to the lead author a compensation committee task he developed, a heavily edited version of which was used in the study.

REFERENCES Anderson, C., Brion, S., Moore, D. A., & Kennedy, J. A. (2012). A status-enhancement account of overconfidence. Journal of Personality and Social Psychology, 108, 718
735. Anderson, C., Willer, R., Kilduff, G. J., & Brown, C. E. (2012). The origins of deference: When do people prefer lower status? Journal of Personality and Social Psychology, 102, 1077
1088. Bales, R. F., Strodtbeck, F. L., Mills, T. M., & Roseborough, M. E. (1951). Channels of communication in small groups. American Sociological Review, 16, 461
468. Berger, J., Cohen, B. P., & Zelditch, M., Jr. (1966). Status characteristics and expectation states. In J. Berger, M. Zelditch, Jr., & B. Anderson (Eds.), Sociological theories in progress (Vol. 1, pp. 29
46). Boston, MA: Houghton Mifflin. Berger, J., Cohen, B. P., & Zelditch, M., Jr. (1972). Status characteristics and social interaction. American Sociological Review, 37, 241
255. Berger, J., & Fisek, M. H. (2006). Diffuse status characteristics and the spread of status value: A formal theory. American Journal of Sociology, 111, 1038–1079. Berger, J., Fisek, M. H., Norman, R. Z., & Zelditch, M., Jr. (1977). Status characteristics and social interaction: An expectation states approach. New York, NY: Elsevier. Berger, J., Norman, R. Z., Balkwell, J. W., & Smith, R. F. (1992). Status inconsistency in task situations: A test of four status processing principles. American Sociological Review, 57, 843
855. Berger, J., Rosenholtz, S. J., & Zelditch, M., Jr. (1980). Status organizing processes. Annual Review of Sociology, 6, 479
508. Berger, J., & Zelditch, M., Jr. (1985). Status, rewards, and influence: How expectations organize behavior. San Francisco, CA: Jossey-Bass. Burke, P. (2003). Interaction in small groups. In J. Delamater (Ed.), Handbook of social psychology (pp. 363
388). New York, NY: Springer. Cohen, E. G., & Lotan, R. A. (1995). Producing equal-status interaction in the heterogeneous classroom. American Educational Research Journal, 32, 99
120. Correll, S. J., Benard, S., & Paik, I. (2007). Getting a job: Is there a motherhood penalty? American Journal of Sociology, 112, 1297
1339. Fisek, M. H., Berger, J., & Norman, R. Z. (1991). Participation in heterogeneous and homogeneous groups: A theoretical integration. American Journal of Sociology, 97, 114
142. Foschi, M., Lai, L., & Sigerson, K. (1994). Gender and double standards in the assessment of job applicants. Social Psychology Quarterly, 57, 326
339. Goetsch, G. G., & McFarland, D. D. (1980). Models of the distribution of acts in small discussion groups. Social Psychology Quarterly, 43, 173
183.

240

JEFFREY W. LUCAS ET AL.

Homans, G. C. (1974). Social behavior: Its elementary forms. Oxford, England: Harcourt Brace Jovanovich. Huberman, B. A., Loch, C. H., & O¨nc¸u¨ler, A. (2004). Status as a valued resource. Social Psychology Quarterly, 67, 103
114. Kern, A. (1995). Darker shades of blue: A case study of failed leadership. In L200, Developing leaders and organizations (pp. 321
342). CGSC/Department of Command and Leadership. Lovaglia, M. J. (1995). Power and status: Exchange, attribution, and expectation states. Small Group Research, 26, 400
426. Lovaglia, M. J., Lucas, J. W., Houser, J. A., Thye, S. R., & Markovsky, B. (1998). Status processes and mental ability test scores. American Journal of Sociology, 104, 195
228. Podolny, J. M. (1993). A status-based model of market competition. American Journal of Sociology, 98, 829
872. Pugh, M. D., & Wahrman, R. (1983). Neutralizing sexism in mixed-sex groups: Do women have to be better than men? American Journal of Sociology, 88, 746
762. Ridgeway, C. (1991). The social construction of status value: Gender and other nominal characteristics. Social Forces, 70, 367
386. Ridgeway, C. (2006). Status construction theory. In P. Burke (Ed.), Contemporary social psychological theories (pp. 301
323). Stanford, CA: Stanford University Press. Ridgeway, C., & Berger, J. (1986). Expectations, legitimation, and dominance behavior in task groups. American Sociological Review, 51, 603
617. Ridgeway, C., & Erickson, K. G. (2000). Creating and spreading status beliefs. American Journal of Sociology, 106, 579
615. Ridgeway, C. L. (2001). Gender, status, and leadership. Journal of Social Issues, 57, 637
655. Sauder, M., Lynn, F., & Podolny, J. M. (2012). Status: Insights from organizational sociology. Annual Review of Sociology, 38, 267
283. Sewell, W. H., Haller, A. O., & Portes, A. (1969). The educational and early occupational attainment process. American Sociological Review, 34, 82
92. Skvoretz, J. (1988). Models of participation in status-differentiated groups. Social Psychology Quarterly, 51, 43
57. Skvoretz, J., & Fararo, T. J. (1996). Status and participation in task groups: A dynamic network model. American Journal of Sociology, 101, 1366
1414. Thye, S. R. (2000). A status value theory of power in exchange relations. American Sociological Review, 65, 407
432. Weber, M. (1978). Economy and society. Berkeley, CA: University of California Press. Willer, R. (2009a). Groups reward individual sacrifice: The status solution to the collective action problem. American Sociological Review, 74, 23
43. Willer, R. (2009b). A status theory of collective action. In. S. R. Thye & E. J. Lawler (Eds.), Advances in group processes (Vol. 26, pp. 133
163). London: Emerald. Zelditch, M., Jr. (2001). Processes of legitimation: Recent developments and new directions. Social Psychology Quarterly, 64, 4
17.

REFERENT NETWORKS AND DISTRIBUTIVE JUSTICE David Melamed, Hyomin Park, Jingwen Zhong and Yue Liu ABSTRACT Purpose
This study examines how the structure of referent networks, or the social network defined by knowing others’ reward levels, affects perceptions of distributive justice. The homogeneity of rewards in the referent network, the amount of inequality in the referent network, and an individual’s reward level are all associated with distributive justice perceptions. Several moderating relationships are also examined. Methodology/approach
We relied on data from a controlled laboratory experiment to test a series of theoretically derived hypotheses. Findings
The study shows that several aspects about the structure of the referent network are important for shaping perceptions of distributive justice. Specifically, the reward heterogeneity and amount of inequality in the network are found to be negatively associated with distributive justice, while reward levels are found to be positively associated with distributive justice. Furthermore, the effect of reward levels on distributive justice is moderated by both (i) the presence of a referential standard for rewards and (ii) the amount of inequality in the network.

Advances in Group Processes, Volume 31, 241
262 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0882-6145/doi:10.1108/S0882-614520140000031006

241

242

DAVID MELAMED ET AL.

Research limitations/implications
While being among the first studies to demonstrate effects of referent networks on perceptions of fairness, it is unclear how group memberships combine with referent network effects and which factors may blur these relationships in uncontrolled environments. Subsequent scholarship on the effect of referent networks on justice perceptions should leverage multiple data sources. Originality/value of chapter
Research on the effects of referents on justice perceptions has focused on particular referent individuals. We recast this issue in terms of referent networks, which highlights the empirical reality that individuals have a variety of sources or alters which could operate as referents. Keywords: Inequality; distributive justice; networks; referent network; referent others

When individuals evaluate the distributive justice of rewards, they may rely on referent others, reference groups, or referential standards (Berger, Zelditch, Anderson, & Cohen, 1972; Buckingham & Alicke, 2002; Goodman, 1974). Referent others are specific others that serve as a basis for social comparison, reference groups are a set of individuals with common attributes that may be used for comparison purposes, and referential standards are normative beliefs about expected rewards. Organizational psychologists have established that referent choice is an important determinant of pay satisfaction and concomitant perceptions of fairness (e.g., Berkowitz, Fraser, Treasure, & Cochran, 1987; Goodman, 1974; Kulik & Ambrose, 1992). This literature, however, tends to focus on particular referents. Focusing attention on specific referents may induce what Markovsky (1988) calls an “anchor.” Anchors provide baselines that may bias perceptions. If, for example, a referent other is used with a very low reward level, the fairness of a focal individual’s reward is likely to seem fairer, relative to the under-rewarded referent. If the referent other is highly rewarded, the same individual’s reward level is likely to be regarded as less fair. By focusing attention on specific referents, social comparisons and perceptions of fairness may be adversely affected by anchoring. Rather than focus on particular referent others, the present research examines how referent networks shape perceptions of fairness. Our take on the role of referents is particularly relational (see also Gartrell, 1985).

Referent Networks and Distributive Justice

243

Rather than ripping individuals from the social structures in which they are embedded (Mayhew, 1980), we leverage those structures, as conceptualized in terms of a social network, to examine how these referent networks shape perceptions of justice. The referent network is an asymmetrical network based on the relation “knows the reward level of.” People know, or at least think they know,1 the reward levels of a variety of individuals, including colleagues, supervisors, neighbors, family members, etc. The members of the referent network may be drawn from a variety of reference groups (e.g., coworkers, neighbors, or individuals with similar attributes). Presently we are interested in only a few structural aspects of the referent network, but subsequent research should investigate how a variety of reference groups get aggregated into a referent network. Are some reference groups (e.g., coworkers) more influential than others (e.g., friends) in evaluations of distributive justice? While groups are known distinctions that shape favoritism, etc. (Tajfel & Turner, 1986), it is unclear how the intersecting social circles (Simmel, 1955) in which a person is embedded combine to affect these evaluations. Determining how reference groups combine in referent networks to shape evaluations of distributive justice will require substantial theoretical and empirical work. The task of this chapter is much simpler
to establish how and why particular structural features of the referent network are related to evaluations of distributive justice. Specifically, in this study we examine how reward levels within referent networks, the amount of reward homogeneity within the networks, the amount of inequality in the networks, and the presence of a referential standard for rewards shapes perceptions of distributive justice. In what follows, we review relevant aspects of the social psychological literature on distributive justice. We then elaborate a series of arguments corresponding to how features of the referent network may be related to perceptions of distributive justice. We test our arguments using data from a controlled laboratory study that manipulated aspects of referent networks while participants evaluated reward information. Our results illustrate that the reward heterogeneity and amount of inequality in the referent network are found to be negatively associated with distributive justice, while reward levels are found to be positively associated with distributive justice. Furthermore, the effect of reward levels on distributive justice is moderated by both (i) the presence of a referential standard for rewards and (ii) the amount of inequality in the network. We conclude with some implications of the work and with directions for further research that overcome some present shortcomings.

244

DAVID MELAMED ET AL.

DISTRIBUTIVE JUSTICE Generally speaking, distributive justice refers to the application of a normative rule to the allocation of benefits to recipients (Hegtvedt, 2006). Early on, research on distributive justice focused on equity as a normative distribution rule (e.g., Adams, 1965; Homans, 1961). As Homans (1961, p. 244) noted, “a person in an exchange relation with another will expect the profits of each to be directly proportional to his or her investments … If the investments of the two … are equal, their profits should be equal, and if their investments are unequal, the one with the greater investment should get the greater profit.” Adams (1965) formalized this conception of equity and extended it by developing a typology of means to alleviating inequity. If two individuals’ input-to-outcome ratio is equal, it is an equitable situation according to Homans (1961) and Adams (1965). As Berger et al. (1972) and Berger, Fisek, Norman, and Wagner (1998) noted, this permits collective states of injustice: both individuals may be either over- or underrewarded compared to some referential standard, such as average wages for a given occupation. To account for this, Berger and colleagues incorporated referential standards or “referential structures” into their conception of equity. They define referential structures as socially validated beliefs that are held in common by actors. To the extent that the distribution of resources violates these referential structures, a state of inequity exists, and restorative behavior is in order. While Berger et al. (1998) Reward Expectations Theory argues that status-valued attributes result in differential reward anticipations, research in non-collectively and task-oriented groups find little evidence of attributes shaping perceptions of fairness (cf. Melamed, 2012a). As Hegtvedt (2006) noted, age and gender “hardly specify what is likely to be considered just” (see also, Hegtvedt & Cook, 2001). Some research finds that women prioritize procedural justice, while men prioritize distributive justice (Belliveau, 2012), but this difference may be a status difference rather than an essentialist difference between the sexes (Clay-Warner, Culatta, & James, 2013). While individual attributes have shown to be ineffective predictors of perceptions of fairness, situational or context effects have been noted with respect to distributive justice. Friends prefer equal distributions, while strangers prefer equitable ones (Hegtvedt & Cook, 2001). Further, being valued by the group in which the distribution occurs decreases perceptions of injustice (Hegtvedt, Clay-Warner, & Johnson, 2003). What is more, distributions of rewards that are deemed legitimate are also viewed as fairer (Hegtvedt & Johnson, 2000; Melamed, 2012b).

Referent Networks and Distributive Justice

245

Beyond the specific alter used by Homans (1961) and Adams (1965), research in organizational/industrial psychology finds mixed results about the effects of referent others on pay satisfaction and perceptions of justice. While most of this literature finds a positive effect of the equity of comparisons between an individual and referent others on pay satisfaction and perceptions of fairness (e.g., Buchanan, 2008; Goodman, 1974; Law & Wong, 1998; Major & Forcey, 1985), Berkowitz et al. (1987) did not find an association between referent others and pay satisfaction. Although Goodman’s study did not define a referent network, it is one of the few studies to examine the differential impacts of multiple referents. Even here, however, only one reference group is used (coworkers) and the overall structural relationship between the respondent and the referents was unexplored.

Distributive Justice and Referent Networks This section presents our arguments linking the structural aspects of the referent network to perceptions of distributive justice. As noted above, many factors have been found to be related to distributive justice. Individual beliefs (see Hegtvedt, 2006), aspects of the context, and referent others each contribute to perceptions of justice. Referent others are the only aspect that is necessarily relational, but the manner with which referent others have been studied has neglected the larger context in which these referent others are situated. By focusing on the referent network, as opposed to specific referent others,2 we can concentrate on how variations in structure might impact perceptions of distributive justice. One structural feature of an individual’s referent network is the amount of reward homogeneity within the network. Referent networks wherein alters have similar reward levels to ego are reward homogeneous. Referent networks where alters have dissimilar reward levels to ego are reward heterogeneous. Homogeneous referent networks promote similarity of social comparisons. Similarity of social comparisons, in turn, should lead to perceiving the distribution of rewards as more fair. Put differently, reward homogeneous referent networks should be perceived as fairer than reward heterogeneous referent networks. By minimizing social comparisons, homogeneous referent networks are more egalitarian, and recent evolutionary theorizing indicates that humans (and other primates)3 are predisposed to be averse to inequality (Dawes, Fowler, Johnson,

246

DAVID MELAMED ET AL.

McElreath, & Smirnov, 2007; Fowler, Johnson, & Smirnov, 2005). We therefore predict the following: H1. Reward homogeneity in referent networks will be positively related to distributive justice. The above mentioned inequality aversion in humans also implies that inequality within referent networks will be negatively associated with perceptions of distributive justice. The inequality of the referent network is not the same thing as the reward heterogeneity of the network. Reward heterogeneity is from the perspective of a particular node within the referent network, whereas the amount of inequality within the network is a global property that may be perceived by each node in the network (Chiang, 2011). Consider an example referent network based on incomes with ego and four alters. Suppose ego makes $40,000 per year, while each alter makes $200,000 per year. From ego’s perspective, everyone else in the network makes a lot more money than he or she does. Looking at the network as a whole, however, four of the nodes have the exact same reward level, while only one of them is “under-rewarded.” The network is relatively reward homogeneous from the perspective of any alter. In principle, inequality within the referent network can range from one node earning one resource more than all other nodes, to one node earning all of the resources and none of others earning any. Irrespective of reward levels, we expect that as inequality increases in the referent network, perceptions of distributive justice will decrease. Our second hypothesis: H2. Inequality in referent networks will be negatively related to distributive justice. Referent networks entail the micro-structural aspects that contribute to perceptions of fairness. They encompass the interpersonal associations at the local level that Homans (1961) and Adams (1965) described. We know from reward expectations theory (Berger et al., 1972) and empirical research (Shepelak & Alwin, 1986) that referential standards also shape perceptions of fairness (see also, To¨rnblom, 1977). While referential standards may be leveraged to aid in determining how one should respond to a given reward, we argue that where an individual stands in terms of resources with respect to the referential standard matters most. Specifically, when a referential standard is known, individuals within referent networks that are rewarded at levels below the referential standard will perceive the distribution of rewards as less fair than individuals within referent networks with an unknown referential standard. When the referential standard

Referent Networks and Distributive Justice

247

is known, it makes salient the fact that they are under-rewarded, despite the referent network. Thus we expect the following: H3. For individuals with reward levels below the referential standard, the presence of a referential standard for rewards will decrease distributive justice. We have no prediction about how the referential standard will affect individuals above the referential standard. We leave that as an exploratory issue. On the subject of reward levels, it stands to reason that individuals at the bottom of the reward distribution within referent networks will be less sensitive to the overall inequality in their network than individuals at the top of the reward distribution. For those individuals at the bottom, the distribution of rewards is already unfair, no matter how unequal it is. For those individuals at the top, however, less inequality means more rewards for everyone since they are already well-off relative to the referent network. Given this, we expect the negative effect of the amount of inequality within referent networks on perceptions of distributive justice to increase as reward levels increase. We predict the following moderating relationship between inequality within the referent network and reward levels: H4. The negative effect of inequality on distributive justice in referent networks will be exacerbated by increasing reward levels. As we made parallel predictions about inequality within the referent network and reward heterogeneity within the referent network, so to do we make parallel moderating predictions. Specifically, within more reward homogeneous referent networks, we expect increasing reward levels to increase perceptions of distributive justice. Insofar as the overall distribution of rewards within the referent network is fair (homogeneous networks), then it will be fairer for ego to receive more relative to the rest of the network. When the overall distribution of rewards within the referent network is less fair (heterogeneous networks), then it will be less fair for ego to receive more relative to the rest of the network. This leads us to our final hypothesis: H5. The negative effect of reward heterogeneity on distributive justice in referent networks will be exacerbated by increasing reward levels. To evaluate the above hypotheses, we designed a controlled laboratory experiment. Experimentation is ideal for evaluating our hypotheses for several reasons. First, we may construct situations that instantiate the key concepts of referent networks, reward homogeneity, and referential standards for rewards. In natural settings with uncontrolled observations, it

248

DAVID MELAMED ET AL.

would be much more difficult to have a reference group without a referential standard for rewards. In this case, the artificiality of experimentation is indeed a strength. Second, we can abstractly investigate key dimensions of network structure that may otherwise be confounded with individual attributes. As Burt (1995) notes, attributes and positions tend to be conflated in observational studies of networks. Finally, the aim of this research is not to mimic any naturally occurring situation, but rather to create procedures and manipulations of variables that operationalize key dimensions of the theoretical and formal concepts (Zelditch, 1969). The test and evaluation of the theoretical argument can then inform understanding of events in natural settings.

METHOD4 Design and Subjects The experiment was a 12 × 2 × 3 mixed experimental design crossing reward levels within a network, the presence or absence of a referential standard for rewards, and the amount of reward homogeneity in the network. The first two factors were between-subjects manipulations and the third factor was a within-subjects manipulation. Participants were randomly assigned to a reward level, and the presence or absence of a referential standard for income, and then asked to assess their reward level, given three different sets of network connections, which were chosen to vary the amount of reward homogeneity within the overall network. In all conditions, participants were led to believe that they were to imagine having an occupation and that they were assigned to imagine being an “archiator” (i.e., a physician). All participants were asked to imagine being an archiator because it was presumed that participants would not know what that occupation is, and would therefore not have a sense of referential income standards like they might for a more common occupation. Participants were undergraduate students from a large public university. A total of 288 participants were randomly assigned to the 24 conditions of the study, yielding twelve cases per cell of the design. Upon completion of the main portion of the experiment, participants completed a post-study questionnaire. One of the questions asked to participants if they knew what an archiator did. Participants that indicated knowing what an archiator did were excluded from our sample, because they had a sense of the referential income standard for their assigned

Referent Networks and Distributive Justice

249

occupation. A total of 56 (16%) participants were excluded based on this question. Once the problematic cases were identified, they were replaced with new participants that were also randomly assigned to conditions. In the end, 288 participants that indicated not knowing what an archiator does were retained for the analyses reported later. In terms of participant sex, 65% of them were female. In terms of race, 72.2% were white, 16.7% were African American, and 11.1% reported “other.” The average age of the participants was 19.5, with a standard deviation of 2.5.5

Procedures Upon arriving at the experimental laboratory, participants were directed to isolated rooms and informed that they would complete a brief study. Multiple participants were scheduled at the same time, but contact between participants was minimized by escorting them into subject rooms upon arrival to the laboratory. The entire experiment, including the consenting and debriefing information, was computer mediated, minimizing any interaction between the participants and the research assistants. The instructions for the study informed participants that we were interested in “pay satisfaction” for various occupations and that they had been assigned to imagine being an archiator. The instructions also indicated that we would ask them several questions about a hypothetical income level, given the income levels of four of their colleagues. They were asked to imagine that their colleagues were also archiators. After these brief instructions, there was an on-screen tutorial illustrating the relevant information (their income level and the incomes of their colleagues) and how to respond to the questions. Upon completion of the instructions, participants were shown the incomes of four other archiators and asked a series of questions assessing how fair their income was, given the distribution of incomes and whether or not there was a referential standard. They repeated this process, with the incomes of their colleagues changing depending on the amount of reward homogeneity in the network. Participants did not actually know that the information about their colleagues was derived from network connections; rather, the information was simply displayed on their screen. After completing the questions about income fairness given each set of colleagues, and hence each level of reward homogeneity, participants completed a brief survey and then they were debriefed. All participants were paid 10 dollars for their time.

250

DAVID MELAMED ET AL.

Manipulations The first between-participants manipulation is the reward level of the position within the network. Following Chiang (2011), participants were assigned to one of 12 network positions. Each position was assigned an income level. The actual values that were used came from a simulated draw off of the distribution of incomes in the United States. Specifically, we assumed incomes were normally distributed and, using the actual mean and standard deviation for the distribution of incomes, the 12 income levels were simulated and rounded to the nearest thousand dollars.6 The assigned incomes ranged from $15,000 to $72,000, with a mean of $40,000 and a median of $39,000 (see Fig. 1). The second between-participants manipulation is whether or not there was a referential standard for incomes. In the conditions with a referential standard, the program displayed the following text while the participants completed the study: “The National Average Income for an Archiator is

(a)

(b)

(c)

Fig. 1. Graph-Theoretic Visualization of the Three Network Structures. (a) Network Structure Based on Reward Homogeneity. (b) Network Structure Based on Random Assortment. (c) Network Structure Based on Reward Heterogeneity.

Referent Networks and Distributive Justice

251

Approximately $40,000.” In the conditions without a referential standard, there was no such information. The within-participants manipulation was the amount of reward homogeneity in the networks. Networks with high levels of reward homogeneity had clusters with similar reward levels. Networks with low levels of reward homogeneity had clusters with dissimilar reward levels. To achieve such clusters, again following Chiang (2011), we used the simulation tool in the Exponential Random Graph Model (ERGM) package for the R environment (Hunter, Handcock, Butts, Goodreau, & Morris, 2008).7 Given the nodal attribute of income, the simulation generates networks with varying degrees of attribute-based similarity. The one constraint on the model was the outdegree distribution: every position in the network was connected to four others. The actual value of the income-based similarity parameter matters very little; several values were attempted and the same networks resulted. Positive similarity parameters result in simulated networks with high levels of reward or income homogeneity. Negative similarity parameters result in simulated networks with low levels of reward homogeneity, or high levels of reward heterogeneity. We also simulated networks using a value of zero for the similarity parameter. This results in a network of random ties, or a network based on random assortment. Thus the withinparticipants manipulation had three levels: reward homogeneity, reward heterogeneity and random assortment. This factor dictates which other archiators’ incomes are displayed at any given time. In homogenous networks, the others’ incomes were as similar to the participants’ as possible (i.e., the closest ones in the distribution); in heterogeneous networks, the others’ incomes were as dissimilar to the participants’ as possible. Fig. 1 visually depicts the three network structures that were employed in the experiment. Fig. 1(a) is the homogeneous network structure, Fig. 1 (b) is the random assortment network, and Fig. 1(c) is the heterogeneous network. Participants were randomly assigned to positions that were designated by income levels. Over three rounds, participants experienced each network structure while remaining in the same position. Thus the distribution of incomes within their ego network changed as a result of the process of association-reward homogeneity, random assortment, or reward heterogeneity. It is obvious that the sequence with which participants experience the within-participants manipulation may have an effect on their perceptions of distributive justice. Experiencing the heterogeneous network structure first, for example, may make the homogeneous network structure seem fairer. To safeguard against this possibility, the design also counterbalanced on the sequence with which the participants experienced the levels

252

DAVID MELAMED ET AL.

of reward homogeneity. Specifically, given that there are three levels of this manipulation, there are six possible sequences of reward homogeneity.8 Two of the twelve participants in each condition of the between-participant manipulations were assigned to each sequence. On the one hand, this is not as controlled as a completely between-subjects design. On the other hand, a completely between-subjects design would have had 72 conditions (i.e., 2 × 12 × 3), making a laboratory experiment unfeasible, and the balanced design that was implemented has more than enough power to statistically control for any sequence effects (see the statistical models below).

Measures The key independent variables are the income level to which the respondent was assigned, whether there was a referential standard for rewards, and the amount of reward homogeneity in the network. Actual assigned incomes were divided by 1,000 for ease of presentation. For every referent network, we also computed the Gini coefficient to capture the amount of inequality within the network. The main dependent variable is a scale assessing the participants’ perceptions of distributive justice. Specifically, participants were asked how ‘just’ and how ‘fair’ the overall distribution of rewards was on nine-point Likert-type scales. These two items are highly reliable (α = 0.93), so they were averaged to generate the main outcome. We also control for an additional factor aside from the sequence of the reward homogeneity factor
the mean reward level within the referent network. We control for the mean of the referent network to offset the possibility that individuals have a preference, or a “universal longing” (Jasso, 1980, p. 9), for equality. If people did have a “universal longing” for equality then nothing else would be able to predict perceptions of distributive justice.

RESULTS Table 1 presents the means and standard deviations for the perceptions of distributive justice by reward levels and network structure. Looking across the columns of Table 1, it is evident that the reward homogenous networks were perceived as the fairest in terms of distributive justice. In most cases, the random assortment networks are fairer than the heterogeneous networks, but not in all cases. Looking down the rows of Table 1, it is harder

253

Referent Networks and Distributive Justice

Table 1.

Means and (Standard Deviations) of Perceptions of Distributive Justice by Reward Levels and Type of Referent Network.

Reward Level

15,000 19,000 21,000 22,000 27,000 29,000 39,000 51,000 55,000 61,000 69,000 72,000

Referent Network Structure

Overall

Heterogeneous

Random assortment

Homogeneous

2.06 (1.90) 2.29 (1.84) 2.73 (1.58) 2.29 (1.53) 2.13 (1.29) 2.92 (1.36) 2.94 (1.68) 2.65 (1.50) 2.65 (1.57) 2.13 (1.15) 2.35 (1.42) 1.56 (0.98)

2.16 (1.63) 2.42 (1.40) 2.58 (1.72) 3.75 (1.50) 2.13 (0.97) 2.58 (1.44) 4.00 (1.84) 3.10 (1.62) 3.92 (2.03) 2.94 (1.39) 3.27 (1.84) 3.06 (2.12)

4.04 (1.74) 5.25 (2.01) 4.63 (1.87) 5.38 (1.48) 5.02 (2.24) 5.60 (1.93) 5.04 (1.60) 5.02 (1.81) 5.15 (1.81) 6.25 (2.09) 6.52 (1.96) 6.19 (1.88)

2.76 (1.96) 3.32 (2.22) 3.31 (1.94) 3.81 (1.95) 3.09 (2.09) 3.70 (2.08) 3.99 (1.89) 3.59 (1.93) 3.90 (2.06) 3.77 (2.39) 4.05 (2.50) 3.60 (2.58)

Note: The distributive justice scale ranged from 1 to 9.

to directly discern patterns. In the homogeneous network structure, the pattern seems to be one of increasing fairness as reward levels increase. This is not the case in the other two network structures, which is descriptive support for hypotheses one and five. To formally evaluate our hypotheses, later we report the results of three linear mixed models. All of these models include random intercepts, which accounts for the nesting of network structures within participants. Table 2 presents the results of the mixed models. Our first hypothesis predicted that more reward homogeneous referent networks would lead to perceiving the distribution of rewards as more fair. We find clear support for this hypothesis in Model 1 of Table 2. Specifically, participants in

254

DAVID MELAMED ET AL.

reward homogeneous networks reported that the distributive justice was 1.2 units higher than those participants in the random assortment networks (on a nine-point scale). Likewise, participants in reward heterogeneous networks reported that the distributive justice was 0.75 units less than those participants in the random assortment networks. Both of these results are controlling the participants’ reward levels, whether there was a referential standard, the amount of inequality in the network, the average reward within their ego network, and the sequence within which participants experienced the network structures. As predicted, structural features of the referent network shape interpersonal social comparisons, which increase distributive justice. Our second hypothesis was that inequality of the referent network would decrease distributive justice. In Model 1 of Table 2, the Gini coefficients have been transformed to z-scores. As such, a standard unit increase in ego network inequality leads to a 0.66 unit decrease in distributive justice, net of the other factors in the model. This is clear support for hypothesis number two. Our third hypothesis predicted that the presence of a referential Table 2.

Linear Mixed Models Predicting Perceptions of Distributive Justice.

Predictor Heterogeneous referent network Homogeneous referent network Reward level (R)

Model 1

Model 2

Model 3

−0.751*** (0.120) 1.180*** (0.222) 0.018*** (0.004)

−0.749*** (0.120) 1.290*** (0.219)

−0.531*** (0.136) 1.064*** (0.223) 0.022*** (0.004)

Reward level below referential standard (=1) Referential standard (=1) (S) Gini coefficient for the referent network (G) R S R G

0.071 (0.141) −0.660*** (0.107)

−0.223 (0.208) 0.461** (0.219) −0.596*** (0.105) −0.669** (0.286)

0.071 (0.141) −0.267 (0.157)

−0.011** (0.003)

Note: The number of participants is 288 and the number of referent networks is 3, for a total sample of 864. All models control for the sequence with which participants experienced the referent networks and the average reward level within the referent networks. ** p < 0.01, *** p < 0.001.

255

Referent Networks and Distributive Justice

standard for rewards would have a negative effect on participants who were rewarded at below the referential standard. Model 2 in Table 2 tests this hypothesis. The key coefficient for evaluating our argument is the interaction term between whether participants had a reward level that was below the referential standard and whether the referential standard was present. Indeed, there is a significant decrease in perceptions of distributive justice when participants are rewarded at levels below the referential standard when the referential standard is present. We left as exploratory how the referential standard would affect those participants above it. To illustrate this process, we reestimated Model 2 from Table 2, except we replaced the dummy reward variable with the linear term. Again, the interaction term was significant (β = 0.016, σβ = 0.007, p = 0.028). To illustrate this result, we’ve plotted the marginal means from the mixed models, setting all of the other factors at their means. This is presented in Fig. 2. Net of ego network inequality, the structure of the network, etc., participants below the referential standard when one is known are predicted to report lower distributive justice values than when a referential standard is unknown. At the same time, participants above the referential standard when one is known are predicted to report higher distributive justice values than when a referential standard is unknown. We interpret this effect as follows: when participants are above a known referential standard and are top earners in their referent network, they are over-rewarded with respect to both the referential standard and their network. When participants are at the high

5

4 No Referential Standard 3

Referential Standard

2 15 19 21 22 27 29 39 51 55 61 69 72 Reward Level (1,000s)

Fig. 2. Perceptions of Distributive Justice Marginal Means by Reward Levels and Whether There was a Referential Standard. Note: All other factors from Table 2 are set at their means.

256

DAVID MELAMED ET AL. 6 5 4

Gini Maximum Gini Median

3

Gini Minimum

2 15 19 21 22 27 29 39 51 55 61 69 72 Reward Level (1,000s)

Fig. 3. Perceptions of Distributive Justice Marginal Means by Reward Levels and Three Values of the Amount of Inequality in the Referent Network. Note: All other factors from Table 2 are set at their means.

end of the distribution of rewards within their network, but there is no referential standard, they are only over-rewarded with respect to one reference point. Our fourth hypothesis predicted that reward levels would moderate the relationship between ego network inequality and distributive justice. Specifically, we expected that the negative effect of inequality on distributive justice would be exacerbated by increasing reward levels. This hypothesis is evaluated in Model 3 of Table 2. Indeed we find a significant interaction between reward levels and the amount of ego network inequality. To illustrate that this single term reflects what we have predicted, we again plotted the marginal means from the mixed model.9 This is presented in Fig. 3. First, notice how at low reward levels there is less variability in terms of how the Gini coefficient affects the fitted distributive justice values. Second, notice how as the reward level increases, the variability of the effect of the Gini coefficient on distributive justice increases. In terms of our argument, when reward levels are low, the amount of ego network inequality should matter less. However, when reward levels are high, higher levels of inequality indicate over-reward and a concomitant unfair distribution of rewards. Finally, in terms of our fifth hypothesis, we have found mixed support. Recall, we predicted that the negative effect of reward heterogeneity in referent networks will be exacerbated by increasing reward levels. To evaluate this hypothesis, we reestimated Model 1 from Table 2 including the

Referent Networks and Distributive Justice

257

interaction between reward level and the homogeneous network dummy, and a second interaction between reward level and the heterogeneous network dummy. In this model, we found that reward levels have a positive main effect on distributive justice and that the positive effect of rewards is stronger in the homogeneous referent networks. We did not, however, find a significant decrease in the effect of rewards in heterogeneous referent networks, as we had predicted. Subsequent sensitivity analyses revealed that including the ego network Gini coefficient suppresses the predicted effects. That is, when the Gini coefficient is removed from the model, the predictions are supported. While we have tried to be innovative with regard to means to experimentally investigate effects of referent networks, the experimental protocols entail a lot of moving parts that need to be controlled for valid inferences to be made. Perhaps an experiment that was constructed with the sole purpose of testing hypothesis five may find support. For now, we have to conclude that we found only partial support. So to summarize our results, the first four hypotheses were supported and the fifth was partially supported. Referent networks have predictable effects on distributive justice. More homogeneous referent networks increase perceptions of distributive justice. Inequality within the referent network decreases perceptions of distributive justice. The moderating effects of referential standards and reward levels were also investigated, and, for the most part, were supported.

DISCUSSION AND CONCLUSION Since most social behavior is embedded in networks of interpersonal relations (Granovetter, 1985), examining justice processes in network structures provides us with a better understanding of its mechanisms. The research reported earlier constitutes a first step towards systematically examining how network structures affect perceptions of distributive justice and social comparisons more generally. The results convincingly indicate that the amount of reward homogeneity in an individual’s network affects perceptions of distributive justice. This is an important finding as it relates to the extant literature. On the one hand, little sociological work has examined how interpersonal contacts shape justice processes (cf. Gartrell, 1985), despite the fact that we recognize that justice processes are fundamentally about relative deprivation (e.g., Homans, 1961). On the other hand, work in organizational psychology has focused on interpersonal relationships,

258

DAVID MELAMED ET AL.

but this research tends to focus on specific referent others (e.g., Goodman, 1974), rather than the referent network within which a person makes social comparisons. Only one structural property of the referent networks was manipulated in the present study. Many other structural features warrant investigation. In a study of perceptions of distributional inequality, Chiang (2011) investigated both reward similarity and degree inequality. Degree inequality refers to the number of ties associated with each node. Some people may know the incomes of several other people, while others may know only a few. Knowledge of others’ incomes affects the number of potential social comparisons, and so may also affect perceptions of distributive justice. Aside from aspects of the local network, the broader network structure has been shown to be related to perceptions of fairness (Gartrell, 1985; Shah, 1998). Individuals in similar structural positions, or individuals with similar tie profiles, have similar perceptions of pay satisfaction (Shah, 1998). Although network positions are amenable to experimental manipulation this is rarely accomplished, with the exception of the exchange tradition in social psychology (e.g., Molm & Cook, 1995). An interesting direction for future research would be to examine not only how network structures shape perceptions of fairness, but also to examine how these, in turn, shape emotional and behavioral outcomes. Distributive justice mediates emotional reactions to rewards (Younts & Mueller, 2001), which in light of the present project, implies that the reward homogeneity of the referent network would also mediate emotional reactions by conditioning perceptions of distributive justice. Likewise, the context in which referent networks are embedded may be an important element to consider, as previous research has found context effects on perceptions of justice (Hegtvedt et al., 2003; Johnson, Hegtvedt, Brody, & Waldron, 2007). The theoretical arguments developed in this chapter have implications beyond the laboratory setting in which they were tested. The amount of reward homogeneity at various levels of analysis is likely to be a predictor of distributive justice. At the organizational level, for example, the amount of reward homogeneity affects the amount of reward homogeneity in an individual’s network by shaping the reward levels of those with whom one is likely to be in contact. This, in turn, should affect perceptions of distributive justice. Likewise, the amount of inequality in a society shapes the baseline probability of interacting with individuals with different incomes. As such, as inequality in society increases, the likelihood of reward homogeneity within any given network decreases. Consequently, as inequality in a society increases, there should be a corresponding decrease in perceptions of distributive justice. This line of reasoning, however, must be careful as

Referent Networks and Distributive Justice

259

other factors, such as the degree of legitimation of inequality (Della Fave, 1980; Stolte, 1983) or length of job tenure (Oldham, Kulik, Stepina, & Ambrose, 1986) have also been shown to affect justice assessments. Carrying this line of inquiry further it will be important, as mentioned in the introduction, to distinguish between the aggregation of multiple reference groups and the structure of the referent network. While participants in our experiment were told that the information in their network came from coworkers, we have no way to determine the extent to which they viewed them as group. Instead, we simply wanted to generate a set of ties in the referent network and examine systematic changes in perceptions of distributive justice resulting from different structures in the network. Subsequent research is needed to determine how people combine information from multiple reference groups, and whether the structural properties of referent networks remain significant predictors once those reference groups are included. In summary, despite the overt role of social comparisons in the early justice literature, there has been very little empirical research on the relationship between networks of relations and perceptions of justice. This study makes an initial first step toward understanding how the structure of relationships affects justice processes. We developed an argument linking reward homogeneity to distributive justice and elaborated the argument with the role of referential standards for rewards and reward levels. An experimental investigation found support for most of the argument, with the evidence fully supporting four of the five hypotheses and partially supporting the fifth. Further experimental and observational research is required to investigate other structural properties of the referential network and to examine whether these patterns are found in real-world networks.

NOTES 1. Here we assume that knowledge of the reward levels in the referent network is known. Future research should investigate how uncertainty in the referent network affects perceptions of justice. 2. As noted above, focusing participant’s attention towards specific referents may introduce justice anchors (Markovsky, 1988). 3. With respect to monkeys, see Brosnan and De Waal (2003). With respect to chimpanzees, see Brosnan, Schiff, and De Waal (2005). 4. Portions of this section are adapted from Melamed, Liu, Park, and Zhong (2013). 5. Sex (χ 21 = 0.66, p = 0.41), race (χ 22 = 2.26, p = 0.32), and age (β = −0.022, σβ = 0.031, p = 0.47) were not associated with excluding cases. Those factors were also

260

DAVID MELAMED ET AL.

not associated with distributive justice. The results of these models are available from the first author upon request. 6. The mean ($39,658.97) and standard deviation ($35,886.26) were computed using the cumulative file of the General Social Survey (Davis & Smith, 1972
2010). 7. A technical overview of ERG models is outside of the purview of this paper. For a gentle conceptual introduction, please see Robins, Pattison, Kalish, and Lusher (2007) and the references cited therein, and for an introduction to estimation, please see Snijders (2002). 8. Participants could have experienced any one of the three network structures first, then any of the remaining two second, and then the remaining structure third. The combinatorics yields six potential sequences. 9. While the interpretation of the interaction is straightforward
as reward levels increase, the effect of the Gini coefficient decreases
it is easiest to convey the relationship visually.

ACKNOWLEDGMENTS Portions of this chapter were presented at the Departments of Sociology at the University of Georgia and the University of North Carolina at Charlotte, and the annual meetings of the Southern Sociological Society. We thank the participants of those events, Karen Hegtvedt, Barry Markovsky, and Scott Savage for helpful comments and suggestions on this research. This research was supported by the College of Arts and Sciences at the University of South Carolina.

REFERENCES Adams, J. S. (1965). Inequity in social exchange. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 2, pp. 267
299). New York, NY: Academic Press. Belliveau, M. A. (2012). Engendering inequity? How social accounts create vs. merely explain unfavorable pay outcomes for women. Organization Science, 23(4), 1154
1174. Berger, J., Fisek, M. H., Norman, R. Z., & Wagner, D. G. (1998). Formation of reward expectations in status situations. In J. Berger & M. Zelditch (Eds.), Status, power, and legitimacy (pp. 121
153). New Brunswick, NJ: Transaction Publishers. Berger, J., Zelditch, M., Anderson, B., & Cohen, B. P. (1972). Structural aspects of distributive justice: A status value formulation. In J. Berger, M. Zelditch, & B. Anderson (Eds.), Sociological theories in progress (pp. 119
146). Boston, MA: Haughton Mifflin Company. Berkowitz, L., Fraser, C., Treasure, F. P., & Cochran, S. (1987). Pay, equity, job gratifications, and comparisons in pay satisfaction. Journal of Applied Psychology, 72(4), 544. Brosnan, S. F., & De Waal, F. B. (2003). Monkeys reject unequal pay. Nature, 425(6955), 297
299.

Referent Networks and Distributive Justice

261

Brosnan, S. F., Schiff, H. C., & De Waal, F. B. (2005). Tolerance for inequity may increase with social closeness in chimpanzees. Proceedings of the Royal Society B: Biological Sciences, 272(1560), 253
258. Buchanan, T. (2008). The same-sex-referent-work satisfaction relationship: Assessing the mediating role of distributive justice perceptions. Sociological Focus, 41(2), 177
196. Buckingham, J. T., & Alicke, M. D. (2002). The influence of individual versus aggregate social comparison and the presence of others on self-evaluations. Journal of Personality and Social Psychology, 83(5), 1117. Burt, R. S. (1995). Structural holes: The social structure of competition. Cambridge, MA: Harvard University Press. Chiang, Y. S. (2011). Judgment of distributional inequality in networks. Social Networks, 33(4), 342
349. Clay-Warner, J., Culatta, E., & James, K. R. (2013). Gender and organizational justice preferences. Sociology Compass, 7, 1074
1084. Davis, J. A., & Smith, T. W. (1972
2010). General Social Surveys, 1972
2006. [machinereadable data file]. Principal Investigator, James A. Davis; Director and Co-Principal Investigator, Tom W. Smith; Co-Principal Investigator, Peter V. Marsden, NORC ed. Chicago, IL: National Opinion Research Center, producer, 2005; Storrs, CT: The Roper Center for Public Opinion Research, University of Connecticut, distributor. 1 data file (51,020 logical records) and 1 codebook (p. 2,552). Dawes, C. T., Fowler, J. H., Johnson, T., McElreath, R., & Smirnov, O. (2007). Egalitarian motives in humans. Nature, 446(7137), 794
796. Della Fave, L. R. (1980). The meek shall not inherit the earth: Self-evaluation and the legitimacy of stratification. American Sociological Review, 45(6), 955
971. Fowler, J. H., Johnson, T., & Smirnov, O. (2005). Egalitarian motive and altruistic punishment. Nature, 433. doi:10.1038/nature03257 Gartrell, C. D. (1985). Relational and distributional models of collective justice sentiments. Social Forces, 64(1), 64
83. Goodman, P. S. (1974). An examination of referents used in the evaluation of pay. Organizational Behavior and Human Performance, 12(2), 170
195. Granovetter, M. (1985). Economic action and social structure: The problem of embeddedness. American Journal of Sociology, 91(3), 481
510. Hegtvedt, K. A. (2006). Justice frameworks. In P. Burke (Ed.), Contemporary social psychological theories (pp. 46
69). Stanford, CA: Stanford University Press. Hegtvedt, K. A., Clay-Warner, J., & Johnson, C. (2003). The social context of responses to injustice: Considering the indirect and direct effects of group-level factors. Social Justice Research, 16(4), 343
366. Hegtvedt, K. A., & Cook, K. S. (2001). Distributive justice: Recent theoretical developments and applications. In J. Sanders & V. L. Hamilton (Eds.), Handbook of justice research in law. New York, NY: Kluwer Academic/Plenum Publishers. Hegtvedt, K. A., & Johnson, C. (2000). Justice beyond the individual: A future with legitimation. Social Psychology Quarterly, 63(4), 298
311. Homans, G. C. (1961). Social behavior: Its elementary forms. New York, NY: Harcourt, Brace, & World. Hunter, D. R., Handcock, M. S., Butts, C. T., Goodreau, S. M., & Morris, M. (2008). ERGM: A package to fit, simulate and diagnose exponential-family models for networks. Journal of Statistical Software, 24(3), 1
29. Jasso, G. (1980). A new theory of distributive justice. American Sociological Review, 45(1), 3
32.

262

DAVID MELAMED ET AL.

Johnson, C., Hegtvedt, K. A., Brody, L. M., & Waldron, K. W. (2007). Feeling injustice, expressing injustice: How gender and context matter. Advances in Group Processes, 24, 149
186. Kulik, C. T., & Ambrose, M. L. (1992). Personal and situational determinants of referent choice. The Academy of Management Review, 17(2), 212
237. Law, K. S., & Wong, C. S. (1998). Relative importance of referents on pay satisfaction: A review and test of a new policy-capturing approach. Journal of Occupational and Organizational Psychology, 71(1), 47
60. Major, B., & Forcey, B. (1985). Social comparisons and pay evaluations: Preferences for samesex and same-job wage comparisons. Journal of Experimental Social Psychology, 21(4), 393
405. Markovsky, B. (1988). Anchoring justice. Social Psychology Quarterly, 51(3), 213
224. Mayhew, B. H. (1980). Structuralism versus individualism: Part 1, shadowboxing in the dark. Social Forces, 59(2), 335
375. Melamed, D. (2012a). Deriving equity from expectations: A cross-cultural evaluation. Social Science Research, 41(1), 170
181. Melamed, D. (2012b). The effects of legitimacy and power on perceptions of fairness. Sociological Focus, 45(2), 125
142. Melamed, D., Liu, Y., Park, H., & Zhong, J. (2013). Referent networks and just rewards. Unpublished manuscript. Molm, L. D., & Cook, K. S. (1995). Social exchange and exchange networks. In K. Cook, G. Fine, & J. House (Eds.), Sociological perspectives on social psychology (pp. 209
235). Boston, MA: Allyn and Bacon. Oldham, G. R., Kulik, C. T., Stepina, L. P., & Ambrose, M. L. (1986). Relations between situational factors and the comparative referents used by employees. The Academy of Management Journal, 29(3), 599
608. Robins, G., Pattison, P., Kalish, Y., & Lusher, D. (2007). An introduction to exponential random graph (p*) models for social networks. Social Networks, 29(2), 173
191. Shah, P. P. (1998). Who are employees’ social referents? Using a network perspective to determine referent others. The Academy of Management Journal, 41(3), 249
268. Shepelak, N. J., & Alwin, D. F. (1986). Beliefs about inequality and perceptions of distributive justice. American Sociological Review, 51(1), 30
46. Simmel, G. (1955). Conflict and the web of group affiliations. New York: Free Press. Snijders, T. A. (2002). Markov chain Monte Carlo estimation of exponential random graph models. Journal of Social Structure, 3(2), 1
40. Stolte, J. F. (1983). The legitimation of structural inequality: Reformulation and test of the self-evaluation argument. American Sociological Review, 48(3), 331
342. Tajfel, H., & Turner, J. C. (1986). The social identity theory of intergroup behavior. In S. Worchel & W. G. Austin (Eds.), Psychology of intergroup relations (pp. 7
24). Chicago, IL: Nelson-Hall. To¨rnblom, K. Y. (1977). Distributive justice: Typology and propositions. Human Relations, 30(1), 1
24. Younts, C. W., & Mueller, C. W. (2001). Justice processes: Specifying the mediating role of perceptions of distributive justice. American Sociological Review, 66(1), 125
145. Zelditch, M., Jr. (1969). Can you really study an army in the laboratory. In A. Etzioni (Ed.), A sociological reader on complex organizations (pp. 528
539). New York, NY: Holt, Rinehart and Winston.

BEYOND NETWORKS IN STRUCTURAL THEORIES OF EXCHANGE: PROMISES FROM COMPUTATIONAL SOCIAL SCIENCE James A. Kitts ABSTRACT Purpose
The research community currently employs four very different versions of the social network concept: A social network is seen as a set of socially constructed role relations (e.g., friends, business partners), a set of interpersonal sentiments (e.g., liking, trust), a pattern of behavioral social interaction (e.g., conversations, citations), or an opportunity structure for exchange. Researchers conventionally assume these conceptualizations are interchangeable as social ties, and some employ composite measures that aim to capture more than one dimension. Even so, important discrepancies often appear for non-ties (as dyads where a specific role relation or sentiment is not reported, a specific form of interaction is not observed, or exchange is not possible).

Advances in Group Processes, Volume 31, 263
298 Copyright r 2014 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0882-6145/doi:10.1108/S0882-614520140000031007

263

264

JAMES A. KITTS

Methodology/approach
Investigating the interplay across the four definitions is a step toward developing scope conditions for generalization and application of theory across these domains. Research implications
This step is timely because emerging tools of computational social science
wearable sensors, logs of telecommunication, online exchange, or other interaction
now allow us to observe the fine-grained dynamics of interaction over time. Combined with cuttingedge methods for analysis, these lenses allow us to move beyond reified notions of social ties (and non-ties) and instead directly observe and analyze the dynamic and structural interdependencies of social interaction behavior. Originality/value of the paper
This unprecedented opportunity invites us to refashion dynamic structural theories of exchange that advance “beyond networks” to unify previously disjoint research streams on relationships, interaction, and opportunity structures. Keywords: Social networks; social exchange; interpersonal sentiments; social interaction; online exchange; relational event modeling

Since the advent of social network analysis, scientists have been defining, measuring, and analyzing social networks in four fundamentally different ways. Some use social networks to refer to substantive role relations, whether represented as a cognitive category (such as friendship, kinship, and marriage) or shared involvement in some higher-order social unit (teammates, officemates, housemates, coauthors, comembers). By contrast, some use social networks to refer to patterns of interpersonal sentiments (liking, respect, trust, esteem, disesteem, hatred). Some define social networks as behavioral interaction (communication, advice, social support, citation, gift, or transaction). Lastly, researchers in sociological exchange theory often think of networks as opportunity structures for interaction (i.e., choice sets of possible exchanges, whether or not exchange is actually realized in any given dyad). All four definitions of social networks
as role relations, sentiments, interactions, and opportunity structures
offer building blocks for distinct theories, and all have supported decades of fruitful research. Any of these interpretations may overlap with the others empirically, as friends or work partners may like one another, interact with one another regularly, and be available for some kinds of exchange. However, two people may like each

Beyond Networks in Structural Theories of Exchange

265

other and regard each other as friends although they hardly ever interact or they may interact regularly without regarding each other as friends or liking each other, and some actors may be available to others as exchange partners without interaction, emotional attachments, or socially recognized friendships taking place. In studying “social networks,” researchers often elide the distinction between these four conceptualizations in their measures, analyses, interpretations, and applications. Overwhelmingly (if implicitly) their use and interpretation of social network analysis assumes these four versions are interchangeable. I aim to show that strong assumptions are required to leap from one version of the networks concept to another, especially with regard to their treatment of non-ties (interpretation of the case where a tie is not observed). Rigorously identifying these boundary conditions for theory extension and application will offer an unprecedented opportunity to constructively integrate our diverse research programs. Most foundation work in social network analysis and theory
including concepts, measures, and methods
has drawn from sociometric study of socially constructed role relations (friends, acquaintances, kin, coworkers, teammates, neighbors). A related approach has measured interpersonal sentiments (e.g., liking, hatred, trust, or esteem). Although there is room for fruitful research on the interplay between interpersonal sentiments and socially constructed role relations, I will generally characterize the research on “relationships” as encompassing these two approaches. Some common measures
such as close friendship
try to capture both role relations and sentiments. I will show that during a time interval on which a graph is defined, these approaches similarly assume temporal continuity (i.e., social ties are continually active on the interval) and temporal stability (i.e., the structure of ties is fixed on the interval). This foundation research on relationships has also been motivated by the assumption that relationships channel social interaction
with many theorists assuming that measured ties are proxies for interaction behavior
so studies of relationships rarely examine social interaction directly. Researchers who study relationships also implicitly assume that just as ties are continuously open for interaction, non-ties are continuously shut; that is, relevant social interaction can never occur outside of social ties. Contrasting with studies of relationships (either role relations or interpersonal sentiments), some scholars have constructed networks of social interaction from observable relational behavior such as communication (e.g., e-mail, phone, face-to-face conversations), contact (e.g., colocation, meetings, sex), or resource exchange (support, advice, gifts, lending, drug

266

JAMES A. KITTS

needle sharing, scholarly citations). Having observed a sequence of behavioral interaction events, scholars typically aggregate those behaviors over time, then set a filter that transforms these event counts into a set of ties and non-ties. Beginning with data that could give us rich insight into fine-grained dynamics of interdependent interaction behavior, scholars collapse those data into categories and interpret the resulting ties as if they are temporally continuous and stable to fit the traditional conceptual lenses for network analysis of relationships. Research on relationships (role relations or interpersonal sentiments) or social interaction has focused almost exclusively on ties, either ignoring dyads where no tie is observed or assuming that non-observed tie means non-tie. By contrast, researchers who view social networks as opportunity structures for exchange have relied on non-ties to carry the theoretical weight: Ties are sites where interaction may or may not take place and emotional attachments may (or may not) appear, whereas non-ties are precisely defined as vacuums, where interaction and sentiments are exogenously prohibited. The porous boundaries between these four conceptualizations of social networks have allowed a healthy diversity of ideas and empirical research, but understanding general processes requires us to ask how these levels connect. I argue that a constructive step is to temporarily suspend use of the reified but vacuous concepts of social ties and networks
which represent some unspecified congeries of aggregated ideas, behaviors, and emotions
and instead directly consider the structural dynamics of interaction. Making sense of these decades of research requires theoretical lenses, methodological approaches, and empirical data that speak to one another, and are collectively suited to investigating dynamics. I will show how tools from Computational Social Science
employing analysis of the fine-grained temporal dynamics of social interaction
may serve as a common language to integrate the various threads of networks research. Although some early work on social interaction has used field observation or time diaries to observe interaction directly, new telecommunications and sensor technologies allow researchers to systematically collect data on interaction behavior with unprecedented volume and granularity. Electronic traces, such as logs of e-mails sent and received, telephone calls, meetings recorded on electronic calendars, exchanges in online commerce or sharing sites, allow us to monitor social interaction in fine time grain. Meanwhile, there is a convergence in statistical methodology that is poised to address the wealth of longitudinal interaction data. Cutting-edge

Beyond Networks in Structural Theories of Exchange

267

analytical methods allow us to abandon assumptions that our observations are independent, and to explicitly model various forms of temporal and relational dependence. This confluence of ideas, tools, and data allows us to consider the dynamic interdependence of interaction behavior for a set of social actors in continuous time. New analytical tools and rich data on behavioral interaction offer promising applications for all areas of social networks research. Indeed, I show how analyzing interaction dynamics can deepen our understanding of socially constructed relationships, interpersonal sentiments, and opportunity structures for exchange. However, these new frontiers also demand attention to scope conditions on theories developed for alternative conceptualizations of networks, such as structural balance theory for interpersonal sentiments or network exchange theory for opportunity structures.

CONVENTIONAL VIEWS OF NETWORKS AS TEMPORALLY CONTINUOUS AND STABLE STRUCTURES Decades of research in social networks has employed the directed or undirected graph as a way to represent a set of relationships (typically assumed to imply interaction) or interaction (typically assumed to imply relationships) or opportunity structures. This graph has often been reified, where substantively vacuous concepts like social ties have allowed slippage across these four levels (role relations, sentiments, interaction, opportunity structures) in conversations across theories or findings. Fig. 1 illustrates a conventional binary and static representation of a network as a graph. We can observe various features of the ties in this network: The tie A-B is reciprocated and the tie B→C is unreciprocated. A triad E-F-I of mutually tied actors is a closed triad, whereas the triplet D-E-I is intransitive, and so on. This typical basic representation includes no weight on the tie (all ties are uniform in strength), no consideration of dynamics (changes over time in either end of the relationship), and there has been no principled way to incorporate a dependence on history. These simplifying assumptions have allowed the use of elegant and powerful tools of social network analysis (Wasserman & Faust, 1994), but have limited our ability to think in more nuanced ways about the structural dynamics of exchange.

268

JAMES A. KITTS

G

E

D

F

C H I B

A

Fig. 1.

Illustration of a Social Network as a Directed Graph.

Sociometric graphs are often constructed by a “relationship” measure (such as “close friends” among students) where ties are interpreted as continuously-available conduits, and non-ties are insurmountable barriers. For example, if Fig. 1 represents a network of best friends in a school class, authors read in the graph that because A is not friends with I (or E), A must “reach” I through a long path, with B always as an intermediary. Conceiving of the network as a flow of information leads us to interpret B’s position as a source of power, because B may be able to withhold or manipulate information at will.1 The vast majority of this usage of the graph metaphor has presumed that the network is temporally continuous; that is, for the time interval in which the graph is defined, each tie is continuously active (available) during any moment within that interval; that is, there is no sequence of activation. Similarly common is the assumption that the graph is stable over that time interval, and this regularity motivates our use of terms like social structure. Lastly, conventional social network analysis often assumes that ties are interchangeable; that is, each friendship or coauthorship presents the same weight or dyadic diffusion potential as any other. These simplifying assumptions give us analytical leverage. For example, under these assumptions we can say that node B is reachable by node D by a path of length 3,

Beyond Networks in Structural Theories of Exchange

269

which is three times longer than the path from F to G. We may then compute a centrality measure for each node (Bonacich, 1987; Freeman, 1978/1979), showing that a particular node lies on the most shortest paths connecting other nodes, or can reach other nodes by the shortest paths, or simply is connected to the most other nodes. We can also say that G and H are structurally equivalent (Friedkin, 1984) or that F’s personal network spans structural holes (Burt, 1992) between {E,I} and {G,H}. Lastly, we can compute graph level metrics allowing us to show that one graph is more centralized or cohesive (Moody & White, 2003) than another based on patterns of ties and non-ties. There are many ways to relax these assumptions, such as allowing that one tie may be more intense than another tie (with faster or more effective communication or influence, realized as a weight on the tie) or allowing that ties may be negative as well as positive, and these complications invite more nuanced measures and models. But the power and elegance of the simple graph metaphor in Fig. 1 makes this representation almost ubiquitous even when more fine-grained data are available. Notably, decades of research have shown the generality and analytical power of social network analysis tools, and they have spread rapidly across theoretical and empirical contexts. I will be especially concerned about the usually implicit assumptions that ties are temporally continuous and stable over the time interval on which the graph is defined. Relaxing the assumption of temporal continuity would make all paths in the graph above ambiguous, as the simultaneity or sequencing of tie activation could make any path impassible. Allowing that the structure changes over time would require proliferating graphs. Either relaxation could make concepts like centrality, structural power, structural cohesion, or structural equivalence unwieldy. A notable feature of the simple graph representation, which is rarely questioned or discussed, is the strong equivalency assumption for non-ties (such as A→I or B→I above), which are valued at exactly zero, representing zero behavioral interaction, zero sentiments, zero opportunity for exchange. These assumptions underlie our network metrics described above, but also imply important scope conditions for those same lenses. Our computing the length of the shortest path between node A and node I (through nodes B
E) is predicated on an assumption that interaction is strictly impossible among non-tied actors in this graph. The treatment of non-ties is a crucial issue with distinct implications for all four representations of social networks, and I will give it special attention.

270

JAMES A. KITTS

FOUR CONVENTIONAL WAYS OF THINKING ABOUT SOCIAL TIES Social Ties as Socially Constructed Role Relations Most of the methodological and theoretical foundations of social network analysis were developed to suit social networks as sets of substantive role relations. These can be represented as a cognitive category (friendship, kinship, coauthorship) or shared involvement in some higher-order social unit (teammates, officemates, housemates). Examples include friendships among students (Kandel, 1978) or members of a fraternity (Newcomb, 1961) or karate club (Zachary, 1977), business ties or intermarriage among Florentine families (Padgett & Ansell, 1993), or coauthors of papers (Moody, 2004). In this usage, a social tie is measured as a labeled role relation operating at a level of abstraction above concrete social behavior. Scholars have often measured role relations as perceptions using surveys. For example, a conventional name-generator survey could ask the respondent, “list the names of your five closest friends.” The role relation has offered comfortable standing for the crucial assumptions of temporal continuity and stability supporting classical network analysis. Perceived relations like best friend, collaborator, spouse, comember, or teammate are plausibly continuous in time and relatively stable, such that we can use a graph to depict the structure of friendships in a college dormitory or shared corporate board memberships in a given month. Some role relations may also imply general patterns of sentiments and interaction behavior. For example, we often assume that friends like each other, talk with each other, spend time together, and trust and support each other, but friendship as a socially constructed category is not reducible to any of these dimensions and the correspondence from the relation label to any particular behavior or sentiment may be weak. For less culturally loaded role relations (such as comember or coauthor), the correspondence to behavioral interaction, sentiments, or opportunity structures is even more unclear. When role relations are measured using selfreported perceptions, the strong equivalency assumption for observed ties deserves critical consideration, as responses may depend on idiosyncratic and culturally or contextually contingent interpretations of the survey question. Let us consider an example: The use of self-reported friendships to represent ties in a network has been extensively applied for children in

Beyond Networks in Structural Theories of Exchange

271

school, where the term “friend” may presumably distinguish a student’s closest alters from less close alters. Scholars have also applied this term to adults, assuming that friendships convey information, job opportunities, or social support; however, there is much ambiguity about how respondents interpret survey questions about their friendships. Indeed, there is mounting evidence that adult respondents’ use of the word “friend” does not map onto social network researchers’ concepts of strong or weak ties. Indeed, researchers find that respondents may employ friend as a generic category for miscellaneous associates who have no other more specific role label. [The label of ‘friend’] is likely to be applied: to an overwhelming majority of nonrelatives in a largely unsystematic way; to associates lacking other specialized role relations; to people of the same age; to people known a long time; and to people with whom respondents had primarily sociable, rather than intimate or material, involvements. (Fischer, 1982, p. 287)

The set of friends can thus be an et cetera category, which often does not include kin, lovers, coworkers, neighbors, people of different ages, or people who engage in material exchange, even when those other people may be more important emotionally and interact more frequently (i.e., stronger ties) than the people respondents actually label as “friend” on a survey. Research has shown us that although it seems easiest to measure a relationship by merely employing a name generator for a role relation, we must be very careful in interpreting such reports as interpersonal sentiments, behavioral interaction, or opportunity structures, and be attentive to the necessity of validating or defending these interpretations. Much previous research has investigated what friendship means
in other words, how to interpret a “1” in a sociomatrix of relations
but a more serious problem for network analysis has hardly ever been acknowledged: When survey respondents fail to mention another person as a close friend (or partner, confidant, etc.) on a name generator, we are even less sure what this “0” in a sociomatrix means. Someone not mentioned as best friend is ostensibly not-a-best-friend, but it is still problematic to interpret such a non-observed-relationship as an observed-non-relationship. Moreover, just as scholars often conflate networks-as-relationships with networks-as-interaction in the positive case (assuming that relationships imply interaction), scholars often commit the same slippage for non-ties (assume that non-observed-relationships imply zero interaction) where this assumption is hardly ever defensible. In many applications of network analysis, researchers interpret a non-tie in a friendship network

272

JAMES A. KITTS

as no-communication, no-liking, no-time-spent-together. This is problematic in the Group Process tradition, where we apply network analysis in small populations. When these network nodes are students in a small class (or coworkers in a project team), where actors spend most of every day together, the interpretation that students can only transmit information to each other through long paths of best friends seems implausible. This is especially true when our measure of ties is a restrictive name generator such as five-closest-friends, and we have no direct observation or validation of non-ties. This kind of measurement error is extremely consequential for most conventional methods of social network analysis, as even a few “false-zeroes” in a sociomatrix can fundamentally distort graph-theoretic metrics like centrality. This problem of interpreting non-reported relationships is highlighted when respondents give discrepant reports about whether or not they are friends with each other (Adams & Moody, 2007; Vaquera & Kao, 2008), resulting in the very common but awkward belief among network analysts that “friendship” is a directed relationship, where person C can be best friends with person D while D has exactly zero relationship with C. Scholars typically assume the one-directional friendship has been measured without error, that the “present” C→D tie works just like any other directed tie and the “absent” D→C tie works just like any other non-tie (Cheadle & Schwadel, 2012; Frank, Muller, & Mueller, 2013; Heidler, Gamper, Herz, & Eßer, 2014; Mouw & Entwisle, 2006).2 With the ubiquitous slip of interpreting relationships as interaction, a scholar would then assume that C spends time with D but D does not spend time with C, C can send information to D but D cannot send information to C, etc. This dilemma gives another reason why the reified relationship graph should not be interpreted as a behavioral interaction graph. In particular, even when relationships plausibly imply social interaction, a non-tie in a relationship graph should not be interpreted as devoid of interaction, exchange, or communication unless these interpretations are also explicitly validated. In summary, role relations were for many decades the easiest and most common measures of social ties, so role relations are well represented in the corpus of classical social network data. Because they operate at a level of abstract concepts, role relations are quite robust to issues of temporal continuity and stability, but they often rely on self-reports, which require some attention to measurement error. There is not much general theory about the social processes underlying role relations. However, the ease of measuring role relations has led researchers historically to regard them as proxies

Beyond Networks in Structural Theories of Exchange

273

for interpersonal sentiments (e.g., applications of structural balance theory), behavioral interaction (e.g., application of social influence or contagion theories), or opportunity structures (e.g., applications of network exchange theory), even though none of these theories has anything to say directly about role relations. Ties in a role relation graph may in some cases correspond roughly to sentiments, interaction, or exchange opportunities
as friends may share positive sentiments and interact regularly, or collaborators may exchange feedback. Non-ties in a role relation graph (non-best-friends, non-coauthors, non-comembers) are rarely informative about sentiments, interaction, or exchange opportunities, and this is a principal obstacle to application of theories about sentiments, interaction, or opportunity structures to networks of role relations. At least, such applications demand attention to how the observe graph satisfies the scope conditions of the theory.

Social Ties as Interpersonal Sentiments Another conceptualization of social networks has focused on interpersonal sentiments, such as liking, love, or respect. This has been seen as an alternative way to ask “Who are your friends?” Notably, direct measure of sentiments gives much more defensible grounds to investigate a theory about interpersonal sentiments, such as work on structural balance (Cartwright & Harary, 1956) or attraction (Smeaton, Byrne, & Murnen, 1989). This measure may be less vulnerable to some of the contaminants for a measure of perceived friendship
such as where individuals share very close positive relationships to their kin and lovers, but may not use the label friends. The sentiment measure may also be more interpretable in the case of disagreement in self-reported ties. Unlike friendship, marriage, or spendingtime-together, interpersonal sentiments may be truly asymmetric so dyadic discrepancies are substantively interpretable. To avoid the ambiguities of role relations like friendship, and to focus on sentiments as an internal directed state rather than behavioral interaction, some researchers have measured aspirations for social interaction. For example, Leinhardt (1972) asked school children “Who would you like to play with?” which is interpretable as a directed sentiment. By contrast, if we ask “Whom do you often play with?” we measure not sentiments but behavioral interaction (which reflect exogenous logistical, spatial, and sociometric factors). Also, the underlying reality for behavioral play must be mutual, whereas the aspirations may be one-sided.3

274

JAMES A. KITTS

A unique strength of sentiment-based survey measures is their capacity to measure negative ties (dislike or disesteem) in a straightforward way, making them properly applicable to theories of network evolution based on structural balance (Cartwright & Harary, 1956; Marvel, Kleinberg, Kleinberg, & Strogatz, 2011). According to this broadly applied theory, social actors feel a drive to change their networks as a result of dissonance implied by having two positive ties to alters who are connected by a negative tie (having two friends who dislike each other), or having a positive and a negative tie to alters who are connected by a positive tie (having a friend who is friends with an enemy).4 Given the challenge of measuring negative relationships in surveys (as respondents are reluctant to describe negativity in their relationships), many scholars in this tradition have eschewed measuring negative ties and have instead employed a conventional name generator for role relations that imply positive sentiments (e.g., best friendship) and interpreted nonmeasured ties as measured-negative sentiment ties. In a representative study, Hallinan (1974) measures such sentiment relations (“best friends”) for students in 51 classes from 14 schools, and (by applying balance theory) implicitly assumes that alters not nominated as best friends must operate as negative sentiment ties. I have already mentioned that interpreting nonmeasured ties as non-ties can often be a consequential mistake, but interpreting non-measured ties as negative ties is a step more extreme in this regard. Non-ties in a graph of relationships (e.g., for individuals not nominated as a friend on the survey) should not be generally interpreted as disliking. This should be read as a challenge to over four decades of studies that have appealed to structural balance theory as an explanation for triad closure or transitivity in positive-sentiment relations (such as friendship) following from Davis (1970) and Holland and Leinhardt (1970) but continuing to today (Frank et al., 2013). It is an even greater challenge to the widespread recent applications of structural balance theory to behavioral interaction networks, such as the phenomenon of triad closure in telephone or e-mail communication networks (Kossinets & Watts, 2009) or Facebook friends (Wimmer & Lewis, 2010). Applications of structural balance theory to positive or neutral social interaction data such as communication similarly assume that non-interaction is equivalent to a negative sentiment tie, an assumption that at least needs to be defended. In a detailed study of triad closure using a combination of longitudinal survey data and wearable sensor data on physical locations, face-to-face conversations, phone conversations, social visits, and work projects, Kitts (2010) identifies boundary conditions in which non-interaction might

Beyond Networks in Structural Theories of Exchange

275

be interpreted as negative sentiments. Outside those narrow boundary conditions, structural balance theory and similar models of relational dissonance should not be applied to positive relationship data or neutral behavioral interaction data. In such cases, we must rely on direct measures of sentiments.

Social Ties as Behavioral Interactions Most foundation social networks research focused on relationships, whether role relations or sentiments (or implicit-sentiment role relations, such as best friends). But many researchers’ substantive interest is in social interaction behavior, as it may transmit diseases, spread innovations, or offer job opportunities. There are reasons for caution in inferring behavioral interaction from observed relationships, as realized behavioral interaction may reflect many different (mostly unmeasured) roles
kin, friend, lover, classmate, teammate, coworker, comember, neighbor
as well as exogenous constraints and inducements due to scheduling and physical space. For these reasons, any role relation (even a strong relationship, such as best friend) will not capture a large portion of social interaction. Respondents may even apply a sentiment-charged label like “best friend” to a rarely seen alter such as a childhood friend. In order to narrow the set of closest friends to an interpretable subset of regular interaction partners, some researchers have included behavioral interaction as part of a selfreport measure of relationships, such as Laumann (1973, p. 264) asking respondents to nominate “the three men who are your closest friends and whom you see most often.” This composite measure narrows possible interpretations and more plausibly captures a set of alters that represent both positive sentiments and interaction. In trying to capture to both dimensions, however, it fails to capture either very well. It is not a measure of the most regular interaction partners (many of whom are not closest friends) or the most positive-sentiment ties (some of whom may not be called friends or may not be seen often). And a non-tie by this measure is hardly interpretable, as it could be a close friend who is not among the most frequently seen or a daily interaction partner who is only a casual friend. Some researchers (Bailey & Marsden, 1999; Ruan, 1998; Uehara, 1990) have avoided ambiguity about role relations and instead employed name generators for specific interaction behaviors. They have asked about help with housework, borrowing money, borrowing household goods, discussing marital problems, discussing feelings of depression, frequently visit socially

276

JAMES A. KITTS

outside the home, or shared social visits, calling these “exchange” name generators. Given the expense and difficulty of employing different name generators to measure networks for specific exchange behaviors, scholars have long hoped for an omnibus sociometric survey question that could capture the concept of strong ties overall, rather than trying to measure each individual exchange behavior. Granovetter’s classic (1973, p. 1361) definition of the strength of ties includes not only the quantity of interaction but the quality of interaction (and associated sentiments): The strength of a tie is a (probably linear) combination of the amount of time, the emotional intensity, the intimacy (mutual confiding), and the reciprocal services which characterize the tie.

Among the most extensive attempts has been the widely-used network measure from the General Social Survey, which asks respondents to think back to the last 6 months and report the names of people with whom they have discussed “important matters” during that time. The measure targets interaction behavior but researchers typically interpret these responses as strong ties (including close friends, and close kin).5 However, just as earlier work had found that we should not see adults’ self-reported friends as strong ties (because of how the label “friend” may be used as a residual category to apply to alters who are not particularly close or important to the respondent), recent empirical research has challenged the interpretation of the standard “core discussion network” name generator as a measure of strong ties: The core discussion network is not a representation of our strong ties; it is a combination of people we are close to, people we are not close to but who are knowledgeable about the matters we regularly find important, and people we are not close to but who are available because of our routine activities. (Small, 2013, p. 481)

In other words, this retrospective aggregation of past interaction reflects respondents’ opportunistic use of available experts (advisors, therapists, accountants, computer support technicians, clergy, physical trainers, medical practitioners, etc.) as well as miscellaneous people who happen to be near us at times when we want a question answered, where in both cases the discussion partners themselves may not be either close or important to us. Using a single survey question to identify a respondent’s social ties in a general way that is robust across cultures, genders, and life stages remains an elusive Holy Grail for social networks researchers. Whether we develop the question to focus on socially constructed relationships, sentiments, or

Beyond Networks in Structural Theories of Exchange

277

social interaction, such a general and robust measure of a social tie has never been found. In some cases, survey researchers (Cornwell & Laumann, 2011; Zachary, 1977) have used a conventional relationship measure but added a second survey question to identify the frequency or intensity of interaction within each tie. That is, respondents’ self-reports of the frequency of interaction are used to represent how strong the observed ties are. Others may ask two independent survey questions, one to measure friend relations and a second for frequency of conversations or advice exchange (Coleman, Katz, & Menzel, 1957). Using two independent graphs allows researchers to observe regular interaction partners who are missed by the “friend” question and differentiate friends who are regular interaction partners from those who are seldom seen. Alternatively, to appreciate the advantages of measuring sentiments directly (avoiding ambiguity of sentiment-charged role relations, and allowing for negative ties)
we can pair a measure of sentiments with an independent measure of behavioral interaction. This will be generally superior to an omnibus relation question or an interaction question alone. It will allow us to distinguish regular interaction partners from parents or childhood friends who not part of the respondent’s day-to-day life, and also allow us to distinguish the level of emotional closeness among regular interaction partners. Oddly, this combination of sentiment and behavioral interaction measures has been used rarely, even in studies collecting sociometric data on several dimensions simultaneously. In a rare and often-cited example, Sampson (1968) measured liking and disliking, esteem and disesteem, positive influence and negative influence, praise and blame for a study of 18 monks in a monastery. As social network lenses spread across scientific disciplines, many scholars are interested in social interaction behavior not as a feature of relationships but as a phenomenon in itself. Increasingly, researchers are employing direct measures of social interaction and constructing networks from the interaction data, but not measuring perceived relationships at all. Such researchers use social interaction as the operational definition of a social tie; that is, our ties are those others with whom we interact. When we are studying the diffusion of HIV on a network of partners in sex or intravenous drug use, a constructed network of those actual interpersonal risk behaviors may be a better focus of our attention than their perceived relationships. Scholars interested in the underlying behaviors of social interaction may simply observe interaction over time and define a network as an

278

JAMES A. KITTS

aggregation of past interpersonal behavior. Qualitative field researchers informally aggregate observed sequences of interactions and interpret these aggregated interactions as measures of ties (Vargas, 2011). For example, Roethlisberger and Dickson’s Bank Wiring Room study (1939) involved observations of various kinds of relational behavior (horseplay, arguments, helping, job trading) among employees. In his classic analysis of these data in The Human Group, George Homans (1950) treated these aggregated observations of interactions analytically as binary relationships. Just as the GSS asks the survey respondent to aggregate over a 6-month time window, the field observer similarly aggregates over a time window, ultimately turning interaction histories into inferred ties that may be analyzed using conventional tools of social network analysis. Unlike the survey data, however, the networks constructed from observed behavior can use a specific and rigorously applied definition of social ties, not vulnerable to differences of construal among the various survey respondents. Real-time applications can use multiple observers and sophisticated coding schemes to record social interaction for later aggregation into networks, and networks can even be constructed and analyzed from ethnographic or archived historical accounts (Heidler et al., 2014). While employing the same temporal aggregation of interaction events, it is possible to cut a long time period into slices, defining a series of panel observations of networks. Moody, McFarland, and Bender-deMoll (2005) proposed ways of aggregating interaction behavior in time and then visualizing the resulting networks as “flip-books” or “movies” that represent a changing network over time. Following from Moody et al.’s “moving windows” approach to converting relational events into networks, Kossinets and Watts (2009) explore thresholds for defining network snapshots based on temporally aggregated e-mail exchanges, then investigate dynamic changes in these networks.

Social Ties as Opportunity Structures Researchers in sociological exchange theory have offered valuable insights into the dynamics of exchange given exogenous opportunity structures (Cook et al., 1983) or restricted access networks (Marsden, 1983). For example, they have investigated the structural foundations of power, often focusing on negotiated exchange in “negatively connected” networks, where alternative exchange partners are mutually exclusive. In this research, there are elegant links between basic constructs, theoretical principles, and empirical findings. Ties and non-ties are unambiguously

Beyond Networks in Structural Theories of Exchange

279

controlled by the investigator in laboratory experiments. In this usage, a tie is conceived as a location where exchange is possible, whether or not it actually occurs. For example, if A and B are tied and B and C are tied but A and C are not tied, then B can choose to interact with A or C (but A and C can only choose to interact with B). In this case, non-ties are precisely defined, whereas ties are dyads where exchange might occur. In the words of Thye, Lawler, and Yoon (2011, p. 407), “the network itself does nothing for individuals except generate a series of opportunities for and constraints on dyadic exchange.” The opportunity structures studied in network exchange theory
properly called “exchange networks” (Markovsky, Willer, & Patton, 1988; Willer, 1999)
may bear little resemblance to social networks as conventionally observed in studies of interaction and relationships. Studies of relationships aim to capture the network of socially constructed links among people (kin, coworkers, best friends, etc.) whereas observational studies of social interaction aim to capture behavioral exchanges that actually occur among individuals. Both approaches focus on the presence of ties and give little attention to non-ties. Neither approach allows observation of impossible partners. For example, in a deep study of the GSS name generator, Bearman and Parigi (2004) demonstrate that reporting zero peers as important discussion partners need not imply that respondents do not have any partners available. Research on role relations, sentiments, or aggregations of observed behavior relies on a strong definition of ties with a weak or implicit definition of non-ties, and hardly ever validates non-ties as unavailable. By contrast, research on opportunity structures relies on a weak definition of ties with a strong interpretation of non-ties: Ties are dyads where exchange might occur, and actors cannot exchange through non-ties. This mismatch creates an obstacle for empirical application of exchange theories that focus so much weight on the assumption that exchange is impossible outside observed network ties. Indeed, empirical situations closest to exchange network studies are where there are complete barriers to interaction in some dyads (but not others). It is difficult to identify such situations, especially in contexts of interest to Group Process scholars, such as networks of friends or discussion partners in organizations.6 This raises a dilemma for empirical application of network exchange research, because many real-world informal social structures appear to be governed by voluntary choices, where alters may be more or less attractive or familiar but none are explicitly prohibited. Rather than a particular alter being available (tied) or strictly impossible (not tied) as in laboratory studies, empirical restricted access networks more often represent a continuum

280

JAMES A. KITTS

of accessibility. For example, alternative partners may vary in attractiveness, resourcefulness, geographic distance, environmental obstacles, regulatory restrictions, convenience of communication, or transaction costs. Indeed, Marsden (1983, pp. 690
691) explicitly noted that geographic distance, organizational structure, or other impediments (even “inertia” or “brand loyalty” in alternative partnerships)7 can block exchange in a particular dyad. Lawler et al. (2006) recently made a distinction between exchange relations that are forced by a lack of alternatives (i.e., the experimental design makes exchange unavailable in some dyads) versus exchange relations that are preferred by individuals because they provide superior terms of exchange. This distinction is easy to perform in the laboratory and may occur in the natural world primarily in fine time grain interactions due to scheduling constraints; for example, if B and C are prom dates on Friday night, then A or D cannot attend the prom with either B or C, so in this sense the A
B, A
C, B
D, or C
D interactions are rendered logistically impossible at that time. (Lawler et al. use a similar example of dyadic conversations occurring in two rooms at the same time.) Where interactions are not simultaneous, such logistical impossibilities are hard to identify. In the natural world it is rare for empirical exchange to be exogenously prohibited in any dyad. In coarse time grain, even explicit B
C commitments (say, B and C are married, which is hardly an exogenous constraint) are notoriously ineffective in prohibiting A
B contacts if the parties so prefer, provided that A and B can schedule an encounter when C is unable to observe their interaction. Actors may choose to refrain from exchanging with some others because of direct terms of exchange, or due to inertia that takes the form of emotional commitment or path dependent routines, or for strategic reasons (e.g., to avoid helping an enemy), all impediments that could be represented in the incentive structure. Even explicit prohibitions, such as state embargos, antitrust regulations, or restraining orders could be interpreted as disincentives rather than strict barriers to exchange, and contraband exchanges such as adultery, illegal drug sales, or political bribes are empirically common. Recognizing that these issues combine to form a continuum of constraints or incentives tends to blur the distinction between induced and enabled exchange relations, and whether a constraint is regarded as a structural disincentive or a prohibition is often a matter of framing. This guides our interpretation of Lawler’s choice-process theory to either simultaneous interactions such as those in the laboratory (where barriers make exchange logistically impossible) or to situations where

Beyond Networks in Structural Theories of Exchange

281

the incentive structure is framed as an externally binding situational constraint.

COMPUTATIONAL SOCIAL SCIENCE: NEW LENSES FOR STUDYING BEHAVIORAL INTERACTION The advent of Computational Social Science (Lazer et al., 2009) has enabled recording of “Big Data” on social activities of millions of people. Among the most easily accessible are online contact lists, such as Facebook “friends,” Twitter “followers,” Google “circles,” or LinkedIn “connections.” Unfortunately, many or most such names on contact lists are not significant as either relationships or social interaction, as evidenced from research on Facebook “friends” (Golder, Wilkinson, & Huberman, 2007) and Twitter “followers” (Huberman, Romero, & Wu, 2009). Use patterns vary, and some such online contacts might indeed interact socially or have relationships with one another deeper than a contact list on the website, but many or most such online contacts apparently never interact (even on the website itself) and some are not even people. To carve through the junk data, scholars interested in relationships may use additional filters to identify significant links on online contact lists that may point to substantive relationships or interaction partners. For example, Wimmer and Lewis (2010) narrow their scope to Facebook friends that are colocated as students at the same university and appear in each other’s tagged photos, a subpopulation of humans who are likely to interact at least occasionally (Lewis, Kaufman, Gonzalez, Wimmer, & Christakis, 2008). Golder et al. (2007) recommend focusing a network analysis of Facebook on the small subset of “friends” who send each other electronic messages or comment on each other’s materials. Whatever the meaning of these easily available user contact lists, industry or organizational partnerships can offer fine-grained privacysensitive data about social interactions. For example, researchers can analyze interactions through time-stamped records created by e-mail servers (Kossinets & Watts, 2009; Quintane & Kleinbaum, 2011), phone call logs (Onnela et al., 2007), radio communication transcripts (Butts, Petrescu-Prahova, & Cross, 2007), shared online calendars (Lovett, O’Neill, Irwin, & Pollington, 2010), or wearable sensors that detect face-toface conversations (Wyatt, Choudhury, Bilmes, & Kitts, 2008, 2011) or physical proximity (Eagle, Pentland, & Lazer, 2009; Ingram & Morris,

282

JAMES A. KITTS

2007).8 Online arenas for dating (Lin & Lundquist, 2013), gaming (Szell & Thurner, 2010), exchange (Cheshire & Cook, 2004), and scholarly citations (Shwed & Bearman, 2010) also provide time-stamped event records that may represent instantaneous transactions organized according to finegrained social dynamics. Recall that traditional network analysis concepts and tools were developed for the study of temporally continuous and stable relationships, which construe networks as practically timeless abstractions (Gibson, 2005). Facing this incongruity between traditional sociometric tools and the world of time-stamped relational event data, CSS researchers who collect finegrained interaction data have the problem of transforming their data to be conformable to tools derived from an age of sociometric surveys, simple graphs, and sociomatrices. For example, given a rich event history of contacts (phone conversations, e-mails, face-to-face conversations) for a set of actors, a researcher might aggregate the events into a simple matrix of counts, and then further apply a threshold filter: For a given period of time, more than n contacts within a period of time may be defined as a tie and fewer than n contacts may be defined as no tie. For example, Onnela et al. (2007) constructed a graph of interaction partners for millions of people using temporally aggregated records of calls on a cellular phone network, and Wyatt et al. (2008, 2011) used wearable sensors to record face-to-face conversations for a cohort of graduate students during a school year, but similarly aggregated rich micro-interaction data over time, and then treated those temporally aggregated interactions as if they were relationships.9 Conaldi and Lomi (2013) observed the individual bug-fixing activities of software developers and Papachristos, Hureau, and Braga (2013) observe individual gang homicides, but all conventionally aggregate these behaviors into temporally continuous and stable social networks. Such aggregation of contacts may seem ad hoc, but see that it is in fact a more extensive and systematic version of the longstanding approach by qualitative field researchers who have tried to identify relations from observing and aggregating interaction histories. It is also analogous to the ways that sociometric surveys ask respondents to mentally aggregate over their own interaction histories to identify their important discussion partners. The new methods for automatically recording social interactions at least allow the aggregation of interaction data to be done in a transparent and rigorous way to avoid idiosyncrasies or inconsistencies in how observers or survey respondents perform this aggregation. Of course, just as we should be cautious about interpreting sociometric relationship data as measuring social interaction, we should be cautious

Beyond Networks in Structural Theories of Exchange

283

about interpreting aggregated interaction data as substantive relationships. Perusing our own records of e-mail or phone communication will illustrate that counts of contact events do not necessarily indicate our closest ties. Also aggregating events into counts destroys valuable information about temporal dynamics and action sequences. Although converting these finegrained interaction dynamics into coarse ties is ubiquitous and unquestioned, it is not inevitable. In fact, the leap from conventional sociometric research to the world of Big (relational) Data invites us to deeply consider issues such as time grain in measurement, time frames of underlying social processes (Kitts, 2009; Quintane, Carnabuci, Robins, & Pattison, 2012), and short-term dynamics of interaction event sequences (Gibson, 2005; Kitts, Lomi, Pallotti, Mascia, & Quintane, 2013). We turn to these next.

BEYOND NETWORKS: CONSIDERING THE STRUCTURE AND DYNAMICS OF INTERACTION There is clear value in studying relationships as socially constructed entities (including actors’ perceptions of their friendships). There is also clear value in studying how opportunity structures for exchange affect important outcomes such as interpersonal power (Emerson, 1972) and commitment (Lawler, Thye, & Yoon, 2000). However, I advocate for a new generation of theoretical development and analysis of interaction dynamics. This includes direct measurement of behavioral social interaction (including time diaries, wearable sensors, and archival records such as telecommunication or online exchange) and should also include direct modeling of the structural and temporal dependencies of this relational behavior. Moving beyond temporally aggregating interaction behavior into ties will require employing and extending tools for dynamic relational data analysis. Fortuitously, there is a convergence in statistical methodology that is poised to address the wealth of longitudinal interaction data: Social network analysis, which models interconnections among actors, is being extended to consider changes over time (Desmarais & Cranmer, 2012; Snijders, Van de Bunt, & Steglich, 2010). Event history analysis, which models rates of events occurring as they may depend on the environment, including other events, is being extended to consider forms of statistical dependence across interconnected actors (Stewart, 2005). This confluence
recently articulated as relational event modeling (Butts, 2008; Stadtfeld, 2012) and applied to radio communications in the WTC disaster (Butts,

284

JAMES A. KITTS

2008), postings in online Q&A communities (Stadtfeld & Geyer-Schulz, 2011), critics’ reviews of new books (De Nooy, 2011), and reciprocity and generalized exchange in patient transfers among hospitals (Kitts et al., 2013)
allows us to consider the dynamic interdependence of interaction behavior for a set of social actors in continuous time. As we consider a dynamic structural alternative to the static social network metaphor, we will observe another blind spot of typical relationship measures, which points to a constructive solution for several dilemmas raised here: Our use of relationships in dynamic theory and analysis is inherently limited because relationships are typically divorced from time. It is difficult or impossible to identify a specific time when a relationship starts and stops, as they are cognitive categorizations of role relations and/ or interpersonal sentiments, not specifically linked to time. Even for the rare exceptions, such as legal marriage, the observable beginning and ending of the relationship (marriage and divorce) may correspond only weakly to the dynamic patterns of behavior and emotions assumed to underlie the network. Similarly, coauthorship is a socially constructed role relation but implies interaction over time. Moody (2004) observes a collaboration outcome
a coauthored paper
but this event identifies a coauthor relationship only after a delay, when interaction behavior may be finished. In this way, a network is constructed out of the set of coauthorship relations in a literature, although the underlying sequence and timing of interactions is still unknown. The socially constructed categories that actors apply to their interpersonal lives are worthy of further study. However, analyzing the fine-grained dynamics of social interaction will give us more purchase on the social processes underlying what we intuitively understand as social networks, rather than aggregating interaction data and interpreting them as social ties. The perspective that I outline here is an important step to realizing the goals articulated by Walker et al. (2000, p. 333) that network exchange theorists “should begin to focus on network processes of self-organization, adaptation, and feedback.” We now have an opportunity to directly theorize, measure, analyze, and interpret patterns in the structure and dynamics of exchange. Extending the same conceptual tools that have proven useful in the study of social networks, we can focus our lenses on dynamic patterns in the structure of interaction (or structural patterns in the dynamics of interaction). Rather than studying reciprocity as a state (i.e., some number of ties are mutual in the graph), we can study the dynamic process by which individuals reciprocate communications, gifts, support, or other goods. The perspectives

Beyond Networks in Structural Theories of Exchange

285

described here will bring our theories and field research on social interaction into closer dialog with ethnographic fieldwork on social interaction and also allow a clearer link to research in controlled laboratory settings. These analytical lenses provide a new way of thinking about ties, as interaction events occurring in continuous time, depending on the history of previous interaction events and on states of the environment. Rather than defining an arbitrary threshold to identify non-ties (if interaction is not frequent enough), we can directly observe and analyze delays between interaction events. In the previous view of networks-as-aggregatedinteraction, authors faced an often-unacknowledged dependence between the time grain of their aggregation (daily, weekly, monthly, yearly) and the structure and dynamics of their networks: If a researcher computes weekly interaction “networks” based on an aggregation of interaction events and applies a threshold (i.e., a tie is more than two conversations in a week), then ties will appear for some weeks and not for other weeks, and thus the apparent network is constantly changing. Using a coarser time grain (monthly or yearly) will result in a denser graph than using a finer time grain (daily or weekly) and will also result in a more stable structure. Such networks can thus be sensitive to arbitrary details of aggregation, and developing the most robust images of the overall structure (by more aggregation) will necessarily destroy most details of the sequence and timing of interaction. For moving beyond the laboratory to observation of natural settings, I have instead argued for capturing and analyzing time-stamped interaction data, directly modeling the interaction event rates for all dyads. This will take advantage of all information we have about histories of exchanges, including delays between exchanges, dyadic and higher-order time dependence in exchanges (i may be more likely to give to j again if j reciprocates quickly versus slowly, or if j has a longer history of exchange with another actor, k). It also allows for characteristic structural sequences of interaction to be observed and analyzed. All this information is lost when we aggregate a history of time-specific social interactions into a matrix of assumed relationships (Moody, 2002). Indeed, producing a static network by aggregating over a sequence of contacts
such as sex in a high school (Bearman, Moody, & Stovel, 2004)
can lose crucial insights into social processes such as diffusion on the network. I have explained that an important reason to move “beyond networks” in considering structural dynamics (or dynamic structures) is that we can consider the nuances of interaction patterns in time. Now I add that the same benefit applies to analysis of opportunity structures under this

286

JAMES A. KITTS

framework. The analytical perspective described here can advance the work in sociological exchange theory by providing a seamless and integrated way to monitor and model the rates of exchange across dyads as well as the constraints on exchange. This avoids arbitrarily defining ties versus non-ties in the observed interaction network (instead viewing them as continuous rates of interaction, or impediments to interaction) and avoids arbitrarily defining opportunity structures based on historically observed exchange events. The latter is a crucial generalization to a world where impediments to exchange may be relative rather than absolute. It may be that impediments make some exchanges relatively difficult or undesirable, but not strictly impossible, and those relative difficulties may be incorporated as continuous variation in the availability of partners (and alternative partners) within a model of exchange events. In other words, the simplifying assumption that drives much of the work in network exchange theory
that exchange relations are either on (available) or off (impossible)
is unnecessary under this framework. A matrix of dyadic geographic distances does represent a structure, and under the assumption that travel is costly could even be interpreted as (an input to) an opportunity structure for interaction. For example individuals who work in different buildings, have different schedules, speak different languages, or companies that operate in different regions or countries, may face higher transaction costs for exchange that make them less attractive (but not impossible) exchange partners. Rather than interpreting the observed history of interaction as an exogenous opportunity structure of ties and non-ties, the framework that I discuss here could be used to implement impediments or facilitators as continuous influences. It is a straightforward extension of formative ideas in network exchange theory (Marsden, 1983) to note that restricted access is actually a continuum, which makes exchange more or less feasible in a given dyad. Importantly, this generalization will bring us a step toward applying network exchange theory to behavioral interaction networks. The theoretical and analytical lenses described here can be applied to discover the interplay of dynamic social interaction patterns with coarser socially constructed relationships, interpersonal sentiments, or opportunity structures over time, all largely unexplored frontiers. For example, individuals may strategically alter their networks of interaction to enhance their own (or neutralize someone else’s) structural power in exchange. This could build on some earlier work that has combined methods
such as survey measures of relationships with time diary measures of interaction histories (Milardo, Johnson, & Huston, 1983; Nezlek, 1993). The age of Big Data offers tremendous leverage to extend this research program, such as

Beyond Networks in Structural Theories of Exchange

287

combining sociometric surveys with direct measures of social interaction using wearable sensors of proximity (Eagle et al., 2009) and conversation detection (Wyatt et al., 2008, 2011) which allow us to model interdependence of exchange behavior over short and long time-spans.10

CONCLUSION Social networks research has employed four qualitatively different approaches to the basic concept of the social tie
where ties represent role relations, interpersonal sentiments, behavioral interaction, or opportunity structures. Basic process theories have been developed primarily for sentiment structures (e.g., structural balance theory), interaction (e.g., social influence or contagion theories), and opportunity structures (e.g., network exchange theory). Although early work in Group Process explicitly targeted the dynamic interdependence of these structures, especially social interaction and sentiments (e.g., Homans, 1950; Newcomb, 1961), there has been much more progress within each domain than integrative work on how they fit together. This limited attention to how our concepts interrelate has impeded theoretical integration across distinct research programs and has occasionally led theorists in one camp to inappropriately generalize their arguments to empirical networks where their theories do not make sense (such as widespread application of structural balance theory to triad closure in neutral interaction networks). We must clarify what we all mean by social ties, and begin the process of identifying scope conditions under which theoretical arguments generalize across different network conceptualizations. Much empirical work has employed measures of role relations (such as coauthors, comembers, or friends), which have historically been the easiest to measure, and often serve as a proxy in any kind of network study. However, the connection of role relations to sentiments, interaction, and opportunity structures is implicit, inconsistent, and often unclear, even for relations such as close friends that ostensibly imply positive sentiments and behavioral interaction. We have seen many reasons why role relations such as friendship are often inadequate as general measures of behavioral interaction (because most social interaction occurs outside the measured relations, and friendship often misses the strongest interaction ties), sentiments (because role relation measures may similarly miss the strongest sentiments), or opportunity structures for exchange (which role relations

288

JAMES A. KITTS

do not even attempt to measure). The fact that conventional relationship measures typically give no attention to non-relationships is a principal obstacle to generalizing to the other network concepts, where identifying dyads with negative sentiments, zero interaction, and blocked exchange may be crucial. Minimally, I have shown that strong assumptions need to be explicitly defended (especially about non-ties) to link theoretical propositions and findings across these levels. I thus begin the process of identifying scope conditions for which role relations may be applicable to general theories about network dynamics. Increasingly, empirical work has constructed networks directly from behavioral interaction, such as communication or participation in shared activities, and these are becoming a new generic way to measure social ties. This tradition often aggregates interaction events (e.g., citations, e-mails, phone calls, or face-to-face conversations) over time and treats the resulting social ties as a proxy in any kind of network study. However, aggregated behavioral interactions also are not general measures of role relations, sentiments, or opportunity structures. Again, I begin the process for identifying scope conditions in which networks of interaction events may be applicable to our general theories about network dynamics. Rather than continuing to organize our work using under-theorized and vague concepts like social ties or even friendships, I have advocated for maintaining distinct focus on social interaction behavior, interpersonal sentiments, and opportunity structures
none of which are generally represented by composite relationship measures. All three concepts play crucial roles in distinct bodies of theory, and can be operationalized clearly. Through studying them independently and jointly, we can understand their unique and interactive dynamics. Traditional tools such as survey research are still important to allow us to simply measure both sentiments and interactions, and remain the bestdeveloped ways to examine negative sentiments. Measures of sentiment can be used fruitfully in tandem with conventional measures of role relations and behavioral interaction. Assessing sentiments and interaction independently would give us a more strictly interpretable measure of strong ties in terms of emotional closeness and frequency of interaction, avoiding many problems of interpreting self-reported role relations (such as friendship) or core discussion partners as an omnibus measure of strong ties. This would also allow us to differentiate non-interaction from disliking, neither of which is measured from a conventional friends or discussion network name generator (and both of which are ubiquitous misinterpretations of nonobserved ties in conventional relationship surveys).

Beyond Networks in Structural Theories of Exchange

289

As for the recent explosion of time-stamped longitudinal relational data (e-mails, phone calls, meetings, wearable sensors measuring colocation and face-to-face conversations) in CSS, I note that early work has aggregated these rich sources of interaction data into simplistic sociomatrices in order to apply classic network theories, concepts, measures, and methods. Aggregating the micro-social dynamics of interaction into coarse ties in networks follows a long history of such aggregation (whether the aggregation is performed implicitly by field observers or by respondents themselves in surveys). It is also motivated by a longstanding interest in durable structural patterns (such as who typically interacts with whom) rather than the fine-grained dynamics of their interaction within that structure. The new technologies of data collection and analysis make this distinction much less important, and allow us to investigate social dynamics within particular encounters using the same lenses that we use to study the evolution of typical-structures over long time-spans. I have advocated going beyond the traditional concept of the social network as a temporally continuous and stable structure. Rather than imposing the coarse-grained lens of locally stable networks to observe the structure of micro-interactions, I argue to generalize the fine-grained lens of micro-interaction research to longer time scales (constituting the network). The universe of time-stamped relational data and the lenses of dynamic relational event analysis make this possible for the first time. I have thus strongly advocated for analysis of the dynamic dependencies in interaction behavior using newly developed eventbased frameworks. In moving beyond “network ties,” we can now look at the short-term and long-term temporal dependencies of reciprocity and generalized exchange (Kitts et al., 2013), deference or dominance behavior leading to social hierarchies (Martin, 2011), and the continuous development of commitment in real-world exchange. Now that we are able to move beyond the simplifying assumption that social ties are temporally continuous and stable binary states, we are ready for a qualitative shift in integrative theory development. For example, I have discussed sociological exchange theory, in which non-ties are dyads where exchange is exogenously blocked by the researcher in experiments, versus ties where exchange is allowed and may or may not occur. I show that this usage does not correspond to any of the traditional methods for conceptualizing or measuring interpersonal networks in natural settings. However, the new CSS lenses for monitoring and analyzing the dynamics of exchange provide a natural next step for carrying valuable insights from exchange theory out of the laboratory and into dialog with other networks research. Both the dynamic regularities in exchange and continuous or

290

JAMES A. KITTS

discrete constraints on exchange can be incorporated into the event-based lens. Barriers or impediments to exchange, such as geographic distance or resource complementarities that affect exchange values, as well as endogenous network processes that facilitate or obstruct exchange, can be explicitly modeled within this framework. In discussing the four versions of social networks, I have often emphasized careful study of the dynamics of interaction behavior not because behavior is inherently more important than role relations, sentiments, or opportunity structures. Rather, it seems that a surge in empirical research on the dynamics of interpersonal behavior is inevitable: The advent of CSS allows us to easily collect copious and fine-grained data on behavioral interaction, whereas sentiments, socially constructed role relations, and opportunity structures may still be challenging to collect for most social contexts. This leaves us with two crucial frontiers: First, to the extent that we take advantage of the new wealth of behavioral interaction data, we can translate and extend some of our general theories for sentiments and opportunity structures by deriving implications for interaction behavior. We should do so carefully, with assiduous attention to scope conditions as discussed here. Second, we can develop fine-grained, longitudinal, scalable measures for sentiments, opportunity structures, and role relations. For example, we could go beyond surveys to employ automatic indicators of affect based on nonverbal behavior (eye gaze, facial expressions, body spacing, posture, gestures, speech prosody, response latency, etc.), biometric indicators (brain imaging, hormone analysis, etc.) or natural language processing of electronic messages. Notably, some online spaces already allow us to observe populations of individuals interacting over time while also evaluating each other, and some may even allow experimental manipulation of opportunity structures by researchers. We must of course scrutinize the generalizability of findings from research in these settings, where both interaction and sentiments can be quite thin, but the opportunity for largescale observational and experimental research is unprecedented. I have addressed a broad range of social networks research, mostly outside the Group Process community. As networks research has exploded into thousands of papers across many disciplines, and the formal tools of network analysis have proven powerful in countless domains, the extension of social ties to any kind of relational data has often led to uncritical application of theories, slowing theoretical integration. Having once been a major part of the development of social network theory and analysis, the Group Process community has lost its prominent role as much networks research moved to scales beyond small groups. However, Group Process remains a

Beyond Networks in Structural Theories of Exchange

291

valuable domain for thinking about how the basic process elements of social networks research
role relations, interpersonal sentiments, behavioral interaction, and opportunity structures
interrelate more generally. For example, Lawler’s theory of relational cohesion (Thye, Yoon, & Lawler, 2002) combines all four elements: An exogenous network exists as a prior condition (implemented as an opportunity structure in the laboratory), interaction occurs in some ties on that network, and then emotions develop as a result of that interaction, resulting in further commitment and objectification of those ties as perceived relationships. Molm’s work similarly examines opportunity structures, interactions, and sentiments in clear experiments. For example, Molm, Collett, and Schaefer (2007) experimentally manipulate “forms of exchange” (which are effectively exogenous constraints on timing of exchange opportunities in the set of participants) to consider the impact on sentiments toward exchange partners. Just as new data sources and methodological advances of CSS allow us to investigate the fine-grained structural and temporal dependencies of the social world at various spatial and temporal scales, research on social exchange is increasingly focusing on temporal dynamics and history dependence in both experimental (Kuwabara & Sheldon, 2012; Molm, Whitham, & Melamed, 2012; Schaefer, 2012) and observational (Kitts et al., 2013; Willer, Sharkey, & Frey, 2012) work. This sensitivity to interdependent dynamics sets the stage for fruitful dialog with the broader community in social networks and CSS, which can also develop large-scale experiments in online spaces (Centola, 2010, 2011) Here I have aimed to at least temporarily suspend our use of the vacuous concept social ties, and instead resume our attention to the interplay of role relations, interpersonal sentiments, behavioral interaction, and opportunity structures, which have for decades formed a tractable foundation for work in the Group Process tradition.

NOTES 1. Whether network centrality is a source of power or weakness depends on our definition of ties, as information conduits (Bonacich, 1987) or exchange opportunities in a negatively connected exchange network (Cook, Emerson, Gillmore, & Yamagishi, 1983). This issue is orthogonal to my focus here. 2. Of course, some relationships (advisor-advisee, patron-client, teacher-student) are directed. Given that friendship is likely understood by survey respondents as a mutual relationship, an alternative approach is to regard the disagreement as reflecting measurement error. Discrepant friendship self-reports might indicate

292

JAMES A. KITTS

unbalanced sentiments within a dyad (not the same thing as a one-directional relationship), but more likely it could indicate different construal of the survey question, different use of the label “friend” (vs. lover, brother, colleague, neighbor, teammate), context effects on the salience of alters during the survey, or truncation effects due to a cap on the number of alters in a name generator. One solution (e.g., South & Haynie, 2004) is to interpret the non-tie as error and reciprocate the relationship. Another solution (e.g., Goodreau, Kitts, & Morris, 2009; Schaefer, Simpkins, Vest, & Price, 2011; Young, 2011) is to study only corroborated mutual friendship reports. 3. We can hardly interpret dyadic disagreements in behavior self-reports as directed sentiments, as disagreements could be due to different construal of words like often and play, differences in recall of social encounters, or differences in popularity that affect the alter’s salience. 4. Modeling research has shown that such patterns can be explained parsimoniously by dynamics at the dyad level (Faust, 2007; Kitts, 2006; Macy, Kitts, Flache & Benard, 2003), without the strong information conditions required for structural balancing processes to operate in triads. However, the assumption that ties are explicitly negative (not merely less-positive) is essential. 5. It is important to note that the GSS is an ego-network study, which measures a small sample of Ego’s discussion partners, with non-ties (to the rest of the population, including other respondents and their alters) undefined. Thus, researchers typically focus on characteristics of the alters in the observed sample, such as their demographic composition. In rare exceptions, researchers (e.g., McPherson, SmithLovin, & Brashears, 2006; cf. Paik & Sanchagrin, 2013) have applied the interpretation from complete-network studies to GSS ego-network data, assuming that the entire population of unmentioned potential alters are non-ties (unavailable for discussing important matters) because they are not mentioned on the name generator. 6. This may be a reason why so many empirical applications of sociological exchange theory have been macro-level studies of organizations or states, where there are more concrete and observable indicators of non-ties as prohibited interaction (Webster & Whitmeyer, 2001). 7. We could also refer to this brand loyalty as “commitment” in the exchange relation (Lawler, Thye, & Yoon, 2006), and for most purposes it makes sense to regard this increasing development of an exclusive exchange relation between A and B as an endogenous process, rather than as an exogenous constraint. 8. Measures of proximity (bluetooth, infrared radiation, localization by GPS, wifi, or cell tower) have high false-positive rates for detecting social interaction because people are often colocated without interacting. Meetings on shared electronic calendars produce many false positives and false negatives (Lovett et al., 2010), as people use calendars as reminders or to-do lists, often miss meetings on their calendar, attend meetings without RSVPing, or otherwise do not fit the locations listed on their calendars. Even accurate measures of colocation do not necessarily capture social interaction. Thus, these methods need to be combined with some other filter or hand-coded to identify which links represent realized social interaction. For example, Wyatt et al. (2008, 2011) combined physical colocation with automatic detection of conversations in audio recordings to identify interaction with a higher level of accuracy.

Beyond Networks in Structural Theories of Exchange

293

9. Audiences seem driven to interpret phone or face-to-face conversations as directed relations, just as they are driven to interpret friendship nominations on a survey as directed relationships. Although it may be meaningful to interpret conversations in some sense as directed (one person speaks first, one person dials the phone or approaches the other to initiate a conversation, and of course each utterance is directed from one party to the other), conversations themselves are generally undirected. Both parties speak and listen. 10. An approach that blurs the distinction I am making here is a fine-grained network panel approach. For example, Almquist and Butts (2013) observe a set of political blogs with snapshots of the network of links at 484 time points. Where ties are composed of time-stamped interaction events (such as phone calls or e-mails) aggregated over a time interval, reducing the interval width will make the resulting networks sparser until their observable structure disappears altogether. By contrast, some ties (web hyperlinks, marriages, corporate board overlaps) are temporally continuous so the network can be captured by an instantaneous snapshot at any time. In that case, increasing the number of snapshots by shortening the interval width will enhance the resolution of panel data, allowing us to observe the changes in the network as the length of time intervals becomes short. The appropriateness of the panel versus event-based approach then depends importantly on the assumption of temporal continuity.

ACKNOWLEDGMENTS This research was supported by the National Science Foundation (BCS0433086, IIS-0433637). Eric Quintane, David Schaefer, James Moody, and Carter Butts provided helpful comments.

REFERENCES Adams, J., & Moody, J. (2007). To tell the truth: Measuring concordance in multiply reported network data. Social Networks, 29, 44
58. Almquist, Z. W., & Butts, C. T. (2013). Dynamic network logistic regression: A logistic choice analysis of inter-and intra-group blog citation dynamics in the 2004 US presidential election. Political Analysis, 21(4), 430
448. Bailey, S., & Marsden, P. V. (1999). Interpretation and interview context: Examining the general social survey name generator using cognitive methods. Social Networks, 21(3), 287
309. Bearman, P. S., Moody, J., & Stovel, K. (2004). Chains of affection: The structure of adolescent romantic and sexual networks. American Journal of Sociology, 110(1), 44
91. Bearman, P. S., & Parigi, P. (2004). Cloning headless frogs and other important matters: Conversation topics and network structure. Social Forces, 83(2), 535
557.

294

JAMES A. KITTS

Bonacich, P. (1987). Power and centrality: A family of measures. American Journal of Sociology, 92(5), 1170
1182. Burt, R. S. (1992). Structural holes: The social structure of competition. Cambridge, MA: Harvard. Butts, C. T. (2008). A relational event framework for social action. Sociological Methodology, 38(1), 155
200. Butts, C. T., Petrescu-Prahova, M., & Cross, B. R. (2007). Responder communication networks in the World Trade Center disaster: Implications for modeling of communication within emergency settings. Journal of Mathematical Sociology, 31, 121
147. Cartwright, D., & Harary, F. (1956). Structural balance: A generalization of Heider’s theory. Psychological Review, 63, 277
293. Centola, D. (2010). The spread of behavior in an online social network experiment. Science, 329, 1194
1197. Centola, D. (2011). An experimental study of homophily in the adoption of health behavior. Science, 334, 1269
1272. Cheadle, J. E., & Schwadel, P. (2012). The ‘friendship dynamics of religion,’ or the ‘religious dynamics of friendship’? A social network analysis of adolescents who attend small schools. Social Science Research, 41(5), 1198
1212. Cheshire, C., & Cook, K. S. (2004). The emergence of trust networks under uncertainty
implications for internet interactions. Analyse & Kritik, 26(1), 220
240. Coleman, J., Katz, E., & Menzel, H. (1957). The diffusion of an innovation among physicians. Sociometry, 20, 253
270. Conaldi, G., & Lomi, A. (2013). The dual network structure of organizational problem solving: A case study on open source software development. Social Networks, 35, 237
250. Cook, K. S., Emerson, R. M., Gillmore, M. R., & Yamagishi, T. (1983). The distribution of power in exchange networks: Theory and experimental results. American Journal of Sociology, 89, 275
305. Cornwell, B., & Laumann, E. O. (2011). Network position and sexual dysfunction: Implications of partner betweenness for men. American Journal of Sociology, 117, 172
208. Davis, J. A. (1970). Clustering and hierarchy in interpersonal relations: Testing two graph theoretical models on 742 sociomatrices. American Sociological Review, 35(5), 843
851. De Nooy, W. (2011). Networks of action and events over time. A multilevel discrete-time event history model for longitudinal network data. Social Networks, 33(1), 31
40. Desmarais, B. A., & Cranmer, S. J. (2012). Micro-level interpretation of exponential random graph models with application to estuary networks. Policy Studies Journal, 40(3), 402
434. Eagle, N., Pentland, A., & Lazer, D. (2009). Inferring social network structure using mobile phone data. Proceedings of the National Academy of Sciences (PNAS), 106(36), 15274
15278. Emerson, R. M. (1972). Exchange theory, part II: Exchange relations and networks. In J. Berger, M. Zelditch Jr., & B. Anderson (Eds.), Sociological theories in progress. Boston: Houghton-Mifflin. Faust, K. (2007). Very local structure in social networks. Sociological Methodology, 37, 209
256. Fischer, C. S. (1982). What do we mean by ‘Friend’? An inductive study. Social Networks, 3, 287
306.

Beyond Networks in Structural Theories of Exchange

295

Frank, K. A., Muller, C., & Mueller, A. S. (2013). The embeddedness of adolescent friendship nominations: The formation of social capital in emergent network structures. American Journal of Sociology, 119(1), 216
253. Freeman, L. C. (1978/1979). Centrality in social networks: Conceptual clarification. Social Networks, 1, 215
239. Friedkin, N. E. (1984). Structural cohesion and equivalence explanations of social homogeneity. Sociological Methods & Research, 12(3), 235
261. Gibson, D. R. (2005). Taking turns and talking ties: Network structure and conversational sequences. American Journal of Sociology, 110(6), 1561
1597. Golder, S. A., Wilkinson, D. M., & Huberman, B. A. (2007). Rhythms of social interaction: Messaging within a massive online network. In C. Steinfield, B. T. Pentland, M. Ackerman, & N. Contractor (Eds.), Communities and technologies (pp. 41–66). New York, NY: Springer. Goodreau, S. M., Kitts, J. A., & Morris, M. (2009). Birds of a feather, or friend of a friend? using exponential random graph models to investigate adolescent social networks. Demography, 46(1), 103
125. Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology, 78(6), 1360
1380. Hallinan, M. T. (1974). A structural model of sentiment relations. American Journal of Sociology, 80(2), 364
378. Heidler, R., Gamper, M., Herz, A., & Eßer, F. (2014). Relationship patterns in the 19th century: The friendship network in a German boys’ school class from 1880 to 1881 revisited. Social Networks, 37, 1
13. Holland, P. W., & Leinhardt, S. (1970). A method for detecting structure in sociometric data. American Journal of Sociology, 76, 492
513. Homans, G. C. (1950). The human group. New York, NY: Harcourt, Brace & World. Huberman, B., Romero, D. M., & Wu, F. (2009). Social networks that matter: Twitter under the microscope. First Monday, 14(1), 1
9. Ingram, P., & Morris, M. W. (2007). Do people mix at mixers? Structure, homophily, and the “Life of the Party”. Administrative Science Quarterly, 52(4), 558
585. Kandel, D. B. (1978). Homophily, selection, and socialization in adolescent friendships. American Journal of Sociology, 84(2), 427
436. Kitts, J. A. (2006). Social influence and the emergence of norms amid ties of amity and enmity. Simulation Modelling Practice and Theory, 14(4), 407
422. Kitts, J. A. (2009). Paradise lost? Age dependent mortality of American communes. Social Forces, 87(3), 1193
1222. Kitts, J. A. (2010). Dynamics of Networks Within Groups. Presented at the Group Process conference Atlanta, GA. Kitts, J. A., Lomi, A., Pallotti, F., Mascia, D., & Quintane, E. (2013). Looking inside interorganizational ties: The dynamics of patient exchange among Italian hospitals. Presented at American Sociological Association Meeting. New York, NY. Kossinets, G., & Watts, D. J. (2009). Origins of homophily in an evolving social network. American Journal of Sociology, 115, 405
450. Kuwabara, K., & Sheldon, O. (2012). Temporal dynamics of social exchange and the development of solidarity: “Testing the Waters” vs. “Taking a Leap of Faith.” Social Forces, 91(1), 253
273.

296

JAMES A. KITTS

Laumann, E. (1973). Bonds of pluralism: The form and substance of urban social networks. New York, NY: Wiley. Lawler, E. J., Thye, S. R., & Yoon, J. (2000). Emotion and group cohesion in productive exchange. American Journal of Sociology, 106, 616
657. Lawler, E. J., Thye, S. R., & Yoon, J. (2006). Commitment in structurally enabled and induced exchange relations. Social Psychology Quarterly, 69(2), 183
200. Lazer, D., Pentland, A., Adamic, L., Aral, S., Barabasi, A.-L., Brewer, D., … Van Alstyne, M. (2009). Computational social science. Science, 323(5915), 721
723. Leinhardt, S. (1972). Developmental change in the sentiment structure of children’s groups. American Sociological Review, 37(2), 202
212. Lewis, K., Kaufman, J., Gonzalez, M., Wimmer, A., & Christakis, N. (2008). Tastes, ties, and time: A new social network dataset using Facebook.com. Social Networks, 30(4), 330
342. Lin, K.-H., & Lundquist, J. (2013). Mate selection in cyberspace: The intersection of race, gender, and education. American Journal of Sociology, 119(1), 183
215. Lovett, T., O’Neill, E., Irwin, J., & Pollington, D. (2010). The calendar as a sensor: Analysis and improvement using data fusion with social networks and location. UbiComp ’10 Proceedings of the 12th ACM international conference on Ubiquitous computing, NY, USA (pp. 3
12). Macy, M., Kitts, J. A., Flache, A., & Benard, S. (2003). Polarization in dynamic networks: A hopfield model of emergent structure. In Dynamic social network modeling and analysis (pp. 162
173). Washington, DC: National Academies Press. Markovsky, B., Willer, D., & Patton, T. (1988). Power relations in exchange networks. American Sociological Review, 53, 220
236. Marsden, P. V. (1983). Restricted access in networks and models of power. American Journal of Sociology, 88, 686
717. Martin, J. L. (2011). Social structures. Princeton, NJ: Princeton University Press. Marvel, S. A., Kleinberg, J., Kleinberg, R. D., & Strogatz, S. H. (2011). Continuous-time model of structural balance. Proceedings of the National Academy of Sciences, 108(5), 1771
1776. McPherson, M., Smith-Lovin, L., & Brashears, M. E. (2006). Social isolation in America: Changes in core discussion networks over two decades. American Sociological Review, 71(3), 353
375. Milardo, R. M., Johnson, M. P., & Huston, T. L. (1983). Developing close relationships: Changing patterns of interaction between pair members and social networks. Journal of Personality and Social Psychology, 44, 964
976. Molm, L. D., Collett, J. L., & Schaefer, D. R. (2007). Building solidarity through generalized exchange: A theory of reciprocity. American Journal of Sociology, 113(1), 205
242. Molm, L. D., Whitham, M. M., & Melamed, D. (2012). Forms of exchange and integrative bonds effects of history and embeddedness. American Sociological Review, 77(1), 141
165. Moody, J. (2002). The importance of relationship timing for diffusion. Social Forces, 81, 25
56. Moody, J. (2004). The structure of a social science collaboration network: Disciplinary cohesion from 1963 to 1999. American Sociological Review, 69, 213
238. Moody, J., McFarland, D., & Bender-deMoll, S. (2005). Dynamic network visualization. American Journal of Sociology, 110(4), 1206
1241.

Beyond Networks in Structural Theories of Exchange

297

Moody, J., & White, D. (2003). Structural cohesion and embeddedness: A hierarchical concept of social groups. American Sociological Review, 68(1), 1
25. Mouw, T., & Entwisle, B. (2006). Residential segregation and interracial friendship in schools. American Journal of Sociology, 112(2), 394
441. Newcomb, T. (1961). The acquaintance process. New York, NY: Holt, Rinehart and Winston. Nezlek, J. B. (1993). The stability of social interaction. Journal of Personality and Social Psychology, 65, 930
942. Onnela, J.-P., Sarama¨ki, J., Hyvo¨nen, J., Szabo´, G., Lazer, D., Kaski, K., … Baraba´si, A.-L. (2007). Structure and tie strengths in mobile communication networks. Proceedings of the National Academy of Sciences, 104, 7332. Padgett, J. F., & Ansell, C. K. (1993). Robust action and the rise of the Medici: 1400,1434. American Journal of Sociology, 98(6), 1259
1319. Paik, A., & Sanchagrin, K. (2013). Social isolation in America: An artifact. American Sociological Review, 78(3), 339
360. Papachristos, A. V., Hureau, D. M., & Braga, A. A. (2013). The corner and the crew: The influence of geography and social networks on gang violence. American sociological review, 78(3), 417
447. Quintane, E., Carnabuci, G., Robins, G. L., & Pattison, P. E. (2012). An investigation of the temporality of structural holes. Academy of Management Best Papers Proceedings, OMT division. Quintane, E., & Kleinbaum, A. M. (2011). Matter over mind? E-mail data and the measurement of social networks. Connections, 31(1), 22
46. Roethlisberger, F. J., & Dickson, W. J. (1939). Management and the worker: An account of a research program conducted by the western electric company, hawthorne works, Chicago. Cambridge, MA: Harvard University Press. Ruan, D. (1998). The content of the general social survey discussion networks: An exploration of general social survey discussion name generator in a Chinese context. Social Networks, 20, 247
264. Sampson, S. F. (1968). A novitiate in a period of change: An experimental and case study of social relationships. Unpublished Ph.D. dissertation, Department of Sociology, Cornell University, NY, USA. Schaefer, D. R. (2012). Homophily through nonreciprocity: Results of an experiment. Social Forces, 90(4), 1271
1295. Schaefer, D. R., Simpkins, S. D., Vest, A. E., & Price, C. D. (2011). The contribution of extracurricular activities to adolescent friendships: New insights through social network analysis. Developmental psychology, 47(4), 1141. Shwed, U., & Bearman, P. S. (2010). The temporal structure of scientific consensus formation. American Sociological Review, 75(6), 817
840. Small, M. L. (2013). Weak ties and the core discussion network: Why people discuss important matters with unimportant alters. Social Networks, 35, 470
483. Smeaton, G., Byrne, D., & Murnen, S. K. (1989). The repulsion hypothesis revisited: Similarity irrelevance or dissimilarity bias? Journal of Personality and Social Psychology, 56, 54
59. Snijders, T. A., Van de Bunt, G. G., & Steglich, C. E. (2010). Introduction to stochastic actorbased models for network dynamics. Social Networks, 32(1), 44
60. South, S. J., & Haynie, D. L. (2004). Friendship networks of mobile adolescents. Social Forces, 83, 315
350.

298

JAMES A. KITTS

Stadtfeld, C. (2012). A stochastic actor-oriented framework for dynamic event processes in social networks. Karlsruhe, Germany: KIT Scientific Publishing. Stadtfeld, C., & Geyer-Schulz, A. (2011). Analyzing event stream dynamics in two-mode networks: An exploratory analysis of private communication in a question and answer community. Social Networks, 33(4), 258
272. Stewart, D. (2005). Social status in an open-source community. American Sociological Review, 70(5), 82342. Szell, M., & Thurner, S. (2010). Measuring social dynamics in a massive multiplayer online game. Social Networks, 32(4), 313
329. Thye, S. R., Lawler, E. J., & Yoon, J. (2011). The emergence of embedded relations and group formation in networks of competition. Social Psychology Quarterly, 74(4), 387
413. Thye, S. R., Yoon, J., & Lawler, E. J. (2002). The theory of relational cohesion: Review of a research program. Advances in Group Processes, 19, 139
166. Uehara, E. (1990). Dual exchange theory, social network analysis and informal social support. American Journal of Sociology, 96(3), 521
557. Vaquera, E., & Kao, G. (2008). Do you like me as much as I like you? Friendship reciprocity and its effects on school outcomes among adolescents. Social Science Research, 37(1), 55
72. Vargas, R. (2011). Being in “Bad” company: Power dependence and status in adolescent susceptibility to peer influence. Social Psychology Quarterly, 74(3), 310. Walker, H. A., Thye, S. R., Simpson, B., Lovaglia, M. J., Willer, D., & Markovsky, B. (2000). Network exchange theory: Recent developments and new directions. Social Psychology Quarterly, 63(4), 324
337. Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and applications. Cambridge: Cambridge University Press. Webster, M., & Whitmeyer, J. M. (2001). Applications of theories of group processes. Sociological Theory, 19(3), 250
270. Willer, D. (1999). Network exchange theory. Westport, CT: Praeger. Willer, R., Sharkey, A. J., & Frey, S. (2012). Reciprocity on the hardwood: Passing patterns among professional basketball players. PLoS ONE, 7(12), e49807. Wimmer, A., & Lewis, K. (2010). Beyond and below racial homophily: ERG models of a friendship network documented on Facebook. American Journal of Sociology, 116(2), 583
642. Wyatt, D., Choudhury, T., Bilmes, J., & Kitts, J. A. (2008). Towards the automated social analysis of situated speech data. Proceedings of the 10th international conference on Ubiquitous Computing, NY, USA (pp. 168
171). Wyatt, D., Choudhury, T., Bilmes, J., & Kitts, J. A. (2011). Inferring colocation and conversation networks from privacy-sensitive audio with implications for computational social science. ACM Transactions on Intelligent Systems and Technology, 2(1), 7.1
7.41. Young, J. T. N. (2011). How do they ‘End Up Together’? A social network analysis of selfcontrol, homophily, and adolescent relationships. Journal of Quantitative Criminology, 27(3), 251
273. Zachary, W. (1977). An information flow model for conflict and fission in small groups. Journal of Anthropological Research, 33, 452
473.