FEELING POLITICS
This page intentionally left blank
FEELING POLITICS EMOTION IN POLITICAL INFORMATION PROCESSING
E...
103 downloads
1478 Views
1MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
FEELING POLITICS
This page intentionally left blank
FEELING POLITICS EMOTION IN POLITICAL INFORMATION PROCESSING
Edited by
David P. Redlawsk
FEELING POLITICS
© David P. Redlawsk, 2006. All rights reserved. No part of this book may be used or reproduced in any manner whatsoever without written permission except in the case of brief quotations embodied in critical articles or reviews. First published in 2006 by PALGRAVE MACMILLAN™ 175 Fifth Avenue, New York, N.Y. 10010 and Houndmills, Basingstoke, Hampshire, England RG21 6XS Companies and representatives throughout the world. PALGRAVE MACMILLAN is the global academic imprint of the Palgrave Macmillan division of St. Martin’s Press, LLC and of Palgrave Macmillan Ltd. Macmillan® is a registered trademark in the United States, United Kingdom and other countries. Palgrave is a registered trademark in the European Union and other countries. ISBN 1–4039–7178–1 Library of Congress Cataloging-in-Publication Data Shambaugh Conference on Affect and Cognition in Political Action (2003 : University of Iowa) Feeling politics : emotion in political information processing / edited by David P. Redlawsk. p. cm. Includes bibliographical references and index. ISBN 1–4039–7178–1 1. Political psychology. 2. Emotions and cognition. 3. Personality and politics. I. Redlawsk, David P. II. Title. JA74.5.S46 2006 320.01⬘9—dc22
2005049549
A catalogue record for this book is available from the British Library. Design by Newgen Imaging Systems (P) Ltd., Chennai, India. First edition: June 2006 10 9 8 7 6 5 4 3 2 1 Printed in the United States of America.
CONTENTS
List of Tables
vii
List of Figures
ix
Notes on Contributors
xi
Preface 1 Feeling Politics: New Research into Emotion and Politics David P. Redlawsk 2 First Steps toward a Dual-Process Accessibility Model of Political Beliefs, Attitudes, and Behavior Milton Lodge, Charles Taber, and Christopher Weber 3 The Measure and Mismeasure of Emotion George E. Marcus, Michael MacKuen, Jennifer Wolak, and Luke Keele 4 Contributions of a Sociological Perspective on Affect to the Study of Political Action Lisa Troyer and Dawn T. Robinson 5 Affect and Politics: Effects on Judgment, Processing, and Information Seeking Linda M. Isbell, Victor C. Ottati, and Kathleen C. Burns 6 Motivated Reasoning, Affect, and the Role of Memory in Voter Decision Making David P. Redlawsk 7 Fear and Loathing in American Elections: Context, Traits, and Negative Candidate Affect Marco R. Steenbergen and Christopher Ellis
xvii
1
11 31
47
57
87
109
vi
CONTENTS
8 The Three Faces of Negative Campaigning: The Democratic Implications of Attack Ads, Cynical News, and Fear-Arousing Messages Ann Crigler, Marion Just, and Todd Belt 9 Racial Cues in Campaign News: The Effects of Candidate Strategies on Group Activation and Political Attentiveness among African Americans Vincent L. Hutchings, Nicholas A. Valentino, Tasha S. Philpot, and Ismail K. White
135
165
10 I Like Him, but. . .: Vote Choice when Candidate Likeability and Closeness on Issues Clash David P. Redlawsk and Richard R. Lau
187
11 The Emotional Calculus of Foreign Policy Decisions: Getting Emotions out of the Closet Nehemia Geva and J. Mark Skorick
209
References
227
Index
259
LIST OF TABLES
3.1 Factor Analysis of Emotional Response, Frequency Response Format 3.2 Emotional Response to Affirmative Action Policy, Experimental Studies 3.3 Emotional Response to Challenging Versus Affirming Affirmative Action Policy, Experimental Studies 3.4 Factor Analysis of Emotional Response, Likert Response Format 6.1 Information Congruency and Processing Time 6.2 Information Congruency and Candidate Memory 6.3 Effects of Memory and OL Tally on Candidate Feeling Thermometer Evaluation 7.1 Predicting Anxiety and Aversion toward Carter in 1980 7.2 Predicting Anxiety and Aversion toward Clinton in 1995 8.1 Predicting October Emotions about Dole 8.2 Predicting October Vote Choice 8.3 Predicting Experimental Effects 8.4 Predicting Consumption of Campaign Information in Different Media Sources 8.5 Predicting Information Seeking 9.1 Predicting Effects of Manipulating Candidate Distance on Support for Presidential Candidates by Race 9.2 Predicting Effects of Manipulating Candidate Distance on Support for Presidential Candidates by Feelings of Hopefulness toward Gore and Race 9.3 Predicting Effects of Manipulating Group Threat Cues on Campaign Learning by Race and Levels of Political Information 9.4 Predicting Effects of Manipulating Group Threat Cues on Campaign Learning by Race
38 40
40 42 96 100 103 120 124 145 146 147 151 152 173
175
177 179
viii
LIST OF TABLES
10.1 Voting in a Primary Election Choosing between “Close” and “Likeable” Candidates 10.2 Vote for a Likeable but Distant Candidate 10.3 Correct Voting and Candidate Likeability 11.1 Pretest Results for Experimental Manipulations 11.2 Inferences and Judgments of the Target Nation as a Function of Information Valence 11.3 The Gap in Inferences and Judgments of the Target Nation between Positive and Negatively Valenced Information Sets in the Different Emotive States
196 200 204 218 220
221
LIST OF FIGURES
2.1 The Structure of Political Beliefs, Attitudes, and Intentions 2.2 Reaction Times for Incongruent/Congruent Prime-Target Pairs of Long and Short SOAs 2.3 Reaction Times by Prime Type at the Short SOA 3.1 Typical Distribution of Experienced Moods 6.1 Effects of Information Congruency on Processing Time 6.2 Information Congruency and Memory Reports 8.1 Respondents’ Preferences for Campaign News Coverage 9.1 Probability that Something about Al Gore Makes Subject Feel Hopeful 9.2 Probability that Something about George W. Bush Makes Subject Feel Afraid 10.1 Variations of Candidate Likeability 10.2 Difficulty of Vote Choice by Type of Candidate Chosen 10.3 Probability of Voting for a Likeable Candidate, Primary Election by Proportion of Issues and Ideology Versus Pictures and Personality Examined Female, High Political Expertise 11.1 The Cognitive Calculus Model of Decision Making 11.2 Processing Time as a Function of Experimental Condition 11.3 Proportion of Items Recalled as a Function of Emotional/Affective Manipulation
19 23 25 32 97 99 139 172 176 194 198
202 213 222 224
This page intentionally left blank
NOTES ON CONTRIBUTORS
Todd Belt is an Assistant Professor of Political Science at the University of Hawai’i at Hilo. His research interests include political psychology, public opinion, political behavior, and campaigns and elections. Kathleen Burns is a graduate student in Psychology at the University of Massachusetts at Amherst and expects to complete her Ph.D. in 2006. Her research interests include social cognition, the relationship between affect and cognition, and stereotyping. Ann N. Crigler is a Professor of Political Science and Director of the Jesse M. Unruh Institute of Politics at the University of Southern California. Her current research focuses on emotions and civic engagement in American politics. Her most recent book is Rethinking the Vote: The Politics and Prospects of American Election Reform (Oxford University Press, 2004). Christopher Ellis is a Ph.D. candidate in Political Science at the University of North Carolina-Chapel Hill. His areas of interest are American political behavior and the links between public opinion and public policy. His current research projects include understanding the role of elite party polarization in shaping mass party coalitions and citizen attitudes toward the major parties. Nehemiah Geva is an Associate Professor of Political Science at Texas A&M University. His research interests are in the areas of foreign policy decision and political cognition (perceptions, information processing, and images in politics). Recent work has included the establishment of the Poliheuristic Theory of Foreign Policy (with Alex Mintz), published in their edited volume, Decision Making on War and Peace: The CognitiveRational Debate and in the American Political Science Review and the development of a cognitive calculus model of political decisions published in the Journal of Conflict Resolution and in International Interactions. His work has also been published in the Journal of Politics and Political Psychology.
xii
NOTES ON CONTRIBUTORS
Vincent Hutchings is an Associate Professor of Political Science at the University of Michigan and a Research Associate Professor at the Institute for Social Research. His research interests focus on the circumstances under which citizens are attentive to political matters and engage in issue voting. He has recently published a book on this topic entitled Public Opinion and Democratic Accountability (2003), from Princeton University Press. His research also examines the ways in which political candidates structure their political appeals to take advantage of the voters’ sympathies and/or antipathies for particular groups along racial, religious, and gender lines. His work has appeared in the American Sociological Review, the American Political Science Review, the Journal of Politics, Political Communication, Public Opinion Quarterly, Journal of Communication and Legislative Studies Quarterly. Linda M. Isbell is an Assistant Professor of Psychology at the University of Massachusetts at Amherst. Her research, funded by the National Science Foundation and the University of Massachusetts, focuses on the relationship between affect and cognition. Her broader research interests include social cognition, political psychology, and sexual harassment. Marion Just is a Professor of Political Science at Wellesley College and an Associate of the Joan Shorenstein Center on Press, Politics and Public Policy at the John F. Kennedy School of Government, Harvard University. She is a coauthor of Common Knowledge: News and the Construction of Political Meaning and Crosstalk: Citizens, Candidates and the Media in a Presidential Campaign, and coeditor of the recent volume, Rethinking the Vote: The Politics and Prospects of American Electoral Reform. Luke Keele is an Assistant Professor in the Political Science Department at Ohio State University. His research interests are at the intersection of American politics and statistical research methods. Specifically, he focuses on macro movements in public opinion, affect in politics, time series, and discrete choice models. Richard R. Lau is a Professor of Political Science and Director of the Whitman Center for the Study of Campaigns, Elections, and Democracy at Rutgers University, where he has been since 1990. Before coming to Rutgers he taught at Carnegie Mellon University; he was also a fellow at the Center for the Study of Democratic Politics at Princeton University for the 2000–2001 academic year. Lau’s interdisciplinary research focuses on information processing and voter decision making, the nature of public opinion and its links to political elites, the effects of political campaigns, and health policy. He is also affiliated with the Institute of Health, Health Care Policy, and Aging Research at Rutgers.
NOTES ON CONTRIBUTORS
xiii
Milton Lodge is Distinguished University Professor of Political Science at Stony Brook University. His primary focus is on experimental studies of the primacy and automaticity of affect toward political leaders, groups, and issues with a special interest in the impact of affect on political judgments and evaluations. He research has been supported by the National Science Foundation and published in all of the major journals of political science and political psychology. Michael B. MacKuen is Burton Craige Professor of Political Science at the University of North Carolina, Chapel Hill. His research has focused on the way that citizens gather and digest information about politics and the economy as well as on the ways that the broader macro-polity connects citizens, politicians, and public policy in a systemic way. This work has been published in scientific journals including the American Political Science Review, the American Journal of Political Science, and the Journal of Politics. His books include Affective Intelligence and Political Judgment (with George Marcus and W. Russell Neuman, 2000) and The Macro Polity (with Robert Erikson and James Stimson, 2002). George E. Marcus is a Professor of Political Science at Williams College and president elect of the International Society of Political Psychology. He is also author of Affective Intelligence and Political Judgment, with W. Russell Neuman and Michael MacKuen (University of Chicago Press, 2000) and The Sentimental Citizen: Emotion in Democratic Politics (Penn State University Press, 2002). He has published numerous articles in the major journals of political science. His current research continues on the role of emotion in democratic politics. Victor Ottati has served as an Assistant Professor at SUNY-Stony Brook (Department of Political Science), and Purdue University (Department of Psychological Sciences). He is currently an Associate Professor within the Department of Psychology at Loyola University, Chicago. Research interests include social cognition, stereotyping, affect and cognition, attitude formation, persuasion, communication, political psychology, crosscultural psychology, and consumer psychology. Tasha Philpot is an Assistant Professor of Government and African and African American studies at the University of Texas at Austin. She specializes in American Politics with particular emphasis on African American Politics, Public Opinion and Political Behavior, Political Communication, and Political Parties. Her research examines the consequences of using racial images in political communication. David P. Redlawsk is an Associate Professor of Political Science at the University of Iowa. His work focuses on voter information processing and
xiv
NOTES ON CONTRIBUTORS
the role of emotions in voter decision making. Papers on these topics have been published in the American Political Science Review, the American Journal of Political Science, the Journal of Politics, and Political Psychology. He is coauthor of the book How Voters Decide: Information Processing During Election Campaigns (Cambridge University Press, 2006) with Richard R. Lau (Rutgers University). Dawn T. Robinson is an Associate Professor of Sociology at the University of Georgia. She is Deputy Editor of Social Psychology Quarterly and Director of the Laboratory for the Study of Social Interaction (LaSSI) at the University of Georgia. She has held elected offices in the American Sociological Association’s sections on Social Psychology and the Sociology of Emotions. Her research focuses primarily on identity and emotion processes in face-to-face interaction. She has just completed a series of National Science Foundation experiments investigating control system implications for interpersonal identity management. J. Mark Skorick is a Tower Fellow with the John Goodwin Tower Center for Political Studies at Southern Methodist University. He has taught courses in international relations, foreign policy, and terrorism at Texas A&M and the University of Kansas. He specializes in foreign policy decision making, U.S. foreign policy, and political psychology. He has published papers in the Journal of Conflict Resolution and International Interactions. Marco R. Steenbergen is an Associate Professor of Political Science at the University of North Carolina-Chapel Hill. His areas of interest are political psychology and quantitative methods. His current research projects focus on the role of ambivalence in electoral behavior, the political psychology of deliberation, and elite political behavior. He is the coauthor of Deliberative Politics in Action (Cambridge University Press, 2004) and his articles have appeared in the American Political Science Review, the American Journal of Political Science, Comparative Political Studies, and Political Psychology, among others. Charles Taber is an Associate Professor of Political Science at Stony Brook University. Research interests include international relations, political psychology, foreign policy, conflict modeling, and computational modeling (AI). Recent work includes development of a descriptive theory of motivated political reasoning (with Milton Lodge) to account for how and why biased processing can so easily and so often overwhelm the objective quality of evidence; in particular how it is that both ordinary citizens and political sophisticates are prone to follow a biased course of information processing when forming and updating their political beliefs and
NOTES ON CONTRIBUTORS
xv
preferences. His work has appeared in the American Political Science Review, and Political Psychology, as well as in edited volumes. Lisa Troyer is an Associate Professor of Sociology at the University of Iowa. Her research, funded by the National Science Foundation, University of Iowa, and Rockefeller Foundation includes the study of group dynamics with a focus on roles of technology and social expectations on innovation in groups. She directs the Virtual Immersive Social Environments Laboratory (VISE Lab) at the University of Iowa. Recent reports of her research have appeared in Advances in Group Processes and Sociology of Organizations. Nicholas Valentino is an Associate Professor of Communications Studies and Political Science, and Research Associate Professor in the Center for Political Studies, at the University of Michigan. He is interested in the impact of political communication between elites and citizens. Recent work has focused on “group priming,” the activation of group attitudes and identities via news or political advertising during campaigns, the role of emotions in the link between campaign communication, and the salience of group attitudes and identities. His current research project integrates many of these themes with a study of mass uses of new information technology. He has published articles in the American Political Science Review, American Journal of Political Science, Communication Research, and the Journal of Communication, among others. Christopher Weber is a graduate student in the Department of Political Science at Stony Brook University. His primary research interests include political communication and, more specifically, examining the role of affect in political advertising and deliberation. Ismail K. White is an Assistant Professor of Government at the University of Texas at Austin. He studies American politics, specializing in African American politics, public opinion, and political participation. The focus of his research is on the formation of African American public opinion and roles that racialized political messages play in shaping racial divisions in the American public. Jennifer Wolak is an Assistant Professor of Political Science at the University of Colorado at Boulder. Her research interests include political psychology, public opinion, and political communication.
This page intentionally left blank
PREFACE
fter many years during which political psychologists have considered affect only in very limited ways and often as an afterthought, new research has begun to examine how affect may be intricately linked to the cognitive system. This work leaves little doubt about the need to understand the role of affect and the emotional processing of information. With the idea that emotions matter, the Shambaugh Conference on Affect and Cognition in Political Action was convened at the University of Iowa, March 7–9, 2003. This conference brought together a group of scholars with new and exciting research programs designed to better our understanding of the role of affect in politics, while not forgetting that thinking matters as well. Representing a wide range of perspectives, and many different methodologies, this group met, discussed, and argued over several days. The results are presented in this volume. As with any project there are many people to thank for their efforts. Bob Boynton worked with me to manage the Shambaugh Conference from which this book originated. His advice and support were invaluable. The conference was a success because of those who participated— discussions were lively, exciting, and fun. I also want to thank Ralph Adolphs, from the Department of Neurosurgery at the University of Iowa for his illuminating lunch session presentation on brain neurophysiology, which gave all of us a better understanding of the biophysical wellsprings of emotion. The staff of the Department of Political Science at the University of Iowa, Karen Stewart, Carole Eldeen, and Wendy Durant, provided logistical and organizational support for the conference. Financial support was provided by the Benjamin F. Shambaugh fund which allows the political science department at Iowa the opportunity to host conferences such as this on a regular basis. Jason Humphrey helped with all the little things that had to be done each day during the conference and should be commended for his organization, efforts, and most of all driving and navigation skills. During much of the time that conference planning and manuscript preparation was in progress, I was a Scholar at the Obermann Center for
A
xviii
PREFACE
Advanced Study at the University of Iowa. I thank the staff at Obermann, Director, Jay Semel, and Lorna Olson for their support and the use of the Center’s facilities. And of course I thank my wife, Aletia Morgan, and my sons, Andrew and Greg Redlawsk, for their patience during the long hours that were spent preparing for the conference and getting this book ready in its aftermath.
CHAPTER 1 FEELING POLITICS: NEW RESEARCH INTO EMOTION AND POLITICS David P. Redlawsk
olitics is about feeling. For all the apparent desire of many political scientists to boil down political action to rational actors assessing their utilities for the options that they face (whether about voting, policy choices, legislative action, etc.) using some cool—unemotional—calculus, new research presented in this book and elsewhere reaffirms a central role for passion in politics. Since politics is concerned at its most basic level with the allocation of scarce resources, and since this means some people get things while others do not, it is not surprising that peoples’ feelings are an important part of any political calculus. This is not a new idea. Madison, after all, placed great emphasis on the need to control the inherent “passions” of the citizenry in the political arena, even going so far as devising a Constitutional system explicitly designed to minimize the role of emotion in politics. Any attempt to explain political action by considering only its cognitive roots is certain to result in only a partial explanation, and a not very good one in the end. Yet, for all its obvious importance, research into the role of emotions in politics has been surprisingly lacking in recent years in political science in general and political psychology in particular. Whether due to the influence of Anthony Downs (1957) and the rational choice paradigm that came after him, or simply a legacy of the enlightenment’s privileging of reason over emotion, the role of emotion in politics has been understudied, despite the clear connections between how people feel about politics and how they act. Rational choice theorists have long argued that judgments should be guided by a kind of Bayesian updating, where
P
2
DAVID REDLAWSK
exposure to new information updates prior beliefs in a “rational” way. Thus, for example, new information that is negative in relation to prior beliefs should move the updated beliefs away from priors, while positive information should strengthen existing beliefs (Green and Gerber 1999). While such updating may in fact be rational, and even normatively correct, arguing that it is a true description of what people do requires a host of untenable assumptions about the cognitive capabilities of human beings (Green and Shapiro 1994). Much of political science has taken these assumptions for granted, leaving not only little room for the interplay of emotion, but little reason to even investigate whatever role emotion might play. Indeed, emotion has long been viewed more as something that gets in the way of good decision making than as integral to information processing. Perhaps because accurately measuring emotional response to political stimuli is very difficult, even political psychologists not necessarily working in the rational choice tradition turned first to the tools of cognitive psychology to understand how people process political information. The cognitive revolution of the past decades led to a great deal of focus (much quite successful) on the cognitive underpinnings of political behavior. Research such as that reported in Lau and Sears (1986) took center stage, greatly influenced by a computer-like view of the operation of the human brain. There was much talk of inputs and outputs, and central processing units. Yet a long line of psychology research from Festinger’s (1957) description of cognitive dissonance, Heider’s (1958) balance theory, and Abelson’s (1959) belief dilemmas posits that cognition is not unbiased; that people have various cognitive and emotional motivations to see the world in particularistic ways. But somehow this recognition that emotional motivations matter a lot did not find its way very far into political psychology. Instead a distinctly cognitive information processing approach developed that talked of “schemas” and “heuristics” and “rational” decision making.1 But it did not talk much about motivation and emotion (Hastie 1988). The result is that a great deal of what political psychologists know about how individuals interact with candidates, leaders, and public policy choices, is about the thinking part of politics. But how can we deny that politics is at least as much, if not more, about feeling? It is not really fair to say that emotions have been completely ignored by political scientists. The survey research design underlying The American Voter (Campbell et al. 1960)—and now a standard paradigm in political behavior research—drew from an existing understanding of attitudes toward political objects as a combination of belief, feeling, and behavioral intentions. And, of course, respondents were explicitly asked to discuss their “likes” and “dislikes” about the political parties in their open-ended
FEELING POLITICS
3
questions. Political scientists who have followed in this tradition have generally assumed that their surveys engage respondents in cognitive processing about issues and candidates and that the outcome of this processing is an evaluation, or feeling, about the object/policy/person in question. Thus while it may not be completely fair to say that feeling has been ignored, emotion has often been conceptualized thinly as the outcome of a cognitive process, rather than as an integral part of decision making. Some researchers have tried to measure emotion and sometimes even to account for it.2 But nonetheless few have attempted to develop a comprehensive understanding of emotions and politics. Even when emotions have been considered in a larger context (e.g., Part I of Kuklinski 2001), it is usually as a small part of a larger, more cognitively oriented enterprise. Feelings, however, most certainly play a critical role in how citizens view politics. While we continue to think and write about thinking, recent work developing theories of affective intelligence (Marcus, Nueman, and MacKuen 2000) and motivated reasoning (Lodge and Taber 2000) have given new prominence to the role of emotion in political psychology. These approaches and others argue for the primacy of affect, and for its interconnectedness with cognition. They show us that one cannot think without feeling. As Damasio (1994, 1999) has discovered in his work with patients who lack capacity for emotional response, information may be processed cognitively, but human beings are generally unable to turn that information into a course of action (or decision) if they are unable to incorporate emotions into their information processing. Damasio calls the brain a “thinking machine for feeling” (1999) in describing just how completely cognition and emotion are intertwined. As the neurophysiology of the brain becomes better understood, it also becomes clear that a complete understanding of how people respond to politics requires that our old conceptions of “reason” and “rationality” give way to a new perspective. Political Science and Emotions: Two Perspectives As in any work in its early stages, much of what has been learned about emotions and politics has generated more questions than have been answered and important debates have risen. A key debate in particular is about the very nature of affective reactions, how they are generated, and their implications for politics. George Marcus and his colleagues (1993, 2000), following closely on Damasio’s work, argue that emotions arise from the structure of the brain, and that two different systems are at work. The first, the dispositional system, operates on routine information, disposing of it without much effort, and generating emotional responses on an “enthusiasm” dimension. The second, the surveillance system, becomes
4
DAVID REDLAWSK
aroused when the environment changes from the expected to the unexpected. Once aroused by something unexpected (read “dangerous”) the surveillance system heightens awareness and prepares us to respond by elevating “anxiety” levels. This process is not driven by cognitive processing of the environment but by an emotional response to an unexpected stimulus. The result is that in this aroused state learning is enhanced, since one needs to understand the nature of whatever threat has been encountered and is thus motivated to find out more about the stimulus. But what happens when attention is drawn to an unexpected stimulus? Does it mean that the resulting decisions are necessarily better than they would be otherwise? The evolutionary perspective that drives affective intelligence would certainly suggest so. After all, presumably the reason we have an emotional system capable of responding before we can even think is to enhance our ability to survive. But the perspective of motivated reasoning, as brought to political science by the work of Lodge and Taber (2000, 2005) may suggest a different answer. The processing of motivated reasoners is a tightly bound combination of emotion and cognition; the term Lodge and Taber use is “hot cognition.” They argue that the result of hot cognition is that as information is acquired it is immediately evaluated for its affective content. Unexpected information—affectively incongruent with prior evaluation—requires extra processing; it is not as easily assimilated as congruent information. So in the initial stages both affective intelligence and motivated reasoning lead to the conclusion that affectively incongruent information—especially that which generates a negative reaction—is more carefully attended to. But what happens next? A “good” decision process presumably would be one that correctly adjusts beliefs to reflect this new information, and while Marcus and colleagues do not directly address this it is clear that they equate more information with better decisions. But Lodge and Taber’s work suggests that processing may be biased by affect, in which case emotionally driven responses might leave something to be desired. They question whether people really are rational Bayesian updaters who accurately perceive the direction of new information and revise their beliefs, attitudes, and behaviors accordingly. Instead, they find people are more likely to stick to their guns, to support their prior beliefs, and thus to allow affect to interfere with updating. A great deal of other research supports this perspective. While Green and Gerber (1999) argue that most, if not all, findings of bias can be explained by the Bayesian model, an array of studies in multiple domains suggests otherwise. From early studies showing housewives rationalizing decisions already made (Brehm 1956) to Tversky and Kahneman’s (1974) seminal work on heuristic biases, to Steenbergen’s (2001) recent evidence of a conservatism bias in updating, a wide array of “mistakes” and “biases”
FEELING POLITICS
5
have been documented. In my own work I have documented clear effects as voters encounter information they dislike about a candidate they like— they take longer to process this information and they fail to adjust their evaluations to account for it (Redlawsk 2002). What causes these biases, these failures to act as cool rational models would have us? Out feelings may be the answer, and our emotions may have serious implications for our thinking. Thus these two lines of research, both promising in their ability to help us understand how voters respond emotionally to politics, are themselves somewhat at odds. Emotional responses may well help humans learn, but at the same time perhaps interfere with the ability to integrate that learning into an accurate evaluation. Whether affective systems are viewed as independent (and indeed primary) drivers of attention and understanding as Marcus proposes, or affect and cognition are seen to work in tandem as in the Lodge and Taber studies, the role of emotion is clearly an important factor in political judgment. And while this volume will not resolve any debates between these two approaches, and while the debate itself is not the primary focus on this book, readers will get a clear flavor of both perspectives from various chapters of this book, many of which adopt either one of the other as their analytic base. The Interdisciplinary Nature of this Research As can be seen readily within this book as well as in the broader developing literature on emotion and politics, research in this area is an example of interdisciplinary association at its best. While political psychologists are of course influenced by the classics of political behavior in developing understandings of citizens and politics, there are obvious connections well beyond this. Work from cognitive and social psychology figures prominently in most of the chapters of this book and given that politics is clearly a social phenomenon, it seems especially appropriate that sociologists also make a significant contribution. Further, much of what political psychology is now beginning to learn about the psychological foundations of political affect and cognition is built squarely on very recent work in neurophysiology, which provides a critical link between the functioning of the brain and its physical systems and how those systems implicate our emotional and cognitive responses to political stimuli. Perhaps of even more interest to political scientists generally is the potential for broad application of the theories now being developed and tested. Where “political behavior” is usually focused on the mass behavior of citizens—often in terms of voting—the political psychology of emotion is developed at a more individual level of analysis and therefore is broadly
6
DAVID REDLAWSK
applicable to situations in which individuals must process information about political conditions, whether we talk about citizens evaluating candidates, or elites addressing beliefs about war and peace. The fact that the study of emotions is the study of individuals means that general theories of how people react emotionally to stimuli have potential to apply in any realm in which people interact with other people, information, and situations. The Plan for this Book The primary goal for this book is to introduce readers to some of the most interesting and up-to-date research projects focused on emotion. These projects fall into different classes: some which strive to develop broad theoretical underpinnings for our understanding of the roles of affect and cognition and some which attempt to understand the practical, political implications of emotional reactions to issues, people, and situations. The book begins by focusing on the broader theoretical questions by presenting the somewhat competing perspectives of motivated reasoning and affective intelligence in chapters 2 and 3. In chapter 2, Milton Lodge, Charles Taber, and Chris Weber make use of a dual-process accessibility model that draws on work by Zajonc (1980, 2000), Bargh (1997), and Fazio (1986) to argue that emotions can be automatically invoked in a process that occurs outside of conscious awareness. Their experiments test the idea that affect is intricately tied to cognition and that the resulting “hot cognition” has real effects on how citizens process political stimuli. Here they summarize key experimental work defining hot cognition and developing its implications, particularly how nonconscious affective responses may bias evaluation, judgment, and choice, with a focus on the important mediating effects of political sophistication. An important argument is that affect cannot readily be measured by simply asking people to recount how they feel about politicians or issues. Because affect is often automatic—that is, recalled spontaneously and without cognitive effort—any cognitive evaluation of affective response must necessarily be reconstructing, and not necessarily accurately, feelings that were nonconscious to start with. In chapter 3, George Marcus, Michael MacKuen, Jenny Wolak, and Luke Steele turn to what is thus a vexing question—how do we measure emotional response? Driven by their affective intelligence perspective they do not follow Lodge and Taber’s lead in rejecting introspection. Instead, after reviewing the approaches used in the psychology and political science literature to date, Marcus and his colleagues argue that what political science needs to do is to become more fine-grained in the questions asked that attempt to tap emotional response. In particular they find that what
FEELING POLITICS
7
they label “aversion” is often distinct from anxiety and routinely ignored in the types of questions asked in typical surveys. Aversion arises when existing beliefs are challenged, rather than when they are confirmed. Thus they offer a significant enhancement to the basic paradigm of affective intelligence and a set of recommendations for future efforts to measure emotion. Sociologists Lisa Troyer and Dawn T. Robinson examine emotions and political identity in chapter 4, arguing that emotions are not only psychological but also constructed and understood in response to social situations. Given that politics is about social evaluation—voters evaluating candidates, citizens considering courses of action, and the like—it is surprising that this sociological perspective is often overlooked by political psychologists. In this chapter Troyer and Robinson outline key microsociological perspectives on emotion and then offer Affect Control Theory [ACT] (Heise 1977, 1987) as a way of thinking about how people adjust their emotional responses to the social environment. They then suggest how this theory may be applied to political action as ACT is expanded to incorporate political identities, behaviors, and settings. Applying a theory such as ACT might give us a substantial cross-discipline understanding of the role emotions play not just in politics but also through a more comprehensive view of social interaction, both political and not. In the final chapter (chapter 5) in this section, Linda M. Isbell, Victor C. Ottati, and Kathleen C. Burns take seriously the argument advanced by Marcus and his colleagues that thinking about emotion in politics must go well beyond the typical valence model. They focus on three key questions in the literature: the extent to which specific negative emotions such as anger, fear, and disgust play differing roles in political judgment; how facial displays as affective cues are incorporated into information processing; and how emotions influence the selection of information. They conclude that while it is clear that affective states strongly influence judgment, selection, and processing, the effects are highly dependent on the extent to which individuals are actually aware of their emotions as they process political information. Beginning with chapter 6 the book turns to a range of work that applies theories of emotion directly to political concerns. David P. Redlawsk extends earlier work on the role of memory in voter decision making by considering why memory might matter even for otherwise online voters. The key appears to be the emotionally driven motivated reasoning process described by Lodge and Taber (2000), which suggests that in the face of unexpected information—affectively incongruent in terms of preexisting preferences—people often make a significant effort to maintain prior evaluations rather than accept the new information at face value. One way
8
DAVID REDLAWSK
of doing this is when learning something negative about a liked candidate is to bolster prior evaluations by recalling reasons why the candidate continues to deserve support. As affectively incongruent information is encountered, memory for previous information must be used at least in part to override the new information. Thus, motivated reasoning, which is based in part on online processing, appears to actually explain some of the reasons that memory matters in candidate evaluation during a campaign emotional commitment toward candidates and campaign information clearly plays an important role. Responding to the call for a more comprehensive measure of emotion in chapter 7, Marco R. Steenbergen and Christopher Ellis examine the effects of a three-dimensional model of affect on candidate evaluation. Adopting the affective intelligence perspective that anxiety and fear (aversion) are distinct negative emotions, Steenbergen and Ellis test two hypotheses designed to understand when negative emotion toward presidents is based more in anger than in fear. Since anxiety appears to be more common than fear, at least in the political realm tapped by survey research, they argue that “for aversion to be felt as a distinct emotion, separate from anxiety, an individual must perceive that a political figure’s actions affront the individual’s most important beliefs and values, or that the figure’s actions (or inactions) cause harm to the person’s wellbeing” (p. 115 of this volume). Voters whose strongest values or well-being are challenged may ultimately fail to show the kind of learning effects shown for anxious voters, becoming less well-informed citizens, rather than more. In chapter 8, Ann Crigler, Marion Just, and Todd Belt examine the implications of negative campaigning for the democratic polity and the ability to arouse fear in campaign messages. In an exemplary instance of using multiple methods, they employ both survey-based panel data and a field experiment in the 1996 election to support their argument that different types of negative campaigning have different influences on voters. The emotions generated by negativity are incorporated by citizens as they build their understanding of the campaign and candidates. Attack advertising may lower positive feelings toward the candidate sponsoring the ads as well as potentially lower turnout likelihood. The news media are not blameless, either, as cynical news coverage decreases citizens’ feelings of trust in government, also resulting in voter demobilization. Yet, in line with Marcus’s affective intelligence argument, Crigler, Just, and Belt find that frightening emotional appeals may well enhance voter learning and lead citizens to seek additional information. In chapter 9, Vincent L. Hutchings, Nicholas A. Valentino, Tasha S. Philpot, and Ismail K. White argue that as voters find themselves more or less distant from candidates on important issues, differing emotional responses develop. By using an experiment in which they manipulate the
FEELING POLITICS
9
relative appeals to the interests of African American voters of George W. Bush and Al Gore, they are able to examine the extent to which emotional reactions to the candidates change and the implications of that change. Hutchings and colleagues find that as the candidates are portrayed as significantly different on racially relevant issues, Gore generates more enthusiasm in African American voters, though not in other groups. Further, when Bush is portrayed as relatively more conservative on these issues, African American voters express more anxiety and become more attentive to campaign appeals. In addition to supporting the thesis that provoking anxiety promotes learning, this chapter gives us a sense of how group-centric issue frames can influence voters beyond those in the targeted group, as the news articles used as experimental stimuli were aimed at African Americans, but influenced others as well. Chapter 10 turns to a different focus of emotional response—candidate likeability. David P. Redlawsk and Richard R. Lau use a dynamic information board to present a presidential election simulation in a laboratory setting. In this study subjects faced two candidates—one who was close to them on the issues but sporting less likeable physical and personality traits and one who, while distant on the issues, was significantly more attractive and with a more positive personality. The question: which would voters choose; the likeable candidate or the candidate closer on issues? As with most research the answer is dependent on the context. In the primary election candidate likeability does not necessarily outweigh issue positions and a number of individual voter factors come into play. For some voters— such as those with less political expertise and education—image and personality appear to be more important considerations than for others. In the general election the findings differ somewhat, as the well-established role of partisanship intervenes. Attractiveness and a good personality do not seem particularly strong in drawing partisans to vote for the other party’s candidate, and in fact, an out-party candidate who is “good” on the issues is more likely to appeal even when less attractive. Up to this point the focus of the studies included here have been domestic politics; primarily candidate evaluation. But the final chapter (chapter 11) turns to foreign policy, as Nehemiah Geva and J. Mark Skorick attempt to integrate a theory of emotions into existing research on foreign policy making. Reacting to the predominance of cognitive and heuristic based explanations for foreign policy decision, Geva and Skorick propose a model that explicitly incorporates emotions such as hatred and fear—often intense and long lasting—into how interstate conflict is viewed by policy makers. Importantly, they take a nuanced view of emotions and affect as distinct experiences. Emotions are part of the physiological response to stimuli, object-oriented, and linked after the fact to cognitive
10
DAVID REDLAWSK
components. Affect, they suggest, better describes the belief aspect of an attitude, tied directly to a cognitive structure, and not associated with physical arousal. In a sense Geva and Skorick’s approach connects to both affective intelligence and the idea of an emotional system as well as motivated reasoning’s hot cognition thesis, with the close connection between believing and feeling. To test this thesis, they report on an experiment designed to tease out the differences in which they find that negative emotion appears to generate more facile processing, as processing time is decreased and recall inhibited, compared to the more cognitively based negative affect condition. While recognizing that their findings could be taken as evidence of a continuum of emotional response from less strong (what they call affect) to extremely strong (emotion) they argue that the differences they find are more substantial than can be accounted for by a simple continuum explanation. It seems somewhat trite to say it, but the study of emotion really is on the cutting edge in political psychology. While the cognitive revolution took us a long way toward understanding how people mentally structure and think about the political world, it is a revolution mostly devoid of feeling. And yet politics is certainly about feeling every bit as much as it is about thinking. More importantly, perhaps, recent work in neurophysiology has shown that emotions may in fact influence everything else we do; what we pay attention to, what gets into our memory, and what we can recall, and how we make decisions. Adolphs (1999) identifies specific parts of the brain that influence how we interact socially and the role emotions play. Politics is clearly in the domain of social interaction. To ignore the important role of emotion means to ignore what might in fact be the most critical component to our understanding of how people process political information and make sense of the political world. The work presented in this book represents important new approaches to these questions; approaches that begin to take us down the road toward a more complete understanding of political information processing. Notes 1. The Symposium on schemas in the American Political Science Review, 85:4 (December 1991) for several different views on the use and value of the schema concept in political science. 2. For a valuable overview of the political science view of emotion, see Marcus (2000).
CHAPTER 2 FIRST STEPS TOWARD A DUAL-PROCESS ACCESSIBILITY MODEL OF POLITICAL BELIEFS, ATTITUDES, AND BEHAVIOR Milton Lodge, Charles Taber, and Christopher Weber
ith the advent of the political behavior movement in political science in the 1950s, in particular with the publication of The American Voter in 1960, beliefs, feelings, and behavioral dispositions were brought to center stage in the prediction and explanation of political behavior. In line with an implicit assumption of human rationality, the social sciences commonly presumed that thoughts, feelings, and behavioral intentions coming to mind consciously determine the lion’s share of behavior. Congruent with this assumption of conscious considerations arbitrating the expression of beliefs and emotions, political scientists commonly ask people to voice their beliefs, report their likes and dislikes, recount feelings and past behaviors, and foretell their intended actions. Because of this reliance on introspection, much of what we know about public opinion and electoral behavior and how we model the expression of political attitudes, beliefs, and behavioral dispositions is based almost exclusively on what respondents say when asked for their present, past, or future beliefs, intentions, and behavior. Conventionally, political scientists, like their fellow social scientists, conceive of cognition, affect, and behavior as conceptually distinct and analytically separable, treating any disassociation of beliefs and feelings from behavior as measurement error, ignorance, or plain irrationality. We believe that this assumption of memory-based processing, and ease of obtaining self-reports, led political scientists (and to a lesser extent social
W
12
LODGE, TABER, AND WEBER
psychologists, who were long wary of introspection) to treat beliefs, attitudes, and behavior as independent, consciously created entities open to introspection—eventually becoming manifest in expectancy value and subjective utility theories (Ajzen and Fishbein 1980; Evans and Over 1996). While theoretically possible and traditional to treat the “holy trinity” of beliefs, attitudes, and intentions as separate, conscious mediators of behavior (Eagly and Chaiken 1993; Kinder 1998), it proves impossible to readily tease them apart in practice. Now, 20 years into the cognitive revolution (Lindsay and Norman 1977; Lachman, Lachman, and Battlefield 1979; Eysenck and Keane 1995), we see that thinking and deliberation are in fact only part of the process in the expression of attitudes and behavior and much of what we say and do is intimately linked to the feelings we have toward an attitude object (Zajonc 1980). Current theorizing in the cognitive sciences, backed by hundreds of well-crafted behavioral studies in social and cognitive psychology and new evidence from the neurosciences, posits a dual-process model that distinguishes between automatic and deliberative processing in the formation and expression of beliefs, attitudes, goals, and behavior (Devine 1989). Automatic processes, in which thoughts, feelings, and intentions come to mind spontaneously, on the order of milliseconds (ms), contrast with more demanding controlled processes employed when the individual has sufficient time, motivation, awareness, and cognitive resources to deliberate (Devine 1989; Fazio et al. 1995). Moreover, this work demonstrates that beliefs, feelings, and behavioral intentions may, if “contiguously activated” (Hebb 1949), become so strongly associated in long-term memory (LTM) that they are effectively “unitized” in a network of interdependencies that come automatically to mind on mere exposure to a “triggering event,” and are only uncoupled in pathological cases (Damasio 1994). What is critical here is that this bundle of automatically triggered beliefs, feelings, and behavioral intentions come to the fore in milliseconds, before any conscious considerations come to mind, and will predictably impact all consciously retrieved judgments, evaluations, intentions, and actions. Unlike Petty and Cacioppo’s (1986) elaboration-likelihood model or the systematic-heuristic distinction in Eagly and Chaiken’s (1993) model, both of which center on the conscious processing of cognitive considerations, our dual-process model focuses on the accessibility of both automatic and conscious beliefs, feelings, intentions, and actions. Without denying that much of everyday life is conscious, it is demonstrably the case (and the focus of this chapter) that automatic processes are primary in the sense that they enter the decision stream earlier than do conscious considerations. Not only that, but these automatic considerations strongly influence the conscious thoughts, feelings, and interpretations that are used to justify and
A DUAL-PROCESS ACCESSIBILITY MODEL
13
explain one’s decisions. With this in mind, a primary component of this model is the affective associations formed on the repeated evaluations of sociopolitical objects. And once formed, the constructed affective tally is believed to be the first component to enter in the information processing stream, and thereupon cascades across subsequent higher-order, conscious considerations. What is not well established is where, when, for whom, and to what degree these automatic processes impact the conscious expression of political beliefs, attitudes, and behavior on which so many of our discipline’s predictions and explanations rest (Karpinski, Steinman, and Hilton 2005). A basic premise of our research program is that Mother Nature did not design separate areas of the brain or evolve specialized mechanisms for the unique expression of political life. This being the case, as political psychologists we see no reason not to adapt theories from the cognitive sciences (Krosnik 2002a; Rahn, Sullivan, and Rudolph 2002) and shamelessly apply research paradigms from social and cognitive psychology to test hypotheses on the automaticity and interdependency of political beliefs, attitudes, intentions, and behavior. With this in mind, our dual-process model draws heavily on: ●
●
●
●
Zajonc’s (1980, 2000; see also Murphy and Zajonc 1993) theorizing on the primacy of affect; Bargh’s (1997) demonstrations of automaticity across a broad range of cognitive, affective, and goal-directed behaviors; Fazio’s (1986) Motivation and Opportunity DEterminants (MODE) model of the attitude-behavior process, which posit two complementary processes, one automatic, the other consciously controlled, where attitude accessibility is determined by the ease of retrieval of associations from LTM; Fazio’s and his colleagues’ (1986) attitude-priming paradigm as a method for discriminating automatic from deliberative processes.
Of critical importance to the model are Zajonc’s experiments suggesting that all perceptual objects are rapidly evaluated by a somewhat independent affect processing system that immediately and spontaneously transmits positivity or negativity to even novel stimuli (Zajonc 1980; Murphy and Zajonc 1993). The linking of affect to cognition readies the perceiver’s approach or avoidance behavior toward attitudinal objects before the slower, cognitive processes come into play (LeDoux 1996, 2003). Similarly, John Bargh, his students, and colleagues (Bargh 1994; Bargh, Chen, and Burrows 1996) have demonstrated experimentally how easily concepts can be primed—even subliminally—with striking consequences for subsequent
14
LODGE, TABER, AND WEBER
thinking, feeling, and doing. For instance, when white subjects are subliminally primed to the concept “African American” they are often prone to act in accord with black stereotypes without awareness of having been primed. The labeling of one mode of processing as “deliberate” emphasizes the reflective, consciously controlled character of one’s responses to an object— whether person, place, event, thing, or idea—which generally (but not necessarily) involves verbal reasoning. Other descriptors of the poles on what is surely a continuum (Fiske and Neuberg 1990) are: controlled versus automatic (Bargh 1997; Fazio 1986); conscious versus unconscious ( James 1890); explicit versus implicit (Greenwald and Banaji 1995). Deliberative processes are cognitively effortful, time-consuming, demanding of attention, and often premised on an intentional memory search for relevant facts and considerations. Conversely, automatic processes—whether the immediate activation of cognitive associations (e.g., Bush is a Republican), or the spontaneous activation of affect (Republicans are evil; Democrats are dumb), or such habitual actions as online processing that operate “mindlessly” (Betsh et al. 2001)—are involuntary, fast, immediate, top of the head, and unlike conscious processes can be activated even when the individual’s conscious attention is focused elsewhere. In sum, people are frequently unaware of the specific situational and contextual factors (call them “primes”) that bring the thoughts, feelings, and intentions to mind that prompts a decision that appears to the actor to be the outcome of a deliberative evaluation of the evidence. To call a process “automatic” it must satisfy four criteria (Bargh 1997): 1. It must be spontaneous. That is, the process or response must be triggered even if the individual is not consciously engaged in making an evaluation. In the experiments we review participants are often engaged in a cognitive task while the researcher indirectly measures the spontaneous effect of affect on their recognition, recall, comprehension, and judgments. 2. It must be unconscious; the ideal research design uses subliminal priming where the “priming” stimulus is presented so quickly, in some studies for as few as 15 ms that participants are not even aware that they were exposed to the stimulus. 3. The response must be uncontrollable. That is, once triggered, the process runs its course without conscious guidance. 4. The process is invoked and carried through while expending little or no cognitive resources. While there are many such studies of automaticity, controversy has surrounded whether automatic processes can be brought under the control of conscious, deliberative processes (Bargh 1997; Bargh and Barndollar
A DUAL-PROCESS ACCESSIBILITY MODEL
15
1996; Bargh, Chen, and Burrows 1996; Bargh et al. 1992; Fazio 1993; Chaiken and Bargh 1993; Devine 1989). On one side befalls the notion that, given certain environmental constraints and individual factors, automatic thoughts, behaviors, and intentions can be overridden (Devine 1989; Fazio 1993); whereas, on the other, the very nature of automaticity implies a process occurring outside of consciousness rendering it implausible to bring under conscious control (Bargh 1999). Rather than taking sides, we concede that in particular instances one’s automatic responses may be eclipsed by more effortful processing—for instance, if the individual is made aware of his/her automatic response and has sufficient time, motivation, and incentive to think more deliberatively. Nonetheless, because many situations do not call for “central” processing, we believe a fair deal of political information processing will be colored by the automatic activation of thoughts, behaviors, and intentions. What is important here is that in many familiar situations (as well as in such uncommon settings as a survey interview) automatic affective responses can, and predictably will, directly impact the expression of subsequent evaluations, judgments, goals, decisions, and actions with little or no pre- or postconscious appraisal. Although by no means do we suggest that people, even the much maligned American voter, are automata, in many cases deliberative processes will be influenced by prior, unconscious factors inaccessible to introspection. Yet in none of our own work, or the cited evidence from psychology, is it claimed that people are compelled to act contrary to their explicit beliefs and goals. People sometimes monitor their behavior, check their beliefs against their actions, and act in accordance with their conscious choices. What is now demonstrably clear from the experimental literature is that automatic processes underlie all conscious processing and are especially powerful determinants of thought and action when one is not paying close attention, or when one’s attentional resources are engaged or distracted, or when one is under time pressure, or when an environmental event is noticed but not recognized as being influential, or—this especially important—when one is unaware of the environmental triggers. The implicit-explicit distinction therefore should not be seen as either/or but as complementary, for all thoughts, feelings, and actions have an immediate automatic component and any pattern of associated thoughts, feelings, or behaviors—if repeatedly coupled together—can become unitized in memory. Our take-home point is that all deliberation is underwritten by automatic memory processes so that there is in fact no such thing as a purely explicit or conscious response. (See Neely 1977 for the classic studies of the automaticity of cognitive associations on a deliberation task.) Experiments comparing implicit to explicit attitudes show that subtle factors, which easily escape direct conscious awareness, can cause one to
16
LODGE, TABER, AND WEBER
deviate from what a conscious appraisal of the pros and cons would dictate (Wilson 2002). We would expect implicit processes to be most influential where one’s behavior is not consequential so as to demand extensive attentional resources or to trigger questions as to “why did I think, feel, say, do that?” It is important to keep in mind that when one’s thoughts and feelings are congruent there is little, if any, impetus to challenge them or question the source. The thoughts, feelings, and reasoning processes that come spontaneously to mind often “feel” right. Why resist congruent facts, figures, and impressions? Precisely because people often do not challenge beliefs, attitudes, or intentions, we suspect that most automatic influences will remain covert, and go unnoticed. We typically become aware of the force of automaticity in our own thinking when given negative feedback or when an automatic response challenges our self-image as a reasonable, rational being. In other words, implicit responding is the default process, and if we are distracted, weakly motivated, or hurried, these responses may go unquestioned. Our most basic expectation is that people will rely on their spontaneously accessible beliefs and attitudes unless confronted by irrefutable evidence, social pressure, or when one’s routine way of thinking leads to more readily apparent failures than successes. We expect implicit processes to be strongest when the motivation to think hard is weak, when time or cognitive resources are wanting, and when the consequences of being wrong are negligible. These situational and contextual factors favoring automaticity appear to characterize the world of politics for most citizens, where typically the direct consequences of one’s political thoughts and actions are distant and indirect, where uncertainty reigns, rumination is often neither called for nor encouraged, where one is easily distracted by rapid-fire TV images, and where the stream of information routinely runs in parallel with congenial cues. It takes concerted effort to change habitual ways of thinking, feeling, and acting. Sometimes of course there is a discernable disassociation of implicit and explicit thoughts and actions. If consciously conflicted, one may deliberatively override the spontaneously “readied response” given the time, attentional resources, sufficient domain-relevant knowledge, and, of course, with the motivation to challenge an uncongenial association. We expect, and experimental evidence routinely finds, disassociations between unconscious and consciously expressed beliefs and attitudes when explicit measures are tainted by social desirability or contaminated by deceit or prejudice. Where, when, how, and for whom deliberative processing successfully overrides the automatic response is the critical, unanswered question that goes to the heart of all discussions of human rationality and the meaning of a responsible electorate.
A DUAL-PROCESS ACCESSIBILITY MODEL
17
If as we claim those beliefs, feelings, intentions, and actions that have been repeatedly linked together in past experience become chronically accessible on mere exposure to a conditional cue (Lau 1989a), then many of our conscious thoughts and deliberative actions are necessarily infused with spontaneously accessed associations that can run independently of conscious guidance. The basic claim of our dual-process model is that those political beliefs, feelings, goals, and behaviors that were contiguously associated in the past are: ●
●
●
“unitized” in LTM and come spontaneously to mind on mere exposure to an environmental triggering event or situational cue; enter into the decision process in real time—that is, within milliseconds, before any conscious considerations; and color the conscious appraisal of candidates, parties, events, issues, and behavioral strategies.
To the extend that some (we claim all) political thoughts, feelings, intentions, and actions have an automatic component that necessarily influences subsequent evaluations, judgments, and choices, then our discipline’s reliance on conscious, introspectively available considerations as mediators of behavior fails to model correctly how most citizens most of the time think, reason, and act. From this dual-process perspective, several questions come to the fore: 1. Under what conditions can we rely on deliberative responses as accurately reflecting underlying beliefs, feelings, and intentions? 2. Under what conditions will beliefs, feelings, and intentions be spontaneously activated in a chain of conjoined thought, feeling, action? 3. When, for whom, and under what situational cues will implicit beliefs, feelings, and intentions impact subsequent beliefs, attitudes, and action? 4. Under what conditions will citizens be able and willing to override their spontaneous responses with conscious control? The Cognitive Architecture Before turning to evidence supporting the notion of political automaticity, let us briefly review the cognitive architecture underlying our dual-process theory of political information processing (Lodge and Stroh 1993; Lodge and Taber 2005; Taber 2003). A cornerstone of any model of political reasoning is the citizen’s preexisting knowledge and predilections. These long-term factors, functionally speaking, require a vast long-term memory (LTM)
18
LODGE, TABER, AND WEBER
for storing facts, beliefs, and predispositions, and a mechanism for “moving” one’s knowledge about leaders, parties, and issues from LTM into working memory (WM) where it can be attended to (Barsalou 1992; Rumelhart and Ortony 1977; Sanford 1986; Simon 1969). Attention is very limited, perhaps to the magic number 7 ⫾ 2 bits or chunks of information (Miller 1957), hence the need for heuristics, habits, and other simplifying mechanisms for thinking and reasoning (Cialdini 2003). LTM is organized associatively, and it is useful to think of knowledge structures metaphorically in LTM as configurations of nodes linked to one another in a network of associations (Anderson 1983; Anderson and Reder 1999); or if you prefer as neurons “bundled” together by weighted connections (Smith, Fazio, and Cejka 1996). Were we able to tap a citizen’s complete political knowledge structure, there might be tens of thousands of nodes (among them one for George W. Bush) with a complex network of associations (perhaps his demographics, stands on issues, perceived traits, and maybe an inferential abstraction or two—e.g., that he is conservative). Links represent beliefs, the strength of which will vary. Moreover, memory objects vary in accessibility—the ease with which a stored concept lying dormant in LTM can be retrieved into WM. Figure 2.1 provides a simplified example of the architecture of one woman’s political knowledge (for a somewhat similar framework, see Greenwald et al. 2002). Note first that the self is the strongest node in the network and that identity (female, black) and self-esteem are the strongest links in the network. Positive and negative affect and basic identity nodes are distinguished in this representation because of their centrality in human information processing. As with more standard semantic network models, beliefs are represented as links among basic memory objects (e.g., “I am intelligent,” “President Bush has ties to big business”). Attitudes appear as links between basic memory objects and positive and/or negative affect. (Note that ambivalence can be represented by allowing links to both positivity and negativity, as with “American” in figure 2.1). The impact of context, or priming, on evaluations may also be depicted: see, for example, that if “jobs” is primed, “business” will be seen in a positive light, while in the context of “greed” “business” is evaluated negatively. Finally, we represent behavioral intentions—as here the intention “to vote Democrat”— as diamond shaped nodes (the traditional shape for decision points in flow charts). But how is information moved from LTM into WM? Spreading activation provides the mechanism. A node in LTM switches from being dormant to a state of readiness with the potential to be moved into WM when it is activated, either as a direct object of thought or because use it is closely linked to an object of thought. The rise time from dormant-state to activation
19
A DUAL-PROCESS ACCESSIBILITY MODEL ++++++++++++
W
hit
Am
e
er
ica
Jobs
Dem
n
Vote
Male
Female
Greedy Business
ME
AIDS in Africa
Bush Bl
Rep eig n
ac
Fo r
k
------------
Figure 2.1 The Structure of Political Beliefs, Attitudes, and Intentions. Source: Adapted from “The Automaticity of Affect for Political Leaders, Groups, and Issues: An Experimental Test of the Hot Cognition Hypothesis” by Milton Lodge and Charles S. Taber, 2005, Political Psychology, 23(3), p. 458. Copyright 2005 by the International Society of Political Psychology.
threshold is almost instantaneous (100–200 ms). Activation also decays quite rapidly so that a given node will drop back to baseline in about a second if there is no further source of activation. Imagine a person reading about President Bush in a newspaper headline. Without perceptible effort, the concept BUSH becomes activated and pops into consciousness. Even more important for our purposes, activation spreads along the network of links to related concepts, thereby “priming” strong semantic associations of BUSH (he is a REPUBLICAN) as well as beliefs (he is pro-business). For a few hundred milliseconds, these associated concepts remain in a heightened state of arousal, with any additional activation likely to push them over threshold and into WM. It may be useful to think of priming via spreading activation as producing preconscious expectations. Consider again the activation of the concept BUSH from a newspaper headline. Concepts associated with BUSH in LTM also receive activation, thereby raising their potential so that any subsequent processing which passes activation to these energized concepts will likely drive them over threshold. If a primed association (perhaps Bush’s Republican label or his stand on gun control) is “expected,” it takes substantially less processing to activate and has a better chance of getting
20
LODGE, TABER, AND WEBER
into WM, of being processed faster, and thereby of “framing” the perception, recognition, and interpretation of subsequent information. Conversely, spreading activation can inhibit the processing of unexpected categories. When a concept is encountered unexpectedly, more bottom-up processing is necessary before it may pass threshold and enter WM. If the word “walnut” were processed initially, this would inhibit the recognition of semantically unrelated concepts (such as REPUBLICAN), which would thereby take more time and effort to process. Extending this paradigm, we believe that affect is stored within this architecture by way of an online-running affective tally stored with the attitude object (Lodge, Steenbergen, and Brau 1995; Bower and Forgas 2001). In other words, the activation of a concept not only spreads through the semantic network of associated concepts, but also results in the spontaneous retrieval of the online affective tally associated with the object. Concordant with the aforementioned primacy-of-affect theory, one’s summary evaluation is believed to be the first characteristic to enter in the information processing stream and serves as a vital source of information used to evaluate the previously encountered object (Lodge, Steenbergen, and Brau 1995). Evidence for the Automaticity of Beliefs, Feelings, Goals, and Action With this brief excursion into the priming paradigm, which dominates the experimental study of automaticity in social and cognitive psychology, let us now turn to a discussion of the automaticity of beliefs, feelings, goals, and action. Which automatic responses are activated depends on their accessibility—a function of the set of preconditions operative in the environment and what is inside the individual’s head at the moment. The key here is that once triggered, once the preconditions come into play, the responses can occur without any further conscious or deliberative guidance. With this in mind, automatic responses—whether thought, feeling, intention to act, or overt behavior—are believed to occur spontaneously, within milliseconds of a triggering event. Thus, automaticity ensues (1) even if the individual is not focusing conscious attention on the object or on the event; (2) regardless of what the person was recently thinking about (i.e., the response can be primed subliminally, outside the individual’s awareness); (3) can occur independent of one’s conscious intention; and (4) once triggered the automatic response will be carried out without conscious monitoring. Many demonstrations of automatic processing rely on the attitude-priming paradigm developed by Fazio et al. (1986), and Bargh and colleagues (1992, 1996).
A DUAL-PROCESS ACCESSIBILITY MODEL
21
Within the attitude-priming paradigm, participants are exposed to a prime and, moments after, a target word. The participant’s task is to press a button labeled “plus” or “minus” to indicate “as fast as possible without making too many errors” whether the target word has a positive or negative connotation. On each trial the name of an attitude object (e.g., COCKROACH) is presented for 200 ms on a computer screen, followed by a 100 ms blank-screen interval. Then a target word is chosen for its unambiguous positive or negative connotation—for example, “delightful” or “disgusting.” The subject’s task is to indicate by a button press whether the target word is “good” or “bad” in meaning. Critical here for demonstrating automaticity is the elapsed time from onset of the prime to the onset of the target, called the SOA—the Stimulus Onset Asynchrony: The logic of the design, as noted by Bargh et al. (1992) is that if the object name activated the evaluation associated with the attitude object, this evaluation (good or bad) would then influence how quickly subjects could correctly classify the target adjective as positive or negative in meaning. If the adjective is of the same valence as the attitude object prime, responses should be faster (i.e., facilitated) relative to a baseline response. Conversely, if the adjective and prime were of opposite valence, responses should be slower. The time from the onset of the prime word to the onset of the target word (300 ms) is a critical feature of this priming paradigm as it is too brief an interval for subjects to develop an active expectancy or response strategy regarding the target adjective that follows; such conscious and flexible expectancies require at least 500 ms to develop, and to influence responses in priming tasks (Neely 1977; Posner and Snyder 1975). Given an SOA (interval from prime to target) of 300 ms, then, if presentation of an attitude object prime influences response time to a target adjective, it can only be attributed to an automatic, unintended activation of the corresponding attitude. (p. 894)
For instance, if COCKROACH were the prime and the target word “Disgusting” we would expect facilitation—a fast reaction time (RT), predictably, on the order of 500–600 ms, to say “Disgusting” is a negative word—a relatively fast response time because the prime and target are affectively congruent. Conversely, if the target word were “Delightful,” we would expect inhibition—a slower latency time to respond (on average around 800 ms) to say “Delightful” is a positive word because the association is affectively incongruent. Note here that this is a nonreactive measure: the participant’s task is to not to say whether the target word “Delightful” describes the prime word “Cockroach” but rather to simply indicate whether the target word is positive or negative, not whether the word is or is not semantically associated. Because (to be demonstrated) many if not all social concepts in LTM are affectively charged and their positive or negative tag is
22
LODGE, TABER, AND WEBER
triggered almost instantly on its mere exposure, prime-target pairings at a short SOA provide a strong test for discerning whether affect is automatically activated along with the concept itself. Affective priming within this paradigm has been demonstrated for hundreds of concepts (Bargh et al. 1992; Fazio 2001). The Automaticity of Political Concepts In a series of experiments employing this attitude-priming paradigm to test the hot cognition hypothesis (i.e., that political attitudes and beliefs are imbued with an affective association), Lodge and Taber (2000) had participants read a campaign brochure of a hypothetical Congressman, William Lucas. In addition to the Congressman’s picture was information detailing his background and experience as well as his strong position on the death penalty (pro for half the participants/con for the others). After reading the brochure, participants were engaged in a classical sentence verification task in which they indicated by a True/False button response whether LUCAS was, for example, a Republican [Yes], a woman [No], pro- [Yes] or anti[No] death penalty. The participants were also engaged in the attitudepriming task at a short SOA of 300 ms where “LUCAS” preceded such target adjectives as “Delightful,” “Disgusting,” “Angry,” and “Sad.” Our interest is in the reaction times for the cognitive True/False responses to the single-word targets “Congressman,” “Democrat,” “Married,” and “Woman,” as well as the mean reaction time to the affective target words in the attitude-priming task. Results indicated that, on average, it took participants about 700 ms to make an affective response to Lucas, about twice as fast as the time to verify the simple factual queries. We interpret this finding as support for our “hot cognition” hypothesis and, what is more, as support for Zajonc’s “primacy-of-affect” hypothesis (Zajonc 1980, 2000; Murphy and Zajonc 1993)—in that affect comes to mind faster than do the cognitive associations thought to cause the affective response. In three more recent experimental tests of the hot cognition hypothesis, Lodge and Taber (2005) employed the attitude-priming paradigm to test for the automaticity of affect toward political leaders, groups, and issues. In each of the experiments, it was hypothesized that reaction times would be faster for affectively congruent prime-target concepts. Critical to the hot cognition hypothesis is that one’s feelings are triggered automatically on the mere presentation of the priming concept; and accordingly, the predicted facilitation and inhibition effects should only emerge in the short SOA (300 ms). In other words, across these experiments we should observe stronger facilitation effects for affectively congruent prime-target pairs (positive/positive and negative/negative) than for incongruent pairs
A DUAL-PROCESS ACCESSIBILITY MODEL
23
(negative/positive and positive/negative). And, this effect should only be observed at short SOA exposures. In statistical parlance, we should observe a significant three-way prime valence ⫻ target valence ⫻ SOA interaction. Results from these studies indicated that facilitation/inhibition effects were, indeed, pervasive in the short SOA condition but not in the long SOA condition. Not only that, but this effect was strongest among political sophisticates providing evidence that the very individuals conventionally believed (or hoped) to be more resistant to emotion-laden responses (i.e., the politically informed), in fact, are most susceptible to emotional reactions. Results from these studies are presented in figure 2.2. Note that each bar represents an average reaction time (in milliseconds) for each of the basic groups defined by the prime by target valence interaction: from left to right, negative primes/negative targets, positive primes/positive targets, positive primes/negative targets and negative primes/positive targets. Primes consisted of political persons (e.g., Clinton, Hitler, Lincoln, Pataki), groups (e.g., Democrat, Republican, Politician), or issues (Antiabortion, Peace, Taxes, Death Penalty); whereas targets consisted of positive or
RT in milliseconds
650 625 600 575 550 Short SOA Neg/Neg
Pos/Pos
Long SOA Pos/Neg
Neg/Pos
Study 1: Short and long SOAs
RT in milliseconds
675 650 625 600 575 Study 2 Neg/Neg
Study 3 Pos/Pos
Pos/Neg
Neg/Pos
Study 2 and Study 3: Short SOAs
Figure 2.2 Reaction Times for Incongruent/Congruent Prime-Target Pairs of Long and Short SOAs. Source: Adapted from “The Automaticity of Affect for Political Leaders, Groups, and Issues: An Experimental Test of the Hot Cognition Hypothesis” by Milton Lodge and Charles S. Taber, 2005, Political Psychology, 23(3), p. 469. Copyright 2005 by the International Society of Political Psychology.
24
LODGE, TABER, AND WEBER
negative affect words (e.g., Cancer, Sunshine, Appealing). The appropriate comparison then is between the first and third bars (for negative targets) and between the second and fourth bars (for positive targets). We take these effects as evidence that affect is triggered automatically on mere presentation of a political attitude object. All well and good except for a potential confound in the research design that challenges our hypothesis that affect is triggered automatically on mere exposure of an attitudinal object. Recall that the target words in this experiment were adjectives, among them such trait descriptors as “Appealing” and “Delightful.” It is conceivable then that the congruence and incongruence effects we observed were cognitively, rather than affectively, based. That is, perhaps our participants made affect-laden trait inferences about the prime objects—for instance, that Clinton’s entertainment value is “Delightful,” the Democrats’ platform “Appealing,” the death penalty “Repulsive.” We know that people are prone to make trait inferences spontaneously, based on little direct evidence, so perhaps their affective responses were cognitively mediated. Perhaps something the Democrats did led our participants to infer that they are “Horrible” or “Marvelous.” Accordingly, the prime “Giuliani” may have activated a network of linked associations that resulted in the interpretation of something he did as “Magnificent” (leading to quicker reaction times). This is the standard memory-based interpretation for attitudinal response. There is, however, an alternative to this cognitively mediated process— the primacy-of-affect hypothesis, which posits that the cognitive and affective systems are somewhat independent, that affect and cognition follow separate pathways in the brain, with feelings following a quick and dirty route that “prepares” a behavioral response before one’s cognitive associations reach conscious awareness (Zajonc 1980). A strong test of this primacy-ofaffect hypothesis within the attitude-priming paradigm is to break any cognitive connection between the attitudinal prime and the target concepts. Thus, to test for an affective contagion effect, we examined whether attitudinal primes again facilitate reaction times to political persons, groups, and issues, but now the target words are nouns that are affectively unambiguous and semantically unrelated to the leaders, groups, or issues (among them, the targets Comedy, Miracle, Rainbow, Toothache). If we find facilitation effects for semantically unrelated but affectively congruent targets (and inhibition for semantically unrelated but affectively incongruent associations), we will have demonstrated the automaticity and contagion of affect for political objects (see figure 2.3). Figure 2.3 graphically represents the relevant interactions for these studies. To discern between affective versus purely semantic associations, in study 2 and study 3 only targets that were semantically unrelated to the political
A DUAL-PROCESS ACCESSIBILITY MODEL
25
RT in milliseconds
725 700 675 650 625 600 Persons Neg/Neg
Groups Pos/Pos
Pos/Neg
Issues Neg/Pos
Study 2
RT in milliseconds
650 625 600 575 550 Persons Neg/Neg
Groups Pos/Pos
Pos/Neg
Issues Neg/Pos
Study 3
Figure 2.3 Reaction Times by Prime Type at the Short SOA. Note: The appropriate comparison is between the first and third bars (for negative targets) and between the second and fourth bars (for positive targets) for political persons, groups, and issues. Source: Adapted from “The Automaticity of Affect for Political Leaders, Groups, and Issues: An Experimental Test of the Hot Cognition Hypothesis” by Milton Lodge and Charles S. Taber, 2005, Political Psychology, 23(3), p. 471. Copyright 2005 by the International Society of Political Psychology.
primes were used (e.g., Rainbow, Rabies, Toothache). As predicted, the three-way interaction for SOA, prime valence, and target valence was highly significant in both studies: at short SOA for both positive and negative targets, congruent primes elicited significantly faster response times than did incongruent primes; whereas at long SOA, there was no significant difference between congruent and incongruent pairs. In short, we find experimental support for the automatic activation of an evaluative tally for a wide range of political persons, groups, and issues. Taken together, support for the hot cognition hypothesis across these studies is striking. Averaged responses across a wide range of political primes show clear evidence of an automatic link in memory between a broad array of political concepts and positive or negative affect. Moreover, these studies eliminate any purely semantic interpretation of these facilitation and inhibition effects.
26
LODGE, TABER, AND WEBER
In the political domain the hot cognition hypothesis predicts these facilitation and inhibition effects to be contingent on the political sophistication of the respondent. Political sophisticates have presumably thought about and repeatedly evaluated most of the political primes, while those whose political knowledge falls below the sample median are thought to be less likely to have formed strong affective links in memory. As a result, nonsophisticates should not display the pattern of facilitation and inhibition that indicates automatic affect. The pattern of sophistication effects supports this hypothesis (Lodge and Taber 2005). Namely, we found that using “hard” as well as “easy” political issues resulted in low-knowledge subjects being less likely than sophisticates to display automatic affect toward political leaders and groups. And, more telling, we found significant sophistication effects for political issues— indicating that the affective association with political issues is much stronger among sophisticates. This overall pattern lends credence to the theoretical expectations underlying the formation of OL (online) tallies in suggesting that sophisticates, because of their interest in politics, have formed crystallized attitudes to a broad set of political issues. Virtually all New Yorkers in the aftermath of the 2000 election, regardless of level of sophistication, would have given some thought to Bush, Gore, Hillary, and Rudy. Similarly, most everyone would have formed an attitude about such mainstream groups as Democrats, Republicans, politicians, and the issues of guns, peace, and taxes. But when presented with the more difficult sample of primes, the “harder” primes—especially the issue primes—apparently required more thought than non-sophisticates were likely to have invested. Automatic Affect and Deliberation Within the social sciences, affective responses are often conceptualized as postconscious considerations to cognitively processed information (Lazarus 1991). For instance, feelings toward candidates, as measured by the National Election Studies (NES), are assessed by a series of open-ended questions where respondents state their likes and dislikes of the candidate and parties (Campbell et al. 1960). Attitudes, then, are classified by valence and intensity and used as predictors for electoral behavior, for example (Kessel 2004). Beginning with Zajonc (1980), research has pointed to the temporal incorrectness of such a deliberation leading to evaluations. Rather, emotional elements associated with the encountered object are believed to be the first to enter the information processing stream and may potentially influence higher-order deliberation. What is more, affective processes are believed to operate relatively independent of semantic processes. In short, cognitive features are believed to come to mind well after the affective
A DUAL-PROCESS ACCESSIBILITY MODEL
27
response and the affective response may influence subsequent processing— via confirmation and disconfirmation biases (Zajonc 1980; Murphy and Zajonc 1993; Lodge and Taber 2005). With this in mind, the deliberative process will often serve as a post-hoc justification for one’s affective response. Yet much of what we as political scientists claim to know about political beliefs and feelings is based on verbal self-reports. In both surveys and experiments, we routinely ask people factual questions, ask for their beliefs, their feelings, intentions, and past behaviors. The questions—the concepts they invoke—determine how memory is searched and consequently what information is retrieved. From this perspective one can think of the questions as primes, causing some thoughts, feelings, goals, and behaviors to become more accessible. How people respond—what they say, how they say it, the speed of their response, and what they don’t say—is a function of what information is available and accessible in memory. While the general principles guiding the role of accessibility and the retrieval of information are well known (Anderson 1983) and found to be more or less the same for facts and feelings (except for the critical provisos that affect comes to mind faster than do cognitive associations and affect decays more slowly than do the contextual details), let us focus on how automatic affective reactions can influence deliberation and the conscious processing of information. Because affect comes to mind automatically at the earliest stages of information processing, we would expect an immediate “primacy-ofaffect” effect on subsequent processing, such that one’s prior attitudes will powerfully constrain how information is initially encoded, how it is stored in LTM, and consequently how it is retrieved, interpreted, and acted upon. In a series of complementary experiments we repeatedly find (Lodge and Taber 2005)—as do others in nonpolitical domains (Ditto and Lopez 1992; Edwards and Smith 1996; Lord, Ross, and Lepper 1979; Monro et al. 2002; Sherman and Kim 2002)—that one’s prior attitudes are quite resistant to change. Even when motivated to be evenhanded, “to leave their feelings aside,” people find it near impossible to view political policy arguments dispassionately (on gun control, affirmative action, federal support for the arts, etc.). Those holding strong attitudes actively counterargue contrary facts, figures, and interpretations, while uncritically accepting attitudinally congruent information—a disconfirmation bias—and they actively seek out supporting information so as to bolster and protect their priors— a confirmation bias. As a consequence, both biases lead to attitude polarization, especially among the sophisticated and those with strong priors. Although largely unrecognized in mainstream political research, the implications of automatically activated affect may be influential beyond effortless “peripheral” processing. On one hand, since affect is the first component to be accessed from memory, the primacy of these responses,
28
LODGE, TABER, AND WEBER
may indeed, be the primary vehicle implicated in motivated reasoning leading to selective attention, information distortions, and recall biases (Lodge and Taber 2005). The extent to which these initial reactions influence deliberative processing, however, requires further clarification. Subsequently, in a recent set of experiments, we tested whether affectladen primes (e.g., scowling and smiling faces) subliminally embedded within campaign advertisements influenced postconscious evaluations of the candidates targeted in these advertisements (Weber, Lodge, and Taber 2005). To examine this, we employed a simple 2 (candidate known/ unknown) ⫻ 2 (subliminal prime present/absent) ⫻ 2 (positive/negative ad tone) design finding that subliminal affective primes strongly influenced candidate evaluations and feelings for the ad-targeted candidates. For instance, following each ad, we asked participants to rate candidates on nine evaluative items. Our expectations were that positive subliminal primes should raise positive evaluations; whereas negative subliminal primes should yield the opposite effects—which, essentially, is what we found. Namely, the affective primes influenced candidate evaluations and feeling thermometer ratings in the expected direction but only when the politician was well known and the ad was negative in emotional tone. We take this as tentative evidence supporting the notion that affective responses are not exclusive to immediate and effortless decisions and that there is a direct correspondence between affect and higher-order cognitive systems. Notwithstanding, further research should clarify whether automatic reactions can be suppressed given environmental and dispositional contingencies, such as time, motivation, and ability to process messages beyond one’s spontaneous response (Greenwald and Banaji 1995; Devine 1989). And, to appease the skeptic—claiming a minimal, even unimportant, role of automatic affective reactions pertaining to political events—a fruitful research endeavor will be to apply the automaticity paradigm to political and social situations. Conclusion Throughout this essay we have voiced skepticism of people’s ability to reliably and veridically access their beliefs, attitudes, or their past or future intentions and actions. Because much, if not most, of our experiences take place outside our conscious awareness, and as our conscious recollections fade from memory, they are automatically replaced by socially constructed beliefs about how and why we as well as others respond emotionally. Much of the evidence presented relied on one or another priming paradigm with the most compelling evidence coming from experiments that exposed participants to primes outside their conscious awareness. Which automatic
A DUAL-PROCESS ACCESSIBILITY MODEL
29
responses are activated depends on the set of preconditions operative in the environment at the moment and what’s going on inside the individual’s head at the moment. The key here is that once triggered, once the preconditions come into play, cognitive, affective, and behavioral responses can occur outside of awareness without any further conscious or deliberative guidance. To the extent that “some” (Fazio 1993) or “all” (Bargh et al. 1992) sociopolitical thoughts, feelings, intentions, and actions invoke an automatic, affective response that biases subsequent evaluations, judgments, and choice, then our discipline’s focus and reliance on conscious, introspectively accessible considerations as the primary, or even independent, mediator of response will consequently fail to model correctly how citizens think, reason, and act. What we need to learn is how, when, and for whom will the spontaneously triggered considerations elicited in interviews, questionnaires, and experiments impact the conscious appraisal of political judgments, evaluations, intentions, and actions. A dual-process model of public opinion, which incorporates both conscious and implicit processes into the evaluation and judgment process, would improve our understanding of how citizens form, update, and express their political beliefs and preferences. The implicit effects cited here reflect how beliefs, feelings, intentions, and behaviors that were contiguously associated in the past or momentarily joined in a situational context can be automatically activated by mere exposure of an environmental trigger or situational cue. Specifically, ●
●
●
With contiguous activation, simple as well as complex responses become automatic (Shiffrin and Schneider 1977); Once automatized, beliefs cannot be readily accessed by introspection (Nisbett and Wilson 1977; Wilson and Schooler 1991); Such responses are constructed online, at the very instant that attention is momentarily fixed on the object or are triggered without conscious awareness of the environmental trigger (Bargh 1997; Greenwald, Pickrell, and Farnham 2002; Lodge and Taber 2005).
With this in mind, we believe that a more thorough review of the psychological literature than carried out to date will show that, ●
●
●
It will prove impossible to tease apart the cognitive from affective from goal expressions of behavior; Both conscious and automatic processes link beliefs to feelings to intentions to action; Affective responses will enter into the decision stream earlier than do the cognitive associations (Zajonc 1980, 2000). And in many cases, this affective component will cascade across subsequent higher-order processes;
30
LODGE, TABER, AND WEBER ●
Similarly, that affect provides the essential motivational thrust to set and carry out goals (Damasio 1994, 1996, 1999; Marcus and MacKuen 1993).
We expect that individual differences (in particular strength of prior attitude and political sophistication) will prove to be most important as citizens with the strongest attitudes and most knowledge are the most likely to have repeatedly connected their beliefs to feelings, their feelings to goals, and thereby will be most motivated to act in accord with their beliefs and feelings, while their knowledge provides the cognitive wherewithal to choose the most rewarding (and thereby likely to be repeated) course of action. However, these very same citizens—those knowing the most and feeling the strongest— are the most prone to confirmation and disconfirmation biases (Lodge and Taber 2005) that keeps them from integrating new, contrary information in an evenhanded way that the good Reverend Bayes would approve. The finding that semantically unrelated primes can produce facilitation and inhibition effects implies an affective contagion effect that seriously challenges conventional ways of thinking about the formation and expression of thoughts and feelings (Fazio 2001). To date there is much discussion, but no consensus, as to how or why this effect works. One possibility, following Zajonc’s (1980, 1984) account of the primacy of affect, is that the cognitive and affective systems are separable and somewhat independent (though perhaps architecturally interrelated as depicted in our figure 2.1). There is some neurological evidence (e.g., LeDoux 1996) that the affect system is easily and swiftly sparked and once activated generates a “quick and dirty” approach-avoidance reaction to the situation ( JUMP, before you know if it is a stick or a snake), with conscious, deliberative appraisal following moments later. From this perspective the automatic affective response is primary and may or may not (depending on individual and situational factors) be overturned by a later conscious, cognitive assessment (Devine 1989; Murphy and Zajonc 1993). As students of political behavior, a key question—alluded to but at this juncture unanswered—is when, how, and to what extent will one’s immediate, spontaneous affective response impact deliberative processes? We suspect that because people are in some sense aware of their feelings before they are cognizant of an object’s meaning, our prior attitudes will prove to be a, if not the single most, powerful determinant of what citizens think and say when they talk to themselves, answer a pollster’s questions, or act in accord with their intentions. From this perspective Antonio Damasio (1999) is right in claiming, “the brain is a feeling machine for thinking” and William James (1890) was correct in believing that “thinking is for doing.”
CHAPTER 3 THE MEASURE AND MISMEASURE OF EMOTION George E. Marcus, Michael MacKuen, Jennifer Wolak, and Luke Keele
motion, after a modest hiatus during the “cognitive revolution,” has reemerged of late to become a subject of significant attention in political science.1 The other contributions in this volume give ample evidence of the added understanding we gain by including emotion into the theoretical and empirical mix. Our entry in this volume turns to a question relevant to most if not all the other research found here: how do we best measure emotional response? We examine two central considerations— identifying which emotions define political responses and determining which kinds of questions are most suitable to assess these emotional reactions. Evaluating the measurement of emotion is important both because of the inherent challenges in securing reliable and valid measures of emotional reactions, as well as the sensitivity of our understanding of emotional reactions to our choice of measures.2 To begin, we look to psychology literature as a source for guidance on how to measure emotion. After briefly reviewing the findings in psychology, we argue that applying conventional wisdom from psychology would distort our understanding of how emotion impacts politics. We next present evidence from both an NES (National Election Study) Pilot study and experimental research on what is currently the best practice for the measurement of emotional response. From this, we develop a series of recommendations about how to measure emotion, considering both matters of question wording and the dimensions of emotional response.
E
32
MARCUS, MACKUEN, WOLAK, AND KEELE
Measuring Emotion—The View from Psychology Early investigations of the measurement of emotion in social psychology began with the task of description, where researchers essentially asked their subjects to describe the emotions they experienced. The strategy was to collect multiple measures of many different emotions, with the expectation that the analysis of this data would reveal the emotions people experience. The focus was on measuring emotional response in isolation from any broader theoretical perspectives. Thus, this literature speaks to the “structure of emotion,” particularly what to name emotions and where to locate factor dimensions (Russell and Carroll 1999a, 1999b; Watson and Tellegen 1999). The general consensus in social psychology is that the variations in mood that people experience can be best described as filling a dense circle (called a circumplex), where two dimensions are needed to adequately describe that structure (Fabrigar, Visser, and Browne 1997; Larsen and Diener 1992; Plutchik and Conte 1997; Remington, Fabrigrar, and Visser 2000; Russell 1980).3 The two dimensions are generally presented as orthogonal.4 Figure 3.1 presents data of the moods that people experience and how they relate to each other, illustrating whether different moods are similarly or disparately experienced.
delighted enthusiastic ♦ peppy ♦ excited ♦ cheerful♦ ♦♦lively♦ happy♦ ♦ glad ♦ ♦ elated stimulated pleased active ♦ ♦ surprised euphoric astonished ♦♦ ♦ warmhearted ♦ ♦ aroused content ♦ intense ♦ tranquil relaxed ♦ ♦ at ease ♦♦ at rest calm
anxious jittery ♦♦ drowsy fearful ♦ nervous ♦ ♦ ♦ ♦ ♦ annoyed ♦ tired distress serene grouchy ♦ ♦ idle ♦ sad still ♦ sluggish passive ♦ ♦ gloomy ♦ unhappy bored blue ♦ ♦ ♦♦ ♦ ♦ quiet miserable droopy ♦ ♦ inactive dull
Figure 3.1 Typical Distribution of Experienced Moods. Note: Data are from Rusting and Larsen (1995).
MEASURE AND MISMEASURE OF EMOTION
33
In this instance, subjects were asked what they were feeling at various points during a period of time, using a mood checklist to mark how much of each emotion they experienced at that moment. This example concerns general emotional reactions but similar results are found when people are asked to report emotional reactions across a wide array of stimuli (Lang et al. 1993). The illustrated example uses 48 terms for emotional states, but the circumplex is generally the pattern that emerges whether more or fewer terms are used. Studies like this one are helpful in determining which emotion “markers,” or mood terms, are equivalent for measurement purposes (such as “happy” and “pleased”), and which are identifying quite different feeling states (such as “delighted” and “relaxed”). In considering the dimensions of emotional response, two different factor solutions have been put forward, each disagreeing on where in the circumplex space to locate two orthogonal dimensions (see Berkowitz [2000] for an excellent summary of this controversy). The first of these accounts, the valence approach, identifies the causal structure with two primary axes: one dimension understood as pleasurable-unpleasurable, and a second, often-ignored dimension labeled as arousal.5 This account describes emotions as a way to classify all manner of things, so that we can approach those that are experienced as pleasing and avoid those that are experienced as distasteful.6 The principal appeal of this model is that it conveniently focuses our attention to the valence quality of emotional experience (Clore and Ortony 2000). The second view identifies the two primary axes as positive affect (high versus low) and negative affect (high versus low), alternatives that are 45-degree rotations one to the other. This model, most often linked with John Cacioppo and David Watson (Cacioppo, Gardner, and Berntson 1997, 1999; Watson et al. 1999), argues that each of the dimensions of affect has a different evaluative purpose. Positive affect assesses the desirable qualities of stimuli, while negative affect is an independent assessment of the strategic novelty of the stimuli (Marcus and MacKeun 1993: Marcus, Neuman, and MacKuen 2000). This model argues that for many (if not most) stimuli, joint activation is possible, where our mood at any given moment will have emotions representing each of these two assessments. Rather than having only an approach-avoidance function as in the valencearousal view, this positive-negative understanding of affect argues that emotions serve multiple functions. Both of these theories have important measurement implications. For the valence model, in as much as it ignores the “arousal” dimension, measurement is simple. It requires only a single instrument per stimulus for assessing emotional response (i.e., do we like, or feel warm, toward [the stimulus] or do we dislike, or feel cold toward [the stimulus]). One need
34
MARCUS, MACKUEN, WOLAK, AND KEELE
only use a “feeling thermometer” or assemble a number of suitable positive and negative feeling terms, such as happy and sad, and go forward.7 On the other hand, for the positive and negative affect view, one must collect two independent concurrent emotional responses, one defined by the positive dimension and the other by the negative dimension.8 Which understanding is most appropriate when assessing emotional reactions to political stimuli? We argue that psychological models of emotions will not necessarily be able to capture several important aspects of emotional response in politics. While political scientists are quite used to relying on the other social science disciplines to formulate and measure concepts most germane to those disciplines’ expertise, the distinctiveness of the political environment makes a direct application of psychological measures unwise. For instance, social psychological models of emotion are concerned with providing universal theoretical accounts, while the particular emotions of politics are quite distinct from those individuals might encounter in their nonpolitical lives. Consider the number of distinct emotional reactions politics can inspire, from the kinds of feelings political candidates evoke, the emotions prospective wars might inspire, or the sentiments that attendance at a political rally might generate. The realm of politics is a venue of heightened conflict and opposing viewpoints, where people disagree not only on different politics, but also about their feelings. So in translating these psychological models of emotional response to the political arena, we see two areas where these models fail to describe the range of emotional experiences in politics. The first concerns the relationship between the two factors, argued to be orthogonal in the social psychology models. The second concerns the focus on only one or two dimensions of emotions. We argue that neither appears to be universal properties of emotion, and if they are treated as such, it will lead to the mismeasurement of emotion in politics. We first explore the difficulties with treating the two dimensions as strictly orthogonal. Studies have shown that the orthogonality of the two dimensions is not a ubiquitous feature of all emotion response (Abelson et al. 1982; Marcus, Neuman, and MacKuen 2000). For over twenty years, it has been noted that early in political campaigns, when the two dimensions are allowed to be non-orthogonal (leaving the relationship between two dimensions to be empirically assessed, rather than fixed a priori), typically there is a mild positive correlation between the two dimensions that becomes negative as the campaign evolves. This dynamic pattern is more evident for new candidates than for incumbents, suggesting that candidates new to the electorate provoke both a positive response and a degree of uncertainty.9 As voters learn about the candidates, the correlations between these reactions change
MEASURE AND MISMEASURE OF EMOTION
35
from positive to negative, such that experience with a candidate tends to lead citizens to feel either enthusiastic or anxious. Thus, some form of emotional inter-dimensional resolution takes place (Abelson et al. 1982; Marcus, Neuman, and MacKuen 2000). This dynamic is a meaningful one, informing our understanding of the nature and purpose of emotional appraisals in response to political stimuli. By ignoring these dynamics and restricting data to strict orthogonality, researchers will be blind to such dynamics. And these emotional dynamics are at the center of politics, since managing the new and the uncertain (appraising anxiety), as well as determining whom to entrust with our support (tied to feelings of enthusiasm), are recurring issues in democratic politics. The failure to take the dynamic character of non-orthogonality into account is a serious problem for any study that focuses on just one dimension of emotion, as in the valence model. In the case of the two-factor model, if the axes are correlated—non-orthogonal—then there may be serious confounding as the reported relationships may be stronger or weaker or even misattributed. Failing to control for correlated and concurrent emotional responses jeopardizes such studies.10 Therefore: Recommendation 1: Do not automatically apply an orthogonal restriction when exploring the dimensionality of emotion measures. Especially with panel data, enabling measurement of emotion reactions over time, the dynamic change in the correlation between factors will provide meaningful evidence of the changing structure of emotional appraisal. Freeing the relationship between dimensions, as is done in structural equation modeling (SEM), enables the relationship between dimensions to be an empirical and substantively meaningful datum, rather than a presumptive feature of the structure of emotion.
The second problematic area of translating the psychologically defined models of emotion to political questions concerns the number of dimensions of emotional response. The two dominant models in psychology conflict on whether emotion is best understood as having one (valence) or two (positive-negative) central dimensions. In the case of politics, however, we argue that neither model is sufficient for capturing the range of emotional experience that define the political realm. As Bodenhausen, Sheppard, and Kramer (1994) show, negative affect may take multiple forms in the circumplex, each with quite different impacts on human judgment. They highlight in particular the role of anger as negative affect, an emotion that is largely ignored in social psychological studies on emotion and the circumplex. In the next section, we elaborate on the political manifestations of negative affect.
36
MARCUS, MACKUEN, WOLAK, AND KEELE
Measuring Emotion—The View from Political Science To explain how emotion operates in the political realm, we have developed the theory of affective intelligence, based on Jeff Gray’s work (Gray 1973, 1987a, 1987b, 1990). At its core is the assertion that multiple independent but dynamically interactive emotional preconscious appraisal systems function to interpret sensory streams. Depending on their evaluations, these emotional appraisal systems shape feelings, consciousness, and certain motivational states, as well as behavior (Marcus 1988, 2002; Marcus and MacKuen 1993; Marcus, Neuman, and MacKuen 2000). The first system, the disposition system, manages the assessment of ongoing learned behaviors, while the second, the surveillance system evaluates the incoming sensory stream to assess the environment for its safety in relying on learned routines. Central to the first system is the emotional reaction of enthusiasm, encapsulating the likes and dislikes individuals hold relative to political stimuli. Importantly, the disposition system also includes aversion, the emotional response for when a person experiences a familiar but hated stimulus. Aversion represents strong negative responses like loathing, hatred, and contempt. While enthusiasm relates to the pursuit of positive goals, aversion relates to neutralizing disliked elements. The second system, the surveillance system, is oriented around the emotion of anxiety, engaged in novel situations, noting occasions where focused engagement with the task at hand might be prudent. To evaluate this theory of emotion, we consider findings from the 1995 NES Pilot study. An advantage of this survey is that a larger than usual set of emotion items were included. (After an initial reliance on a larger set of emotion items, the number of items in the NES was reduced to four items, which remains the current convention.) Most of the emotions included were chosen to measure the two dimensions of the positive and negative affect dimensions identified by Watson and others. However, six items meant to tap aversion, angry, bitter, resentful, disgusted, hatred, and contempt, were also added. While these emotions are often ignored in social psychological studies of mood, they are of particular importance in the realm of politics. If the valence model is correct, we would expect one of the two dimensions would emerge with pride/hope defining one end of a bipolar dimension and hatred/contempt defining the other. If the positive and negative affect view of emotion were sustained, then we would expect afraid, anxious, and worried would define one dimension of negative affect, and proud, hope, and enthusiastic, defining a second positive dimension. Neither would explicitly provide for the additional six items. This group of items then offers a rich opportunity to test the usefulness of social psychological
MEASURE AND MISMEASURE OF EMOTION
37
accounts, and the emotions they define as central, to politics. It also enables us to see if the emotion of aversion forms a separate measurable response, one not measured by either the valence model or two-dimensional measures. Just under five hundred subjects were interviewed in this study. Two different measurement approaches to ascertaining emotional responses were explored in a split-half design. Half the sample was asked a battery of emotional items relative to the presidential candidates following the traditional NES question wording, while the other half of the sample offered emotional reactions in an alternate Likert format. The wording of the traditional frequency of emotional response question is as follows: 11 Now we would like to know something about the feelings you have toward some people in politics. I am going to name a political leader and I want you to tell me whether something about that person, or something he has done, has made you have certain feelings like anger or pride. Has Bill Clinton— because of the kind of person he is, or because of something he has done— ever made you feel angry? How often would you say you’ve felt angry—very often, fairly often, occasionally, or rarely?
In addition to angry, emotional reactions assessed included bitter, resentful, disgusted, hatred, and contempt, as well as reactions of afraid, anxious, worried, hopeful, and enthusiastic. Analyzing the NES frequency format items shows the multidimensionality of mood quite clearly. When respondents are asked to evaluate emotional reactions to presidential candidate Robert Dole, we find a two-dimensional result, shown in table 3.1.12 Here, two dimensions, one positive and one negative emerge (eigenvalues of 5.36 and 2.19). The results fail to support the valence-arousal model, as the positive valenced terms are on one dimension orthogonal to the negative terms, rather than forming a single valence dimension. The coexistence of two dimensions undermines the view that emotion serves primarily to enable appraisals of approach or avoidance. The positive factor is defined by proud, hope, and enthusiastic, while the second negative factor is equally well described by all the negative emotion terms. The additional items related to aversion (angry, bitter, resentful, disgusted, hatred, and contempt) do not suggest that aversion is to any degree distinguishable from the other negative items (afraid, anxious, and worried). Our surmise that aversion is a distinct emotional phenomenon, requiring its own measurement, is unsupported by these results. The emotional reactions evoked by presidential candidate Robert Dole confirms research that indicates
38
MARCUS, MACKUEN, WOLAK, AND KEELE
Table 3.1 Factor Analysis of Emotional Response, Frequency Response Format Emotional Reaction
Senator Robert Dole Factor 1
Afraid Anxious Worried Angry Bitter Resentful Proud Hope Enthusiastic Disgusted Hatred Contempt
Factor 2
.820 .704 .855 .777 .726 .833
President William Clinton Factor 1
Factor 2
.797 .758 .749 .501 .357 .350
.582 .678 .639 .857 .895 .862
.798 .634 .607
.425 .782 .712
Factor 3
.864 .839 .849 ⫺.397
.466
Note: Principal components factor analysis, varimax rotation. Source: Data from 1995 NES Pilot, Form A.
that emotion has a two-dimensional structure (Abelson et al. 1982; Bruce 1994; Marcus 1988). We next analyze emotional reactions to President William Clinton. In this case, we find three dimensions of emotional response, in contrast to the two dimensions shown in evaluations of Dole. The principal components analysis of the measures for Clinton, again with varimax rotation, reveals three dimensions (eigenvalues of 4.61, 1.82, and 1.24). Three of the “negative” items, afraid, anxious, and worried, clearly define one of the affect dimensions. Similarly, proud, hopeful and enthusiastic define a positive affect dimension. The third dimension is composed of the aversion items: angry, bitter, resentful, hatred, and contempt. In this case, respondents have distinct emotional responses of anxiety and aversion, confirming other studies where a second negative affect dimension related to aversion also occurs (Conover and Feldman 1986). Considering emotional reactions to both candidates, we find that while sentiments of aversion, hatred, and contempt at times reflect a general dimension of negative affect, as in the case of Dole, other situations and stimuli, such as the incumbent candidate Clinton, produce emotional reactions where aversion is differentiated as a distinctly different negative emotional response. To demonstrate that distinct aversive reactions are not confined to this single case, we also analyze experimental data to explore the structure of emotion (MacKuen et al. 2001a). In this study, subjects were presented with the website of an online newspaper, featuring an article on affirmative
MEASURE AND MISMEASURE OF EMOTION
39
action in universities.13 The article was framed in one of two ways, either as reporting that affirmative action programs in university admissions were being greatly strengthened or instead being abandoned. Participants were randomly assigned to one of these two pages, and, from data obtained in a prior questionnaire, we know that they will find that first page either affirming or confronting their own position on that issue. After reading the initial stimulus article, participants could then go from page to page, as in any website, to read more, or go directly to a final questionnaire that asked about reactions to the policy, including a battery of emotional response items. For instance, we asked about the level of emotion experienced by the participant, from a lot to none at all: Would you say the affirmative action policy makes you feel: VERY enthusiastic, SOMEWHAT enthusiastic, NOT VERY enthusiastic, NOT AT ALL enthusiastic.
We assessed several emotion reactions in addition to enthusiasm, including proud, hopeful, anxious, afraid, uneasy, angry, contempt, bitter, and disgust. We have three items each to define the two conventional dimensions, enthusiasm and anxiety, and four items to define aversion. For participants reading an article that confirmed their issue position, we would predict that only two dimensions emerge, one positive dimension and one negative dimension defined by both the anxiety and the aversion items. However, when confronted with an issue that challenges them, people may respond with anxiety if the issue is new or unexpected or with aversion if the stimulus is a familiar challenge. When people are confronted with an issue where distinct negative impressions have been established, the negative emotion of anxiety will carry a different character than the negative emotion of aversion. But if the issue presented is one in line with an individual’s own policy desires, this separate dimension of disgust or strong dislike will not be raised. Thus, for those sent to a news article diametrically opposed to their issue position, we predict that the four aversion items will form a third dimension of emotional response separable from anxiety. Table 3.2 presents the principal components analysis for all subjects. Table 3.3 displays the same analysis separating the reactions of those who initially saw a confirming webpage from those presented with a confronting webpage. The results shown in tables 3.2 and 3.3 support our expectations. Not all political stimuli engage distinct aversive reactions. Those who read about policies in line with their policy preferences reveal two dimensions of emotional response, positive and negative. But for those confronted by a stimulus that is familiar or unfavorable, negativity assumes two dimensions—anxiety as well as aversion.
40
MARCUS, MACKUEN, WOLAK, AND KEELE
Table 3.2 Emotional Response to Affirmative Action Policy, Experimental Studies Emotional Reaction Proud Hopeful Enthusiastic Contempt Bitter Angry Disgusted Anxious Uneasy Afraid
Factor 1
Factor 2
— — — .746 .743 .739 .637 .746 .690 .679
.875 .894 .899 ⫺.347 ⫺.457 ⫺.491 ⫺.590 — ⫺.509 —
Note: Principal components factor analysis, varimax rotation. Source: Data from 1995 NES Pilot, Form A.
Table 3.3 Emotional Response to Challenging Versus Affirming Affirmative Action Policy, Experimental Studies Emotional Reaction Proud Hopeful Enthusiastic Contempt Bitter Angry Disgusted Anxious Uneasy Afraid
Affirming Condition Factor 1
.810 .734 .761 .585 .599 .724 .677
Challenging Condition
Factor 2
Factor 1
Factor 2
.832 .897 .880
⫺.329
.866 .825 .857
⫺.316 ⫺.384 ⫺.534
.826 .869 .854 .712
⫺.414
⫺.339
.437
⫺.477
Factor 3
.881 .542 .717
Note: Principal components factor analysis, varimax rotation. Source: Data from studies conducted Spring 2001 and Spring 2002.
In sum, we find that aversion is a dynamic phenomenon, drawn out by some stimuli but not others.14 This highlights the importance of including questions to assess emotional reactions of aversion. While excluding aversion emotions in measurement may not be problematic in every instance, their omission can lead to oversimplified assessments of emotional reactions. Previous research shows that aversion has powerful effects
MEASURE AND MISMEASURE OF EMOTION
41
distinct from those of anxiety (MacKuen et al. 2001b) and sadness (Bodenhausen, Sheppard, and Kramer 1994). As such, identifying the presence of aversion is consequential for understanding the nature of people’s emotional responses.15 Therefore: Recommendation 2: Unless one has independent evidence that aversion is not a germane response to the focus of investigation, one should include measures of aversion in addition to measures of anxiety. If aversion is absent, these latter items will act as if they are measures of anxiety. If aversion is present, they will differentiate and form a second “negative” dimension of emotional appraisal.
How to Ask Citizens about Emotional Reactions to Politics Next, we consider the optimal strategies for asking people to describe their emotional reactions to politics. Within psychology, David Watson has studied different approaches to measuring emotion by self-report.16 He considers whether formulating the response as a measure of frequencies (“how often have you felt angry about the stimulus”) produces the same responses as when the question is posed as intensity (“how angry would you say you feel about the stimulus”). Watson’s findings (1988b) suggest that subjects do not differentiate how often they feel and how much they feel about some stimulus (e.g., themselves, their friends, their political allies, or opponents). Here, different measurement approaches, considering the frequency of emotional experience versus measuring the intensity of emotional experience, are essentially equivalent.17 Therefore: Recommendation 3: Researchers may adopt a framing response of either “how often have you felt angry about. . .” or “how angry do you feel about. . .” Asking subjects to judge intensity of an emotion produces the same results as asking subjects how often they have felt an emotion.
We also are interested in what emotional markers are most appropriate to capture emotional reactions. In considering the factor loadings of emotion items across both the survey and experimental data presented, we find that these markers are generally reliable indicators of categories of emotional response. Following from Watson’s research as well as these findings, we suggest: Recommendation 4: For measures of enthusiasm and anxiety, we recommend using the markers identified by Watson as reliable. For enthusiasm, this includes enthusiastic, hopeful, and proud (to which could also be added
42
MARCUS, MACKUEN, WOLAK, AND KEELE
interested and elated ). For anxiety, this includes anxious, worried, and afraid. For measures of aversion, we recommend hatred, contempt, bitterness, and resentful.
The NES Pilot study also provided an opportunity to examine the effect of alternate question-wording strategies in the measurement of emotional response. In the second measurement strategy of the split-half design, people were asked to report how often they felt each emotion about prospective presidential candidates. Unlike the standard frequency format used in the other half of the sample, where respondents were asked how often they felt each listed emotion, respondents in this second format were given items with a semantic differential, or Likert scale of labeled opposites. Here, respondents are provided with five semantic labels that identify different degrees of positive, neutral, and negative responses to some stimuli. In this instance, the emotion pairs included enthusiastic/indifferent, hopeful/discouraged, anxious/calm, upset/relaxed, and angry/comfortable. The Likert scales juxtapose two distinct mood terms into a single question rather than asking subjects to rate how intensely or frequently they experience one specific emotion. What follows is a sample item in this format: Now we would like to know something about the feelings you have toward some people in politics. I am going to name a political leader and I want you to tell me, when you focus on that person, how he makes you feel. How would you say Bill Clinton makes you feel at this moment. . .Does (Bill Clinton/he) make you feel very enthusiastic, somewhat enthusiastic, neither enthusiastic nor indifferent, somewhat indifferent, or very indifferent?
The results of principal components factor analysis, shown in table 3.4, reveal that the Likert response format items unambiguously define a single Table 3.4 Factor Analysis of Emotional Response, Likert Response Format Emotion Term
Enthusiastic—indifferent Hopeful—discouraged Anxious—calm Upset—relaxed Angry—comfortable Eigenvalues
Senator Robert Dole
President William Clinton
Factor 1
Factor 1
⫺.753 ⫺.882 .704 .858 .845
⫺.820 ⫺.867 .792 .809 .887
3.290
3.491
Note: Principal components factor analysis, varimax rotation. Source: Data from 1995 NES Pilot, Form B.
MEASURE AND MISMEASURE OF EMOTION
43
dimension. This is in contrast to questions’ structures concerning the frequency of emotion items, which reveal two or three dimensions (see tables 3.1 and 3.2). Therefore, we can clearly conclude that this is a measurement artifact. Use of Likert (bipolar) measures that juxtapose different emotions within the same question tend to produce single dimensions, while other frequency measures of emotion reveal multiple dimensions. Therefore: Recommendation 5: Avoid using Likert scales in measuring emotional response. Such measures will “flatten out” subject’s self-report into a single valence response. To date, the evidence is that this is a measurement artifact rather than a meaningful subjective self-report.
One obvious implication of this research is that feeling thermometers, like the Likert scales, will collapse the subject’s emotional self-report to some summary state that is itself a confounded index of two, or even three, underlying discrete emotional feeling states (Marcus 1988). Whether such measures should be used at all depends on evidence that there is some value in using confounded measures. The common claim advanced is that such measures provide a “global” summary report of a person’s evaluative state (Greene 2002; Rahn et al. 1990). However, most studies do not include distinct measures of the three primary dimensions of emotional evaluation (enthusiasm, anxiety, and aversion) and are therefore unable to test whether the two formats reveal different results. Conclusion The aforesaid recommendations reflect the current state of understanding on what self-report methods best suited to measuring emotional response in survey and questionnaire-based research. We find that emotion is especially sensitive to the form of measurement. Subtle and seemingly inconsequential changes can have major consequences for the responses our subjects are likely to offer. Some of forms of measurement obscure the nature of emotional response, while other reveal. Nonetheless, it is our hope that as more research is conducted we will learn how to better measure emotion. Hopefully the recommendations put forward here will encourage still more research on the measurement of emotion. Notes 1. This hiatus of interest in emotion was unusual since for during most of the history of political thought on politics the role of emotion was at the forefront of concern, respecting its force and centrality (Nussbaum 2001).
44
MARCUS, MACKUEN, WOLAK, AND KEELE
2. We restrict our focus to subjective measures of emotional self-report, the approach most useful to political psychologists and political scientists. Other measurement approaches to the study of emotion use psychophysiological measures, such as facial EMG and galvanic skin response. While these are useful they have their limitations, not the least of which is their restriction to laboratory settings. 3. For a brief period, the view was advanced that mood could be reduced to a single dimension, once measurement error was formally incorporated (Green, Goldman, and Salovey 1993). That in turn led to a spate of articles pro and con, however, at the end of the day, all of the participants now seem to agree that two dimensions are minimally required (Marcus 2000). 4. The depiction of the axes as orthogonal follows from the near universal reliance on factor analysis to find the underlying structure of covariance matrixes. While factor analysis allows, if one chooses, for non-orthogonal solutions, only the Tellegen and associated psychologists have entertained the possibility, let alone explored it (Tellegen, Watson, and Clark 1999a, 1999b). 5. Here, the far points of the pleasurable-unpleasurable dimension mark the same thing, with more potent forms of pleasure and more potent forms of displeasure. 6. We intentionally use the word “distaste” for in this view, emotions, like the sense of taste or smell, have a simple attractive or avoidance function. 7. While most researchers in this school acknowledge that at least two dimensions are required to account for the response to all stimuli, it is also often the case that researchers view collecting responses on only one dimension (typically happy-sad rating) as sufficient (Barrett and Russell 1998; Russell and Barrett 1999). 8. Watson has presented the most comprehensive review of which emotion terms are best suited to this end (Watson 1988b, 1994, 1997). 9. In 1995, a Washington Post/ABC news poll included measures of positive affect (enthusiasm) and negative affect (anxiety) and found that these two measures were positively correlated for a group of relatively new individuals, e.g., Newt Gingrich (⫹.18), Sen. Phil Gramm (⫹.29), and Gen. Colin Powell (⫹.41). See also, Zevon and Tellegen 1982. 10. And such examples of failure to control for alternative influences are the case not only in psychology but also in studies of emotion in political science (Rahn 2000). 11. The obvious advantage is we obtain more levels of an emotion by shifting from yes or no (two response options) to frequency (normally four or five levels, ranging from “all the time” down to “never”). Alternatively one can ask for the intensity of emotion (again typically securing four or five levels). As noted earlier, the work of David Watson shows that either intensity or frequency can be used to good effect. 12. In this and other reports of factor scores, we suppress values less than .30 to focus on the key loadings. Full analyses are available from the authors. 13. We combine data from two sessions of the same experimental study, conducted in spring 2001 and spring 2002. In the first session, 120 undergraduates participated, while 93 undergraduates participated in the second session.
MEASURE AND MISMEASURE OF EMOTION
45
14. An alternative formulation is offered by Tellegen and colleagues (Tellegen, Watson, and Clark 1999a, 1999b) conceive of negative affects as hierarchically organized. Such a concept is a viable alternative and there is considerable neuroscientific evidence that lends support to that conception (Gray 2004). Notwithstanding that alternative conception, our measurement recommendation stands to treat anxiety and aversion as distinct factors of emotion that warrant measurement. 15. In politics, stimuli that provoke aversion are likely to be far more frequent than in most social venues. Greene’s work on the affective component of partisanship takes the unusual approach of including measures for positive affect and for aversion, but none for anxiety (Greene 2002). This leaves it unclear whether his data is measuring anxiety (as would be the result of including aversion measures if no subjects actually experience aversion) or aversion, or both. Hence, the nature of the affective aspects of partisanship remain more obscure than they otherwise would have been had recommendation 2 been followed. 16. In addition to developing the PANAS measurement instrument (Positive Affect and Negative Affect Scale), he has published considerable work on response formats, item-item reliability and temporal stability (Watson 1988a, 1988b; Watson and Clark 1991, 1994, 1997). 17. Though this remains open to other interpretations, such results suggest that intensity might be the better response phrasing because the equivalent results suggest that subjects do not accurately recall how often they experienced a specific emotion, but rather rely on their current assessment.
This page intentionally left blank
CHAPTER 4 CONTRIBUTIONS OF A SOCIOLOGICAL PERSPECTIVE ON AFFECT TO THE STUDY OF POLITICAL ACTION Lisa Troyer and Dawn T. Robinson
ffect, the generalized feelings of actors in situations, has long been recognized by sociologists as both fundamental to and emergent from social life (for reviews of sociological work on affect and emotion, see Gordon 1990, and Smith-Lovin 1995). Four themes characterize contemporary sociological theorizing on affect: identity, control, action, and meaning. Goffman (1963, 1959) suggested that successfully enacting a social identity requires managing affective responses in a manner consistent with others’ expectations for that identity. For example, enacting the identity of a friend might involve physical gestures that convey feelings, such as smiling and prolonged eye contact, which facilitate defining both the social situation and the identities of the interactants within it (e.g., a meeting of friends). Once a situation and its interactants are defined the interaction can proceed. Likewise, Parsons and Shills (1962) argued that in any social encounter, individuals must assess both the nature and degree of affect that should be displayed, and that this depends upon their relationships to one another. Hochschild (1979) carries the notion of the norms of affect further through the explicit concept of “feeling rules,” the idea that the emotions felt by occupants of particular positions are subject to fine-grained norms. Thus, for example, an actor who claims the identity of a politician is expected to feel (and exhibit) compassion and caring for his constituents, but not romantic passion for them. Note, however, that the same actor in
A
48
TROYER AND ROBINSON
a different context may lay claim to the identity of a paramour, making displays of romantic passion toward a partner appropriate. As this example suggests, an actor may claim different identities in different contexts and an actor’s behavior signals underlying feelings that may be more or less appropriate to the identity the actor is claiming within a context. An important insight of this work is that feeling rules, social identities, action, and social contexts are interdependent and dynamic, and that consonance across the dimensions and situations enables social action. Microsociological perspectives on affect tend to view violations of feeling rules as problematic for social interaction. The meaning of situations and our sense of who we are and who others are (i.e., the identities of those in the situations) are suddenly called into question when feeling rules are violated (e.g., Goffman 1959, Kemper 1991). The sudden decay of meaning can paralyze social interaction. Consider, for example, the discovery of a politician’s romantic passion for constituents. Upon such a discovery, political action may be waylaid while observers seek to make sense of the untoward situation. Consequently, these perspectives suggest that when such events occur, actors often seek reparative actions to bring the situation and identities back into alignment. Reparative actions can range from feigning ignorance of the violation (as in historical cases when the press avoided publishing the reports) to expressing shame or embarrassment— affective states that acknowledge and take responsibility for the violation (e.g., when a politician issues a public apology for an indiscretion), to redefinition of the action, actors, and/or the situation (such as when the indiscretion results in the surrender of the political office, which essentially strips the actor of the political identity). These insights on the meaning of social situations and the key role that affect plays in determining meaning represent important breakthroughs in sociological understandings of social life. Yet, for years, they took the form of discursive descriptions of social action and lacked an underlying calculus that is critical to predictive theories. The lack of predictive utility, in turn, limits the value of the theoretical ideas for analyzing domains of social life, such as political life. Recently, however, the development of affect control theory (Heise 1977, 1987) has provided the needed rigor and calculus to generate a formal model of identity, control, action, and meaning with a high degree of predictive utility. We turn now to an overview of affect control theory. Affect Control Theory Three assumptions underlie affect control theory: (1) individuals seek to enact behaviors that generate feelings appropriate to the situation; (2) if individuals cannot maintain situation-appropriate feelings through behavior,
A MICROSOCIOLOGICAL PERSPECTIVE
49
then they adjust their perceptions of the situation; and (3) affective responses signal the relationship between their experiences and their definitions of the situation. The theory portrays individuals as engaged in an ongoing assessment of the meaning of the social interactions in which they are embedded and as agents who seek to maintain transituational consistency in the meanings that interactions evoke for them. Thus, the affective meanings individuals experience serve as “signals” to individuals that determine their subsequent perceptions and actions to facilitate consistency in their experiences as social interaction evolves. Affect control theory conceptualizes social events in terms of their constituent elements: (1) actors who assume identities in a situation (e.g., police officer, criminal, citizen); (2) the behaviors of the actors (e.g., arrest, steal, ask for help); (3) the object-actors to whom the actions are directed, which are also identities (e.g., police officer, criminal, citizen); and (4) the setting in which the event takes place (e.g., home, store, courtroom). These elements correspond to the linguistic structure through which individuals both describe and understand social situations. For instance, one might view a situation in which one person roughly strips an item from another’s arms after the person has exchanged money for the item within a building containing a variety of items and describe it as, “Criminal steals from citizen in a store.” Further, the observer might see the other approach another in a blue uniform and cap inside the building pleading for help and describe the observation as, “Citizen asks for help from police officer in store.” This case-grammar structure of events takes the form of Actor-BehaviorObject-Setting, or ABOS, in affect control theory terms. Each element of the ABOS structure has meaning that is independent of the actual event. Drawing on psychometric work by Osgood and colleagues (e.g., Osgood, Suci, and Tannenbaum 1957), Heise (1987) proposes that the meaning of each ABOS element is defined in three dimensions: Evaluation, Potency, and Activity (EPA). Evaluation reflects contrasts such as good versus bad. Potency captures meaning in terms of contrasts like powerful versus powerless. Activity involves contrasts like noisy versus quiet. Each dimension (Evaluation, Potency, Activity) is measured on a scale ranging from ⫺4.5 to ⫹4.5, with zero reflecting a neutral perception of the element on the dimension. Using evaluation as an example, a value of ⫹1.0 represents an element that is slightly good, ⫹2.0 one that is quite good, ⫹3.0 is extremely good, and ⫹4.5 reflects something infinitely good. On the negative side of the scale, ⫺1.0 reflects something that is slightly bad, ⫺2.0 something that is quite bad, ⫺3.0 is extremely bad, and ⫺4.5 represents something infinitely bad. Again, affect control theory proposes that shared cultural attitudes toward all elements of an ABOS event can be captured (independently of
50
TROYER AND ROBINSON
an event in which they are embedded) by using average ratings from culture members in the three-dimensional EPA meaning space. Thus, using survey methods, researchers have found that in the United States the concept “mother” (which may be either an actor or an object) has an EPA rating of ⫹2.3 (Evaluation), ⫹1.9 (Potency), 0.0 (Activity).1 This suggests that in our culture “mother” means something that is quite good, quite powerful, and neither noisy nor quiet. In contrast, “child,” has EPA ratings of ⫹1.9, ⫺1.1, ⫹2.5. These EPA values suggest an identity that is quite good, slightly powerless, and very lively. The action of “whipping” has an EPA rating of ⫺2.2, ⫹0.4, ⫹0.9: An act that is quite bad, very slightly powerful, and slightly lively. These independent EPA ratings of each element are referred to as fundamental sentiments. The fundamental sentiments we have toward an actor, behavior, object, or setting reflect our culturally shared affective associations with that element, and are quantified in the EPA dimensions. The independent elements of mother, child, and whipping may come together in an observable event, “Mother Whips Child.”2 The event elicits feelings in observers toward the actor (mother, in this example), action (whipping), and object of the action (child). For instance, the observation of a mother hugging a child generates positive feelings toward the mother, while the observation of a mother whipping a child generates negative feelings toward the mother. The feelings reflect combinatory meanings of the actor who takes on the action. Since whipping is quite bad and noisy, it elicits a negative impression of the actor engaging in the action. This negativity is exacerbated by the fact that it is being inflicted upon a nice, weak person (child). Following the work of Gollob (1968), Heise and colleagues (Heise 1969, 1979; Smith-Lovin 1987) developed mathematical models (i.e., impression-formation equations) to predict the feelings toward an element that an event evokes.3 These models mathematically combine the independent EPA ratings of elements comprising an event to generate revised EPA ratings for each element embedded in the event. Thus, for example, the impression-formation equations predict that the EPA ratings for “mother” in the event “Mother Whips Child” become ⫺1.2, ⫹1.3, ⫹0.5. That is, the meaning attached to the mother who whips a child shifts to somewhat bad, fairly powerful, and slightly lively. These feelings are referred to as “transient impressions” (“transient” because the next event that occurs will alter those feelings). Transient impressions of the child also result from the “whipping event.” The EPA ratings for the child will shift to ⫹0.8, ⫺1.5, 2.0 (slightly good, quite weak, and quite lively). The interpretation corresponding to this shift is that the object (child) of a bad action (whipping) loses goodness (i.e., we tend to attribute less goodness to targets of bad
A MICROSOCIOLOGICAL PERSPECTIVE
51
actions and hence deserving of the bad action). Also, the powerlessness of the object of a powerful action increases. Liveliness declines because recipients of activity seem themselves less active. The impression-formation equations afford researchers predictive power: The theory predicts how meanings assigned to people in social interactions shift as interaction unfolds. For each event, there is a fundamental sentiment associated with actors (the initial independent EPA ratings assigned to each) as well as a transient impression that results from the event. The sum of the squared differences between fundamental sentiments and transient impressions operationalize “deflection” in affect control theory. Conceptually, deflection corresponds to the amount of affective disruption produced by an event. This quantification of deflection permits researchers to predict both the feelings that participants and observers are likely to experience as the result of the current event as well as likely future behaviors. With respect to the feelings that result from experiencing events, Heise and MacKinnon (1987) have shown that deflection corresponds to how likely individuals perceive an event to be. A deflection value of 8.0 or less is within the range of normalcy. That is, an observer of or participant in such an event would feel as though it was common and expected. Events generating deflection values of 9.0–16.0 feel unusual (the deflection for the event “Mother Whips Child” is about 15.0); those with deflection values of 17.0–24.0 feel extraordinary. Experiencing an event with a deflection of 25.0–44.0 yields feelings of incredibility, and those involving deflections of 45.0 and higher feel so unbelievable that they elicit feelings commensurate with a supernatural experience. Thus, deflection values of 9.0 or higher correspond to the level of salience reflecting an uneasiness or anxiety that may motivate the kind of conscious introspection, information searching, and/or deliberative processing examined by other contributors to this volume (e.g., Crigler, Just, and Belt; Lodge and Taber; Marcus et al.; Redlawsk). As such, affect control theory provides systematic tools for predicting when the threshold of salience that motivates deliberative processing of events is reached. The capacity of affect control theory to predict future events arises from three factors—the control model, an empirical corpus of cultural meanings, and the theoretical equations. First, as noted earlier, the theory assumes that actors seek stability in meanings and will take actions to lend such stability to their social experiences. Thus, individuals seek actions that minimize the deflection between fundamental sentiments and transient impressions resulting from events. When it is not possible to effectively minimize deflection—as in when the deflection is so extreme that there are no appropriate actions available, or when situational constraints limit actions— actors may redefine elements of events (i.e., the actors, behaviors, and
52
TROYER AND ROBINSON
objects) in a manner consistent with their fundamental sentiments about the elements. Second, nearly three decades of research by affect control theorists has yielded a set of “cultural dictionaries” of EPA ratings for thousands of actor-identities, behaviors, object-identities, and settings. Third, the impression-formation equations described earlier capture important information about how our affect toward particular identities and behaviors change as a result of combination in specific events. Together, these tools allow researchers to predict both how actors respond emotionally in culturally situated social interactions (which are experienced as ABOS events), and the subsequent reactions to events that are most likely. The most likely reactions (and hence future events) are those that most reduce deflection arising from the event. As this suggests, social interaction is itself a dynamic process and affect is a mediator—it is the appraisal of one event that determines subsequent events. The predictive capacity of affect control theory to model the dynamic unfolding of events is perhaps best illustrated in events that seem highly unlikely. Consider, for example, a trial involving an aggressive attorney in which the interaction is experienced by observers as “Attorney Terrorizes Victim.” This generates a deflection 9.0, representing an unusual event. For the attorney to restore meaning to the situation and retain the identity of an attorney, reparative actions are required. Affect control theory predicts that the next behavior on the part of the attorney that would maximally reduce deflection is soothing the victim (which reduces deflection to 2.0, in the range of normalcy). Such predictions of affect control theory are amenable to testing in different research settings (including experimental and field settings). As this example illustrates, one way that actors maintain fundamental meanings is through engaging in reparative actions when events create impressions that are misaligned with our expectations. When institutional structures constrain reparative action (as when face-to-face interaction with a senator is not possible) or when deflection is extreme, a secondary way to restore meaning is through redefinition. Consider, for instance, the event “Judge Tortures Child.” Analysis of this event in affect control theory yields a deflection of 13.0—a fairly unusual event. Unusual events violate expectations, generating feelings of uncertainty regarding responses and forthcoming events. For a participant and/or observer to continue in the interaction, some sense of what has occurred (or will occur next) is needed. Affect control theory can determine how the judge might be redefined to lend greater certainty to the situation. For example, affect control theory predicts that the observer might decide that the perpetrator of the torture cannot possibly be a judge (since torturing behavior is so unusual for someone with the identity of a judge). The observer might redefine the judge as
A MICROSOCIOLOGICAL PERSPECTIVE
53
an “ogre” (which reduces the deflection to 1.0). Note that the initial identity attributed to an actor serves as a powerful heuristic, constraining interpretation of the actor’s behaviors and reactions to the deflection. This is akin to the role of categorization (e.g., a candidate’s partisanship) as a heuristic, which Isbell et al. (in this volume) describe as biasing subsequent judgments of candidates. Affect control theory adds value to such insights by specifically identifying the likely interpretations that the heuristic will engender in a particular situation. As a normative model of affective dynamics in social life, affect control theory quantifies and systematically represents the cultural knowledge that political professionals and political observers already take for granted. Thus, for example, while it was no accident that the Willie Horton/ revolving door attack ads were incredibly damaging for Dukakis during the 1988 presidential campaign, affect control theory could have predicted as much by modeling how someone with the goodness, power, and liveliness of a governor could be transformed into someone with the goodness, power, and liveliness of an actor who would Free a Murderer. More importantly, it could predict the reparative actions a governor might take in such an event to avoid becoming redefined as someone who would Free a Murderer. Yet the use of affect control theory is currently limited in the domains of social life to which it has been applied. Of particular interest to us is the theory’s lack of attention to political life. In the final section of this chapter, we briefly outline a research agenda to remedy this gap in the development and application of affect control theory and suggest how filling this gap may advance theory and research in political psychology. An Affect Control Theory Research Agenda for the Study of Political Life The limited social domains of affect control theory are reflected in the dictionaries of EPA ratings for actors, behaviors, objects, and settings, which have been developed in response to researchers’ interests. Thus, for example, the dictionaries contain a range of actors, behaviors, and objects that correspond to deviant and criminal situations, an early area of interest to Heise (1968), the theory’s founder. Interestingly, there are relatively few terms corresponding to political domains of social life. Researchers have yet to collect EPA ratings for such common political identities as politician, voter, protestor; and behaviors such as campaigning, voting, and demonstrating. Because of this, the utility of affect control theory for the study of political life is currently limited. This limitation, however, represents an opportunity for both affect control theory and political psychologists.
54
TROYER AND ROBINSON
An affect control research agenda for the study of political life starts with developing an inventory of political identities, behaviors, and settings, for which EPA ratings could be collected. In the past, affect control theorists have relied heavily on samples of undergraduates as sources of EPA ratings, making the acquisition of the EPA ratings efficient.4 Additionally, researchers would ideally collect in-context EPA ratings for the items in an event setting. This requires the construction of a second inventory in which the political identities, behaviors, and settings are combined to represent events (e.g., “Protestor Demonstrates Against Politician,” “Lobbyist Bribes Politician”). After in-context EPA ratings have been acquired for these events, impression-formation equations can be estimated.5 With the dictionary of EPA ratings for political events compiled and the impression-formation equations estimated, the toolkit for generating predictive claims on political dynamics would be complete.6 We believe that the application of affect control theory to political life would be useful for addressing a number of questions relevant to political psychologists. We inventory a few questions that political psychologists may find relevant: ●
●
●
●
●
How does the presentation of political events affect perceivers’ feelings toward the institutions and actors involved in the events? How do different campaign advertisements (e.g., portraying different kinds of behaviors by candidates) affect voters’ feelings toward political candidates? What kinds of behaviors by voters toward candidates are those feelings most likely to elicit? What kinds of behaviors on the part of political figures are most likely to restore trust and confidence in the figure among constituents following an event that undermines the trust and confidence? What kinds of protest behaviors are most likely to evoke empathy and support versus alienation and anger on the part of observers of the behaviors?
By expanding affect control theory to incorporate more political identities, behaviors, and settings, it is possible to generate predictive claims that would address these questions. These claims could then be tested through experimental, archival, and/or field research. Through this process of theoretical refinement and testing greater insight on the microdynamics of political life will become available, substantially advancing the field of political psychology.
A MICROSOCIOLOGICAL PERSPECTIVE
55
Conclusion The systematic development of affect control theory over recent decades has provided sociologists with both a theoretical framework and methodological toolkit for investigating social action and its link to affect. Yet, the identities, actions, and situations that are currently subsumed by the theory do not adequately represent the political domain of social life. This gap defines an important opportunity to both expand the theory’s research agenda and to enhance our understanding of political life. Furthermore, it is an opportunity that may be best realized through collaborations between sociologists and political psychologists. Clearly researchers in both areas are inherently multidisciplinary in their theoretical ideas and methods. Their multidisciplinary experiences are likely to be an asset to forging successful collaborations that generate new insights for political science, psychology, and sociology. Notes 1. The EPA ratings and other quantitative values we report were drawn from the 1978 U.S. female dictionary available in Java Interact 2 (Heise 2001), at: http://www.indiana.edu/~socpsy/ACT⬘/. 2. For the sake of parsimony, we omit the “S” or “Setting” component of the ABOS case-grammar in this and subsequent examples we present. 3. See Smith-Lovin (1987) for a thorough discussion of the iterative regression methods that generated the impression-formation equations. Briefly, E, P, and A ratings for thousands of A, B, O, and S elements (out-of-context ratings) and E, P, and A ratings of each ABOS element of events (i.e., incontext ratings) have been empirically generated through nearly three decades of survey research. Each of these ratings are regressed on the independent E, P, and A ratings of A, B, O, and S elements out of context. Nonsignificant terms are dropped from the analyses, and the regressions are repeated until only significant terms remain. 4. Critics of research that relies on undergraduate participation argue that the generalizability of the research is limited because the data generating the results reflect unique properties of college students. For affect control theory, however, the acquisition of EPA ratings from college students may be ideal, since the theory proposes that meaning (as derived from EPA ratings) is culturally shared. Thus, it is important to acquire the ratings from cultural experts. By many accounts, college students represent cultural experts. 5. Existing impression-formation equations could be applied to a dictionary of political elements. Yet, these equations were generated from dictionaries and events that lacked representation of political experience. Assessing whether the mathematical models for impression formation are the same for events in the political domain as they are for events in other domains would itself be a
56
TROYER AND ROBINSON
contribution to the affect control theory research program. Moreover, it would ensure more accurate predictive power for political experiences, and as such, we consider this an important step in the research agenda. 6. Current developments are facilitating such elaborations of affect control theory. Recently, Heise launched the Project Magellan (Heise 2001) with one goal of building the content of the dictionaries. Project Magellan allows researchers to easily customize EPA surveys and administer them via the Web.
CHAPTER 5 AFFECT AND POLITICS: EFFECTS ON JUDGMENT, PROCESSING, AND INFORMATION SEEKING* Linda M. Isbell, Victor C. Ottati, and Kathleen C. Burns
olitical figures and events often elicit strong emotional responses in citizens. These responses have the power to impact judgments and information processing, as well as the types of information that individuals seek out. Recent examples of political events that have elicited strong emotional reactions are easily accessible. The fiasco in Florida during the presidential election of 2000 led many voters to experience anger at the outcome of the election and disgust at the process whereby it was decided. The terrorist attacks on the United States on September 11, 2001, led citizens to experience a collective sense of fear and anxiety, along with sadness for the loss of life and anger at Osama bin Laden for masterminding the attacks. Along with these negative emotions was a sense of enthusiastic patriotism in the United States. Positive affective reactions, however, tend to be more general than negative reactions. That is, while positive reactions may be experienced as general positivity, negative feelings are typically more differentiated and may be experienced, for example, as fear, anger, sadness, disgust, or guilt (e.g., Averill 1980; Ellsworth and Smith 1988). This chapter examines the influence of affect on judgments, information processing, and information seeking. In contrast to our prior treatments of
P
* The preparation of this chapter was supported in part by a grant from the National Science Foundation (SBR 0132254) awarded to the first author.
58
ISBELL, OTTATI, AND BURNS
the role of affect in politics (e.g., Isbell and Ottati 2002), we move beyond valence and consider the effects of specific types of negative affect (e.g., anger, fear) on judgment and information processing style. We also examine the extent to which affective cues (e.g., happy versus serious facial displays) influence information processing without triggering an affective state in the perceiver, and consider whether these cues need to be consciously perceived to be effective. Finally, we extend the role of affect in information processing to a new domain: information seeking. In investigating these issues, we rely heavily on research and theory in social psychology and draw implications for political information processing. We also point to gaps in the literature and make suggestions for future directions. Effects of Affective States on Political Judgment: Assimilation and Contrast Effects Political candidates frequently use feel-good campaign techniques involving positive affect-eliciting stimuli (e.g., beautiful scenery, the American flag). Televised political commercials are designed by high-priced advertising firms that specialize in producing visual imagery that arouses feelings of warmth, excitement, and joy. Political strategists operate under the assumption that these feel-good campaign tactics will translate into more favorable attitudes toward the candidate. That is, strategists assume that positive feelings will produce an assimilation effect on attitudes toward the candidate (see Thorson, Christ, and Caywood 1991a, 1991b). An assimilation effect emerges when a positive affective state elicits a positive evaluation of an object, whereas a negative affective state elicits a negative evaluation. Assimilation effects have been obtained in numerous political judgment tasks, including those that require participants to form an impression of a political candidate (Ottati and Wyer 1993). However, affective states do not invariably produce assimilation effects on political judgment. In some cases, affective states produce a contrast effect (e.g., Ottati et al. 1989; Ottati and Isbell 1996) wherein positive affective states elicit a relatively negative evaluation of an object and negative affective states elicit a relatively positive attitude. An obvious example of the assimilation effect occurs when emotional reactions to a political candidate influence voters’ global attitudes toward a candidate. Correlational research suggests that positive emotional reactions to a political leader (e.g., “George W. Bush makes me feel proud”) are associated with positive attitudes toward a leader, whereas negative emotional reactions are associated with more negative attitudes (Abelson et al. 1982; Ottati 1997; Ottati, Steenbergen, and Riggle 1992; Ragsdale 1991). Yet, it is unclear whether emotional reactions determine attitudes toward a candidate independent of cognitive reactions to the candidate. Although
AFFECT AND POLITICS
59
some studies (e.g., Abelson et al. 1982; Ottati, Steenbergen, and Riggle 1992; Ragsdale 1991) support this view, Ottati (1997) found that cognitive and affective reactions served as redundant predictors when he specifically assessed “emotionally relevant” cognitions. Ottati (1997, 2001) has emphasized, however, that this later finding does not preclude the possibility that emotion directly influences global attitudes toward a candidate. This is because a direct effect of emotion on global attitudes may be justified or rationalized by “emotionally relevant” cognitions that have nothing to do with the original emotion-eliciting process. These cognitive justifications might fully account for the statistical relation between emotions and global attitudes, even if emotions directly influence attitude judgments. To more clearly isolate the influence of affect on political judgment, researchers have manipulated the perceiver’s affective state. When experimentally manipulating mood, the perceiver’s affective state is elicited by a contextual cue (e.g., happy or sad music or movie clips), not a feature of the object to be judged. This ensures that the affective state is not elicited by the individual’s prior cognitive appraisal of the object, and thereby enables the researcher to more clearly isolate the independent influence of affect on judgments of the object. This strategy also mimics real-world conditions in which contextual cues (e.g., balloons, American flag) are designed to trigger positive affective reactions to a political leader. In these experiments, participants complete two ostensibly unrelated tasks: the mood induction procedure and a subsequent judgment task (e.g., Ottati and Isbell 1996). After the mood induction, participants receive information about a political object (e.g., a political candidate) and then report their attitude toward that object. An assimilation effect on attitude judgments occurs when attitudes toward the object are more favorable in the positive mood condition than the negative mood condition. Psychological Mediators of Mood-Induced Assimilation Effects A variety of psychological mechanisms potentially mediate mood-induced assimilation effects. These include mood-congruent selective encoding, selective retrieval, interpretation, and elaboration, as well as mood misattribution. Of these, only the mood misattribution process involves a relatively direct influence of affect on political judgment (Clore 1992; Niedenthal 1990; Schwarz 1990). The other process mechanisms involve more extensive cognitive activity. In these cases, mood functions as a prime that activates material in memory that is congruent with the affective state (Bower 1981, 1983, 1991; Branscombe 1988; Singer and Salovey 1988). Once activated, these mood-congruent concepts guide the encoding, interpretation, elaboration, and retrieval of information pertaining to the object.
60
ISBELL, OTTATI, AND BURNS
Mood-Congruent Selective Encoding Bower and his colleagues (Bower 1981; Bower, Gilligan, and Monteiro 1981; see also Bower and Forgas 2000; Forgas and Bower 1987; Forgas 1992) were the first to document selective encoding of mood-congruent information. Happy or sad moods were induced in participants by hypnotic suggestion in order to examine how mood would impact learning and memory of material contained in a written text. Results indicated that participants recalled more facts about mood-congruent incidents in the text than mood-incongruent incidents. This effect emerged when mood was manipulated prior to presenting the written text, but not when mood was manipulated afterward, suggesting that this effect was the result of selective encoding when individuals originally received the information. According to Bower and his colleagues (Bower 1981; Bower, Gilligan, and Monteiro 1981; see also Bower and Forgas 2000, 2001), the mood state served to activate mood-congruent concepts in memory, which in turn guided the encoding process. Although yet to be documented in the political domain, it is possible that an analogous effect would emerge if one were to substitute a political text for the text used in this earlier research. That is, individuals in a positive mood might be more likely to encode positively valenced political events whereas individuals in a negative mood might be more likely to encode negatively valenced political events. An effect of this nature may have occurred as a result of the 9–11 tragedy. The selective encoding hypothesis suggests that the outrage and sorrow stemming from this event might have led the American public to selectively encode more negative information contained in other news stories that followed the 9–11 incident. As a consequence, this process may have produced pessimistic views of political developments in a variety of domains (e.g., the economy, unemployment). Mood-Congruent Selective Retrieval Mood also elicits selective processing when individuals retrieve information for purposes of reporting a judgment (Isen et al. 1978; Isen 1984; see also Bower 1991; Bower and Forgas 2000; Erber 1991; Forgas and Bower 1988a, 1988b; Lloyd and Lishman 1975; Mayer, Gayle, and Harman 1990; Teasdale and Russel 1983). In one study, Isen and her colleagues manipulated mood by giving people a free gift at a shopping mall (Isen et al. 1978). Then, as part of an ostensibly unrelated task, these individuals rated the performance and overall service record of their automobile and television set. Compared to individuals who had not received a gift, these individuals rated the items more positively. Isen et al. (1978) proposed that individuals in a good mood were more likely to selectively retrieve positive
AFFECT AND POLITICS
61
information pertaining to the product items, and therefore rated them more positively. In this research, mood was not manipulated until after the product information had been encoded. This finding suggests a slightly different prediction associated with the September 11 tragedy. Specifically, negative affect precipitated by the tragedy might have led the public to recall more negative events that preceded 9-11 than would otherwise have been the case. Note that this selective retrieval effect should produce selective recall of negative events that preceded the 9-11 tragedy, whereas the previously described selective encoding effect should produce selective recall of negative events that followed the 9-11 tragedy. Selective retrieval effects could have important political consequences. For example, following 9-11, political analysts who reconstruct and evaluate past political events may have been likely to arrive at a skeptical and negative construal of earlier events. Mood-Congruent Interpretation By definition, ambiguous information pertaining to a target person (e.g., “Donald is well aware of his ability to succeed”) can be interpreted in terms of more than one concept (e.g., “confident”, “conceited”). The concept that is most accessible is most likely to be used as a basis for interpretation (Higgins, Bargh, and Lombardi 1985; Wyer and Srull 1989). As a consequence, priming one of these concepts can guide the interpretation of subsequently presented ambiguous behavior (Higgins, Bargh, and Lombardi 1985; Wyer and Srull 1989). Mood can function as a prime that increases the accessibility of mood-congruent concepts that, in turn, affect the interpretation of ambiguous information (Bower and Forgas 2000; Clark and Waddell 1983; Bower 1981, 1991; Schiffenbauer 1974). This effect may frequently emerge when citizens interpret political statements. Ambiguous statements enable politicians to navigate through perilous waters without alienating large segments of the voting public (Page 1976; Shepsle 1972). Thus, it appears that the average American citizen is provided with a great deal of ambiguity upon which to project the most accessible interpretation. Effects of this nature might be triggered by affective states that are elicited for highly personal or idiosyncratic reasons. On the other hand, these effects might also be elicited by relatively impersonal events that influence the affective state of the entire nation. For example, negative affect elicited by a severe downturn in the national economy might produce negative interpretations of other political events that are unrelated to the economy. Although effects of this nature have not been adequately tested, the implication is that negativity in one domain of politics might trigger a negative construal of a completely unrelated political domain.
62
ISBELL, OTTATI, AND BURNS
Mood-Congruent Elaboration Related to the mood-congruent interpretation effect is the mood-congruent elaboration effect. This effect has been documented in persuasion research in which individuals generate cognitive responses to a persuasive communication. Specifically, some research has demonstrated that individuals in a happy mood generate more pro-arguments, whereas individuals in a sad mood generate more counterarguments (Petty et al. 1993; Wegener, Petty, and Klein 1994). The result is greater persuasion in the direction advocated in happy mood conditions than in sad mood conditions. Political speakers might be well advised to scrutinize this phenomenon carefully. It is conceivable that an appropriately timed joke (which elicits a positive mood state) can serve to “soften up” an otherwise critical audience and allow a political speaker to be more convincing or persuasive. Mood-congruent encoding, retrieval, interpretation, and elaboration effects result in a mood-congruent impression of the political object. These processes might operate in combination to produce an assimilation effect on judgment of the object. To the degree that this is the case, one can say that processing of specific, substantive information pertaining to the object is “infused” with mood-congruent cognitive biases (Bower and Forgas 2000, 2001). This “affect infusion” effect cannot emerge unless the individual actually comprehends, interprets, elaborates upon, or retrieves substantive information en route to deriving a judgment. Thus, some have suggested that these effects are most likely to occur when capacity and motivation to scrutinize substantive information are high (Bower and Forgas 2000, 2001; Forgas 1995; Forgas, Ciarrochi, and Moylan 2000). Mood Misattribution Another process that can produce an assimilation effect on judgment is mood misattribution. According to the affect-as-information (AAI) model (Schwarz and Clore 1983; see Schwarz and Clore 1996; Wyer, Clore, and Isbell 1999 for reviews), individuals are not always fully aware of the actual source of their affective states. As a consequence, affect elicited by a contextual cue can be misattributed to an object and produce an assimilation effect on judgments of that object. In a test of this model, Schwarz and Clore (1983) telephoned individuals and asked them to report their life satisfaction. As expected, participants who were contacted on sunny days reported higher life satisfaction than those contacted on rainy days. However, this effect was eliminated among individuals who were reminded of the actual source of their mood by asking them to report their local weather conditions prior to rating their life satisfaction. Thus, the critical mediator of this effect appears to be the misattribution of irrelevant affect to the object being judged (see Clore and Parrot 1991, 1994; Gasper
AFFECT AND POLITICS
63
and Clore 1998; Keltner, Locke, and Audrain 1993; Schwarz and Clore 1988; Schwarz, Bless, and Bohner 1991 for similar evidence). Unlike the processes described earlier, this simple misattribution process presumably requires minimal cognitive capacity or motivation (see Clore et al. 2001). In accordance with this conceptualization, political research has demonstrated that mood is most likely to influence political judgments when cognitive capacity and motivation are low. For example, Ottati and Isbell (1996) found that individuals who lack the capacity to carefully process political information evaluated a political candidate more favorably following a happy mood manipulation than a sad one. Similarly, Isbell and Wyer (1999) found that individuals who lack the motivation to carefully process political information misattributed their affect when judging a political candidate. Psychological Mediators of the Contrast Effect Contrast effects occur when positive affective states elicit a negative evaluation of an object and negative affective states elicit a positive evaluation (e.g., Ottati et al. 1989; Ottati and Isbell 1996; see Martin 1986; Martin, Seta, and Crelia 1990; for related evidence outside the political domain). There are at least two psychological processes that might account for this reversal of the assimilation effect: overcorrection and standard of comparison. Overcorrection Political experts have been found to engage in overcorrection when trying to eliminate mood-induced biases in their judgments of a political candidate (Ottati and Isbell 1996; Isbell and Wyer 1999). As a consequence, Ottati and Isbell found that experts rated a candidate more negatively when in a happy mood than when in a sad mood. This effect presumably emerges because individuals are not sensitive to the subtlety of the original assimilation tendency and therefore “overdo it” when trying to correct for this tendency (see Petty and Wegener 1993). Ottati and Isbell (1996) argue that this correction process requires awareness that the original assimilation tendency exists, and the capacity to correct for it (see Martin 1986; Martin, Seta, and Crelia 1990; for related evidence outside the political domain). Political experts, who are assumed to process information about a candidate in a relatively efficient manner, are more likely to possess the resources needed to engage in this correction process. Isbell and Wyer (1999) have demonstrated that this correction process also requires that the individual be motivated to correct for the biasing influence of mood. The findings reported by Ottati and Isbell (1996; see also Isbell and Wyer 1999) raise a sobering question for political strategists who advocate
64
ISBELL, OTTATI, AND BURNS
that politicians rely on feel-good campaign tactics. Voters high in political expertise may recognize that such techniques can produce an unrealistically positive evaluation of a candidate. In attempting to arrive at a more “accurate” evaluation of the candidate, these individuals might overcorrect for this initial bias, and evaluate the politician more negatively than would otherwise be the case. Standard of Comparison Standard of comparison effects may also be responsible for contrast effects. Anyone who has returned to a bitter cold winter after spending a week vacationing in the Bahamas has experienced a standard of comparison effect. In comparison to the warmth of tropical beaches, winter wind chills can feel ferocious and even cruel. In this instance, the standard of comparison effect produces a more negative reaction to the cold weather than would otherwise be the case. Affective states are most likely to elicit a standard of comparison effect when the affect-eliciting cue is thematically similar to or falls within the same category as the object being judged (Abele and Gendolla 1999; Schwarz et al. 1987). For example, sitting in an unpleasant room might elicit a standard of comparison effect when participants subsequently rate their satisfaction with their housing, but not when they rate their life satisfaction (Schwarz et al. 1987). On the other hand, some evidence suggests that an affective state can serve as a standard of comparison for a subsequent affective state, even when the two states are elicited by two unrelated events (Berkowitz 1993; Clore, Schwartz, and Conway 1994; Erber and Erber 1994; Parrott and Sabini 1990; Sedikides 1994). Thus, one news event might serve as a standard of comparison when individuals emotionally respond to an unrelated, subsequent news event. For example, a news story that emphasizes the suffering of diseased or malnourished children might seem especially tragic when preceded by a news story that covers Mardi Gras in New Orleans. Given that the news media select the order to present their stories, they have significant power to influence the impressions and attitudes that viewers form of different political events and issues. Assimilation and Contrast for Specific Affective States Up to this point, our discussion of affect and political judgment has made a general distinction between positive and negative affect without considering more fine-grained distinctions in human emotional experience. Although both positive and negative affect can be differentiated into distinct subtypes (see Ortony, Clore, and Collins 1988; Clore and Isbell 2001), the experience of different types of negative affect (e.g., sadness, anxiety or fear, anger,
AFFECT AND POLITICS
65
disgust) may be more highly differentiated (e.g., Averill 1980; Ellsworth and Smith 1988). The following discussion considers the effects of more specific affective states on political judgment. Happy and Sad Affective States A discussion that specifically focuses on happy and sad affective states is largely redundant with the previously cited work. This is because most research regarding the impact of “positive” and “negative” mood states uses a “happy” versus “sad” mood manipulation. Thus, when comparing happy individuals to sad individuals, there is evidence that happy individuals are more likely to encode and/or retrieve positive information, interpret ambiguous information in a positive manner, generate positive cognitive elaborations, and mistakenly believe that a political object has elicited positive affect when in fact it was elicited by a contextual cue. Similar processes occur among sad individuals for negative material. These effects are observed not only when using an experimental manipulation of mood, but also when using an individual difference approach that compares mildly depressed individuals to more happy individuals (e.g., Watkins et al. 1992; see Bower and Forgas 2000 for a review). Thus, basic research regarding the effects of mood on judgment has provided extensive coverage of assimilation effects elicited by happy and sad affective states. Angry Affective States All negative affective states are not the same, however. Research indicates that mood-congruent selective encoding effects predictably differ when comparing anger and sadness. For example, Gilligan (1982) hypnotized participants and put them into a happy, sad, or angry mood. Participants then read descriptions of various incidents and were asked to imagine themselves in happy (e.g., finding a $20 bill), sad (e.g., death of a pet), or anger-provoking (e.g., someone cuts in front of you in line) situations. Later, under neutral mood conditions, participants were unexpectedly asked to recall the incidents. Analyses of the recall data revealed not only a mood-congruent bias when comparing the happy condition to the two negative mood conditions, but also a differential pattern of recall when comparing the sad and angry conditions. That is, whereas angry individuals recalled the anger-provoking events better, sad people recalled the sad events better (see Calder and Gruder 1988 for additional evidence of this effect). In addition to eliciting selective encoding effects of this nature, anger elicits a tendency for people to attribute negative outcomes to individuals rather than situational forces (Keltner, Ellsworth, and Edwards 1993; Lerner, Goldberg, and Tetlock 1998; see Averill 1983 for related work).
66
ISBELL, OTTATI, AND BURNS
Whether one focuses on selective encoding or biased attribution, anger serves a unique function by activating anger-relevant cognitions that guide subsequent social information processing. These processes constitute a form of “mental contamination” that can subsequently provoke additional anger as a result of the way in which a target person or situation is construed. Research regarding the displaced aggression hypothesis is consistent with this viewpoint. The displaced aggression hypothesis indicates that anger elicited by one stimulus can increase aggressive behavior toward another stimulus. A recent meta-analysis confirms that this effect is robust and strong (Marcus-Newall et al. 2000). Displacement of anger is especially likely to occur when the initial provocation of anger is followed by exposure to a target person who behaves in an ambiguous manner, or when the target person is encountered within an unpleasant setting (Pederson, Gonzales, and Miller 2000). Under these conditions, aggression directed toward the target person exceeds that which is justified by the target’s behavior (Axelrod 1984). Recent conceptualizations of the displaced aggression effect emphasize that the initial anger-provoking stimulus primes anger-related cognitions that produce selective attention to negative cues and a negative interpretation of ambiguous actions performed by the subsequently encountered target person (Berkowitz 1990, 1993; Pederson, Gonzales, and Miller 2000). This negative impression of the subsequently encountered person triggers the aggressive behavioral response. The displaced aggression hypothesis, which is specific to anger, has been used to address politically relevant phenomena involving the link between economic conditions and racial discrimination (Hepworth and West 1988). Hovland and Sears (1940) focused on a disturbing example of this tendency, namely the link between economic conditions and the lynching of blacks in the United States. They found that, from 1882 to 1930, the number of lynchings rose when economic conditions declined. It appears that minority group membership (perhaps coupled with a minor or ambiguous social infraction) is sufficient to trigger an overwhelming and deadly form of displaced aggression (Marcus-Newall et al. 2000). The implications of this finding are disturbing. One hates to speculate how a poorly performing economy might serve to amplify violence or discrimination against minority groups in present-day society. Fearful or Anxious Affective States Research that focuses on the effect of anxiety or fear on social judgment has adopted both an individual difference and experimental approach. The former approach involves comparing high trait-anxious individuals to low trait-anxious individuals. This research suggests that high trait-anxious people selectively attend to and encode threatening information (Broadbent
AFFECT AND POLITICS
67
and Broadbent 1988; MacLeod and Cohen 1993; Mathews and Macleod 1985, 1994, but see Mogg, Mathews, and Weinman 1987; Ciarrochi and Forgas 1999). They also exhibit a negative selective retrieval bias, respond more quickly to threatening stimuli, and are more likely to interpret ambiguous information in a threatening manner (Broadbent and Broadbent 1988; Burke and Mathews 1992; MacLeod and Cohen 1993; see also McNally 1994 for related evidence in panic disorder). When asked to judge event likelihood, trait anxiety produces an assimilation effect. That is, high trait-anxious people report that negative events are more likely to happen to them (Butler and Mathews 1987). Similar effects emerge for individuals high in neuroticism (Zelenski and Larsen 2002). Manipulation of state anxiety elicits a similar assimilation effect on risk judgments (e.g., the likelihood of having something stolen, Gasper and Clore 1998; Lerner and Keltner 2000, 2001). In accordance with the affect-as-information approach, this effect is reduced when participants are reminded of the actual source of their anxious feelings (Gasper and Clore 1998). Interestingly, this later tendency only emerges for individuals scoring low in trait anxiety. Among individuals high in trait anxiety, state anxiety produces an assimilation effect on risk judgments even when they are reminded of the actual source of their mood. According to Gasper and Clore (1998), this is because these individuals may find anxiety to be a very familiar affective experience that is highly relevant to many of their judgments. Thus, such seemingly relevant affect is difficult to discount. These findings have potentially interesting applications within the political domain. High trait-anxiety individuals may be more likely to report that negative political events are likely (e.g., an economic downturn, an impending terrorist attack). In addition, events that provoke widespread anxiety may increase public estimates of the likelihood that other negative events will occur in the future. This later effect should emerge even if the initial anxiety-provoking event is logically or semantically unrelated to the negative event being forecasted. For example, a sudden economic downturn might arouse anxiety that increases the perceived likelihood of a terrorist attack. The affect-as-information model suggests that this effect should be less prevalent among individuals who are reminded of the actual source of their anxious state. Further research is needed to explore whether these effects actually occur in the political domain. According to Butler and Mathews (1983, 1987; see also Gasper and Clore 1998), effects of anxiety on likelihood estimates are quite strong for personally relevant risks (e.g., having something stolen from oneself ) but considerably weaker for impersonal risks (e.g., tension between the United States and Japan). In the political domain, risks may be either relatively personal or impersonal. For example,
68
ISBELL, OTTATI, AND BURNS
one’s perception of the risk of being involved in a terrorist attack as a result of strained relations between the United States and countries in the Middle East may vary depending on one’s geographic location. That is, during the 9-11 attacks, those who lived in New York City may have been more likely to perceive terrorist threats to be personal risks, whereas those who lived in more remote locations may have perceived the threats to be relatively more impersonal. To the extent that this is true, anxiety is likely to influence these individuals’ risk likelihood estimates differently. Perhaps more importantly, anxiety might elicit an assimilation effect on policy judgments. Consistent with this possibility, Lerner et al. (2003) have demonstrated that fearful moods make individuals prefer a precautionary terrorism policy. In contrast, angry moods lead individuals to prefer a more vengeful terrorism policy. It is not clear whether this effect is mediated by differential risk estimates among anxious and angry individuals. Nevertheless, it is possible that anxiety and anger influence a variety of policy beliefs. This is an area in which further research is needed. Disgusted Affective States Calder and Gruder (1988) have demonstrated that feelings of disgust elicit a selective pattern of information processing that can be distinguished from that which is elicited by feelings of anger. In one study, participants were induced to feel angry or disgusted as they read a review of a restaurant that described either anger-provoking or disgusting attributes of the restaurant. Angry participants responded more negatively to the review containing anger-provoking statements, whereas disgusted participants responded more negatively to the review that contained disgusting restaurant attributes. In addition, more intense affective reactions were associated with more frequent recall for restaurant descriptions that were congruent with those affective reactions. These findings suggest that people make a distinction between anger and disgust, and are predisposed toward accepting and learning information that is congruent with these states. Future Directions There are several directions for future research in this area. For example, “affect infusion” has yet to be fully documented in the political domain. To address this gap in the literature, future research in political psychology needs to investigate those conditions in which affective states elicit selective encoding, biased interpretation, biased elaboration, and selective retrieval effects when individuals process political information. If “affect infusion” is enhanced under conditions that enhance systematic processing (see Bower and Forgas 2000, 2001; Forgas 1995; Forgas, Ciarrochi, and Moylan 2000),
AFFECT AND POLITICS
69
it is possible that effects of this nature will be more pronounced for political experts than political novices. Additional work that specifically focuses on correction for anger-induced biases might also be promising. In some cases, individuals may be aware of their initial tendency to displace aggression. Given adequate levels of capacity and motivation, these individuals may actively attempt to correct for this tendency. Imagine, for example, an individual who experiences tremendous anger in response to the 9-11 event. This anger might initially prompt a prejudicial reaction to Arab Americans who bear no personal responsibility for the 9-11 tragedy. Recognizing the inappropriateness of this initial reaction, the individual might attempt to correct for the anger-induced bias. Lastly, it might be useful for researchers to explore standard of comparison effects that may emerge in the political domain. For example, perhaps negative political events (e.g., 9-11) make subsequent events (e.g., a minor upturn in the economy) appear more positive than would otherwise be the case. We encourage future research to explore such possibilities. The Influence of Affect on Information Processing Style Affect not only has predictable and reliable effects on judgment, but also influences the manner in which individuals process information. A significant body of research (see Wyer, Clore, and Isbell 1999, for reviews) suggests that affect influences the extent to which individuals rely on heuristic versus systematic information processing strategies. Systematic processing is characterized by a tendency to base attitudes and judgments on a careful evaluation and elaboration of information (e.g., a candidate’s specific issue positions, individuating behaviors), whereas heuristic processing is characterized by a general tendency to base attitudes and judgments on peripheral cues, such as general category memberships (e.g., a candidate’s political partisanship, an individual’s race; see Eagly and Chaiken 1993). Individuals who lack the motivation or ability to process information carefully are likely to rely on more heuristic information processing strategies. Although it is possible that individuals who rely on heuristic cues may do so without processing any additional information, this may be fairly rare. For example, when forming an impression of a political candidate, it is unlikely that voters will abandon processing altogether after learning the candidate’s partisanship. Most voters will acquire at least a small amount of information about a candidate beyond some initial category information. Under these conditions, heuristic cues (e.g., partisanship) may bias subsequent systematic processing and be reflected in later judgments (see Chaiken and
70
ISBELL, OTTATI, AND BURNS
Maheswaran 1994). Thus, heuristic cues may play a powerful role in judgments of a candidate because (1) individuals fail to attend to information beyond these cues due to ability or motivation deficits, or (2) individuals’ processing of subsequent individuating information (e.g., policy positions) is biased by these cues. Effects of Happiness and Sadness on Information Processing Style Most of the research on affect-induced differences in information processing style has focused on the effects of happiness and sadness. This work demonstrates that happiness is associated with a more heuristic processing style, whereas sadness is associated with a more systematic style. For example, individuals in happy moods are more likely to use categorical information as a basis for their judgments than those in sad or neutral moods (e.g., Abele, Gendolla, and Petzold 1998; Bless, Schwarz, and Wieland 1996; Bodenhausen, Kramer, and Susser 1994; Isbell 2004). In a series of studies, Bodenhausen and his colleagues ( Bodenhausen, Sheppard, and Kramer 1994) found that individuals relied on a defendant’s category membership to a greater extent when making guilt judgments if they were feeling happy, rather than sad or neutral. In a similar vein, Park and Banaji (2000) found that individuals in happy moods were more likely than those in sad moods to incorrectly recall that specific group members possess stereotypic traits. Further, happy people evaluated an out-group more negatively when in direct competition with them than those in neutral moods (Dovidio, Gaertner, and Loux 2000). In competitive partisan politics, positive affect could lead to greater stereotyping of members of the other political party. The persuasion literature also provides evidence for affect-induced differences in information processing style. Happy individuals are relatively uninfluenced by the strength of arguments contained in persuasive messages, but instead are influenced by peripheral cues, such as communicator attractiveness or expertise (Bless et al. 1990; Bless et al. 1996; Bohner et al. 1992; Mackie and Worth 1989; Schwarz, Bless, and Bohner 1991; Sinclair, Mark, and Clore 1994). In contrast, sad individuals tend to be influenced by argument quality rather than peripheral cues. Thus, taken together, much research suggests that individuals in happy and sad moods process information differently. Although affect scholars generally agree on this conclusion, they do not agree on what causes these effects. We turn our attention to this question before examining the impact of frequently neglected specific negative affective experiences (e.g., anger, fear) on information processing style.
AFFECT AND POLITICS
71
What Accounts for the Effects of Happiness and Sadness on Information Processing Style? Early research attributed differences in happy and sad individuals’ processing to deficits in happy individuals’ ability (Mackie and Worth 1989) or motivation (e.g., Schwarz 1990; see also Wegener, Petty, and Smith 1995) to process information in a careful and detailed manner. According to the cognitive capacity hypothesis (Isen 1987; Mackie and Worth 1989; Worth and Mackie 1987), positive affect activates a large body of highly interconnected information in memory, resulting in a reduction of cognitive resources to process new information. This may lead happy individuals to rely on less effortful processing strategies. In support of this hypothesis, Mackie and Worth (1989) found that happy participants were as influenced by argument strength as unhappy participants when they were given ample opportunity to think about the message content, but did not distinguish between strong and weak arguments when processing time was limited. In contrast to the ability hypothesis, mood maintenance and repair hypotheses (e.g., Bodenhausen, Kramer, and Susser 1994; Bodenhausen, Sheppard, and Kramer 1994; Isen 1987) maintain that happy individuals are sensitive to the possibility that systematic processing may adversely influence their mood and thus they do not process information carefully. In contrast, those experiencing sadness are motivated to process information carefully in an attempt to enhance their moods. These hypotheses generally predict that individuals in happy moods rely on heuristic criteria as a basis for judgments unless their processing motivation is increased in some way. Consistent with this possibility, research demonstrates that happy participants process information carefully if they are explicitly instructed to do so (Bless et al. 1990), are held accountable for their judgments (Bodenhausen, Kramer, and Susser 1994), or are led to believe that detailed processing will enhance or maintain their mood (e.g., Wegener, Petty, and Smith 1995). An Affect-As-Information (AAI) Conceptualization Schwarz (1990) proposed a different explanation for affect-induced processing differences that relies on the informational value and meaning of affective cues. According to the AAI conceptualization and theories of emotion (Ortony, Clore, and Collins 1988; see also Roseman 1991), individuals continuously and often nonconsciously appraise situations with regard to their current concerns and goals. These appraisal processes result in conscious feedback in the form of affective feelings, which serve to provide individuals with information about the situations in which they are experienced (Clore 1992; Ortony, Clore, and Collins 1988; Schwarz and Clore
72
ISBELL, OTTATI, AND BURNS
1983; Wyer, Clore, and Isbell 1999). Thus, affective feelings are functional and serve as feedback that directs individuals’ judgments and information processing (see Damasio 1994 and Adolphs and Damasio 2001 for related evidence from cognitive neuroscience). Affective cues provide individuals with information about the nature of a situation (Bless and Schwarz 1999; Schwarz 1990, 2002). Sadness indicates the presence of a problem and signals that careful and detailed processing is required in order to eliminate it. In such situations, heuristic processing strategies are likely to be risky. Happiness in contrast indicates that the situation is safe and benign and does not engender any particular processing motivation. By default, individuals in happy moods tend to rely on heuristic processing. Consistent with this notion, Bless and his colleagues (Bless 2000, 2001; Bless et al. 1996) found that happiness increases the use of general knowledge structures (e.g., stereotypes, traits, scripts, a candidate’s partisanship) that have served individuals well in the past. However, as noted earlier, under conditions in which happy individuals are motivated by situational or other concerns, they tend to process information carefully. In these situations, happy individuals’ processing is likely to be guided by and potentially biased by general knowledge structures. In contrast, sad individuals should be more heavily influenced by specific and detailed information (e.g., a candidate’s specific issue positions). Thus, happy and sad individuals may process similar amounts of information, but their moods may differentially bias the manner in which they process information. Consistent with the idea that happiness results in processing that is guided by general knowledge structures, Bless and his colleagues found that happy participants were more likely than sad participants to rely on easily accessible knowledge (a restaurant script; Schank and Abelson 1977) when processing information, which allowed them to outperform sad participants on a secondary task (Bless et al. 1996). In an impression-formation study, Bless, Schwarz, and Wieland (1996; see also Krauth-Gruber and Ric 2000) found that happy participants relied on a target’s category membership when individuating information was consistent with or unrelated to the category information, but relied on individuating information if it was inconsistent with the category membership. Sad participants relied on the individuating information regardless of its consistency with the category information. Further, Isbell (2004) found that happy participants relied on abstract traits as a basis for judging a target regardless of whether they received this information before or after more specific behaviors the target performed. In contrast, relatively unhappy participants based their judgments on the behaviors rather than the traits, regardless of the order in which they received this information. These findings may have political implications. That is, individuals exposed to positive affect-eliciting stimuli (e.g., the American flag, patriotic
AFFECT AND POLITICS
73
music) may be more likely to judge a candidate on the basis of global information, such as his partisanship or general personality traits. In contrast, those not exposed to positive affect-enhancing stimuli or exposed to stimuli that elicit sadness may be more likely to judge a candidate on the basis of detailed information (e.g., specific political issue positions). For example, a candidate who delivers an unfavorable message to a group of citizens whose party matches the candidate should use positive affecteliciting stimuli. This would help to increase the likelihood that citizens will evaluate him on the basis of his partisanship rather than on the basis of his unfavorable message. In contrast, if the candidate were speaking before citizens who do not share his political partisanship, he should avoid using positive affect-enhancing stimuli because it may decrease the likelihood that this information would be used as a significant basis for evaluation. Biases in information processing as a function of happiness and sadness have been detected in processing measures as well as judgment measures. For example, Isbell (2004) found that happy participants who received trait information before behavioral information attempted to reconcile the inconsistencies between the two and, consequently, recalled more trait-inconsistent than trait-consistent behaviors (for a discussion of inconsistency resolution processes in impression formation, see Srull and Wyer 1989 and Wyer and Srull 1989). Under conditions in which the trait information was received after the behaviors, happy participants relied on the trait information as a retrieval cue and recalled more trait-consistent than trait-inconsistent behaviors (see Wyer, Bodenhausen, and Srull 1984). Sad participants, in contrast, failed to rely on the abstract trait information regardless of when they received it. Although this research reveals that happy moods promote greater use of stereotypes, traits, and other general knowledge structures relative to sad moods, this finding is not the result of sad individuals not attending to this type of information. In the studies by Isbell discussed earlier, sad and unhappy participants were just as likely as happy ones to recall the abstract target information later. Thus, they were aware of this information and may have either discounted it (as the processing effects suggest in the study by Isbell) or corrected their judgments for its influence. Research by Lambert and his colleagues (Lambert et al. 1997) found that sad individuals might correct their judgments for the influence of stereotypic information under conditions in which it seems inappropriate to rely on it. Lambert et al. found that sad participants relied on a physical attractiveness stereotype when evaluating a female job candidate under conditions in which physical attractiveness was an important criterion for the job, but not under conditions in which it was not. Taken together, the research just reviewed provides strong evidence that happy individuals do not necessarily ignore individuating information
74
ISBELL, OTTATI, AND BURNS
or solely rely on the first information that they receive. Similarly, it is also the case that sad individuals do not necessarily ignore global, abstract information altogether but instead consider its appropriateness for the task at hand. In general, however, happy participants appear to engage in more elaborative processing than sad participants and attempt to relate global, abstract information to more specific information. In this respect, happy participants process information at least as systematically as sad participants; however, their processing tends to be biased by the global information that they receive. How do Specific Types of Negative Affect Influence Information Processing Style? The vast majority of research examining the influence of affect on information processing style has focused on happiness and sadness and has generally assumed that differences in valence are responsible for differences in information processing style. Recent research suggests that this is not necessarily the case (e.g., Bodenhausen, Sheppard, and Kramer 1994; Bohner and Weinreth 2001; Tiedens and Linton 2001). For example, in several studies, researchers found that anger promotes heuristic processing whereas sadness and fear promote systematic processing (e.g., Bodenhausen, Kramer and Susser 1994; Bohner and Weinerth 2001; Tiedens and Linton 2001). Such findings highlight limitations of valence-based approaches. Affective experiences vary along a number of dimensions other than valence (Ortony, Clore, and Collins 1988; Roseman 1984; Smith and Ellsworth 1985), and these other dimensions may play a critical role in how affect influences processing (see Tiedens and Linton 2001). For example, anger and fear both result from appraisals of an event as unpleasant and brought about by someone/something other than the self. Anger, however, differs from fear in that it is associated with a sense of certainty about how a situation will unfold, whereas fear is associated with uncertainty (Lerner and Keltner 2000, 2001; Tiedens and Linton 2001). A small number of studies suggest that anger and fear may be associated with correspondingly greater and lesser degrees of reliance on stereotypes and other general knowledge structures when processing information. These differences may be due to variations in feelings of certainty that individuals experience when in different affective states. This possibility is consistent with the AAI model in that feelings of certainty (like feelings of pleasantness) may linger and carry over to influence subsequent processing tasks (Clore 1992). Angry Affective States In an impression-formation study, Bodenhausen and his colleagues (Bodenhausen, Sheppard, and Kramer 1994) found that individuals in
AFFECT AND POLITICS
75
angry moods were more likely to use stereotypes than individuals in sad moods. Further, in two persuasion studies, they found that angry participants’ attitudes were more heavily influenced by peripheral cues, such as source trustworthiness and source expertise, whereas sad participants’ attitudes were not significantly affected. In a specific comparison of high versus low certainty emotions, Tiedens and Linton (2001) found that individuals experiencing a high certainty emotion (i.e., anger) were more likely to be persuaded by a heuristic cue than those experiencing a low certainty emotion (i.e., worry). Many political events are likely to elicit different negative affective reactions in citizens depending on how they are framed and the specific aspects of the events on which individuals are led to focus (see Ortony, Clore, and Collins 1988). For example, the 9-11 terrorist attacks on the United States are likely to elicit anger if one focuses on the fact that they occurred as a result of security failures. In contrast, if one focuses on the loss of life, sadness is likely to be elicited. When discussing this event, a politician may systematically frame his discussion in a way that increases the likelihood that individuals will experience a particular negative emotion. To the extent that anger is elicited, the politician’s subsequent message is likely to be processed heuristically. In contrast, if sadness is elicited, the message is likely to be processed more carefully and systematically. Fearful or Anxious Affective States Early research on the relationship between anxiety and performance demonstrates that the relationship is not linear, but rather is an inverted-U (e.g., Hebb 1955). That is, as anxiety increases, performance increases up to a point at which further increases in anxiety lead to performance decrements. Since increased performance is typically viewed as resulting from increased systematic processing, it is surprising that this early research has not previously been connected to research on the influence of affect on information processing style. The AAI model would suggest that these results are due to the informational feedback associated with anxiety. Anxiety signals the presence of a problem in the environment, indicating that careful and systematic processing is necessary. At moderate levels of anxiety, it is likely that individuals can achieve this level of processing; however, at high levels of anxiety, the affective experience itself may interfere with cognitive activity and lead to decrements in performance. Thus, these results suggest that anxiety may only increase systematic processing when it is experienced at moderate levels. Tiedens and Linton (2001) suggest that the uncertainty associated with fear and anxiety leads to increased systematic processing more than emotions that are associated with greater certainty, such as anger. In one study, Tiedens and Linton found that participants experiencing fear (a low
76
ISBELL, OTTATI, AND BURNS
certainty negative emotion) relied on a stereotype as a basis for impression formation more than those experiencing a high certainty negative emotion (i.e., disgust). Consistent with this research, work by Weary and Jacobson (1997; see also Edwards et al. 2000; Hildebrand-Saints and Weary 1989) found that individuals who feel chronically uncertain (i.e., depressed) process information more systematically than those who feel certain (i.e., nondepressed). Research in the political domain demonstrates that anxiety is associated with differences in information processing style. Using National Election Survey data from 1980 to 1996, Marcus and his colleagues (Marcus, Neuman, and MacKuen 2000) found that voters who felt anxious (fearful) about their own political party’s candidate during an election relied on different criteria when making their vote choices than voters who felt complacent. Consistent with predictions derived from the AAI model, anxious voters weighted the candidate’s specific policy issues more heavily than complacent voters, whereas complacent voters weighted partisanship significantly more than anxious voters. Thus, these results suggest that affect may influence the manner in which political information is processed as well as the bases on which candidates are evaluated (see Nadeau, Niemi, and Amato 1995 for related evidence). Citizens who feel anxious and fearful in response to political events associated with terrorism may be more likely to move beyond partisanship and consider detailed candidate information when making their vote choices. However, as individuals’ anxiety increases to more extreme levels (e.g., if the United States was being attacked directly), it may interfere with their ability to engage in cognitive activity and, consequently, may lead to more heuristic information processing. Fear has also been linked to systematic processing in persuasion research. For example, in one study, Bohner and Weinerth (2001) examined the effects of experimentally induced fear on smokers and nonsmokers’ processing of either a strong or weak antismoking message. Bohner and Weinerth found that fear led to more systematic processing compared to baseline (non-manipulated affect) conditions among nonsmokers, but only when the affect manipulation was low in salience. When it was high in salience, participants discounted the informational value of the affective cues and processed the message in an equally systematic manner regardless of their affect. This finding is consistent with the AAI conceptualization that maintains that affective cues influence judgment and processing only when they are perceived as relevant. When the informational value is discounted through an affect attribution manipulation, affect no longer influences judgments and processing (see Sinclair, Mark, and Clore 1994 for evidence in happy and sad moods).
AFFECT AND POLITICS
77
Interestingly, smokers in Bohner and Weinerth’s study processed the antismoking persuasive communications less systematically in fear conditions than in baseline conditions when the salience of the affect manipulation was low. The authors contend that the impact of fear on information processing style depends on the extent to which individuals perceive a message to be propaganda. If this is the case, individuals in fearful moods are likely to take it less seriously and process it less systematically. When the salience of the affect manipulation was high, smokers processed the messages in an equally systematic manner regardless of their affect. The results of the persuasion study conducted by Bohner and Weinerth (2001) have implications for political information processing. As we live in a post–9-11 world where the threat of terrorism creates fear and anxiety among American citizens, many individuals are likely to be more systematic political information processors. However, if one believes that the security threat level was moved from yellow (elevated threat) to orange (high threat) as a political tactic that may help increase public support for the war in Iraq, then one may view this as propaganda and process information less systematically. If an individual perceives the change in threat level to be genuine, then processing is likely to become more systematic. As noted earlier, however, it is possible that if the threat level was raised to red (severe threat), levels of fear and anxiety would become very high and people would no longer process information systematically regardless of whether they viewed the message as propaganda. Guilty Affective States Few studies have investigated the information processing consequences of guilt. Bohner and Weinreth (2001) conceptually replicated the persuasion and fear study described earlier using guilt. Their findings suggest that guilt and fear have similar effects on processing. In an interesting series of studies, Devine and her colleagues (Devine et al. 1991; Devine and Monteith 1993) investigated the impact of guilt (arising from experiencing stereotyperelated discrepancies) on information processing. When stereotype-related discrepancies were produced in low prejudice individuals, they were likely to experience guilt, which warned them to behave in non-prejudicial ways. Monteith (1993) found that these feelings of guilt led low prejudice individuals to process group-relevant information more carefully and slowly than high prejudice individuals. These findings reflect that guilt may lead low prejudice individuals to attempt to correct for the influence of stereotypes when processing group-relevant information. Within the political domain where social issues (e.g., affirmative action, busing) are often affectively charged and relevant to specific social groups (e.g., African Americans), these results suggest that individual differences in prejudice
78
ISBELL, OTTATI, AND BURNS
may influence the extent to which individuals experience guilt and, consequently, the manner in which they may process issue-relevant information. Low prejudice individuals may form their attitudes about social issues on the basis of more careful and systematic processing than high prejudice individuals. Can Affective Cues Influence Processing without Triggering an Affective State? The AAI conceptualization suggests that different information processing styles result from different types of informational feedback. This feedback is likely to come from affective feelings, but may also come from other sources. That is, it is possible that one may encounter affective cues or informational feedback without consciously experiencing any corresponding affect and this information may directly influence processing style. Consistent with this possibility, researchers have found that affective cues and feedback provide individuals with information about processing requirements and can influence their information processing style without influencing their affective state. For example, Friedman and Foerster (2000) hypothesized that muscular feedback provided by arm flexion (which is associated with approach tendencies, and thus, positive feedback) should lead to less careful, systematic processing than arm extension (which is associated with avoidance tendencies, and thus, negative feedback). Consistent with this expectation, participants who completed GRE questions during arm extension performed better than those who completed them during arm flexion. Similarly, Soldat and his colleagues (Soldat, Sinclair, and Mark 1997) found that students who completed GRE questions on blue paper (which conveys sadness) performed better than those who completed them on red paper (which conveys happiness). Soldat and Sinclair (2001) found additional evidence that paper color conveys affective information that directly impacts information processing style. In one study, participants received either strong or weak persuasive arguments printed on either blue or red paper. Consistent with expectations, the blue paper promoted more systematic processing than the red paper. Participants who received the arguments on red paper were equally persuaded by strong and weak arguments, whereas those who received the arguments on blue paper were more persuaded by strong than weak arguments. These results suggest that candidates would be wise to consider the types of visual information they provide citizens in their campaign materials. Some colors and visual images may decrease the likelihood that voters will attend carefully to their substantive messages.
AFFECT AND POLITICS
79
Processing differences have also been found with other affective cues. For example, in one study (Soldat and Sinclair 2001) participants were instructed to read strong or weak arguments to a confederate who displayed either smiling, supportive, or serious facial expressions. Participants who viewed the positive expressions processed the arguments less systematically than those who viewed the serious expressions. These results are consistent with Butler and Baumeister’s (1998) finding that supportive audiences tend to lead to performance decrements because they cause performers to be overly cautious and “choke” under pressure. It is also possible that affective cues associated with the audience’s reactions in this study led to less careful processing and consequently, to performance decrements. The research just reviewed suggests that facial feedback from an audience influences the extent to which a speaker or performer processes information in a careful and systematic manner. In contrast to this work, research by Ottati and his colleagues (Ottati, Terkildsen, and Hubbard 1997) demonstrates that a speaker’s facial expressions may influence the extent to which an audience processes information in a systematic manner. They found that a political candidate elicited more systematic processing in perceivers when he expressed political information while conveying a serious facial expression than while conveying a happy expression. Specifically, Ottati and his colleagues found that individuals exposed to the happy expressions relied on the candidate’s ideological orientation as a basis for their evaluations of him, whereas those exposed to the serious expressions relied on their agreement with the candidate on the specific issue positions that he was presenting. Thus, the happy facial expressions provided perceivers with information that the communication was not of a serious or important nature and thus, indicated that careful and systematic processing was unnecessary. In contrast, the serious expressions conveyed that the topic of the communication was of some importance and thus, promoted more careful processing. Do Affective Cues Have to Be Consciously Perceived to Influence Processing? Wyer, Clore, and Isbell (1999) suggested that affective cues may spontaneously and automatically activate procedural knowledge in memory without conscious mediation. That is, affective cues associated with happiness may automatically activate a heuristic information processing style, whereas cues associated with sadness may automatically activate a systematic information processing style. Consistent with this expectation, Colcombe, Isbell, and Clore (2001) found that subconsciously priming individuals with smiling faces prior to an impression-formation task led them to rely on stereotypes when forming their impression of a target. In contrast, those
80
ISBELL, OTTATI, AND BURNS
primed with frowning faces relied to a greater extent on specific and detailed behavioral information about the target. Soldat and Sinclair (2001) found similar effects in a persuasion study in which either smiling or serious facial expressions were subliminally presented to participants on a computer screen as they read either strong or weak arguments. Participants exposed to the smiling faces processed more heuristically than those exposed to the serious faces. Processing differences were apparent not only in participants’ attitudes, but also in the thoughts they listed in response to the message they received. That is, the serious faces led to more positive thoughts about the message when the arguments were strong, but more negative thoughts when the arguments were weak. In contrast, happy faces led to similarly positive and negative thoughts regardless of argument strength. Importantly, the affective cues used in Colcombe, Isbell, and Clore’s and Soldat and Sinclair’s studies did not influence participants’ moods. The impact of subliminally presented affective cues on political attitudes became an issue of significant interest for a brief time during the 2000 presidential campaign. During the campaign, the Republican National Committee ran an advertisement for George Bush in which the word “Bureaucrats” and various parts of the word (including the word “RATS”) flashed across the screen when discussing Democratic Presidential Nominee Al Gore’s prescription drug plan. Bush was accused of relying on subliminal tactics to influence voters. Regardless of whether the negatively valenced word “RATS” was presented deliberately, it may have influenced viewers’ processing of the political message contained in the advertisement. That is, the word “RATS” may have activated a systematic information processing style that may have led viewers to scrutinize the advertisement more carefully than they might have otherwise. In addition, it is also possible the negative evaluation of “RATS” may have been directly associated with Gore and led to more negative evaluations of him. The research on the influence of affective cues on information processing style has potentially widespread implications for political information processing beyond those already discussed. Politics is inherently affective and candidates and political events are frequently surrounded by affective cues. These cues may directly impact perceivers’ judgments of a candidate, as discussed earlier, or they may activate a heuristic or systematic processing style and influence the way in which perceivers process politically relevant information. Further, these effects may be the result of either “felt” affect or may be the result of affective cues providing individuals with information about the processing requirements of the situation. Given the number of ways in which affective cues might influence judgment and processing, it is clear that these affect-induced effects often emerge and thus, are likely to play a significant role in political judgment and processing.
AFFECT AND POLITICS
81
Future Directions The previously presented evidence provides strong support for the idea that the informational qualities of stimuli as well as bodily feedback can lead to different styles of information processing. Unlike the effects of subjectively experienced “felt” affect discussed earlier, these effects are the result of the activation of cognitive concepts that contain evaluative meaning and value (Clore and Colcombe 2001). Researchers have failed to move beyond valence in examining the information processing consequences of affective cues that provide affective information, but do not elicit affective feelings. Thus, it is currently unknown to what extent affective cues associated with specific types of negative stimuli (e.g., fear, anger) might produce results similar to those predicted for the subjectively experienced “felt” affect. Future research may investigate this question by simultaneously activating appraisal dimensions in addition to valence. For example, negatively valenced stimuli that activate a sense of uncertainty (characteristic of fear) should be more likely to lead to systematic information processing than negatively valenced stimuli that activate a sense of certainty (characteristic of anger). Research investigating these possibilities in the political domain promises to be particularly valuable and exciting. The Influence of Affect on Information Seeking The research discussed so far has focused on the influence of affect on judgment and information processing. Most of this research has examined the perceiver as a recipient of target information, rather than one who is actively seeking out information from the environment. Yet natural information processing situations usually involve information seeking. This is particularly true in the political domain in which individuals are bombarded with vast amounts of information that they may attempt to synthesize to form impressions of political candidates and make a vote choice. Given that it is impossible to process all of the available information, individuals are likely to selectively seek information to use as a basis for their impressions. The way individuals feel is likely to influence the types of information that they seek and thus, may have a profound impact on the impressions they form. How do Happy and Sad Moods Influence Information Seeking? The research on the effects of happy and sad moods on information processing style suggests that happy moods are often associated with an increased reliance on abstract, global information, whereas sad moods are
82
ISBELL, OTTATI, AND BURNS
associated with an increased reliance on specific and detailed information. Although this research demonstrates clear differences in processing, it fails to examine whether affect influences the types of information that individuals seek out. Several studies conducted by Isbell and her colleagues have begun to investigate the influence of affect on information-seeking behavior under conditions in which individuals possess (1) no prior target information, but can choose either global (traits) or specific information (behaviors) about a target, and (2) prior target information (e.g., a stereotype) in which individuals can solicit either expectancy-confirming or expectancy-disconfirming information. Although these studies were conducted outside of the political domain, they may have important implications for political information processing. No Prior Target Information Isbell, Burns, and Haar (in press) investigated the extent to which differences in affect predict information preferences when selecting information about an unknown target. In one study utilizing self-reported happiness, participants completed an impression-formation task on the telephone. Participants were told that to form an impression of a target, they would receive two different types of information and would be asked to select the type they would like to hear first. Participants were told that some target information was general and consisted of adjectives that others used to describe the target (general information) whereas other information was specific and detailed and described behaviors that the target performed (specific information). The general information described the target as either an introverted librarian or an extroverted sales representative and the detailed behavioral information described an equal number of introverted and extroverted behaviors. Consistent with expectations, relatively happy participants were significantly more likely than unhappy ones to select the abstract trait information first. In a subsequent impression-formation study, Isbell, Burns, and Haar examined the effects of experimentally manipulated happiness and sadness on preferences for abstract (trait) information and concrete (behavioral) information using an information search paradigm. After writing about either a happy or sad life event, participants formed an impression of a target on the basis of information they selected from two labeled boxes presented on a computer: (1) “general trait information” and (2) “specific behaviors that the person performed.” An equal number of traits and behaviors were available for selection. Participants could select as much information as they desired. Consistent with the AAI conceptualization and the results of the earlier study, happy participants were more likely to select a trait initially, whereas sad participants were more likely to select a
AFFECT AND POLITICS
83
behavior. Although happy and sad participants selected equal amounts of information overall, when the traits were positive (i.e., nondiagnostic) happy participants selected relatively more traits than sad participants. In contrast, when the traits were negative (i.e., diagnostic), happy and sad participants selected a similar proportion of traits. This finding suggests that sad participants may rely on trait information when they perceive it to be diagnostic. This conclusion is consistent with the results of Lambert et al.’s (1997) study discussed earlier in which sad participants relied on a stereotype as a basis for judgment under conditions in which the stereotype provided appropriate judgment-relevant information. Given that researchers have generally failed to investigate the impact of affective cues on information seeking, it is not surprising that this work has not been applied to the political domain. This work could have interesting political implications, however. At the beginning of the campaign, voters often have zero to minimal information about the candidates. In a field of many candidates, it may be useful to sort them initially on the basis of their partisanship (global information) before narrowing down to the specific issue stances. Happy people’s information search strategies may be the most adaptive for initial information to prevent information overload. Prior Target Information (stereotypes) Oftentimes individuals have some initial or prior information about a target available before they seek additional information. For example, most people immediately become aware of the race of someone they are meeting for the first time. This may occur before any other information is known and thus raises an important question. That is, to what extent is an individual’s information-seeking behavior guided by initial or prior target information and how does affect influence this behavior? Research ( Johnston 1996; Johnston and Macrae 1994) demonstrates that individuals often engage in confirmatory hypothesis testing in which participants select stereotypeconfirming rather than disconfirming or irrelevant questions to ask a target. This tendency leads individuals to preserve their stereotype-based evaluations of the group. Consistent with the AAI conceptualization, individuals in happy moods should be more likely than those in sad moods to be influenced by accessible stereotype information and thus, should be more likely to attempt to maintain their stereotypic beliefs by seeking stereotype-consistent information. Isbell and her colleagues (Isbell, Burns, and James, 2004) found evidence consistent with this possibility. In one study, Caucasian participants were experimentally induced to feel either happy or sad and then were led to believe that they would conduct a job interview with either a Caucasian or an African American student as part of a job preparation program.
84
ISBELL, OTTATI, AND BURNS
Participants were presented with a list of 16 questions that they could ask the applicant and were told to select 8 for the interview. Half of the available questions were designed to elicit negative information from the target (e.g., “What are some of your weaknesses?”) and half were designed to elicit positive information (e.g., “What are some of your strengths?”). Given that research (e.g., Devine 1989; see also Bargh, Chen, and Burrows 1996) suggests that the evaluative implications of the African American stereotype are generally negative, Isbell, Burns, and James hypothesized that happy participants would be guided by the evaluative implications of the stereotype when selecting questions, whereas sad participants would not. Consistent with this hypothesis, happy participants selected a greater number of negative questions when the target was black than when she was white, whereas sad participants selected an equal number regardless of her race. In the political world, citizens may have the opportunity to ask the candidates questions during “town meetings” or other open forums. This research suggests that candidates from stereotyped groups (e.g., Joseph Lieberman, Elizabeth Dole, Jesse Jackson) may find themselves at a disadvantage simply by the types of questions they are asked and thus, by the types of information that become available to voters who are forming impressions of them. How do Specific Types of Negative Affect Influence Information Seeking? Although research has generally failed to examine the influence of different types of negative affect on the types of information that individuals seek out, it is likely that the effects will parallel those found in research on information processing style. That is, anger and fear may lead to differences in subjective experiences of certainty and thus, may influence individuals’ confidence in global, abstract information. As a result, angry individuals may be likely to seek global target information and information that confirms their hypotheses and stereotypes. In contrast, fearful individuals may seek more detailed and specific information and may be more likely to adopt diagnostic hypothesis testing strategies. Consistent with this possibility, Marcus and MacKuen (1993) found that voters who experience anxiety are likely to learn more about political issues than those who experience enthusiasm and are less likely to rely on habitual or heuristic cues, such as partisanship. These results suggest that anxiety leads voters to seek issue-related information, whereas enthusiasm does not. Future research is necessary to explore the viability of the proposed hypotheses both within the political domain as well as outside of it. For example, angry moods may lead individuals to seek out information that is consistent with a candidate’s
AFFECT AND POLITICS
85
partisanship (e.g., a biased sample of issue positions), whereas anxiety may lead individuals to seek information that diagnoses a candidate’s partisanship (e.g., a representative sample of issue positions). Future Directions Given that research on the influence of affect on information selection is relatively new, there are several additional ways in which it could be extended. For example, to our knowledge, no research has examined the extent to which affective cues (e.g., smiling or frowning faces, red or blue paper) might influence information search in the absence of an affective state. Future research should investigate this possibility. We anticipate that the influence of affective cues on information selection will be similar to those that emerge when individuals experience subjectively “felt” affect. Further, these results are expected to emerge regardless of whether the affective cue is consciously perceived. The current political world is very affectively charged, which makes for an interesting climate when seeking information in order to be an informed citizen and voter. This research could have many important implications and should be explored. Finally, as just discussed, more research needs to be conducted on how specific negative affective states influence information-seeking strategies. Conclusions Overall, affect influences judgment, information processing, and information seeking in different ways depending upon the specific affective state and the level of awareness surrounding it. Affect sometimes produces assimilation effects on judgment wherein affective feelings match the valence of the evaluation of an object, or the opposite can occur producing contrast effects. Information tends to be processed in systematic ways when people are in sad or fearful/anxious moods whereas those in happy or angry moods tend to process information more heuristically. It is noteworthy and interesting that two negatively valenced emotions, such as anger and fear/anxiety, can have drastically different effects on these processes. Clearly, this is a very rich area for future research. The field of information seeking is also an emerging area of research. It is important to study the influence of affect on political judgments and processing using an active-search paradigm in order to better generalize and understand these processes in the “real” world. In these uncertain times, we have a prime opportunity to learn more about the functions of negative emotions, especially fear/anxiety stemming from living in a world where a terrorist attack may occur at any time. It will
86
ISBELL, OTTATI, AND BURNS
be fascinating to observe how these affective experiences will impact citizens’ judgments, information processing, and information seeking in future elections. We can learn from the past in thinking about these issues. Voters in the 1964 election turned to Democratic presidential candidate Lyndon B. Johnson after two showings of his infamous advertisement featuring a nuclear bomb and a child playing with a flower. With the continued threat of nuclear weapons and weapons of mass destruction, it will be interesting to examine the extent to which politicians will use anxietyprovoking messages in an attempt to win popular support. As the previously discussed research indicates, this tactic could turn voters into more systematic information processors who focus on details when seeking information and making judgments. Affect has infused politics throughout history and will continue to do so. For this reason, it is essential that researchers continue to study the influence of affect on political information processing.
CHAPTER 6 MOTIVATED REASONING, AFFECT, AND THE ROLE OF MEMORY IN VOTER DECISION MAKING David P. Redlawsk
or much of the history of voting research, citizens have been viewed as less than well informed, paying little attention to politics and maintaining only a limited grasp on issues. When election day rolls around, votes are cast either in accordance with a party identification uninformed by issue content or simply based on group affiliations or the nature of the times (Campbell et al. 1960). This lack of competence is clearly evident in voters’ inability to respond effectively to open-ended survey questions designed to elicit opinions on the candidates of the day. More recently, however, political psychology has challenged this conventional wisdom. Presaged by Fiorina’s (1981) idea of party identification as a rational retrospective evaluation of past party performance, and supported by a well-developed psychology literature on person-perception (Wyer and Srull 1980, 1986; Hastie and Park 1986), Lodge, McGraw, and Stroh (1989; Lodge, Steenbergen, and Brau 1995) argue that candidate evaluation is an online process, during which candidate affect is developed as information is encountered over time. Affect—represented by the online tally—is maintained in memory while the information that informs it can be safely discarded. When the time comes to vote, citizens need merely query the tally, with no memory search for information about the candidates. Thus, it should not be surprising that efforts to find extensive issue content in the memories of citizens should fail. Indeed, if voters process campaigns online, they could be considering far more information than they can recall for survey researchers. Voters could even be well-informed democratic citizens.
F
88
REDLAWSK
The basic online model posits no direct role for memory in candidate evaluation (Lodge, McGraw and Stroh 1989; Lodge, Steenbergen, and Brau 1995). Studies in realms other than politics seem to confirm this, as special efforts have to be made to find laboratory tasks that actually link memory and evaluation (Hastie and Park 1986; Lichtenstein and Srull 1987). It appears clear that person-evaluation tasks, of which candidate evaluation certainly is one, are routinely proceed online. If candidate evaluation is a simple person-evaluation task, then it should not matter how many candidates are being evaluated. There should be no relationship between memory and evaluation whether one or many candidates are under consideration. Lodge and colleagues argue, in fact, that in a multicandidate election voters need only query the tallies for each candidate and make a comparison (Lodge, Steenbergen, and Brau 1995). The online process does not change. However, most of the studies reported by Lodge and his colleagues include the evaluation of only one political figure with information on the figure provided on an easily processed single sheet of paper. Subjects read statements about a “congressman” and form an impression.1 Such an experimental design certainly tests the basic online model of evaluation, but it does not test whether the model applies in a political campaign environment where voters not only form evaluations but also choose from among a group of contenders. Using a significantly different experimental design based on dynamic process tracing (Lau 1995; Lau and Redlawsk 1997, 2001a, 2001b, 2006; Redlawsk 2001, 2002, 2004), I tested the basic relationship between memory and evaluation in a simulated presidential election campaign (Redlawsk 2001). Subjects faced either two or four candidates in a primary election. After learning about the candidates, subjects voted for one candidate, evaluated all candidates using a 0–100 feeling thermometer, and completed a memory recall test asking them to list everything they could remember about each candidate. The findings were unexpected. When subjects evaluated candidates for the purposes of voting in a multi-candidate environment, memory had clear effects on candidate evaluation, the direction of the vote, and the ability of subjects to vote correctly. Indeed, these memory effects were strong even when the online tally was included in the analysis. Whether subjects were online processors or memory processors in the experiment, memory mattered in ways not explained by the online model.2 In discussing the reasons for the unexpected importance of memory, I focused on the nature of the decision task, suggesting that voters look for information in order to directly compare candidates based on attributes (Redlawsk 2001, p. 52). Given the here-today-gone-tomorrow nature of candidate information, voters could make those comparisons effectively
ROLE OF MEMORY IN VOTER DECISION MAKING
89
only if they made use of memory. While a plausible speculation, evidence to directly support this proposition is not available. Thus, following Steenbergen and Lodge (1998) I tentatively suggested that information processing during a political campaign is a hybrid process and that the pure online model can not describe it completely. Since the initial work on online processing, researchers have developed a more comprehensive view of voter information processing, suggesting that voters guided by an evaluation goal operate as motivated reasoners (Kunda 1987, 1990; Lodge and Taber 2000; Redlawsk 2002) directly influenced by the affect associated with the information they encounter. Lodge and Taber (2000) see motivated reasoning as resting on three bases: (1) hot cognition, where all information is affectively laden; (2) online evaluation, where new information is evaluated immediately, rather than stored for later consideration; and (3) a “How-do-I-feel?” heuristic mechanism for calculating the evaluation. The process of encountering information is one of instant evaluation, comparison to existing affect, and generation of an update to the existing affect. If new information is as expected, for example learning something positive about a liked candidate, processing of that information is quick and easy. On the contrary if new information is not as expected, motivated reasoning suggests a deeper processing of information in such a way that may provide a means of reconciling the lack of memory effects in the Lodge online studies with the strong memory effects I have found. In particular, motivated reasoners take longer to process information incongruent with their affective expectations, showing signs of a disconfirmation bias as they expend cognitive effort to overcome new information which does not fit in order to maintain their existing evaluation (Lodge and Taber 2000; Redlawsk 2002). During this “stop-and-think” process motivated reasoners may work through the unexpected incongruency in any number of ways including counterarguing, discounting the source or validity of the information, or bolstering existing affect by recalling reasons that they feel as they do in the first place (Lodge and Taber 2000). All of these reasoning processes, but especially the last, may require the use of memory in some form. The accessing and processing of memories may account for the additional time spent on incongruent information. Even though motivated reasoning assumes the use of online processing, memory may play an indispensable role in decisions made by motivated reasoners, especially when they encounter affectively incongruent information. If memory does indeed play a role during this process, evidence of the use of memory should be found in online processing subjects tasked with evaluating and voting for political candidates in a campaign simulation.
90
REDLAWSK
Hypotheses The processes described by motivated reasoning are not conscious ones. We are, of course, unable to observe directly what happens when new affectively laden information about a political candidate is encountered and must infer processing from the voter’s overt behaviors and later recollections. The disconfirmation bias reported in previous studies is a good example of this. Subjects spend more time examining incongruent information so we infer that they must be processing that information in some way different from the normal assimilation of congruent information. Lodge and Taber (2000) show that subjects spend longer on incongruent issue statements in the domain of affirmative action and gun control, and that that patterns of “thoughts” subjects report after viewing the arguments supports the idea that subjects bolstered their existing affect upon encountering incongruent arguments. In other work I have shown that encountering disliked information about liked candidates has a similar effect for online processors motivated toward evaluation but not for memory processors motivated toward accuracy (Redlawsk 2002). Hypothesis 1 looks at this process for both liked and disliked candidates for motivated reasoners: Hypothesis 1: Voters encountering new information that is affectively incongruent with existing candidate evaluations will “stop-and-think” about the information, thus taking longer to process it compared with affectively congruent information.
Processing incongruent information is effortful. As a result of that effort, memory links between candidates and the incongruent information may be developed and strengthened. Srull and Wyer (1989) argue that inconsistent behaviors are especially memorable under some circumstances because the process of thoughtfully considering those behaviors strengthens memory links in a way that consistent behaviors do not. Fiske and Taylor (1991) suggest a similar result when schema-inconsistent information is encountered. Hastie (1988) finds that extra attention is given to material that is inconsistent with an existing impression. Together these findings suggest that incongruent information is particularly memorable and may be readily available to memory. However, in particularly complex decision tasks and when no time constraints are placed on recall, other studies find that inconsistent material is less memorable (Wyer and Martin 1986; Bodenhasen and Wyer 1985). Thus, hypothesis 2: Hypothesis 2: Voters will be more likely to recall affectively incongruent information encountered during the election campaign than congruent information.
ROLE OF MEMORY IN VOTER DECISION MAKING
91
Hypothesis 2 appears to contradict the expectation that motivated reasoners recall affectively congruent information during the stop-and-think process when they process incongruent information. But recalling memorable incongruent information does not necessarily mean that congruent memories will be crowded out. In fact, it may be that the stop-and-think process has two effects: first to make memorable the incongruent information encountered and second to strengthen the memory links between the candidate and the congruent information recalled to bolster existing affect. If so, motivated reasoners who encounter incongruent information about candidates may have stronger links to all kinds of memories about the candidate, leading to hypothesis 3: Hypothesis 3: Encountering and processing incongruent information will increase the likelihood of reporting both congruent and incongruent memory for a candidate.
My final hypothesis is a direct test of the bolstering proposition. As voters encounter incongruent information they may recall previous congruent information in order to overwhelm the new information. Motivated reasoners are invested in their existing affect toward the candidate and can be expected to do what it takes to maintain that affect, even if it means finding ways to discount incongruent information. One of the best ways to do this is to simply overwhelm the new incongruent information with lots of reasons as to why the existing affect should not be changed. The more often this has to happen the stronger the links to the original reasons why the candidate is liked or disliked become. Rehearsal of memory strengthens it (Collins and Loftus 1975; Srull 1981). Thus, hypothesis 4: Hypothesis 4: Encountering and processing incongruent information will increase the likelihood of reporting congruent memories for a candidate.
Support for Hypothesis 4 would be strong evidence that motivated reasoners are using memory to help overcome incongruency, and may well help explain why memory matters in an election campaign even when voters are making online evaluations.
Method Dynamic Process Tracing The dynamic process tracing methodology used in this study has been described in detail elsewhere (Redlawsk 2001, 2002, 2004; Lau and
92
REDLAWSK
Redlawsk 1997, 2001a, 2001b, 2006; Lau 1995) and thus will only be summarized here. Subjects in this study were exposed to a simulated presidential primary election campaign in which there were either two candidates or four candidates running in the subject’s party. These in-party candidates were joined by either four or two candidates from the other party, so that there were always six candidates running in the two party primaries. The dynamic process tracing system presents information about the candidates on a computer screen in a design that allows the available information to change over time. Subjects learn about the candidates by clicking on boxes on the screen behind which are detailed statements about positions on issues, candidate personality and experience, interest group endorsements, polls, and more. As the campaign progresses the nature of the information changes, so that early in the campaign voters can learn a lot about personality and horserace information, but less about issue positions. Later come endorsements, polls, and the issues. In order to mimic the sometimes confusing and often unmanageable campaign environment the dynamic information board has the potential to overwhelm subjects with information. In this, the system models a real election campaign with its ever-changing information environment. While subjects are participating in the election campaign, the system tracks the type of information examined, the order of information search, the time spent on each item, and more. Subjects The study on which this chapter is based included 99 subjects recruited from central New Jersey in the fall of 1994.3 One of the experimental manipulations randomly placed half of subjects into a memory processing/ accuracy motivation condition, while the other half processed the campaign online. This chapter reports only on the 50 subjects who were in this latter condition, since motivated reasoning requires online processing as one of its underlying premises. Memory processors are in fact different, and do not show evidence of motivated reasoning (Redlawsk 2002); thus to include them in these analyses would be incorrect. The 50 online subjects examined 2,044 unique items of information for individual candidates from both their party and the other party during the campaign, for an average of just over 40 items per subject. Of these, 768 items could be coded as either congruent or incongruent with the overall evaluation of the candidate. Items that could not be coded were either viewed as affectively neutral by subjects, not recalled as having been examined at all, or if recalled, subjects could not say how the particular piece of information made them feel, so affect could not be ascertained. In order to focus
ROLE OF MEMORY IN VOTER DECISION MAKING
93
analyses on the congruency of information with respect to overall candidate affect, any item coded as neutral was dropped from the analysis. In addition, the first two items examined for each candidate were dropped to allow some time for affect to develop, since subjects started with no knowledge of any of the candidates.4 Procedure Subjects first completed a fairly standard political attitudes questionnaire and were provided with instruction on using the dynamic process tracing simulation on the computer. They then registered as either Democrat or Republican prior to the primary election, and could only vote for the candidates from within their chosen party. The candidates themselves were invented, but designed to realistically represent the ideological spectrum of their respective parties. By creating mock candidates crucial control was retained over the differences between subjects in prior knowledge of actual politicians. No subject knew anything about any of the candidates before the campaign began. Three manipulations were embedded in the primary. As noted earlier, the first was the processing manipulation. Instructions designed to stimulate memory processing were embedded in the general instructions one-half of the subjects read at the beginning of the simulation. The remaining subjects, in the online condition, were given the same set of general instructions without the specific paragraph warning memory-based subjects that they would be expected to remember what they saw. The second manipulation varied the number of candidates faced during the primary. Subjects in the high task demand manipulation faced four candidates in their party, with two candidates in the other party. Subjects in the low task demand condition faced two candidates in their own party and four in the other. Again half of subjects were randomly placed into each condition. The third manipulation, the poll interruption, was designed to test whether the time taken to make a decision would vary according to when the decision was made. One-third of subjects were interrupted after six and one-half minutes and asked to indicate their vote if the election were held at that point. After their preference was recorded they were returned to the campaign. A second third were interrupted with the same poll question after 13 minutes. The final third were not interrupted by a poll and did not express any preference until the campaign was over. When the election was over subjects voted for 1 candidate in their party and evaluated each of the 6 candidates on a standard 101-point feeling thermometer. Subjects then took a memory test (unexpected for those in the online condition) consisting of six pages, each headed by the name of a
94
REDLAWSK
candidate, the order of which was randomly determined for each subject. Subjects were instructed to list everything they could remember about each candidate, no matter how trivial. They were then asked to assess the affect associated with each memory, whether the memory made them feel “good,” “bad,” or “neutral” toward the candidate. Memories tagged as “good” or “bad” were considered affectively charged memories for these analyses. After an exercise to establish whether subjects would have changed their vote with more complete information, that is, whether they voted “correctly” (Lau and Redlawsk 1997), an extensive debriefing and cued recall procedure began. Subjects were shown the script of all information they examined during the campaign and asked to recall for each card they had viewed what they were thinking while learning the information, and to evaluate each bit of information as to whether its contents made the subject feel good, bad, or neutral about the candidate. Finally, the purposes of the study were explained and subjects were allowed to ask questions. The total time required for each subject was about two hours. Defining Information Congruency A key measure in this analysis is information congruency, the extent to which the affect associated with a new piece of information is congruent with existing affect toward the candidate to which the new information refers. This requires operationalizing affect toward candidates and affect toward new information. Because subjects were asked to rate each of the six candidates on a standard feeling thermometer, a simple measure of candidate affect is available.5 A liked candidate is one for whom the subject gave a feeling thermometer rating above the mean rating given to all six candidates. A disliked candidate is one receiving a rating below the mean.6 Affect toward new information about the candidates is measured by using a cued recall procedure in which subjects were shown the script recording their information search. Information that subjects said made them feel good about a candidate was coded positive and that which made them feel bad was coded negative, while neutral information was dropped from this analysis. Information congruency then simply crosses the 2 measures, so that congruent new information (in which affect for the new information matched affect for the candidate) was coded as 0 and incongruent coded as 1.7 Results Hypothesis 1 Little effort is needed for a motivated reasoner to update the online tally when there is nothing unexpected or threatening in the new information
ROLE OF MEMORY IN VOTER DECISION MAKING
95
that is encountered. However, as hypothesis 1 suggests, new affectively incongruent information may be a different story. In this case voters may stop and think in order to make sense of the unexpected negative information about a liked candidate or positive information about a disliked one. An initial OLS model (not shown) was constructed to test hypothesis 1, where the time spent reading a piece of information about a candidate is the dependent variable and the affective congruency of each information card is the independent variable of interest.8 Controls include the task demand and poll interruption experimental conditions, individual subject measures of political sophistication, age and education, and a measure of each subject’s reading ability.9 These covariates, all of which are constants within subjects, also serve the role of “dummy variables” to control for within-subjects effects created by the use of subject-candidate pairings for observations. In addition to individual subject differences the number of words per information card is controlled, since it obviously takes longer to read longer cards. Finally, an indicator of whether a piece of information was selected by mistake is included, since subjects generally spent very little time on information they did not intend to examine. This initial model failed to provide any support for the hypothesis and, in fact, showed no effects at all for information congruency. Because previous analysis of these data had shown significant effects for information congruency (Redlawsk 2002), a closer examination was made of both potential sophistication effects and the effects of candidate affect. The earlier analysis included only liked candidates; thus incongruent information was always negative. Because this chapter examines both liked and disliked candidates, it seems likely that an interaction between candidate affect and information congruency is present but not tested for in the initial model. In addition, following Lodge and Taber (2000), an interaction term for political sophistication and congruency is also included in the revised model, as is the three-way interaction between congruency, candidate affect, and sophistication. The revised model is reported in table 6.1. As expected key indicators of processing time include subjects’ reading ability and the number of words that need to be read. However, no main effects are seen for information congruency, candidate affect, or political sophistication. Instead, there is a complex interaction between all three terms. The nature of this interaction is made clear by figure 6.1, showing the mean adjusted processing times for political sophisticates and political novices by candidate affect. For both sophisticates and novices, and both liked and disliked candidates, the mean adjusted processing time for congruent information, ceteris paribus, is virtually the same. But incongruent information creates a very different story. Political novices show the same basic effects for both liked and disliked
96
REDLAWSK
Table 6.1 Information Congruency and Processing Time Information congruency (1 ⫽ Incongruent) Candidate affect (1 ⫽ Like) Political sophistication Information congruency * Candidate affect Information congruency * Political sophistication Information congruency * Political sophistication * Candidate affect Number of words in card Card chosen by mistake Age in years Education Reading ability Task demand condition Poll interruption condition Constant Adjusted R2 ⫽ .449
⫺.097
(.862)
⫺.139
(.670)
⫺.114 1.792
(.269) (1.123)
⫺3.005 ***
(.667)
3.317 **
(1.066)
.121 *** ⫺2.776 * ⫺.006 .295 * .037*** ⫺.478 .801 *** ⫺7.672 ***
(.007) (1.511) (.013) (.158) (.003) (.414) (.236) (1.574)
Note: Dependent variable is adjusted reading time for information. Table entries are unstandardized OLS regression coefficients, standard errors in parentheses, n ⫽ 543. * p ⬍ .1, ** p ⬍ .01, *** p ⬍ .001
candidates—incongruent information takes longer to process. In particular, when novices encounter positive information about a candidate they dislike, they stop and think to a much greater degree than when they encounter negative information about a candidate they like. Political sophisticates, on the other hand, appear to more or less ignore positive information about disliked candidates, spending less time processing it compared with congruent information. But they show the same effects as novices when it comes to processing incongruent information about liked candidates. Both groups show the effects that would be predicted if motivated reasoning processes are operating. While complicated, there is clear support for the underlying expectations of hypothesis 1, especially for liked candidates. Hypothesis 2 Hypothesis 2 predicts that incongruent information is more likely to be retained in memory and available for reporting than congruent information, due to both its novelty and the stop-and-think process it generates. The unit of analysis here is the same as for hypothesis 1: each piece of
ROLE OF MEMORY IN VOTER DECISION MAKING
Adjusted read time in seconds
11
97
Political novices
10
9
Disliked candidate
8
7
6
Liked candidate
5 Congruent
Incongruent Information congruency
Political sophisticates
Adjusted reading time in seconds
8
7
Liked candidate
6
5
Disliked candidate
4
3 Congruent
Incongruent Information congruency
Figure 6.1 Effects of Information Congruency on Processing Time.
information viewed. Across all 50 subjects 768 information cards were identified as either congruent or incongruent and 155 (20.2 percent) of those were represented in the memories subjects listed when asked what they could recall from the election.
98
REDLAWSK
A simple test of hypothesis 2 is to estimate the likelihood of a particular congruent or incongruent piece of information being recalled from memory. At first, this looks promising. A simple cross tabulation examining liked and disliked candidates separately and dividing subjects into novices and experts finds nothing much for novices for either liked or disliked candidates. Experts, though, are slightly more likely to report memories for congruent information about candidates they dislike (23.5 percent of congruent information viewed was recalled in the memory test) than incongruent information (20 percent), but the difference is not statistically significant. However, for liked candidates, incongruent information is significantly more likely to be reported (35.5 percent) than congruent information (20.9 percent) as was predicted (phi ⫽ .105, p ⬍ .10). However, many other factors affect the likelihood of memory recall. In particular, the time at which a piece of information is viewed may be correlated with recall, with information viewed toward the end of the simulation more likely to be recalled than information viewed early on. In this experiment, 13.8 percent of items viewed during the first third of the campaign were reported as memories, while 20.6 percent of the middle third and 22.6 percent of the last third were recalled (phi ⫽ .080, p ⬍ .10). An attempt was made to construct a logistic regression model with memory as the dependent variable and information congruency, expertise, time when viewed, and other controls as predictors. No model was identified that reached statistical significance, and in all versions of the model, the congruency variable itself was never significant. Hypothesis 2 cannot be supported. There is no evidence in these data that incongruent information is more likely to be recalled than congruent. The predicted advantage for incongruency, which might arise from the extra time spent on such information (Hastie 1988; Srull and Wyer 1989) is not found. It may be that as other research suggests, the incongruency advantage is eliminated in complicated decision environments (Fiske and Neuberg 1990). It seems likely that the dynamic information board environment with multiple candidates qualifies as complicated. Hypotheses 3 and 4 During the extra time motivated reasoners take processing incongruent information they are presumably making an effort to account for the unexpected. One result might be that the more incongruent information encountered the stronger the memory links between a candidate’s name and information about the candidate. Hypothesis 3 suggests that these stronger links should result in the greater likelihood of reporting candidate memories. To examine this hypothesis the unit of analysis is subject-candidate
ROLE OF MEMORY IN VOTER DECISION MAKING
99
% of candidates with memories
pairings, so that the focus is on the number of memories per candidates across all subjects. The analysis includes 212 candidates across the 50 subjects. At least 1 affectively charged memory was recalled for 74 of the candidates, with no such memories reported for the remaining 138.10 An initial look at the data provides support for the basic proposition that encountering incongruency enhances memory. Figure 6.2 displays the percentage of candidates for whom memories were reported by whether incongruent information about those candidates was encountered. Examining memories for all types of information viewed (positive, negative, and neutral) subjects reported at least 1 memory for just under 50 percent of the candidates for whom only congruent information was encountered. But memories were reported for 57 percent of candidates where subjects learned both congruent and incongruent information. The difference is clearer when examining only memories for affectively charged information. Memories for both kinds of affectively charged information were reported for 24.5 percent of congruent-only candidates, while 43.9 percent of candidates where both congruent and incongruent information was encountered generated at least one memory. And if only memories for congruent items are examined, encountering incongruent items appears to increase the likelihood of reporting congruent memories (24.5 percent for congruent only, 37.8 percent for congruent and incongruent information). Table 6.2 reports a series of logistic regression models predicting the likelihood that a given piece of candidate information will be recalled from memory. All of the models include a series of controls. These include affect toward candidates, since subjects typically report more memories for liked
60 50 40 30 20 10
Incongruent/Congruent
0
Congruent only All memory
Affective memory
Congruent memory
Figure 6.2 Information Congruency and Memory Reports.
100
REDLAWSK
candidates than disliked, along with an indicator of party status, since again subjects focus more and report more memories for in-party candidate compared to out-party. Individual subject differences including age in years and political sophistication, and a measure of cognitive (reading) ability, as well as the experimental manipulations are also included. Column1 tests hypothesis 3, the likelihood that any items viewed will be recalled from memory, congruent or incongruent. As subjects view more congruent information the greater is the likelihood that at least one memory will be Table 6.2 Information Congruency and Candidate Memory Predictors
% Incongruent information % Congruent information Any incongruent information (1 ⫽ yes) Any congruent information (1 ⫽ yes) Total information viewed Candidate affect (1 ⫽ Like) In-party candidate Age in years Political sophistication Reading ability Task demand condition Poll interruption condition Constant ⫺2LL Chi-Square Model Cox and Snell Pseudo R2 Correctly classified
Total Memory .039** (.012) .044*** (.009)
.159** (.049) .345 (.228) 1.190* (.479) ⫺.020 (.015) .206 (.250) .005 (.004) .017 (.427) .912*** (.266) ⫺6.489*** (1.512) 166.180 108.089 10 df p ⬍ .001 .399 79.7%
Congruent Memory
Incongruent Memory .082*** (.018)
.045*** (.009) 1.022* (.471)
.130** (.048) .525* (.235) .827* (.479) ⫺.017 (.015) .254 (.254) .006 (.004) .063 (.429) .837** (.269) ⫺6.521*** (1.543) 161.124 106.388 10 df p ⬍ .001 .395 82.1%
.866 (.991) .255*** (.081) ⫺.033 (.007) .390 (.774) ⫺.009 (.022) .029 (.376) ⫺.004 (.007) ⫺.970 (.699) .344 (.366) ⫺6.492** (2.483) 82.327 54.628 10 df p ⬍ .001 .227 94.3%
Note: Table entries are logistic regression coefficients, standard errors in parentheses, n ⫽ 212. * p ⬍ .10, ** p ⬍ .01, *** p ⬍ .001.
ROLE OF MEMORY IN VOTER DECISION MAKING
101
recalled. Subjects also clearly show a greater likelihood of reporting memories for candidates for whom they view greater amounts of incongruent information.11 This effect persists even controlling for the total amount of information viewed during the campaign, which includes congruent, incongruent, and affectively neutral information. These results support the expectations of hypothesis 3. However, a stronger test of the motivated reasoning process is needed, since it could simply be that viewing more information in general leads to a greater likelihood of reporting memories. The more interesting prediction is the expectation of hypothesis 4 that because encountering incongruent information generates bolstering by recalling congruent information from memory, congruent memory should be strengthened upon encountering incongruent information. The second data column of table 6.2 reports the test of this hypothesis. As the amount of congruent information increases, the likelihood of reporting a congruent memory for the candidate increases as well, controlling for the total amount of information encountered. But, controlling for this effect it turns out that encountering any incongruent information also increases the likelihood of reporting a congruent memory. The effect is not simply a result of encountering more information in general. The third data column of table 6.2 shows the effects of encountering congruent information on the likelihood of reporting memories for incongruent information. While viewing greater amounts of incongruent information increases the likelihood of reporting incongruent memories, as would be expected, encountering congruent information does not add anything to the model. Thus there is some confidence that the effect shown in column 2 is the effect expected if incongruent information enhances congruent memory, as suggested by hypothesis 4. A Role for Memory in Motivated Reasoning In a purely online process memory should play little or no direct role in evaluation (Lodge, McGraw, and Stroh 1989; Hastie and Park 1986). There is no particular advantage to maintaining memory stores of information already incorporated into the online tally. Yet elsewhere I have found evidence that in a campaign environment voters who use memory effectively make better decisions and that the affective value of memories can be used to predict the direction of the vote better than the online tally itself (Redlawsk 2001.) To explain this puzzle, I initially speculated that the desire to compare multiple candidates in a campaign environment places a value on memory, when information flows over time and effective use of memory might facilitate direct comparisons on the same set of issues. The series of studies by Lodge and his colleagues, which established online
102
REDLAWSK
processing in political science, did not attempt to model a campaign environment requiring a choice between competing candidates. While the speculation about the campaign environment seems reasonable (and is supported in more recent work, see Lau and Redlawsk 2006), a more complete understanding of the role of affect informed by motivated reasoning suggests another possibility. Memory may matter for online processors because the motivated nature of affective processing activates memories for the candidate under consideration when encountering incongruent information. The lack of memory effects in the initial studies of online processing that involved only a single political figure and a simple information environment may well be because the information processing requirements for evaluating a single politician are minimal and probably do not generate the “stop-and-think” process. Further, motivated reasoning suggests by its very nature that evaluation occurs over a period of time as new information is encountered. The original online studies by Lodge and colleagues did not have “time” element. Subjects simply examined a sheet of paper listing a series of statements about a politician. In such an environment memory will be of minimal importance, since all possible information is available and easily accessed at all times. Even if incongruent information were to be encountered, the subject need only quickly scan the sheet of paper to remind herself of the politician’s other positions with no need to resort to memory. To examine whether encountering incongruent information enhances the role of memory, table 6.3 replicates the analysis examining the relative importance of the online tally and memory in predicting candidate evaluation previously reported (Redlawsk 2002, Table 3). OLS regressions were estimated in which feeling thermometer scores representing global evaluations for each of the four candidates included in all experimental conditions were regressed on measures of the online tally and the affective value of the memories for each candidate.12 In this new analysis, subjects who viewed little or no incongruent information are analyzed separately from those viewing significant amounts of incongruent information. The results are interesting. In three out of four cases, memory is more important for those encountering incongruent information than it is for subjects who encountered only congruent information. Likewise, the online tally is more important in three out of four cases when little incongruency is encountered. The overall mean values for the standardized regression coefficients show a clear pattern: memory matters more when incongruency is encountered; the online tally is stronger when information viewed is congruent with candidate affect. However, the fact that memory is still stronger than the online tally even in the low incongruence environment suggests that the operation of motivated reasoning can explain only part of the reason memory matters in
ROLE OF MEMORY IN VOTER DECISION MAKING
103
Table 6.3 Effects of Memory and OL Tally on Candidate Feeling Thermometer Evaluation Candidate Evaluated Liberal Democrat
Moderate Democrat
Moderate Republican
Conservative Republican
Mean
Minimal Incongruent Information Online tally Memory
.44 .18
.28 .45*
.43* .43*
.02 .58*
.29 .41
.26 .33
.21* .48*
.17 .60**
.12 .56
Significant Incongruent Information Online tally Memory
⫺.16 .84***
Note: Table entries are standardized OLS coeffficients. Dependent variable is feeling thermometer rating. * p ⬍ .1, ** p ⬍ .01, *** p ⬍ .001.
multi-candidate environments. It may well be that the importance of memory is enhanced by both motivated reasoning and the candidate comparison process originally proposed to explain the findings. Discussion Lodge and Taber (2000; Taber, Lodge, and Weber, this volume) show that even in sub-second automatic response congruency matters. Affectively incongruent information is hard to process. Congruent information is easily assimilated, but incongruency sets in motion a more detailed level of processing as attempts are made to somehow reconcile the incongruency with previous evaluations. This stop-and-think process has consequences that can be seen not only at the sub-second level but also at a macro level. Motivated reasoners take longer to process campaign information when it surprises them, as can be seen in the clearly longer times required to read affectively incongruent candidate stands. Once the stop-and-think process begins memory processing is activated, despite the default tendency to process person-evaluation information online (Hastie and Park 1986). This memory processing leaves traces that can be discovered. Lodge and Taber (2000) and I (Redlawsk 2002) have speculated that processing incongruent information takes longer because of efforts to make sense of that information within the context of existing affect. This may occur in any one of several ways. Voters may counterargue the new information, developing reasons why it should be discounted, such as denigrating the
104
REDLAWSK
source, or minimizing its importance. Voters may also seek to bolster existing affect by recalling the reasons for the current evaluation. This study shows clear evidence of such a bolstering process. When voters in the campaign simulation encountered incongruent information they were more likely to report more congruent memories. This is exactly what would be expected if bolstering were occurring. Recalling reasons to support the existing evaluation is a memory process, and as each reason is recalled the links between it and the candidate are strengthened, increasing the likelihood that it will be recalled in the post-campaign memory test. The effect of this bolstering process may be to actually strengthen the existing evaluation even in the face of information that should cause the voter to at least think about reconsidering the existing candidate evaluation. This can be seen as a natural consequence of online processing. Each time previously stored information about a candidate is accessed from memory the online evaluation counter is updated, just as if it were the first time the information had been encountered. Thus each time the motivated reasoner accesses memory for congruent information to bolster an existing evaluation in the face of incongruency the evaluation is incremented, and the existing affective feeling is strengthened. Empirical evidence of this effect has been reported in experimental studies of both issue and candidate evaluation (Lodge and Taber 2000; Taber, Glather, and Lodge 2001; Redlawsk 2002). Normatively this process means that voters may not be model Bayesian processors, updating their priors in accordance with either positive or negative new information. The result may explain the obvious observation that voters are willing to forgive quite a lot of the candidate they already prefer, while discounting positive information for candidates they oppose. The problem comes, however, when failing to take new information into account results in lower quality decisions (Redlawsk 2002). Online processing by itself cannot account for the memory effects reported elsewhere (Redlawsk 2001). But taken as part of the tripartite process of motivated reasoning, the memory findings fit into place. Online processors do evaluate candidates “on the fly” and do update an evaluation counter that is easily queried when an evaluation is required. This is well established. But some of the information encountered during online processing is stored in memory and is accessible during and after the decision process. These memories become an important factor in evaluation and decision making when routine processing is interrupted by the unexpected. Affectively incongruent information is by definition unexpected, thus causing the motivated reasoner to stop and think and to use whatever is in memory to help make sense of a suddenly more complicated decision environment.
ROLE OF MEMORY IN VOTER DECISION MAKING
105
Notes 1. One study (Lodge, Steenbergen, and Brau 1995) did use two political candidates who were ostensibly running for the same office. Subjects viewed the information about these candidates on a one-page campaign fact sheet that showed a column for each candidate on a set of issue positions. The sheet was easy to process in a way that real election information usually is not, and subjects were not asked to “vote” for one of the candidates after learning the information. 2. Further analyses continue to show the robustness of this basic finding (Lau and Redlawsk 2006). Memory matters not only in predicting the direction of the vote but also its accuracy—i.e., the extent to which voters cast a correct vote (Lau and Redlawsk 1997). 3. Subjects were recruited primarily through organizations that were invited to provide members to participate in the study, in return for the member donating the $20 payment to the organization. These organizations included a local YMCA, a senior citizen center, a day care center, and other community organizations. Recruiting was done by the organization using parameters provided by the researchers. Subjects were told they would be participating in a study of campaigns that would include a mock presidential primary election campaign. 4. Analyses which retain the first two items show effects similar to those reported in this chapter, though not quite as strong. 5. While authors elsewhere in this volume (Marcus et al.; Isbell, Ottati, and Burns; Steenbergen and Ellis) make persuasive cases that affect is not fully described by a valence-oriented scale such as the feeling thermometer, for the purposes of this study we need only know the generalized direction of affective feelings towards candidates, i.e., whether subjects liked or disliked any particular candidate. Thus the feeling thermometer, as used here, is adequate to the task. 6. Using a mean split to differentiate between liked and disliked candidates adjusts individually for each subjects’ own propensity to rate candidates highly or lowly. Three candidates (one for each of three subjects) received a rating exactly at the subject’s mean candidate rating and cannot be assigned either like or dislike status. These candidates are not included in subsequent analyses. One potentially difficult assumption has to be made in this analysis because feeling thermometer ratings are collected at the end of the campaign. Candidate evaluation at the end is assumed to approximate evaluation during the campaign. This is justified on several grounds. First, in 85.2% of the cases where subjects were interrupted by a poll during the campaign (two-thirds of subjects were in this condition) the final vote was for the same candidate preferred in the poll. Thus in those cases it is clear that affect for those candidates was positive from early on in the campaign. Second, in all of the remaining cases where the vote changed, affect for the candidate preferred in the poll remained positive, but some other candidate was seen more positively, that is, no subject rated a candidate preferred in the poll below the mean of all candidate evaluations. Third, the process of using a mean split
106
REDLAWSK
7.
8.
9.
10.
11.
results in an analysis that is not sensitive to small changes in affect. Finally, in the extensive debriefing process, subjects generally reinforced the sense of the experimenter that disliked candidates (who are generally out-party candidates) are disliked from early on, and liked candidates are liked from early on. While a more direct measure of affect early in the campaign for both liked and disliked candidates would be preferable, the results of these analyses are, if anything, less strong than they would be if data were less “noisy.” As noted in Redlawsk (2002), there is some reasonable concern about the ability of subjects to accurately recall affect. To minimize problems, subjects were given an opportunity to express “don’t remember” when asked about their recall. Across all subjects about 17% of the items viewed were not recalled at all, and are not included here. Further, Thomas and Diender (1990) show evidence that affect can be recalled reasonably accurately. In the current dataset a direct test of accuracy is possible with a limited subset of items (about 10% of the total). For these items, accuracy ranges between 56% and 83%, and averages 67% (Redlawsk 2002, fn. 10.) The adjusted processing time measure was calculated using the time spent reading an item as recorded by the computer. The computer began measuring reading time when a subject accessed a piece of information about a candidate and ended when the subject clicked on a button to return to the scrolling information list. The raw measure thus contains both actual reading time and the physical time required to move the computer mouse and click on the button. To adjust for this, the raw reading time measure was regressed on the number of words in the candidate information card. These regressions were carried out within subjects. The resulting constant represents the average time required for each subject to read a card with zero words, thus approximating the time needed to handle the physical task of accessing the DONE button to end the card. This constant derived for each subject was subtracted from the raw reading time for each piece of information that subject accessed. This new adjusted reading time measure is used as the dependent variable in this analysis. Political sophistication is an index of political behavior, political interest, and political knowledge, all collected as part of a pre-experiment questionnaire. Reading ability is measured by the amount of time subjects took to read the provided instructions, as recorded by the computer; longer time means slower reading speed. The dependent variable for this analysis includes only the presence of memories for items viewed for which subjects reported either positive or negative affect, and could thus be coded as either congruent or incongruent items. Given the analysis in support of hypothesis 1, a model was developed including interaction terms between political sophistication, congruent and incongruent information, and candidate affect to test whether the patterns found in processing information also pertained to recall. The addition of the
ROLE OF MEMORY IN VOTER DECISION MAKING
107
interaction terms provided no improvement to the model (2diff ⫽ 4.583, 6 df, n.s.) 12. For a complete description of the online tally measure see Redlawsk (2001). The measure is designed to assess the affect subjects develop for each candidate as the campaign progresses. It is calculated by assessing the affect attached to each piece of information that subjects actually examined based on attitudes expressed in the pre-experiment questionnaire. The affect associated with memories is calculated based on subject reports of how each memory made them feel about the candidate with whom it is associated.
This page intentionally left blank
CHAPTER 7 FEAR AND LOATHING IN AMERICAN ELECTIONS: CONTEXT, TRAITS, AND NEGATIVE CANDIDATE AFFECT* Marco R. Steenbergen and Christopher Ellis
ffect-based models of political behavior have become increasingly complex. Not long ago, political psychology was dominated by bipolar conceptions of affect such as the familiar feeling thermometers. Such conceptions assume that movement toward one pole (e.g., positive affect) of necessity implies movement away from the other pole (e.g., negative affect). When researchers discovered that people can simultaneously experience positive and negative affect toward the same object and that these experiences had distinctive effects (Cacioppo and Berntson 1999; Cacioppo, Gardner, and Berntson 1997; Cacioppo et al. 1993; Gray 1982, 1987a, 1987b; Marcus, Neuman, and MacKuen 2000), bipolar, unidimensional conceptions of affect gave way to a two-dimensional model. This model, which distinguishes between positive and negative affect, is now widely accepted in political psychology (Abelson et al. 1982; Conover and Feldman 1986; Marcus and MacKuen 1993; Marcus, Neuman, and MacKuen 2000; Ottati, Steenbergen, and Riggle 1992).1 It may be time, however, to revisit the two-dimensional model. Recent research in psychology and political science suggests that it is sometimes necessary to discriminate between two different types of negative affect: anxiety (e.g., feeling afraid, anxious, uneasy, or worried) and aversion (e.g., feeling angry, bitter, contemptuous, disgusted, hateful, loathing, or resentful; Marcus et al., this volume; Marcus, Neuman, and MacKuen 2000; Morgan and Heise 1988; Nabi 1999; Nabi 2002). Each of these types of negative affect has unique effects on behavior and decision-making
A
110
STEENBERGEN AND ELLIS
processes. Further, they may also have unique antecedents. In this view, the two-dimensional model of affect may be insufficient: instead, we may need a three-dimensional model of emotions, encompassing positive affect and two distinct types of negative affect. In this chapter, we explore some aspects of this three-dimensional model. We ask what the distinctive predictors of anxiety and aversion are. Since anxiety is the more typical form of negative affect in the context of political evaluation (Marcus, Neuman, and MacKuen 2000), we are particularly interested in finding circumstances that cause aversion to be experienced distinctly from anxiety. We argue that there are specific and systematic conditions under which people are likely to experience aversion toward an object that is distinct from the experience of anxiety. Specifically, we argue that aversion is likely if a political candidate possesses traits that, in a particular context, pose a threat to a person’s well-being or are an affront to that person’s core beliefs and values. We illustrate our ideas through an analysis of affective reactions toward President Carter in 1980 and President Clinton in 1995. The Structure of Emotions Before we delve into our research question it is useful to provide a brief background on the three-dimensional affect model. One can think of this model as bridging two traditions in psychological affect research: circumplex models and functional emotion theories. Circumplex models postulate that there are two dimensions of the experience of emotion: valence and arousal. Valence pertains to the quality of the emotion (pleasant or unpleasant), while arousal describes the intensity of the emotion (mild or intense). Thus the circumplex model argues that people can simultaneously experience both positive and negative emotions, and can experience each with different intensity (Russell 1980; Tellegen, Watson, and Clark 1999a, 1999b; Watson and Tellegen 1985). The two-dimensional model of political affect directly corresponds to the circumplex model. Whereas the circumplex model emphasizes common dimensions that all emotions share, functional emotion theories stress the unique character of discrete emotions. Functional theories argue that emotions have inherently adaptive functions and that each emotion is associated with a distinctive goal that motivates a specific course of action. Thus sadness, anger, guilt, disgust, loathing, anxiety, pride, and enthusiasm, to name just a few examples, are conceptually distinct emotions with unique effects on motivation, information processing, and behavior (Frijda 1986; Izard 1977; Izard and Bartlett 1972; Lazarus 1991; Ortony, Clore, and Collins 1988; Scherer 1984; Tomkins 1962).
FEAR AND ANGER IN CANDIDATE EVALUATION
111
The three-dimensional model brings some of the nuances of functional theories into the circumplex model. It agrees with the circumplex model that emotions can be represented with reasonable accuracy in a small number of dimensions. It also agrees with functional theories, however, in arguing that two dimensions are insufficient to capture the qualities of different emotions. Rather, three dimensions (one of positive affect and two of negative affect) are required in many circumstances. Evidence for this three-dimensional structure comes from a number of directions. Morgan and Heise (1988) used factor analytic methods to demonstrate the existence of a third dimension. Analyzing a large battery of affect words, they not only identified the valence and arousal dimensions of the circumplex model, but also an additional dimension which they called potency or dominance (Mehrabian 1980). This third dimension differentiated between feelings of anger (an intense, highly potent, and mobilizing emotion) and fear (an emotion associated with discomfort, apprehension and, often, powerlessness). In this model, fear and anger are distinct classes of emotions with distinct effects on behavior and decision making. The distinction between fear and anger in the context of politics is most explicitly delineated in the Affective Intelligence Model (AIM). Marcus et al. (2000; this volume) show that emotional responses toward candidates sometimes take on a three-dimensional structure with enthusiasm, anxiety, and aversion dimensions. Anxiety is associated with the AIM’s surveillance system. It causes individuals to become more aware of their surroundings, in particular, novel or threatening circumstances (Gray 1982; Ohman 1993). In this model, anxiety stimulates a desire to more fully understand and analyze the source of a potential threat, thus promoting active learning and reasoned thought and decreasing reliance on habits and dispositions (Marcus and MacKuen 1993; Marcus, Neuman, and MacKuen 2000). Aversion is associated with the disposition system. It is an externally directed emotion that reinforces a desire to react against the cause of a negative stimulus or to punish those who could control or moderate the stimulus (De Rivera 1981). In the AIM, aversion stimulates, and is the product of, dispositions and habits (Marcus, Neuman, and MacKuen 2000) such as prior opinions and stereotypes (Bodenhausen, Sheppard, and Kramer 1994; Kunda 1990). But in many cases, anxiety (e.g., fear) and aversion (e.g., anger) are not always experienced as separate emotions. Oftentimes, an object triggers both of these emotions simultaneously and they act as a single, negative affect dimension, just as the circumplex model suggests. It is therefore important to understand when and why anxiety and aversion are experienced as distinct emotions. Psychologists have paid considerable attention to this
112
STEENBERGEN AND ELLIS
topic, but political psychologists have paid far more attention to the differential consequences of anxiety and aversion than to their differential antecedents. In this chapter, we use the insights from psychology so that we can begin to identify political conditions and processes that cause anxiety and aversion to be experienced as separate emotions. Antecedents of Anxiety and Aversion: Insights from (Political) Psychology Both anxiety and aversion are part of a broader class of negative emotions felt in response to people, events, groups, or other stimuli that a person perceives to be inconsistent with and threatening to his or her goals, beliefs, or desires (Smith and Lazarus 1993).2 Whether someone feels anger, fear, or simply negative affect in response to a perceived threat depends on familiarity with the source of the threat, personal importance of the belief or goal that is being threatened, and the individual’s perception of the cause and controllability of the threat. Anxiety Anxiety and fear are generally felt in response to unfamiliar, threatening circumstances that an individual perceives to be out of his or her control (Averill 1988; Epstein 1972; Frijda 1986; Izard 1977; Ohman 1993). The sources of the threat are ambiguous and uncertain, and the individual may be unable to interpret exactly how a threat will impact his or her personal well-being. Since anxiety-producing events have both ambiguous origins and uncertain effects, they are often perceived as arbitrary and novel disruptions to security and established order (Bowlby 1973; Foa and Kozak 1993; Gray 1987b).3 Anxiety-producing stimuli are not necessarily attributable to any particular cause and, as a result, individuals may feel as if they have few options available to remedy or cope with the threat. Thus anxiety is often associated with a feeling of helplessness or loss of control (Kemper 1991; Shaver et al. 1987; Smith and Lazarus 1993). Not only do individuals feel that they lack control over the anxiety-producing stimulus, they believe it is out of the control of others as well. Hence, anxietyproducing circumstances are often blameless (Lazarus 1991). The inability to assign blame is an important aspect of the feelings of helplessness and uncertainty that accompany anxiety (Gal and Lazarus 1975). Since anxiety can be felt in response to any unusual or threatening circumstance, it is a common—and often, rational—reaction (Lazarus 1991; Marcus, Neuman, and MacKuen 2000). In the AIM, anxiety is a function of the surveillance system that monitors the environment for novel or
FEAR AND ANGER IN CANDIDATE EVALUATION
113
threatening conditions. Feelings of anxiety are triggered if there is a mismatch between the current environment and the circumstances for which an individual’s (unconscious) habitual behaviors are well suited (Marcus and MacKuen 1993; Marcus, Neuman, and MacKuen 2000). The role of the surveillance system is to monitor the environment for any signs of novelty or threat, and to shift attention to reasoned learning and choice. The experience of anxiety, in general, leads to decreased reliance on habitual predispositions and routines and a greater desire to gather and process new information to evaluate a particular stimulus. Aversion Perhaps the most critical distinction between anxiety and aversion is the attribution of cause for the source of a threat ( Jones 1971; Weiner 1982). Whereas the experience of anxiety is often associated with general threats to well-being or novel circumstances outside of anyone’s control, aversion is a reaction to a diminishment of one’s well-being or an attack on one’s beliefs and goals for which others are perceived to be responsible (Kemper 1991; Lazarus 1991; Smith and Kluegel 1982; Weiner 1982). Aversion occurs if a person believes that he or she has been wronged or intentionally caused harm, or if an individual perceives that another person could have done something to ameliorate or eliminate a negative event but did not do so (Averill 1982; Nabi 1999, 2002). While anxiety is felt in response to a general or uncertain threat in one’s environment, aversion is felt in response to a perceived affront to an individual’s goals or beliefs (De Rivera 1981; Hampton 1978). The belief that others cause harm or control the sources of a harmful event or threat is a central component of aversion. Aversion is also likely to be felt when one’s core beliefs, values, or lasting attachments are attacked. Anger is associated with threats to one’s “ego identity” (Lazarus 1991), especially if these threats are severe and direct (Frijda 1986; Nabi 1999; Tangney et al. 2001). Familiarity with the source of a threat also distinguishes between anxiety and aversion, with aversion more likely to be experienced if the threat stems from a stimulus with which the individual is well acquainted (Oatley 2000). Because aversion is most often reserved for severe threats or affronts, it is a more forceful and less common emotion than fear. As discussed earlier, Morgan and Heise (1988) find that the distinction between words used to represent “fear” and those used to represent “anger” rests primarily on their potency. Highly potent emotion words, that is, words that are intense and empowering, signify the presence of anger and the desire to punish the source of anger (“fight”). In contrast, less potent emotion words, that is,
114
STEENBERGEN AND ELLIS
those associated with a sense of helplessness, signify fear and are often indicative of a “flight” motive. The AIM agrees with these conceptions of aversion, suggesting that aversion is rare and is felt only in response to familiar “punishing” stimuli. Typically, aversion indicates “moral disapprobation.” Because it is reserved for familiar stimuli, the AIM argues that aversion is governed by the disposition system, the emotional subsystem that controls habitual behavior in response to known circumstances (Marcus et al., this volume; Marcus, Neuman, and MacKuen 2000). Aversion in this guise is different from anxiety: rather than a reaction to a novel and uncertain threat, aversion is formed from a reaction to a reprehensible object or person with which one is already familiar. Negative Affect Oftentimes, feelings of anxiety and aversion toward a stimulus cannot and should not be differentiated (Berkowitz 2000). These feelings may simply be part of a broader category of negative affect produced by unpleasant circumstances. In other words, for some stimuli, under some conditions, only the negative valence of the emotion—not the specific type—may be relevant. This is often the case when dealing with stimuli that are negative but present a relatively distant or secondary threat to an individual’s well-being, or stimuli that are perceived to have little personal relevance. The tendency for anxiety and aversion to behave as a single dimension of negative affect is especially true when dealing with stimuli that are unfamiliar to an individual, since that individual may not be very engaged with the stimulus (either cognitively or affectively). Individuals may still experience negative affect, having a general feeling of dislike toward a particular stimulus, but the unfamiliarity with and relative unimportance of it makes feeling more potent, targeted, emotions such as anger unlikely. Marcus and his colleagues suggest that while a general feeling of anxiety and negative affect is quite common in politics, feelings of true aversion are relatively rare and are only felt in response to familiar stimuli in certain contexts (Marcus et al., this volume; Marcus, Neuman, and MacKuen 2000). When aversion is not experienced as a distinct emotional reaction to a stimulus, the AIM treats all negatively valenced emotions as a single dimension. The model labels this dimension “anxiety,” reflecting the belief that aversion is a relatively rare phenomenon. The AIM suggests that anxiety can be felt in response to any novel, unusual, or potentially threatening stimulus, while aversion is reserved for more severe and familiar affronts to one’s well-being or core values and beliefs. Negative affect, viewed as anxiety in the broad sense, is expected to cause the suspension of disposition
FEAR AND ANGER IN CANDIDATE EVALUATION
115
and to promote reasoned thought, judgment, and choice (Marcus and MacKuen 1993; Marcus, Neuman, and MacKuen 2000). Hypotheses As our discussion has shown, anxiety is the more common of the two negative emotions in politics, including the context of candidate evaluation. For aversion to be felt as a distinct emotion, separate from anxiety, an individual must perceive that a political figure’s actions affront the individual’s most important beliefs and values, or that the figure’s actions (or inactions) cause harm to the person’s well-being. Aversion is the product of more than simply being unsettled by a politician’s leadership style, issue positions, or political party; it comes about when politics is perceived as a direct attack on one’s well-being and identity. We now consider some contexts in which this may be the case. Aversion and the Economy No issue has a stronger impact on one’s well-being as the state of the economy. Thus we might expect that a weak economy, which could jeopardize one’s job and standard of living, is an important context in which individuals experience aversion toward a political figure. But this is only true if economic malaise is attributed to the actions or inactions of this figure. If a person believes that politicians did not do much to cause or cannot do much to improve a threatening economic situation, then this person is likely to experience anxiety but, in the absence of blame, not aversion. Aversion toward political figures is also unlikely if people blame themselves for economic hardships, as is quite common in the United States (Brody and Sniderman 1977; Lau and Sears 1981). Thus aversion toward a political figure should occur only if that figure is held responsible for the economic threat. Holding a politician responsible for economic malaise entails two things. First, an individual must feel that an economic problem can be controlled through properly designed policies. Second, he or she must feel that a politician’s actions do not constitute such properly designed policies. That is, these actions have either caused the problem or have failed to contribute to its solution. While it is conceivable that citizens, especially political sophisticates, assess specific policies in order to evaluate if they were proper in terms of addressing the problem, it is more likely that citizens evaluate the actions or inaction of politicians in terms of global traits. These traits serve as heuristics about the quality of actions taken by the politician (Abelson et al. 1982; Barber 1985; Kinder 1986; Markus
116
STEENBERGEN AND ELLIS
1982; Rahn et al. 1990; Rapoport, Metcalf, and Hartman 1989). Positive trait judgments mean that the person is satisfied with the politician’s conduct, while negative trait judgments mean that the person is dissatisfied. In the context of economic policy, the most important trait should be leadership ability, as this signals the ability or inability of a politician to give direction to policy and to solve economic problems. In the specific context of the American economy, aversion resulting from a poor economic climate will most likely be felt toward the incumbent president. There are two reasons for this. First, presidents are familiar targets and, as we have seen, there is a positive correlation between aversion and familiarity. Second, if anyone is believed to have the power to control economic outcomes, it is the president (or, more generally, the executive). Indeed, it is this belief of presidential control over the economy that underlies economic voting in elections (Fiorina 1981; Lewis-Beck 1988). Thus, our first hypothesis is as follows: Hypothesis 1: Aversion toward an incumbent president is likely to be experienced if the economy is weak, if a citizen thinks that the government has control over economic outcomes, and if that citizen believes that the president is a weak leader.
Aversion and Moral Conduct A second cause of aversion is when a target’s actions and beliefs are a direct affront to our own cherished beliefs and values. Feeling distant from a political figure on the issues may stimulate anxiety or negative affect, but since specific political issues (and politics in general) are of limited salience to most, it is not necessarily the case that disagreeing with a political figure on issues will lead to the experience of aversion as a distinct emotion. But if a politician’s actions are thought to flout values (e.g., of morality, equality, or freedom) that a person feels strongly about, then that person may feel under attack and aversion will be the likely result. In this circumstance, the politician is perceived to attack values and principles that are essential to the individual, being a central part of that individual’s belief system and identity. Our focus in this chapter is on the moral conduct of political figures. As Lakoff (1997) argues, morality plays a critical role in American political discourse, more so perhaps than other values. Allegations of moral misconduct by a politician are likely to be perceived as an attack on the “moral fabric” of American society, especially if this politician is seen as a role model, as most Americans see, for example, the president.4 Thus perceived immorality of a president may well trigger feelings of aversion. This should be particularly true for people who hold traditional moral values, as these
FEAR AND ANGER IN CANDIDATE EVALUATION
117
individuals are likely to place the greatest weight on morality. Hence, our second hypothesis is: Hypothesis 2: Aversion toward the president (or another political figure) is likely to be experienced if he or she is perceived as immoral. This aversion should be the greatest among individuals who embrace traditional moral values.
Data and Methods To test our hypotheses, we use data from the 1980 and 1995 American National Election Studies (NES). The 1980 data provide a good way to test the first hypothesis, since the presidential elections in that year took place in the context of a recession. The 1995 data allow us to test the second hypothesis, since allegations of moral misconduct by President Clinton received a great deal of attention around this time, largely due to the Whitewater investigation. Apart from their excellent match to the hypotheses, these NES data also contain extensive batteries of affect items, which allow us to construct more reliable indicators of anxiety and aversion.5 Our analytic approach is different from most studies on affect. As discussed earlier, much of this research is based on the circumplex model and induces the structure of emotions from data reduction techniques such as factor analysis. Thus, whether aversion is distinct from anxiety depends on whether negative affect items for political figures break down into two factors that can be interpreted as anxiety and aversion. We refrain from this approach, in part because it has been done elsewhere (Marcus et al., this volume; Marcus, Neuman, and MacKuen 2000) and in part because the NES does not always contain sufficient items to perform a reliable factor analysis. As is well known, how many factors can be extracted depends on the number and variety of items (Gorsuch 1983). When the number of affect items is small, or when they are predominantly drawn from one affect category, as is often the case with the NES, then it is almost impossible to extract more than a positive and negative affect factor. Our approach is different. Instead of inducing anxiety and aversion dimensions, we impose them on the data by considering the semantic content of affect words, relying on the criteria laid out by Marcus et al. in this volume. We conceive of anxiety in terms of the affect words “afraid,” “anxious,” “uneasy,” and “worried.” Similarly, we conceive of aversion in terms of the affect words “angry,” “bitter,” “contempt,” “disgusted,” “hatred,” and “resentful.” By operationalizing anxiety and aversion in this manner, we can see if their antecedents are consistent with our hypotheses. Of course, we should remain mindful that these two dimensions may not operate independently. Thus we use Zellner’s (1962) seemingly unrelated
118
STEENBERGEN AND ELLIS
regression approach, allowing the residuals for anxiety and aversion to be correlated. Results We present our results in two parts. First, we perform an analysis of the 1980 NES data to test our first hypothesis concerning the relationship between aversion and the state of the economy. Next, we perform an analysis of the 1995 NES data to test our second hypothesis concerning the relationship between aversion and perceptions of morality in politicians. Affective Responses toward Carter in 1980 Context The 1980 presidential elections occurred during one of the worst economic recessions in recent history. The unemployment rate hovered around 7.1 percent, an increase of about 22 percent compared to 1979 and one of the highest unemployment rates since World War II. At the same time, inflation in 1980 was at one of the highest levels since World War II, measuring over 12 percent (and only slightly down from 1979). To make things worse, GDP growth was negative (⫺.3 percent) in 1980. The economy was contracting and Americans felt it on a daily basis through job loss or through the prices they had to pay in stores. In this context, President Carter was seeking reelection against Ronald Reagan, a battle that he lost and that would bring the Republicans back into the White House for the next 12 years. The Republicans clearly blamed the Carter administration for the poor economy. Reagan’s rhetorical question, “are you better off now than four years ago,” sought to persuade Americans that economic outcomes could be controlled by the president and that Carter had failed the American people in this domain. If our first hypothesis is true, Americans who accepted the Republican argument should have felt aversion toward Carter. Thus the 1980 election provides an excellent test bed for hypothesis 1, although it should be kept in mind that the effective sample size is small. Dependent Variables The 1980 NES contained several affect items for President Carter. Among these items are two measures of anxiety (“afraid” and “uneasy”) and two measures of aversion (“angry” and “disgusted”). The items have a dichotomous format: Think about Jimmy Carter. Now, has Carter—because of the kind of person he is, or because of something he has done—ever made you feel [EMOTION]?
FEAR AND ANGER IN CANDIDATE EVALUATION
119
In the third wave of the 1980 panel, on which we focus here, the two anxiety items have a tetrachoric correlation of .798, yielding a standardized scale reliability of .888.6 The two aversion items are correlated at .831, producing a standardized reliability of .908. The correlation between the anxiety and aversion dimensions is .507.7 Both of these dimensions are scaled to a 0–1 range, with higher values indicating higher levels of anxiety and aversion, respectively. Key Predictors and Moderator Variables The key predictor consists of assessments of Carter’s leadership ability. This is a four-item composite consisting of assessments of whether the phrases “weak” (reversed), “inspiring,” “provides strong leadership,” and “solve[s] our economic problems” describe Carter “extremely well,” “quite well,” “not too well,” or “not well at all.” In the third wave of the 1980 panel, these items form a reliable scale (Cronbach’s alpha ⫽ .854). We have scaled this predictor to a 0–1 range, with high scores indicating more positive assessments of Carter’s leadership qualities. Hypothesis 1 stipulates that perceptions of political control over the economy are a key condition for experiencing aversion. We assess such perceptions via the following question: “Do you think something can be done to deal successfully with the problem of unemployment, or do you feel that we’ll have to continue living with unemployment because not much can be done about it.” This question was asked in the second wave of the 1980 panel and has two response options—something can be done, chosen by a majority of the respondents, and nothing can be done. Given the high level of unemployment that characterized the economy in 1980, this seems a valid indicator of the attribution of responsibility. Thus, respondents who felt that something could be done about unemployment and who felt that Carter was lacking in leadership should register the highest levels of aversion toward the president. In addition to these variables, our seemingly unrelated regression model also considers other traits that may produce affective respondents to Carter (the respondent’s assessment of Carter’s morality and power hungriness), the distance from Carter’s issue positions, prospective evaluations of the economy, partisanship, and ideology. All of these measures come from the same wave as the affect measures (except for issue distance, which comes from the second wave) and are scaled to a 0–1 range, as described in the appendix. We also include demographic variables, which we leave in their original metric and which include age, education, income, race (1 ⫽ white), and sex (1 ⫽ female). Findings To what extent do the 1980 NES data support our first hypothesis? Let us begin simply by considering the impact of perceptions of Carter’s leadership.
120
STEENBERGEN AND ELLIS
The second and third columns in table 7.1 show the estimates from a seemingly unrelated regression analysis in the full sample. This table reveals a powerful effect of leadership on both anxiety and aversion. The effect on aversion is striking, however, when compared to other predictors. For a unit increase in the rating of Carter’s leadership qualities, aversion is expected to drop by .7 points or 70 percent of its range. Clearly, in the context of the 1980 economy, anger was much more likely if people
Table 7.1 Predicting Anxiety and Aversion toward Carter in 1980 Predictors
Issue distance from Carter Prospective economy Morality of Carter Leadership of Carter Power hungriness Carter Partisanship Ideology Age Education Income Female White Constant Pseudo R2 Residual correlation
Full Sample (N ⫽ 230)
Little Control Over Economy (N ⫽ 52)
Some Control Over Economy (N ⫽ 174)
Anxiety
Aversion
Anxiety
Aversion
Anxiety
.369* (.152) ⫺.022 (.074) ⫺.147 (.120) ⫺.369** (.138) ⫺.200* (.082) .198* (.079) ⫺.055 (.118) .002 (.002) .019⫹ (.011) .006 (.004) .075⫹ (.045) .042 (.104) .280 (.193)
.480** (.155) .009 (.075) ⫺.071 (.123) ⫺.706** (.141) ⫺.245** (.084) .011 (.081) ⫺.209⫹ (.120) ⫺.005** (.001) ⫺.012 (.011) ⫺.004 (.004) ⫺.034 (.046) .099 (.106) 1.354** (.197)
.268 (.261) .138 (.146) ⫺.518⫹ (.263) ⫺.353 (.301) .145 (.165) .329⫹ (.194) ⫺.090 (.213) .004 (.003) .022 (.021) ⫺.004 (.009) ⫺.050 (.089) .011 (.202) .309 (.387)
.712* (.344) .137 (.192) .240 (.347) ⫺.185 (.397) ⫺.380⫹ (.217) .020 (.256) .156 (.281) ⫺.012** (.004) ⫺.020 (.028) ⫺.003 (.012) ⫺.030 (.118) .152 (.266) .826 (.510)
.331⫹ (.182) ⫺.104 (.087) ⫺.073 (.133) ⫺.299⫹ (.159) ⫺.312** (.098) .186* (.087) ⫺.073 (.135) .001 (.002) .026* (.013) .008 (.005) .109* (.052) .007 (.125) .312 (.224)
.463** (.175) ⫺.009 (.083) ⫺.096 (.128) ⫺.789** (.152) ⫺.228* (.094) .036 (.084) ⫺.310* (.130) ⫺.005** (.002) ⫺.007 (.013) ⫺.004 (.005) ⫺.029 (.050) .109 (.120) 1.417** (.215)
.330
.366
.388
.301
.345
.430
.302**
.417**
Aversion
.248**
Note: Table entries are maximum likelihood seemingly unrelated regression estimates with estimated standard errors in parentheses. The top seven predictors have all been scaled to a 0–1 range. * p ⬍ .05, ** p ⬍ .01, ⫹p ⬍ .10 (two-tailed).
FEAR AND ANGER IN CANDIDATE EVALUATION
121
viewed Carter’s leadership as weak. Although perceiving Carter to be a poor leader during these poor economic times led to both anxiety and aversion, its effects are particularly powerful in predicting aversion. Table 7.1 reveals another pattern. Anxiety and aversion share only a limited number of common predictors, including leadership as well as perceptions of Carter’s power hungriness and issue distance. Once these predictors have been controlled for, the residual correlation between these two affective dimensions is a modest .3. This suggests that anxiety and aversion possess sufficient unique qualities that people can experience them independently. This is just as the three-factor model assumes. While table 7.1 provides initial evidence for hypothesis 1, it does not explicitly consider the role of causal attribution. The idea of causal attribution suggests that aversion will be most commonly felt among citizens who feel that the shortcomings of a particular political figure are somehow responsible for the plight that they face. In this case, perceptions of poor leadership should cause anger only in those individuals who believe that something can be done to improve economic outcomes. To examine this idea, we split the sample into two groups, those who think that little can be done about unemployment (the “little control over economy” group) and those who think that something can be done (the “some control over economy” group). The results are reported in the last four columns of table 7.1. In the “little control over economy” group, we see that perceptions of Carter’s leadership no longer predict aversion. However, the story is much different for individuals who believe that unemployment can be controlled. For this group, leadership is highly significant and one of the strongest predictors of aversion. Further, its effects on aversion are nearly three times as large as its effects on anxiety for this group. This shows that aversion is indeed contingent on assignment of blame.8 Only if an individual believes that something can be done about the economy will perceived weak leadership during a time of economic crisis produce aversion.9
Summary These analyses show that aversion toward President Carter was a function of evaluations of his leadership qualities during a time of economic crisis. However, this relationship was contingent on perceptions of economic control: only for those who believed that something could be done about unemployment was aversion a function of perceptions of Carter’s leadership. Thus the results lend support for our first hypothesis. Concern about one’s well-being can lead to aversion toward a political candidate. But only if one is able to link this threat to well-being to the actions or inaction of a political figure is aversion likely to manifest itself.
122
STEENBERGEN AND ELLIS
Affective Responses toward Clinton in 1995 Context The year 1995 was marked by extensive news coverage of the Whitewater scandal. This scandal centered about the allegation that President Clinton and his wife had illegally diverted funds from the Arkansas Whitewater Development Corporation to Clinton’s gubernatorial campaign fund in the 1980s. A formal investigation of these allegations had begun in 1994, but it picked up speed in 1995 when the new, Republican-dominated Congress decided that it was not satisfied with earlier investigative work and started a new set of hearings. At the same time Kenneth Starr, who in 1994 had become the Independent Council in charge of the Whitewater investigation, interviewed the Clinton’s and subpoenaed a large number of their friends and associates. This started a drawn-out process of ever more encompassing accusations of Clinton that would culminate in the attempt to impeach him in 1998. The Whitewater investigations followed on the heels of the 1992 campaign, which had also raised issues about Clinton’s morals, that time on grounds of alleged extramarital affairs. The new allegations probably rekindled memories of 1992, but even if they did not, Clinton’s critics—mostly, but not exclusively Republicans—were quick to point out that Whitewater was yet another piece of evidence that the president was an immoral man. Moreover, the critics argued that Clinton’s alleged morality should concern the American people, since the president is a role model for all of society and, especially, new generations. Americans who accepted these arguments likely felt that Clinton’s actions posed an affront to traditional moral values. To the extent that citizens cared deeply about these traditional values, the perception of Clinton as immoral should have triggered aversion toward Clinton. Dependent Variables The 1995 NES contained an unusually large battery of affect items, which (for Form A) break down into three dimensions for Clinton—enthusiasm, anxiety, and aversion (Marcus et al., this volume; Marcus, Neuman, and MacKuen 2000). The anxiety dimension is composed of three items: afraid, anxious, and worried. The aversion dimension has six items: angry, bitter, contempt, disgusted, hatred, and resentful. The question format for these items was as follows: How often would you say you’ve felt [EMOTION]—very often, fairly often, occasionally, or rarely?
The respondents could also volunteer that they had never experienced the emotion in question.10 We summed all of the anxiety items and all of the
FEAR AND ANGER IN CANDIDATE EVALUATION
123
aversion items, rescaling the resulting variables to a 0–1 scale with higher values indicating higher levels of anxiety/aversion. The average polychoric inter-item correlation for the anxiety items is .619, producing a standardized reliability of .830. The average polychoric inter-item correlation for the aversion items is .556, producing a standardized reliability of .883. The correlation between these two scales is .564. Key Predictors and Moderator Variables The key predictor of anxiety and aversion is Clinton’s perceived morality. Respondents were asked “does the phrase ‘moral’ describe Bill Clinton extremely well, quite well, not too well, or not well at all?” We reversed the coding on this item and scaled it to a 0–1 range. The key moderator variable is moral traditionalism. The 1994 panel component of the 1995 data contained four agree-disagree type items tapping this construct: (1) “the newer lifestyles are contributing to the breakdown of our society,” (2) “the world is always changing and we should adjust our view of moral behavior to those changes” (reversed), (3) “this country would have many fewer problems if there were more emphasis on traditional family ties,” and (4) “we should be more tolerant of people who choose to live according to their own moral standards, even if they are different from our own” (reversed). These four items form a reasonably reliable scale (Cronbach’s alpha ⫽ .689). For purposes of the analysis, we use a median split to distinguish between those who score low on moral traditionalism and those who score high. In addition to these variables, our models include perceptions of Clinton’s leadership, economic evaluations, issue distance from Clinton, partisanship, and ideology, all of which are scaled to a 0–1 range (see appendix). We also include the following demographics, which retain their original metrics: age, education, income, race (1 ⫽ white), and sex (1 ⫽ female). Findings Columns two and three in table 7.2 show the results from a seemingly unrelated regression analysis of anxiety and aversion toward Clinton. These results show that perceptions of Clinton’s morality are indeed related to aversion toward him. Those who questioned Clinton’s morality tended to feel more aversion. Of the predictors, morality has the most powerful impact on aversion. Importantly, questions about Clinton’s morality seem to have had an impact only on aversion: there is no evidence of a relationship between morality and anxiety. Instead, anxiety appears to have been the product of issues, leadership, and partisanship. This makes theoretical sense: anxiety toward Clinton seems to be a byproduct of “standard” political factors (being of a different party than Clinton, not sharing his
124
STEENBERGEN AND ELLIS
policy positions, and the like). But Clinton’s perceived immorality moves beyond “politics as usual” to represent an affront to one’s own values. In this circumstance, aversion is the result. Several other things can be noted about the results. As with the 1980 data, there are relatively few common predictors of anxiety and aversion.11 Partisanship emerges as a common predictor, with Republicans feeling both more anxious and more averse toward Clinton than Democrats. Similarly, perceptions of Clinton’s leadership predict both anxiety and aversion, something that we also observed for Carter in 1980. Beyond this, there are no common predictors.12 Table 7.2 Predicting Anxiety and Aversion toward Clinton in 1995 Predictors Full Sample (N ⫽ 109)
Issue distance from Clinton Prospective cconomy Morality of Clinton Leadership of Clinton Partisanship Ideology Age Education Income Female White Constant Pseudo R2 Residual correlation
Low Moral Traditionalism (N ⫽ 63)
High Moral Traditionalism (N ⫽ 46)
Anxiety
Aversion
Anxiety
Aversion Anxiety
Aversion
.210** (.077) ⫺.046 (.057) .119 (.099) ⫺.339** (.096) .135* (.064) .020 (.089) .001 (.001) .017 (.014) ⫺.003 (.004) .070⫹ (.038) ⫺.021 (.062) .075 (.159)
.028 (.066) ⫺.041 (.049) ⫺.204* (.086) ⫺.179* (.083) .151** (.055) .025 (.077) .001 (.001) .006 (.012) ⫺.005 (.003) ⫺.054⫹ (.033) ⫺.005 (.054) .282* (.138)
.154 (.108) ⫺.105 (.088) .136 (.166) ⫺.372** (.138) .070 (.088) ⫺.046 (.105) ⫺.002 (.002) .001 (.018) ⫺.005 (.005) .009 (.054) ⫺.037 (.074) .454* (.220)
.072 .090 (.086) (.119) ⫺.021 ⫺.059 (.070) (.078) ⫺.170 .242⫹ (.133) (.127) ⫺.227* ⫺.472** (.110) (.161) .121⫹ .205* (.070) (.104) .063 .007 (.084) (.171) .001 .004* (.001) (.002) .015 .032 (.015) (.020) ⫺.005 .005 (.004) (.006) ⫺.007 .165** (.043) (.057) ⫺.004 .008 (.059) (.121) .222 ⫺.318 (.176) (.232)
.014 (.119) ⫺.082 (.078) ⫺.300* (.127) ⫺.021 (.161) .281** (.104) ⫺.156 (.171) .002 (.002) ⫺.016 (.020) ⫺.001 (.006) ⫺.093 (.057) ⫺.078 (.121) .427⫹ (.232)
.371
.166
.285
.344**
.437 .552
.539
.466 .236
Note: Table entries are maximum likelihood seemingly unrelated regression estimates with estimated standard errors in parentheses. The top seven predictors have all been scaled to a 0–1 range. * p ⬍ .05, ** p ⬍ .01, ⫹p ⬍ .10 (two-tailed).
FEAR AND ANGER IN CANDIDATE EVALUATION
125
While the evidence so far suggests that perceived immoral conduct triggers aversion, we have not yet considered the role of moral traditionalism. Moral misconduct should cause aversion only if it is an affront to core values. Moral traditionalists should thus be the most likely to experience aversion because traditional morality is so important to them. Those who score low on moral traditionalism should experience much less aversion over perceived moral misconduct, either because they care less about morality or because their standards of morality are more lenient. The last four columns in table 7.2 break down the analysis by moral traditionalism, using a median split of the moral traditionalism scale. The evidence in this table indeed hints at the possibility that moral traditionalism moderates the effect of perceptions of Clinton’s morality on aversion. In the group that scores low on moral traditionalism, aversion is only weakly related to Clinton’s morality and this effect is not statistically significant. In this group, the experience of aversion (and, for that matter, anxiety) is best predicted by perceptions of Clinton’s leadership. However, when we turn to those who score high on moral traditionalism we see that morality becomes a very powerful predictor of aversion. Not only is this predictor statistically significant, it has the strongest impact of all of the significant predictors.13 This is precisely what the second hypothesis predicts.14 Summary The analysis of the 1995 NES data shows that aversion toward President Clinton was a function of evaluations of his morality. However, this seems true the most for moral traditionalists who should have been particularly offended by allegations that Clinton had behaved unethically. Thus the results lend some degree of support to our second hypothesis. More generally, they suggest that aversion can be the result of an affront to deeply held values and beliefs. Conclusions Models of candidate evaluation increasingly recognize the importance of emotions (Abelson et al. 1982; Conover and Feldman 1986; Marcus and MacKuen 1993; Marcus, Neuman, and MacKuen 2000; Ottati, Steenbergen, and Riggle 1992). Most research to date has worked off the circumplex model, which distinguishes between positive and negative affect (Abelson et al. 1982; Cacioppo and Berntson 1999; Cacioppo, Gardner, and Berntson 1997; Cacioppo et al. 1993; Conover and Feldman 1986; Marcus and MacKuen 1993; Ottati, Steenbergen, and Riggle 1992; Russell 1980; Tellegen, Watson, and Clark 1999; Watson and Tellegen 1985). However, researchers argue increasingly that it may be necessary to differentiate
126
STEENBERGEN AND ELLIS
between anxiety and aversion within the class of negative emotions (Marcus et al., this volume; Marcus, Neuman, and MacKuen 2000; Morgan and Heise 1988; Nabi 1999, 2002). In this chapter, we have taken up the suggestion that there are three affective dimensions. We have analyzed the differential predictors of anxiety and aversion, focusing in particular on the conditions under which aversion is likely to be felt toward political candidates. We argue that aversion toward political figures is likely under (at least) two conditions: when there exists a threat to a person’s (or society’s) well-being that a citizen attributes directly to the poor performance of a political figure, or when a political figure’s behavior poses a direct affront to a citizen’s system of core beliefs. Using these general preconditions for aversion, we have developed more specific hypotheses that call attention to the interplay between context and candidate traits. Specifically, we have considered the impact of perceptions of leadership when there is an economic malaise, as well as the impact of perceptions of morality when there is a scandal. In 1980, when the American economy was doing poorly, aversion was likely among those citizens who viewed President Carter to be a poor leader and who felt that something could be done to control economic conditions. Conversely, in 1995, when President Clinton’s own behavior was called into question, those who were predisposed to care a great deal about traditional moral values and who perceived Clinton to be immoral were most likely to feel aversion. These findings suggest a more general framework for studying aversion. In this view, aversion is a function of a citizen’s own perceptions of a political leader and the political context itself. Certain contexts prime certain traits. Citizens have expectations of what traits politicians should display in, for example, times of economic crisis and a politician that breaks this mold may meet with aversion. But the key is that this only happens if citizens believe that the politician bears responsibility or that his or her behavior violates a core belief or value. Thus citizens who do not think that politicians can control the economy will probably not feel aversion, even if they recognize that the economy is weak and rate a politician’s leadership poorly. And citizens who do not care much about morality are unlikely to respond with anger and disgust at the politician whose morality is compromised or called into question. In light of these contingencies, it is easy to see that aversion is not a common emotion, at least not in the American context (Marcus, Neuman, and MacKuen 2000). In a political culture that places a lot of emphasis on individual rights and responsibilities, Americans often blame themselves rather than specific politicians for poor economic circumstances. 1980 may represent a special case in which a poor economy, a president who was often perceived to be a weak leader, and an opposition candidate who
FEAR AND ANGER IN CANDIDATE EVALUATION
127
worked hard to tie the poor economy to the actions of the president all combined to make aversion particularly likely. In terms of affront, politics is often so far removed from people’s lives that they may not realize that certain policies and actions undermine their core beliefs and values. For most citizens, the simple realization that a politician holds issue beliefs different from one’s own may cause some unease, but not necessarily aversion. Unlikely does not mean impossible, however. Clever challengers can do a great deal to mobilize aversion. By playing the blame game, for example, they can convince people that the incumbent has caused a problem or, at least, has prevented its solution. It is also common for candidates to call attention to important values and how opponents have violated those values. Thus challengers may be able to draw the connections that citizens may not make on their own, facilitating aversive reactions. It may be possible for aversion to be felt in many contexts, but it may rely on the actions of opposing candidates and parties to suggest that some specific person or party can be blamed for poor times or that a politician’s actions go beyond the realm of politics and policy “as usual” to affront citizens’ core beliefs and society’s values about what the proper conduct of political figures should be. In this light, it will be useful to analyze the strategies and effects of campaign advertisements. We know a good deal about how campaigns use audiovisual cues to manipulate emotions, in particular anxiety (Boynton and Nelson 2003; Brader 1999; Nelson and Boynton 1997); now it is time to understand better how politicians mobilize anger.15 To this effect, it is necessary that surveys like the NES include more extensive batteries of negative affect items, so that it will be easier to discriminate between anxiety and aversion and to assess when such discrimination occurs and what its consequences are (see also recommendation 3 of Marcus et al., this volume). Further research in the mobilization of aversion is necessary because anxious and aversive reactions have distinctive implications for information processing and decision making. Specifically, while anxiety is thought to stimulate a desire for learning, information gathering, and open-minded processing of new information, aversion may cause citizens to rely on partisanship, ideology, and other predispositions, inhibiting the incorporation of new information into the decision-making calculus. If aversion is experienced alongside anxiety, it may offset the learning effects that the latter produces, thus triggering the biases of motivated reasoning (MacKuen et al. 2001b). More importantly, however, we need to understand aversion because it may be mobilized in order to stimulate mass hatred. In this chapter, we have looked at relatively benign forms of aversion—voters who are mad at a politician and who may punish this politician by voting him or her out off office. But history has demonstrated that aversion is not always so
128
STEENBERGEN AND ELLIS
benign. For example, Adolph Hitler came to power through the mobilization of aversion: we know all too well the scapegoating and hatred that this aversion produced. More commonly, mobilizing particular segments of the population around aversion toward a particular social group (and the idea that an opposing party or candidate’s policies violates these citizens’ core belief systems by supporting this group) has meaningful consequences for political mobilization and political decision making. These (and other) analyses suggest that aversion is much rarer than anxiety in the context of politics, and stimulating true aversion toward a candidate, political figure, or group is not easily done. But when aversion occurs, it may have particularly powerful consequences. Thus knowing why people become angry may have great political significance. This chapter has provided some insights in this matter, but much more research remains to be done. Appendix: Variable Descriptions 1980 NES The variable labels refer to ICPSR Study Number 7763. Anxiety A composite of VMP3055 and VMP3059. This variable is scaled to have a 0–1 range with higher values indicating greater levels of anxiety toward Carter. Aversion A composite of VMP3053 and VMP3057. This variable is scaled to have a 0–1 range with higher values indicating greater levels of aversion toward Carter. Issue Distance from Carter The mean absolute difference between the respondent and the respondent’s placement of Carter on defense spending (VMP2137 and VMP2138), aid to minorities (VMP2148 and VMP2149), government spending (VMP2159 and VMP2160), relations with Russia (VMP2170 and VMP2171), and inflation versus unemployment (VMP2181 and VMP2182). The distance is rescaled to a 0–1 range, with higher values indicating greater distance between the respondent and Carter. Prospective Economy VMP3093. This variable is scaled to have a 0–1 range with higher values indicating greater optimism about the state of the economy in the next 12 months.
FEAR AND ANGER IN CANDIDATE EVALUATION
129
Morality of Carter VMP3366. This variable is scaled to have a 0–1 range with higher values indicating more positive impressions of Carter’s morality. (In some analyses we use VMP2290, scaled in the same manner.) Leadership of Carter A composite of VMP3368, VMP3371, VMP3372, and VMP3373. This variable is scaled to have a 0–1 range with higher values indicating more positive impressions of Carter’s leadership qualities. (In some analyses we use a composite of VMP2292, VMP2295, VMP2296, and VMP2297, scaled in the same manner.) Power Hungriness of Carter VMP3370. This variable is scaled to have a 0–1 range with higher values indicating the perception that Carter is not power hungry. (In some analyses we use VMP2294, scaled in the same manner.) Partisanship VMP2212. This variable is scaled to have a 0–1 range with 0 indicating strong Democrats and 1 indicating strong Republicans. Ideology VMP2125. This variable is scaled to have a 0–1 range with 0 indicating extreme liberals and 1 indicating extreme conservatives. Age VMP0325; the respondent’s age in years. Education VMP0334; the respondent’s highest level of education. Income VMP0543; the respondent’s family income before taxes in 1979. Female VMP0557; the respondent’s sex, dummy-coded so that 1 is female and 0 is male. White VMP0558; the respondent’s race, dummy-coded so that 1 is white and 0 is non-white. Control over Unemployment VMP2207; a dichotomous variable distinguishing between those who think something can be done about unemployment and those who do not.
130
STEENBERGEN AND ELLIS
1995 NES The variable labels refer to ICPSR Study Number 6636. Anxiety A composite of V952023, V952024, and V952025. This variable is scaled to have a 0–1 range with higher values indicating greater levels of anxiety toward Clinton. Aversion A composite of V952026, V952027, V952028, V952032, V952033, and V952034. This variable is scaled to have a 0–1 range with higher values indicating greater levels of aversion toward Clinton. Issue Distance from Clinton The mean absolute difference between the respondent and the respondent’s placement of Clinton on jobs/standard of living (V940930 and V940931), aid to blacks (V940936 and V940937), government services/spending (V940940 and V940941), and federal health insurance (V940950 and V940951). The distance is rescaled to a 0–1 range, with higher values indicating greater distance between the respondent and Clinton. Prospective Economy V940910. This variable is scaled to have a 0–1 range with higher values indicating greater optimism about the state of the economy in the next 12 months. Morality of Clinton V952072. This variable is scaled to have a 0–1 range with higher values indicating more positive impressions of Clinton’s morality. (In some analyses we use V940920, scaled in the same manner.) Leadership of Clinton V952074. This variable is scaled to have a 0–1 range with higher values indicating more positive impressions of Clinton’s leadership qualities. (In some analyses we use V940921, scaled in the same manner.) Partisanship V952263A. This variable is scaled to have a 0–1 range with 0 indicating strong Democrats and 1 indicating strong Republicans. Ideology V940848. This variable is scaled to have a 0–1 range with 0 indicating extreme liberals and 1 indicating extreme conservatives.
FEAR AND ANGER IN CANDIDATE EVALUATION
131
Age V941203; the respondent’s age in years. Education V941209; the respondent’s highest level of education. Income V941404; the respondent’s family income before taxes in 1993. Female V941434; the respondent’s sex, dummy-coded so that 1 is female and 0 is male. White V941435; the respondent’s race, dummy-coded so that 1 is white and 0 is non-white. Moral Traditionalism A composite consisting of V941029, V941030, V0941031, and V941032, scaled so that higher values indicate more moral traditionalism. Notes * The title is an allusion to Hunter Thompson’s famous book Fear and Loathing: On the Campaign Trail, 1972 (San Francisco: Straight Arrow Books, 1973). The research in this chapter is based in part on the 1980 and 1995 National Election Studies: (1) Miller, Warren E., and the National Election Studies. NATIONAL ELECTION STUDIES, 1980 Major Panel Study [dataset]. Ann Arbor, MI: University of Michigan, Center for Political Studies [producer and distributor], 1999; (2) Rosenstone, Steven J., Warren E. Miller, Donald R. Kinder, and the National Election Studies. NATIONAL ELECTION STUDIES, 1995 Pilot Election Study [dataset]. Ann Arbor, MI: University of Michigan, Center for Political Studies [producer and distributor], 1999. These materials are based on work supported by the National Science Foundation under Grant Nos.: SBR-9707741, SBR-9317631, SES-9209410, SES-9009379, SES-8808361, SES-8341310, SES-8207580, and SOC7708885. Any opinions, findings, and conclusions or recommendations expressed in these materials are those of the authors and do not necessarily reflect those of the National Science Foundation. 1. For a contrasting view see the work of Green (Green and Citrin 1994; Green, Goldman, and Salovey 1993). 2. While it is important to note that both anxiety and aversion can be internally directed (Lazarus 1991), for purposes of this analysis our focus is on externally triggered emotions since these are more relevant for politics.
132
STEENBERGEN AND ELLIS
3. Gehm and Scherer’s (Gehm and Scherer 1988) cross-national study of emotional antecedents suggests that their may be cultural differences in the experience of anxiety. They found that the novelty and unpredictability of a threat was an important predictor of anxiety only in economically well-off countries. The authors attribute this to the fact that individuals in richer countries have a greater reliance on and need for predictability and that threatening, anxiety-producing events are more common in poorer countries. 4. Again, the president is a familiar figure, and familiarity is important to the experience of aversion. 5. Moreover, both of these NES studies contain a panel component. This allows us, in a limited way, to address criticisms that evaluations of candidate traits such as leadership and morality are endogenous to candidate affect. 6. The standardized reliability is calculated using the Spearman-Brown prophecy formula. 7. An exploratory factor analysis of the affect items for Carter did not yield separate anxiety and aversion factors, although this is not unsurprising given that there were only seven affect items in total, three of which tapped positive affect. 8. When we interact beliefs about the controllability of the economy with perceptions of Carter’s leadership abilities we observe a significant difference (at the .10 level) between the simple slopes of leadership for the “some control over economy” and “no control over economy” groups. 9. It is possible that we have misspecified the direction of causality: it is not the negative perception of Carter’s leadership that, under certain conditions, causes aversion, but aversion, felt for whatever reason, that causes a bleak view of Carter’s leadership. In this view, affect drives inferences and rationalizations of Carter’s traits, and placing these traits on the right-hand side of our regression equations is therefore suspect from a statistical viewpoint. We can address this criticism, albeit imperfectly, by using lagged values of the trait variables, which is possible by virtue of the panel design of the 1980 NES. Doing so does not greatly affect the results. Considering all respondents, the effect of wave 2 perceptions of Carter’s leadership on wave 3 aversion is –.685 (p ⬍ .01); this is only slightly weaker than the result in table 6.1. Leadership perceptions remain statistically insignificant (b ⫽ .048, ns) among those who feel that the economy cannot be controlled. In contrast, perceptions of leadership ability have a strong and significant effect among those who believe that something can be done about the economy (b ⫽ ⫺.762, p ⬍ .01). 10. The NES also asked a dichotomous version of the affect items, in which respondents had to indicate if they had ever felt a particular emotion toward Clinton. We replicated the analyses using this version and the results are essentially the same as those reported for the five-point scales. 11. Once these predictors are taken into account, the residual correlation between anxiety and aversion is only .344. This corroborates factor-analytic
FEAR AND ANGER IN CANDIDATE EVALUATION
12.
13.
14.
15.
133
evidence that these emotions are distinct (Marcus et al. 2000, this volume), just as the three-factor model suggests. Sex is statistically significant at the .10 level for both affective dimensions, but its effect runs in opposite directions: women are more likely to feel anxious—a result that we also observe in the 1980 data—but less likely to experience aversion toward Clinton. An interaction between moral traditionalism and perceptions of Clinton’s morality does not quite yield a significant difference in slopes, although this is hardly surprising given the small sample size. Note that in the group of moral traditionalists, perceptions of Clinton’s morality are also marginally significant in the prediction of anxiety, but has the wrong sign. We do not have a satisfactory explanation for this. As before, there is a question whether judgments of morality might be endogenous to candidate affect. We thus estimate a separate model that uses lagged (1994) values of the trait evaluations to predict affective reactions in 1995. The results remain supportive of the second hypothesis. Perceptions of Clinton’s morality have a significant effect on aversion (b ⫽ ⫺.167, p ⬍ .05). While this effect is weaker than when we use concurrent values of morality and affect, it is still among the strongest effects in the model. Given that the 1994 evaluations of Clinton’s morality precede most of the investigative activity concerning the Whitewater scandal, their effects should be weaker. Breaking down the sample by moral traditionalism, we again see no significant effect of morality on aversion when moral traditionalism is low (b ⫽ ⫺.094, ns), but a significant effect when moral traditionalism is high (b ⫽ ⫺.227, p ⬍ .10). It would also be interesting to determine when fear appeals produce aversion as a byproduct (Dillard et al. 1996).
This page intentionally left blank
CHAPTER 8 THE THREE FACES OF NEGATIVE CAMPAIGNING: THE DEMOCRATIC IMPLICATIONS OF ATTACK ADS, CYNICAL NEWS, AND FEAR-AROUSING MESSAGES Ann Crigler, Marion Just, and Todd Belt
egative campaigning” is widely deplored by journalists and the public. Commonly, the term negative campaigning refers to critical statements or visuals of political candidates, institutions, processes, or policy issues that may occur during an election campaign. These negative messages are ubiquitous in campaigns. They are found not only in political advertisements, but also in the news, talk shows, Internet Web pages, and late-night TV comedians’ monologues. Negative campaigning ranges from personal attacks among candidates to cynical statements regarding the motives of candidates and the self-serving intentions of actors in the policy-making process. At its worst, negative campaigning is viewed as “mudslinging,” especially when it involves personal attacks. But negative campaigning is not all bad. The dissemination of diverse points of view is vital to democratic governance and some criticism is an inevitable outcome of legitimate contending priorities. Labeling all criticism as “going negative” constrains essential campaign discourse ( Jamieson and Waldman 2000). After all, challengers are expected to run against the incumbent’s record. Criticism, therefore, should be expected to play an essential part in campaigning (Kahn and Kenney 1999; Mayer 1996). In this study we set out to determine what, if any, constructive role negative campaigning can have in American electoral politics.
“N
136
CRIGLER, JUST, AND BELT
The academic literature on the effects of negative campaigning has produced mixed results (Lau et al. 1999; Lau and Pomper 2004; Pentony 1998). While some researchers have found, for example, that negative campaigns can reduce citizens’ trust in politics (Cappella and Jamieson 1997; Patterson 1993), as well as decrease voter turnout and polarize partisans (Ansolabehere et al. 1994; Ansolabehere and Iyengar 1995), others maintain that exposure to negative advertising actually increases both information and participation in elections (Brians and Wattenberg 1996; Finkel and Geer 1998; Freedman and Goldstein 1999). Some studies have found that negative campaigning has little effect on turnout and have offered theoretical challenges to the notion of voter demobilization (Bartels 1996; Freedman and Goldstein 1999; Garramone et al. 1990; Thorson, Christ, and Caywood 1991b). Finally, researchers have even found that negative information—by stimulating voter anxiety—can have positive democratic effects. It appears that people are more inclined to seek further information when they feel that their circumstances or values are threatened (Marcus and MacKuen 1993; Marcus, Neuman, and MacKuen 2000). Researchers may be reaching different conclusions about negativity not only because they have different concerns or use different methods, but also because they are focusing on the impact of different kinds of negative campaigning (Lau and Pomper 2001, 2004; Richardson 2001) on different people (Kahn and Kenney 1999; Lau and Pomper 2001, 2004) and on different candidates (Lau and Pomper 2002). In short, scholars may come to contradictory conclusions because they are studying fundamentally different processes. In reviewing the literature, we find that negative campaigning includes at least three broad areas of communication: ● ●
●
candidate attacks on other candidates, typically found in political ads; cynical news about politics, candidate strategies, or motives, typically found in news coverage, which emphasizes the horserace; fear-arousing messages from candidates or in the press that may be perceived as threatening or frightening, whether or not that is the intent of the source. These messages are found in both news and ads.
This study demonstrates that these three kinds of negative campaign communication arouse varying emotional responses in the public. Analyses of a panel survey and experiment conducted during the 1996 presidential election show that emotions aroused by negative communication become part of a citizen’s dynamic construction of the campaign. Different types of negative communication influence citizens’ assessments of the candidates,
THREE FACES OF NEGATIVE CAMPAIGNING
137
their level of political trust, and their attention to the campaign. In some circumstances, the emotional response to attack advertising may cause the ads to backfire on the instigators. Feelings aroused by cynical news may demobilize voters, while fear-arousing messages can help to engage the public in the campaign. We discuss the political implications of these effects and what, if anything, can be done to enhance the constructive effects and moderate the destructive impact of negative campaigning. Theoretical Orientation Central to this analysis is the converging body of scholarship that has emerged around the role of emotion in political learning and candidate assessment (Conover and Feldman 1986; Lanzetta et al. 1985; Lodge and Tabor 2000; Marcus and MacKuen 1993). Much of the recent research emphasizes the bi-dimensionality of emotions. Marcus and Rahn (1990), in reviewing the considerable psychological literature about the structure of emotions, show that two distinct and graded emotional moods underlie a variety of common emotional states (Marcus 1988; Marcus and Rahn 1990; Marcus and MacKuen 1993; Marcus, Neuman, and MacKuen 2000). These two moods, which form the axis of a circumplex model of emotions, best describe the two distinct emotional responses that are usually associated with democratic politics (Marcus 1988; Marcus et al., this volume; Marcus and Rahn 1990; Marcus and MacKuen 1993; Marcus, Neuman, and MacKuen 2000). While not mutually exclusive, these two dimensions have been labeled in different ways, but generally fall into a positive, enthusiasm dimension and a negative, anxiety dimension (Gray 1981; Tellegen 1985). Enthusiasm is necessary to gain voters’ “support and to create active interest in the election” (Marcus and MacKuen 1993, p. 673). Anxiety, on the other hand, “stimulates attention toward the campaign and political learning and discourages reliance on habitual cues for voting” (Marcus and MacKuen 1993, p. 672). This study tests the role that emotions play in American presidential elections by examining different kinds of negative stimuli—attack advertising, cynical campaign news, and worrisome economic information. Employing the concepts of emotional arousal described by Marcus and colleagues to help explain why different kinds of negative campaign messages have differential impacts on voters, we argue that the principal mediation of negative campaigning lies in the emotions that are aroused among voters, and that some of these emotional responses diminish incentives to engage in political activity while others encourage democratic participation.
138
CRIGLER, JUST, AND BELT
The Three Faces of Negativity The analysis of negative campaign messages is particularly timely as observers have found an increasing level of negativity in both campaign advertising and news. The use of negative advertising in campaigns for president has, on the whole, increased since the first widely televised campaign of 1952. The prevalence of negative ads, however, has fluctuated greatly during this period—while only 12 percent of the prominent campaign ads were negative in 1960, negative ads launched on an upward trajectory to a high of over 80 percent negative in 1988, declining somewhat to 60 percent in 1996 (West 2005, pp. 59–65). Patterson (1993) shows a corresponding increasingly negative tone in campaign news in this same period. It is important to recognize, however, that even though cynicism has increased in the news, the overwhelming majority of news messages are neutral ( Just, Crigler, and Buhr 1999).
Attack Advertising Negative television ads go back a long way in presidential campaigns. The 1964 “Daisy” ad was pulled by the Johnson campaign after only a few airings, because it was regarded as too frightening. In 1968 there were attack ads on both sides, including the anti-Agnew “Spiro Agnew for Vice President/Man Laughing” and the anti-Humphrey “Why is this man laughing?” spot. More recently, the 1988 anti-Dukakis “Willie Horton” and “Revolving Door” ads were heavily criticized for carrying a racist message. While attack ads frequently contain unsubstantiated criticism of opponents, West’s analysis of the most prominent political commercials during the period from 1952 to 2000, indicates that attack ads contained some of the most substantive appeals in campaign spots (West 2005, p. 65). As Kahn and Kenney (1999) have found, people are able to distinguish between negative information that is presented in an appropriate way and mudslinging that tends to be harsh and unsubstantiated. These unsubstantiated attacks tend to demobilize the public; however, as the proportion of “legitimate” criticism increases, voters are more likely to turn out. Our own survey data show that citizens believe that ads should criticize opponents, but citizens also make distinctions about the kinds of criticisms that are acceptable. They approve of attacks on candidate records and qualifications and disapprove of attacks on personal background (see figure 8.1). Assessments of the impact of negative advertising techniques have been mixed. In experimental studies, the use of political attack advertising has been associated with a decline in both voter turnout and citizens’ feelings of political efficacy (Ansolabehere et al. 1994; Ansolabehere and Iyengar 1995);
139
THREE FACES OF NEGATIVE CAMPAIGNING 80 70 Percent
60 50 40 30 20 10 0
s
on
ue
Iss
siti po
e nc
p Ex
e eri
uld
Wo
e
ter
ffic
do
o in
c ara
Ch
lity
na
o ers
’s p
ate
id nd
Ca
More
mp
n aig
Ca
tac
tics
ily
lls
Po
nd
Ca
’s ate
fam
id
Less
Figure 8.1 Respondents’ Preferences for Campaign News Coverage.
whereas in survey analyses, research suggests that reported exposure to both positive and negative political ads is associated with increased voter turnout (Brians and Wattenberg 1996; Freedman and Goldstein 1999; Goldstein 1997; Kahn and Kenney 1999; Lau, Pomper, and Mumoli 1998; Wattenberg and Brians 1999). The survey finding that ads increase turnout remains disputed (Ansolabehere, Iyengar, and Simon 1999). In a metaanalysis of the literature on negative advertising effects, Lau and his colleagues found no overwhelming evidence that negative ads have a detrimental effect on turnout or are more effective than positive ads (Lau et al. 1999). Some studies have indicated that negative ads have the potential to improve public knowledge about substantive political issues and are more memorable than positive ads (Basil, Schooler, and Reeves 1991; Patterson and McClure 1976, p. 116; West 2005, p. 65). While some research has shown that attack ads influence voters to form negative impressions of presidential candidates (Kaid 1997; King, Henderson and Chen 1998; Lau 1985), other research points to the fact that attack ads can backfire on the sponsor, resulting in a “boomerang” effect (Capella and Taylor 1992; Garramone 1984; Hitchon and Chang 1995; Hitchon, Chang, and Harris 1997; Kahn and Geer 1994; Lau, Pomper, and Mumoli 1998; Martinez and Delegal 1990; Merritt 1984; Roddy and Garramone 1988). These two effects might produce different emotional responses. Attack ads might cause voters to worry about the qualifications of the attacked candidate or voters may lose respect for a candidate who is perceived as attacking rather
140
CRIGLER, JUST, AND BELT
than participating in constructive discourse. Lastly, attack ad research has uncovered a “polarizing effect” in which negative advertising reinforces attachments of self-identified Democrats and Republicans to the candidate representing their party (Ansolabehere and Iyengar 1995; Finkel and Geer 1998; Lau and Pomper 2001). Partisanship may explain the different kinds of emotional responses to attack ads. Cynical News The second type of negative communication studied here, negative campaign news, has provoked less scholarly disagreement. Most observers deplore the kind of negative news that takes a consistently cynical view of politicians and the political process, in or out of campaigns. Researchers have demonstrated that campaign news shows a growing emphasis on the horserace (Brady and Johnston 1987, p. 184). The media’s emphasis on candidate strategies often results in less issue coverage while increasing the level of cynicism in news about politics and politicians ( Just et al. 1996; Patterson 1993). A particularly common campaign news theme is to portray virtually any candidate’s action or proposal as simply a ploy to appeal to certain segments of the electorate. Additionally, adwatches—news media critiques of political advertisements—tend to characterize candidates’ claims as manipulative and deceitful and also contribute to cynical news coverage (Kaid 1996; McKinnon Melton, and Kaid 1999). The growing negative tone of news toward candidates and public officials has been blamed for decreasing electoral participation and trust in government (Cappella and Jamieson 1997; Patterson 1993).1 Exposure to increasing cynicism in news may result in decreasing respect for candidates. Fear-Arousing Messages Campaigns are good environments for studying the third aspect of campaign negativity—emotional appeals, as voters receive a barrage of emotion-laden messages about politics from both the media and from the candidates themselves. There is no question that candidate and partisan communications are designed to be emotional: to reinforce the enthusiasm of their supporters or potential supporters, to reassure those who are concerned about a candidacy, and to arouse or reinforce concern about opponents ( Jamieson 1992; Kaid and Johnston 1991). Emotional appeals in ads are often conveyed audio visually through the use of ominous music, scary visuals, or both (Kern 1989; Richardson 2001). A subtle example of the fear theme is the 1984 Reagan “Bear in the Woods” ad. Even without the intent to frighten the audience, however, the news conveys many
THREE FACES OF NEGATIVE CAMPAIGNING
141
frightening messages as well. Most stories about crime, riots, and war carry a negative emotional punch. In any campaign, some candidates may be more successful than others in arousing emotions of enthusiasm, reassurance, or anxiety about their opponents. The 1996 presidential election campaign represents just such an imbalance of persuasive resources, both in the voters’ initial feelings about the candidates and in the emotions elicited by their communications with the voters. The incumbent president, Bill Clinton, aroused more positive and negative feelings in the electorate than his opponent, Bob Dole, did; but Dole pursued a far more negative campaign than Clinton did. The differential attempt by candidates in 1996 to arouse anxiety allows us to test the impact of this strategy on voters.2 It is important to note that citizens as well as candidates do not approach the campaign in the same way but are differentiated by “where they are coming from” in terms of partisan and emotional commitments. Some voters have strong initial predispositions toward candidates that encourage their involvement and attentiveness to the campaign and their receptiveness to new information. Moreover, voters are more likely to develop early ties to candidates in elections between incumbents and other well-known competitors, such as the 1996 presidential election campaign when both contenders were well known.3 Given the public stature of the major party candidates in 1996—the incumbent president Bill Clinton and the Senate majority leader Bob Dole—it is not surprising that about half the public held strong vote preferences six months before election day. Using multiple methods, we tease out the role of these initial positions in order to investigate the role of different types of negative messages during the presidential campaign of 1996. Methods and Data The current research is part of a project on Campaign Discourse and Civic Engagement,4 and relies on two data sets: a panel survey conducted in five waves and a video field experiment carried out as the second wave of the panel. The panel survey design is suited to examining the interactions of negative campaign material and citizens’ emotional evaluations of the candidates over time.5 The sample was made up of a 1,002-person national RDD (random digit dialed) sample of U.S. citizens who were at least 18 years of age plus an additional sample of New Hampshire residents.6 The first wave (wave one) of the panel survey went into the field at the conclusion of the primary elections in mid-June 1996. A total of 1,457 respondents completed the telephone questionnaire in an average time of 37.2 minutes per
142
CRIGLER, JUST, AND BELT
respondent. Subsequent waves of the panel were conducted in late July (wave two), in early September after the national party conventions (wave three), in the last three weeks of October (wave four), and the final round after election day (wave five). In order to maximize response rates, we used a variety of cash and gift incentives, introductory or reminder letters, and multiple telephone recontact attempts (response rates were wave one, 73 percent; wave two, 56 percent; wave three, 79 percent; wave four, 87 percent; wave five, 75 percent).7 Here we focus on waves one and four of the panel (the telephone surveys), wave two (the video experiment), and wave five (the postelection mail questionnaire). The survey instruments included both closed and open-ended questions to assess citizens’ involvement in politics, evaluations of the presidential candidates, exposure and attention to media about the campaign, and levels of political knowledge, as well as a series of demographic characteristics. The candidate evaluation questions covered emotional, evaluative, and issue topics. Respondents’ emotions were attached to the two major candidates specifically. The two questions used were: Has Clinton/Dole ever made you feel (adjective)? How often would you say you have felt (adjective)? The adjectives included: angry, hopeful, afraid, worried, respectful, and enthusiastic.8 The measure of cynicism used was a combination of the two standard American National Election Study measures of agreement with the questions: (1) Public officials care a lot about what people like me think, and (2) You can trust the government in Washington to do the right thing. Exposure to negative advertising over the course of the campaign was measured on the basis of whether or not the respondent could recall any of Clinton or Dole’s advertisements and whether they recalled these ads as being mostly attacks on the other candidate or not. Results from ad recall measures are similar to other measures of ad exposure, such as daytime television viewing habits (Freedman and Goldstein 1999, p. 1198). Exposure and attention to news about the campaign were measured in detail for many possible media sources. Both frequency of use and the amount of attention paid to the news were assessed for network evening news programs and local newspapers. Frequency of use was also measured for television news and news type programs generally, as well as for local television news. The panel survey specifically addressed the interaction of campaign communications, the role of positive and negative emotions, involvement in the campaign, and vote choice. The purpose of the video experiment (the second panel wave) was to examine the impact of specific messages within the context of the ongoing campaign.9 While the panel survey explored the real-world complexity of the presidential race, the video experiment tested specific hypotheses about
THREE FACES OF NEGATIVE CAMPAIGNING
143
the impact of emotionally charged campaign messages on citizens’ evaluations of candidates, cynicism about politicians, and information-seeking behavior in the context of the ongoing election. The political ads and the news stories that were used in the video experiments were selected or manipulated to contain content that would be either positive (producing feelings of confidence or enthusiasm) or negative (producing feelings of anxiety). Three kinds of affective stimuli were presented to subjects in different combinations: positive versus negative advertising (to test impact of candidate attacks), issue versus strategy news reports (to test impact of media cynicism), and either recency or primacy of news about the economy (to test impact of anxiety-stimulating information). Videotapes were mailed to all of the wave one survey respondents for whom addresses were known and who had said that they had access to a television and VCR. Each of the tapes for the eight experimental conditions contained two television news stories: one was an issue story about problems in the economy and what the candidates would do about them, and the other was a strategy game-oriented story about campaign promises and the candidates’ “rapid response strategies” to attacks. The original stories were taken from NBC and CNN news and contained negative or anxious language. For example, the original economic story focused on which candidate would benefit from economic trends and ended by saying, “and the bond market may panic, no matter who wins.” Using the same video footage, the voice-overs for the news stories were rewritten in order to make them less cynical and more substantive. The revised (less cynical) economic story still expressed concern about the economy, but the emphasis shifted to the ideas that Clinton and Dole had put forward for solving the nation’s economic problems.10 Two political advertisements, one for Clinton and one for Dole, were presented on each tape. Subjects in half the conditions viewed attack ads and half saw positive ads.11 The structure of the video field experiment allowed us to assess the impact of specific campaign messages. In addition to the attitudinal measures, a behavioral measure of information seeking was employed. At the end of the tape, the narrator told viewers that if they wanted more information about Bill Clinton, Bob Dole, or any of the candidates’ issue positions, they could fill out the enclosed, stamped postcard and send it along with their answer sheet. Participating respondents watched the stimulus tape in a natural setting, in fact, in their own homes. Of course, that meant that the researcher was not present to make sure that the respondent actually viewed the tape, that the respondent filled out the questionnaire without assistance, or that the respondent answered the questionnaire after and not before viewing the tape. We compensated for some of the limitations of the design by relying
144
CRIGLER, JUST, AND BELT
on an audio-only questionnaire at the end of the videotape. That way, the subjects could not view the questions on the tape without actually watching the tape (audio cannot be heard on “fast-forward”). The answer sheet contained only response categories, so that it did not give away the questions in advance. Because we provided no incentive to subjects to fake their answers (the incentive was included with the tape), we think it likely that we avoided that potential pitfall. We accepted that the trade-off for avoiding false answers was a lower rate of compliance. About half of our wave one sample returned the video questionnaire, which was a good response rate, given the magnitude of the task. The final tally of 787 subjects contained equal-sized treatment groups, and was representative of the total sample.12 Both the telephone-administered panel survey and the video experiment were designed to test whether affective, behavioral, or attitudinal conditions of respondents at one point in time change after exposure to some stimulus: the campaign itself in the case of the panel survey and the video exposure in the case of the experiment. The limitation of the panel survey was that it could only capture the broad brush-strokes of change. The smaller ups and downs are lost in the gaps between panel waves. Even attention to particular campaign events was difficult to gauge depending on the timing of the wave, the time spent in the field for each wave, and the predictability of the event’s timing. The overall impact of prolonged exposure to communication experiences could only be inferred from the repeated measures. The video experiment offered us a tool for assessing the impact of a particular communication experience, and was supplemented with evidence from the more long-term telephone interview data. Negative Advertising From the very beginning of the 1996 campaign, Republican candidate Bob Dole faced an uphill battle. Dole faced stronger than anticipated primary competition from candidates Lamar Alexander, Steve Forbes, and Pat Buchanan. Four days before the New Hampshire primary, Dole began airing attack ads against Alexander (Sabato 1997, p. 43). After losing the New Hampshire primary, Dole stepped up his attack ads, particularly targeting Buchanan. Dole’s negative campaign continued through the general election. By election day, Dole’s advertising was 71 percent negative, compared to 27 percent of Clinton’s.13 How did Dole’s attack strategy affect voters? Did it depress support for Clinton, as advertising lore maintains? Or did the attack ads disadvantage Dole, as the boomerang thesis would predict? The study allowed us to investigate all three reported effects of negative campaign ads—impact on candidate support, demobilization of voters, and potential political polarization.
THREE FACES OF NEGATIVE CAMPAIGNING
145
At the end of the 1996 campaign, respondents in the panel study were asked to describe each candidate’s ads. Fifty-two percent of respondents recalled that Dole aired mostly attack ads, compared to 12 percent for Clinton. Recalling attack ads contributed to respondents’ feeling significantly less positive and more negative about Dole controlling for a range of attitudinal and demographic factors (see table 8.1). There were no parallel effects on feelings about Clinton among the small group of respondents who thought the incumbent’s ads were mostly attacks (12 percent). Respondents who recalled Dole’s ads as mostly negative were significantly less likely to vote for him (see table 8.2). The data show, first, that the public is pretty accurate in assessing which candidate is “going negative.” Four times as many respondents described Dole’s spots as attacks as described Clinton’s ads that way. Second, the results support the “boomerang” hypothesis: negative ads redound against the candidate who uses them (Garramone et al. 1990). Controlling for initial emotional response to Dole and partisanship, describing Table 8.1 Predicting October Emotions about Dole Predictors
Dependent Variables Positive Emotions
July positive emotions about Dole
.489*** (.036)
July negative emotions about Dole Recall of attack ads July cynicism Approval of Clinton on economy Party identification Dole cut taxes Clinton protect Medicare R2 n
Negative Emotions
⫺.479** (.099) ⫺.239** (.081) ⫺.321*** (.089) ⫺.027 (.065) .376** (.118) ⫺.690*** (.105) .590 583
.495*** (.036) .270* (.110) .149 (.092) .409*** (.098) ⫺.103 (.071) ⫺.195 (.130) .513*** (.117) .553 591
Note: Multiple Regression Analyses, Standard Errors in parentheses. * p ⬍ .05, ** p ⬍ .01, *** p ⬍ .001. Source: All variables are from wave four survey (October 1996) except as noted. Models also control for frequency of network TV news and newspaper use, gender, income, age, and race.
146
CRIGLER, JUST, AND BELT
Table 8.2 Predicting October Vote Choice Predictors
July Clinton feeling thermometer July Dole feeling thermometer Recall of attack ads July cynicism Approval of Clinton on economy Party identification Dole cut taxes Clinton protect Medicare Pseudo R2 Log likelihood n
Dependent Variable Vote Choice (Positive ⫽ Dole) ⫺.028*** (.006) .025*** (.007) ⫺.276* (.116) ⫺.010 (.111) ⫺.350** (.101) .186* (.073) .480** (.150) ⫺.520*** (.128) .782 ⫺75.120 535
Note: Probit Analyses, Standard Errors in parentheses. * p ⬍ .05, ** p ⬍ .01, *** p ⬍ .001. Source: All variables are from wave four survey (October 1996) except as noted. Model also controls for frequency of network TV news and newspaper use, gender, income, age, and race.
Dole’s ads as attacking resulted in respondents having less positive and more negative emotions about him and decreased intention to vote for him. The impact of associating Dole with the use of negative ads in our data is impressive, since it takes into account other influential factors, such as party identification, perceived candidate issue positions, approval of Clinton’s handling of the economy, media consumption habits, and a number of demographic variables.14 The data show that Dole hurt himself by concentrating on attack ads, especially as Clinton was not perceived as similarly on the attack. While this evidence does not prove that a negative ad strategy always harms a candidate, it suggests that an imbalance of negative attacks is likely to hurt the candidate who is perceived as the more negative. Apart from the effects on candidate support, evidence from experiments shows that attack ads can demobilize voters (Ansolabehere and Iyengar 1995). The literature on emotions and politics helps to explain this finding (Lau 1985; Marcus, Neuman, and MacKuen 2000). As we noted earlier,
THREE FACES OF NEGATIVE CAMPAIGNING
147
Dole’s attack advertising decreased positive emotions about him. Since enthusiasm for candidates gives voters a greater stake in the campaign (Marcus and MacKuen 1993), decreasing positive emotions about candidates can discourage voting. The experiment wave of the panel study provided an opportunity to examine the immediate effect of attack advertising on subjects’ expected likelihood of voting. The experimental design grouped attack ads together in conditions separate from positive ads, permitting an examination of the influence of ad negativity. The result indicates that subjects who viewed the negative ads were significantly less likely to say that they intended to vote in the coming election than those who saw the videotapes with positive ads (78 percent versus 83 percent, see table 8.3). Viewing attack ads significantly decreased
Table 8.3 Predicting Experimental Effects Predictors
Dependent Variables Likelihood of Voting
July likelihood of voting Cynicism Video: Attack ads
4.235*** (.431) ⫺.201 (.107) ⫺.719** (.246)
Video: Strategy/game news stories Undecided Approval of Clinton on economy Party identification Dole cut taxes Clinton protect Medicare Pseudo R2 Log likelihood n
.142 (.097) ⫺.079 (.078) .213 (.136) .141 (.121) .231 ⫺294.659 630
Cynicism
.966*** (.074)
.323* (.149) .508** (.154) ⫺.057 (.064) .095* (.047) ⫺.118 (.083) ⫺.037 (.075) .122 ⫺855.386 635
Note: Ordered Logit Analyses, Standard Errors in parentheses. * p ⬍ .05, ** p ⬍ .01, *** p ⬍ .001. Source: Dependent variable is from wave two experimental questionnaire ( July–August 1996), independent variables are from wave one survey (July 1996). Model also controls for frequency of network TV news and newspaper use, gender, income, age, and race.
148
CRIGLER, JUST, AND BELT
intention to vote even controlling for stated vote intention in the first wave of the survey one month before and for the subjects’ level of political cynicism. The results of the experiment, conducted during an actual campaign, support the findings in other researchers’ laboratory experiments, which pointed to the demobilizing effect of attack advertisements (Ansolabehere et al. 1994, 1999; Ansolabehere and Iyengar 1995). Although our subjects reported that they were less likely to vote immediately following exposure to negative ads, the passage of time or other factors attenuated the effect. In the preelection wave of the panel survey, subjects exposed to negative ads in the experiment were no less likely than those exposed to positive ads in the experiment to say that they intended to vote. It may be that, over the course of the campaign, voters became inured to candidate attacks. Another explanation is also possible— that as the campaign wore on and the whole population was exposed to negative ads, the experimental subjects simply did not stand out from other disaffected citizens. Whether or not negative ads demobilize voters in the long term, scholars are concerned that negative advertising polarizes the electorate by activating latent partisan support (Ansolabehere and Iyengar 1995). We find no evidence in our field experiment to support this conclusion. Individuals who saw negative ads were not any more likely to develop more negative opinions about either of the candidates. Instead, we found the opposite effect. Viewing positive ads reinforced support for Dole among Republicans (F (2,354) ⫽ 5.857, p ⬍ .01), although there was no similar effect for Clinton among Democrats. Positive ads were more polarizing than negative ads in this case. It is possible that because positive Dole ads were less common, that Republican subjects were especially grateful for the reinforcement, while positive ads added nothing new for Clinton voters. Our data indicate that the conventional wisdom (see Ansolabehere and Iyengar 1995; Bartels 1996; Lau 1985) that negative ads necessarily have greater impact than positive ads is questionable. We find effects for both positive and negative ads. Negative ads, if employed disproportionately to the opposition, can produce a boomerang effect, and may, at least temporarily, demobilize the electorate. Although we find no polarizing effect from negative advertising, we do show significant reinforcing support for positive ads. Media Negativity and Cynicism Turning from ads to news, the experiment addresses an ongoing scholarly debate about the impact of strategy/game-oriented campaign coverage on voters. The cynical news stories that subjects viewed in the field experiment
THREE FACES OF NEGATIVE CAMPAIGNING
149
were taken from actual stories that aired during the campaign. The stories stressed the candidates’ venal motivations and strategic advantage for supporting policy proposals. If strategy/game-oriented stories make voters more negative about politics, we would expect subjects’ political cynicism to increase after viewing the negative campaign stories embedded in the experiment. That is exactly what happened. Controlling for the effect of party identification and a range of demographic and attitudinal factors, subjects who viewed the strategy/ game news stories became significantly more cynical than those who viewed substantive news (see table 8.3). Interestingly, the effect of cynical news coverage on voter cynicism about politicians did not equally affect all types of subjects (as suggested by Kahn and Kenney 1999). Undecided respondents became significantly more cynical than decided voters after viewing strategy/game news. Of the undecided ones who watched strategy/game news stories, 45 percent became more cynical, while only 35 percent of decided respondents became more cynical after viewing these same videos (F (1,371) ⫽ 3.966, p ⬍ .05). Perhaps voters who have already decided to support a candidate have already factored in the candidate’s ambition and are, therefore, less likely to be moved by strategy-oriented news stories. Interestingly, in our experiment Republican subjects became more cynical about government after viewing the strategy/game stories than Democrats or Independents (F (2,356) ⫽ 5.538, p ⬍ .05). Ideological opposition to government may account for the differentially increased political cynicism among Republicans stimulated by cynical news. The increased cynicism of undecided voters exposed to strategy-oriented news is particularly troublesome. First, undecided voters are one of the main targets of campaign communication. Second, increasing cynicism among any significant portion of the electorate carries a long-term threat to the democratic practice. Finally, if cynical campaign news equally diminishes support for all contenders, electoral outcomes may not be affected—but we have no reason to assume an even-handed result from an unintended behavior. Emotions and Engagement Other researchers have found that emotional appeals perceived as threatening or frightening can stimulate information-seeking behavior (Neuman, Marcus and MacKuen 1996; Marcus, Neuman and MacKuen 2000). We investigated that proposition in two different ways. The panel survey assessed how emotions about the candidates were related to information seeking over the course of the campaign, while the experiment tested whether anxiety-laden campaign messages provoked similar short-term response.
150
CRIGLER, JUST, AND BELT
The panel survey assessed respondents’ feelings about the candidates in July and asked for self-reports about information-seeking behaviors in September and October, namely attention to campaign news on television and in newspapers, attention to the Democratic and Republican Party conventions, and attention to the presidential candidate debates. Early support, similar to (and often coinciding with) strong partisanship, is known to correlate with attention to a variety of campaign communication events and processes—conventions, debates, news watching (Holbrook 1996, Ch. 5). In order to insure that attention to campaign material was not the result of early attachment to one of the candidates, the models control for whether or not the respondent had decided upon a candidate in July. The effect of emotions about the candidates on seeking campaign information was varied by candidate and type of communication, but we found that positive emotions were as likely to stimulate attention as negative emotions. Both positive and negative feelings about Dole and positive feelings about Clinton prompted greater attention to television news later in the campaign Both positive and negative feelings about Clinton promoted more reading about the campaign in newspapers later in the campaign (see table 8.4). Negative feelings about Clinton also led people to seek information through exposure to the Republican Party convention. Having warm feelings toward the party standard bearer stimulated viewing of that party’s convention. Those who had positive emotional attachments to Clinton were more inclined to watch the Democratic Party convention, while those with positive emotional attachments to Dole were more inclined to watch the Republican Party convention. This is to be expected, as the functions of party conventions have evolved over the years from forums to do party business to televised platforms for promoting the candidacy of the party’s nominee. Lastly, strong positive and negative feelings about Dole (but not Clinton) stimulated voters to pay greater attention to the televised debates. The asymmetry in the effect of emotions on attention to debates may be related to Dole’s underdog status. By the end of the summer, voters who were either attracted or repelled by Dole were more stimulated to watch his performance in the debates than those who had not developed any feelings about him were. To summarize, both positive and negative feelings about candidates appeared to stimulate attention to different kinds of campaign communications and events. The literature suggests, however, that negative emotions arouse greater voter involvement in the campaign (Marcus and MacKuen 1993; Marcus, Neuman, and MacKuen 2000). Our findings modify the anxiety/information-seeking hypothesis. We find that both positive and negative emotions can lead to greater attention to news and information-rich campaign communications, such as conventions and debates.
THREE FACES OF NEGATIVE CAMPAIGNING
151
Table 8.4 Predicting Consumption of Campaign Information in Different Media Sources Predictors
Dependent Variables Television
TV news usage
.770*** (.050)
Newspaper news usage Positive emotions about Clinton Negative emotions about Clinton Positive emotions about Dole Negative emotions about Dole Undecided Party identification Pseudo R2 Log likelihood n
Newspaper
Democratic Convention .280*** (.045)
Republican Convention .290*** (.045)
Presidential Debates .320*** (.045)
.737*** (.040) ⫺.010 (.021)
.043* (.021)
.084*** (.121)
.070* (.021)
⫺.016 (.021)
.080*** (.022)
.014 (.021)
.058** (.021)
.006 (.021)
.002 (.023)
.012 (.023)
.068** (.023)
.045* (.022)
.060** (.023) .040* (.020) ⫺.171 (.118) .010 (.037) .126 ⫺1,467.949 1070
.013 .024 (.021) (.020) ⫺.175 ⫺.281* (.121) (.119) .007 ⫺.085* (.038) (.038) .175 .051 ⫺1,359.025 ⫺1,266.642 1069 1071
.035 (.021)
.022 .064** (.020) (.020) ⫺.341** ⫺.199 (.120) (.117) .045 .026 (.038) (.037) .052 .037 ⫺1,260.243 ⫺1,473.106 1071 1065
Note: Ordered Logit Analyses, Standard Errors in parentheses. * p ⬍ .05, ** p ⬍ .01, *** p ⬍ .001. Source: Dependent variables are from Wave four survey (October 1996), independent variables are from wave one survey ( July 1996). Models also control for gender, income, age, and race.
The experiment examined both the emotional impact of communications and the effect of emotional arousal on short- and long-term engagement in the campaign. One aspect of the experiment was designed to test whether different kinds of negative communications—attack advertising, cynical news, or anxiety-provoking subject matter—could produce immediate information-seeking behavior. Information seeking was operationalized as a postexperiment action of mailing a postcard requesting specific information about one or more of the candidates or issues in the campaign. Overall, about 35 percent of the subjects who completed the experiment returned the information request card. There was no difference in the information-seeking behavior either for those subjects who were exposed to the negative versus positive ads or cynical versus substantive news. However, there were significant differences in information seeking for the subjects who saw the economic
152
CRIGLER, JUST, AND BELT
(fear-arousing) story last. Forty percent of these subjects returned their information request cards, compared to only 30 percent of those who saw a campaign process story last (F (1,785) ⫽ 6.832, p ⬍ .01). In substantive and strategic/game-oriented (i.e., cynical) versions of the economic story, the hook was that the bond market had tumbled even though unemployment was down. Both versions of the economic story contained disturbing information about the lack of improvement in real wages for people at the bottom end of the income distribution and evidence about the increasing gap between rich and poor Americans. In the strategic/ game-oriented version (i.e., the story that actually aired in the news) these economic problems were interpreted as helping or hurting Clinton’s or Dole’s election chances, while in the substantive version, these problems were set against each candidate’s proposals for improving the economy. The message that there were problems in the economy was present in both versions of the story and was expected to arouse anxiety about economic conditions. The known effect of recency on response explains the greater information-seeking behavior on the part of subjects who viewed the anxiety-laden economic story last. This effect was produced even when controlling for subjects’ media habits, decisiveness of vote choice, party identification, and other demographic influences (see table 8.5). Table 8.5 Predicting Information Seeking Predictors
Video: Fear-arousing story last TV news usage Newspaper news usage Undecided Party identification Pseudo R2 Log likelihood n
Dependent Variable Information Card Returned .226* (.098) .053 (.037) ⫺.058 (.030) ⫺.045 (.100) .016 (.025) .021 ⫺455.408 711
Note: Ordered Probit Analysis, Standard Errors in parentheses * p ⬍ .05. Source: Dependent variable is from wave two experimental questionnaire ( July–August 1996), independent variables are from wave one survey ( July 1996). Model also controls for gender, income, age, and race.
THREE FACES OF NEGATIVE CAMPAIGNING
153
In addition to arousing anxiety, the economic story may have given subjects an insight into the importance of the campaign that encouraged further efforts to gain information. There is evidence that when people achieve some understanding of a problem they are more likely to be attentive and to learn new information (Neuman, Just, and Crigler 1992). If people think a campaign is about issues that have an impact on their daily lives, they are more likely to pay attention to what is going on. Furthermore, the economic story on the experimental video exposure may have stood out in people’s minds because it was not typical of the hoopla and game-oriented news that is so prevalent in campaigns ( Just et al. 1996; Kerbel 1995; Patterson 1993). Both the anxiety induced by the economic story and the novelty of a policy-rich news story in the video tape appear responsible for stimulating the information-seeking behavior observed in the experiment. Discussion Three types of campaign negativity have been shown to have different effects on the electorate. The first type of campaign negativity, the wide perception that one candidate uses more attack advertising than another, seems to have harmful effects on the attacking candidate. As predicted by the “boomerang” thesis, the challenger, Bob Dole, damaged his campaign by the heavy use of attack advertising. Going negative early and attacking, often resulted in a decrease in voter’s emotional attachments to Dole. In the end, it appears that Dole’s attack ads cost him votes. The experiment also appeared to offer support to the thesis that attack advertising demobilizes the electorate. In their responses immediately after the experiment, subjects who saw attack ads were significantly less likely than other subjects to say that they intended to go to the polls in November. The demobilizing effect of the experimental manipulation, however, disappeared over time as the experimental exposure was swallowed up by the campaign. In any case, we also found no survey evidence of increased engagement in the campaign or information-seeking behavior attributable to viewing attack advertising. The experimental results show, however, that positive ads for Dole reinforced the emotional attachment of Republican subjects to Dole. The experimental and survey evidence suggest that politicians can help their campaigns by airing positive ads and by avoiding a one-sided negative campaign. Being stigmatized as the negative candidate can really hurt a campaign. Our data, however, show no clearly discernable, long-term impact of negative ads on citizens’ engagement in the campaign. The second type of campaign negativity, cynical news coverage, had a detrimental effect on experimental subjects’ level of trust in government.
154
CRIGLER, JUST, AND BELT
The current style of strategy-oriented campaign news coverage appears to be one of the culprits for increasing voter cynicism. Subjects who viewed videotapes that questioned the motives of politicians and framed their actions in terms of the “horserace,” expressed more distrust for government than those who were not exposed to strategy-oriented news. The third type of campaign negativity, issue-based, fear-arousing communications, can actually benefit democratic participation. Concern about the economy, for example, increases information-seeking behavior during an election campaign. This effect may account for the increased attention to news and increased voter turnout in the 1992 presidential campaign, which took place during an economic downturn. Subjects in our 1996 experiment who viewed worrisome and information-rich news content about the economy were more likely than other subjects to seek further information about the candidates and the issues. Anger, worry, and fear about a candidate also stimulated some respondents to pay greater attention to campaign media. Respondents in the panel survey who registered early negative feelings about one of the candidates were more likely to pay attention to campaign news on television and in newspapers during the general election campaign than people with similar media habits, partisanship, and decisiveness of vote. This finding supports the anxiety/informationseeking research by Marcus and his colleagues. In contrast to Marcus and his colleagues (2000), however, our survey data show that positive feelings about candidates also stimulate campaign involvement. Respondents who felt positive emotional ties to candidates early in the campaign were more likely than other citizens with the same media habits to pay attention to campaign news and to watch televised party conventions. Also, emotional attachment to the challenger, either positive or negative, stimulated interest in the candidate debates. The results underscore the complexity of response to campaign communications— while the type and tone of campaign material may have a detrimental effect on the way candidates and government are evaluated the voters’ enthusiasm as well as anxiety may stimulate interest, involvement, and informationseeking behavior in a political campaign. Conclusion Our data illustrate some of the advantages of using multiple research methods—a panel survey and experimental designs—to understand the interactive nature of campaigns. By assessing the emotions of a panel of respondents both early and late in the campaign, it was possible to evaluate the impact of exposure to campaign media on feelings about the candidates. By incorporating an experiment into the panel design, we were not
THREE FACES OF NEGATIVE CAMPAIGNING
155
only able to gauge the immediate impact of different kinds of negativity in news and campaign ads on a large and representative sample of voters but also to test the long-term effect of such exposure. In addition, the panel study provided an avenue for triangulating the experimental results with evidence from voters’ interactions with actual campaign communications. This study has provided a clarification of the effects of campaign negativity through an individual treatment of the three types of negativity most prevalent in political communications: attack advertising, cynical news messages about strategies and motives, and threatening or frightening news. While it would be easy to assert that candidates would do well to moderate their use of attack advertising, we are aware that some attacks may carry the issue-oriented, anxiety-producing messages. In our study, fear-arousing messages stimulated voter interest and involvement in the campaign. The experiment demonstrated that when citizens have cause for concern about the economy or the polity, they are aroused to seek information and participate in the campaign. We note that economic anxiety news permeated the 1992 presidential campaign—an election marked by a significant increase in voter turnout. Because economic concerns were not reinforced in the subsequent news of the 1996 campaign, we did not find any longterm information-seeking or participation effects from exposure to that kind of information in the 1996 experiment. The results, however, support the view that real economic distress reported in the news can stimulate participation in election campaigns and we may see that effect in upcoming elections. The effects of campaign negativity examined here must be considered within the wider framework of how individuals derive meaning from campaign communications. Voters are incorrectly conceived of as passive receivers of campaign communications who necessarily respond in similar fashion to a given message. Indeed, the results of the multivariate models presented here indicate that changes in voters’ attitudes toward the candidates, their levels of political cynicism, and their emotions are all related to a range of preexisting positions, party identification, media consumption habits, and demographic characteristics. For example, positive messages can reinforce the enthusiasm of partisans, while negative messages can increase campaign attention among undecideds. Some of the effects of negative campaign communication are temporary and some are long lasting. The jury is still out on the impact of attack advertising. Our data illustrate the differences in research outcomes when different methods are employed. Our experimental data showed demobilizing effects from negative ads, while our panel survey data showed no impact at all. In our study, the only long-term impact of negative ads redounded to candidates, not voters. It seems unlikely that moral suasion will effectively limit negative
156
CRIGLER, JUST, AND BELT
ads. Candidates generally “go negative” because their political consultants believe attack ads are more effective than positive ads or because campaigns believe that they have been attacked and must respond. Indeed, the “unanswered attack” is considered a losing strategy in game theory as well as in political practice. Some evidence is accumulating, however, that there are ways to respond to attacks that mute the appearance of negativity both to audiences and academic observers. Although Clinton’s 1996 ads contained many attacks on Dole, they also included statements contrasting Dole’s record with Clinton’s. Some researchers label these ads “comparative” ( Jamieson and Waldman 2000; contrast with the approach of Kaid 1997; Pinkleton 1997). Our data show that most members of the public did not regard Clinton’s “comparative” ads as evidence of “attack.” Whatever the intrinsic merit of comparative ads, they appear to help candidates avoid the “boomerang” effect of negative campaigning (see also Pinkleton 1997). It appears that the lessons of the negative 1996 campaign carried over to the 2000 presidential primary, where candidates competed against each other for the mantel of the most positive candidate. The high-road response to attack advertising seems to be played out most effectively not in the advertising context (where complaining about attacks may give the attack itself more exposure or make the complainant appear to be wimpy or whiney), but rather in candidate speeches, debates, press conferences, and position papers. The answer to negative ads may lie in other kinds of positive speech. Our evidence confirms what others have found, namely that strategy/game-oriented campaign news has little to recommend it. Our experiment showed that strategy-oriented news increases voter cynicism and political detachment. Journalist negativity arises primarily from news values employed across a broad spectrum of news outlets and formats (Graber 1984; Just et al. 1996; Kerbel 1995; Patterson 1980; Robinson and Sheehan 1983; Sigal 1973). News observers and practitioners have used moral suasion to try to change the professional judgments that result in negative news (Committee of Concerned Journalists 1997). Given the present decline in overall audience for news and market pressures arising from competition in an increasingly diverse communication environment, journalists feel under siege. As in the case of attack advertising, it may be more feasible to concentrate on efforts to provide voters with campaign messages that dilute the cynical tone of news coverage rather than to rely on changes in journalists’ values. Enhancing messages that come directly from candidates (such as candidate debates, free candidate airtime, town meetings, and promotion of candidate sites on the World Wide Web) may help to moderate the effects of cynical reporting.
THREE FACES OF NEGATIVE CAMPAIGNING
157
With respect to emotional arousal, our data confirm that reports of bad economic conditions can stimulate civic engagement. No one wishes to solve the problem of declining civic engagement by promoting a poor economy. It is possible, however, that news devoted to the substantive interests of voters could equally inspire greater engagement of the electorate. Certainly that is what voters say they want. When people believe there is something important at stake, they are likely to take steps to promote their interests. Likewise, arousing strong feelings about a candidate, both negative and positive, appears to motivate long-term information-seeking behavior in a campaign. The evidence here shows that it is not only the fear of the opponent but also the hope for the preferred candidate that attracts the attention of voters to campaign communication. It is heartening to think that the campaign can involve not only angry voters but optimistic ones as well. Candidates that inspire hope engage citizens in the campaign ( Just, Crigler, and Belt 2004). Candidates who lack the ability to establish an emotional bond with voters seem doomed to fail. The variety of voter responses to attack advertising, cynical news, and anxiety-producing campaign messages underscores the role of citizens in constructing the meaning of candidates and the campaign and the centrality of emotions in campaigns. The study illustrates the three faces of “negative campaigning” in both stimulating and discouraging engagement with the campaign and suggests that there are ways that we can use these effects to enhance the democratic process.
Appendix: Wording of Questions in Surveys and Field Experiment and Coding Scheme Measures of Perceptions and Emotions about Candidates Emotions We would like to know something about the feelings you have toward the candidates for president. I am going to name a candidate and I want you to tell me whether something about that candidate or the candidate’s positions on the issues has made you have certain feelings like anger or hope. The first (or next) candidate I am going to ask you about is [candidate] (Bill Clinton or Bob Dole). Has [candidate] ever made you feel [adjective] (angry, hopeful, afraid, worried, respectful, or enthusiastic)? (0 ⫽ No; If Yes, continue). How often would you say you have felt [adjective] (1 ⫽ Rarely; 2 ⫽ Occasionally; 3 ⫽ Somewhat Often; 4 ⫽ Very Often)?
158
CRIGLER, JUST, AND BELT
Feeling Thermometers I’d like to get your feelings toward some of our political leaders and other people who have been in the news these days. I’ll read the name of a person and I’d like you to rate that person using something we call the feeling thermometer. You can choose any number between 0 and 100. The higher the number, the warmer or more favorable you feel toward that person; the lower the number, the colder or less favorable. You would rate the person at the 50-degree mark if you felt neither warm nor cold toward them. If we come to a person whose name you don’t recognize, you don’t need to rate that person. Just tell me and we’ll move on to the next one. The first person is [candidate] (Bill Clinton or Bob Dole). How would you rate him using the thermometer? (actual number 0–100) Clinton’s Job Approval on the Economy Do you approve or disapprove of the way Bill Clinton is handling the economy? Do you (approve or disapprove) strongly or not strongly? (1 ⫽ Disapprove Strongly; 2 ⫽ Disapprove Not Strongly; 3 ⫽ Neither Approve nor Disapprove; 4 ⫽ Approve Not Strongly; 5 ⫽ Approve Strongly) Dole on Taxes As president, do you think [candidate] (Bill Clinton or Bob Dole) would be more likely to lower taxes or do you think there wouldn’t be any difference between them? Do you think [candidate chosen] would be much more likely or somewhat more likely to lower taxes than [other candidate]? (1 ⫽ Clinton Much More Likely; 2 ⫽ Clinton Somewhat More Likely; 3 ⫽ No Difference; 4 ⫽ Dole Somewhat More Likely; 5 ⫽ Dole Much More Likely) Clinton on Medicare As president, do you think [candidate] (Bill Clinton or Bob Dole) would be more likely to protect Medicare benefits or do you think there wouldn’t be any difference between them? Do you think [candidate chosen] would be much more likely or somewhat more likely to protect Medicare benefits than [other candidate]? (1 ⫽ Dole Much More Likely; 2 ⫽ Dole Somewhat More Likely; 3 ⫽ No Difference; 4 ⫽ Clinton Somewhat More Likely; 5 ⫽ Clinton Much More Likely) Attack Ads Do the ads about (Clinton or Dole) talk about who he is, mostly explain his views on the issues, or mostly attack the other candidates? (1 ⫽ Talk About Who He Is; 2 ⫽ Explain Views on the Issues; 3 ⫽ Mostly Attack)
THREE FACES OF NEGATIVE CAMPAIGNING
159
Measures of Demographic Information and Party Identification Party Identification Generally speaking, do you usually think of yourself as a Republican, a Democrat, an Independent, or what? [if Democrat or Republican] Would you call yourself a strong (Democrat or Republican) or not a very strong (Democrat or Republican)? [if Independent or other party] Do you think of yourself as closer to the Democratic Party or the Republican Party? (1 ⫽ Strong Democrat; 2 ⫽ Not Strong Democrat; 3 ⫽ Independent Leaning Democrat; 4 ⫽ Pure Independent; 5 ⫽ Independent Leaning Republican; 6 ⫽ Not Strong Republican; 7 ⫽ Strong Republican. Income To get a picture of people’s financial situation, we need to know the general range of incomes of all people we interview. Now, thinking about (your/your family’s) total income from all sources (including your job), did (you/your family) receive $5,000 or more in 1995? Was it $15,000 or more? Was it $25,000 or more? Was it $35,000 or more? Was it $50,000 or more? Was it $75,000 or more? Was it $100,000 or more? (1 ⫽ under $5,000; 2 ⫽ $5,000–$14,999; 3 ⫽ $15,000–$24,999; 4 ⫽ $25,000–$34,999; 5 ⫽ $35,000–$49,999; 6 ⫽ $50,000–$74,000; 7 ⫽ $75,000–$99,999; 8 ⫽ $100,000⫹) Age What is the month, day, and year of your birth? (actual value by year) Race Would you mind telling me your race or ethnic origin? Are you white, black, or African American, Hispanic, American Indian or Alaskan native, Asian, Pacific Islander, or some other race? (0 ⫽ Nonwhite; 1 ⫽ White) Gender Determined by sampling technique (0 ⫽ Female; 1 ⫽ Male) Measures of Cynicism Officials Care Do you agree strongly, agree, disagree, or disagree strongly that public officials care a lot about what people like me think? (1 ⫽ Disagree Strongly; 2 ⫽ Disagree; 3 ⫽ Agree; 4 ⫽ Strongly Agree) Trust in Government Do you agree strongly, agree, disagree, or disagree strongly that you can trust the government in Washington to do the right thing? (1 ⫽ Disagree Strongly; 2 ⫽ Disagree; 3 ⫽ Agree; 4 ⫽ Strongly Agree)
160
CRIGLER, JUST, AND BELT
Measures of Voting Likelihood of Voting So far as you know, do you expect to vote in the presidential election this coming November or not? (0 ⫽ No; 1 ⫽ Yes) Vote Choice [If expect to vote] Who do you think you will vote for in the election for president? [Probe if DK/RF] We all know the election is some time away and people are not certain at this point for whom they will vote. Still, who do you think you will vote for in the election for president in November? (1 ⫽ Bill Clinton; 2 ⫽ Bob Dole) Decisiveness of Vote Choice Would you say that your preference for [candidate] is strong or not strong? (1 ⫽ Strong; 2 ⫽ Not Strong) Measures of Media Consumption Televised News ( July) How often do you watch news or news-type programs? (1 ⫽ Never or Only Occasionally; 2 ⫽ Once or Twice a Week; 3 ⫽ Three or Four Times a Week; 4 ⫽ Five or Six Times a Week; 5 ⫽ Every Day) Newspaper News ( July) Many people don’t have time to read the entire newspaper. They normally read only certain sections such as the sports pages, the business pages, or the news pages. How often do you read the news pages of your daily newspaper? (1 ⫽ Never or Only Occasionally; 2 ⫽ Once or Twice a Week; 3 ⫽ Three or Four Times a Week; 4 ⫽ Five or Six Times a Week; 5 ⫽ Every Day) Network TV News about the Campaign (October) On which of the television network news programs do you watch campaign news most often (World News Tonight on ABC, The NBC Nightly News, The CBS Evening News, or some other network news program)? (1 ⫽ Network). How many days in a typical week do you watch campaign news on [network] (1 ⫽ None; 2 ⫽ Once or Twice a Week; 3 ⫽ Three or Four Times a Week; 4 ⫽ Five or Six Times a Week; 5 ⫽ Every Day) Newspaper News about the Campaign (October) Many people don’t have time to read the entire newspaper. They normally read only certain sections such as the sports pages, the business pages, or the
THREE FACES OF NEGATIVE CAMPAIGNING
161
news pages. How often do you read campaign news about the presidential campaign in your daily newspaper? (1 ⫽ None; 2 ⫽ Once or Twice a Week; 3 ⫽ Three or Four Times a Week; 4 ⫽ Five or Six Times a Week; 5 ⫽ Every Day) Attention to Party Conventions In talking to people about the political party conventions last summer, we found that many people watched the coverage of the conventions on television and some did not. How much of the (Democratic or Republican) party convention did you watch (1 ⫽ None; 2 ⫽ A Little of the Coverage; 3 ⫽ Some of the Coverage; 4 ⫽ All of the Coverage)? Attention to Presidential Debates Not everyone had a chance to see all of the presidential and vice-presidential debates. How about you? How many debates did you have a chance to watch, or didn’t you have a chance to watch any? (1 ⫽ One; 2 ⫽ Two; 3 ⫽ Three; 4 ⫽ Four) Notes 1. While the trends in negative news and lack of trust in government have occurred simultaneously, the jury is still out as to whether there is a cause and effect relationship. 2. Skaperdas and Grofman (1995) found that frontrunners and weaker candidates employ very different campaign strategies. Frontrunners tend to engage in more positive campaigning. Weaker candidates, in contrast, tend to be more negative especially toward the frontrunner. This is borne out in the 1996 presidential campaign. 3. A third party challenge was launched by another well-known figure, Ross Perot, who had contested the presidency against Clinton in the immediately preceding presidential campaign of 1992. 4. Funding for this project was provided by the Pew Charitable Trusts, the Ford Foundation, and the National Science Foundation Grant #SBR9601418. 5. Authors may be contacted for a summary of the panel design. 6. The New Hampshire sample was not significantly different from the national sample in terms of the variables examined here, and therefore the two samples were combined for this analysis to produce a total sample of 1,457. 7. Additional information on response rates is available from the authors. 8. Candidate feeling thermometer and trait items from the NES (i.e., really cares about people like you, provides strong leadership, knowledgeable, trustworthy, and gets things done) were employed. In addition, respondents were asked to compare the candidates’ positions on such issues as: crime, taxes, Medicare, abortion, and reducing the deficit. Whereas in other campaigns the line between the candidates’ issue positions was fairly clear, in the
162
CRIGLER, JUST, AND BELT
9. 10.
11.
12.
1996 campaign it was difficult to delineate between the positions of Clinton and Dole. The communications of the two campaigns were deliberately designed to blur the distinctions as each candidate moved toward the center to capture votes. See Wattenberg and Brians’ (1999) findings on the importance of realworld experiments for assessing the impact of campaign negativity. In order to make the revised stories as realistic as possible, Marvin Kalb, a former correspondent for NBC news, copyedited the new text and read the new voice-overs. The positive ads set out each candidate’s policy positions on a range of issues, including taxes, welfare, and crime. Clinton’s ad was taken from his State of the Union message to Congress. Dole’s ad was made up of speech clips, voice-overs, and text on screen. The negative Clinton ad was based on a news clip that showed Dole sliding back and forth behind Newt Gingrich, as a kind of eminence grise. The ad voice-over emphasized the Republican leaders’ cooperation in the budget debate. The negative Dole ad also focused on budget issues, and contained an amusing juxtaposition of Clinton’s changing projections of when he planned to end the federal deficit. The videotapes were professionally edited and produced so that the final version of each tape lasted less than 15 minutes and contained a news story and two advertisements, followed by another news story. Each videotape concluded with a set of oral instructions and questions and a printed answer sheet that was mailed along with the videotape. While the questions were being asked in the audio track, the video displayed the question number that corresponded to the question on the answer sheet. The questions repeated many of the vote preference, candidate thermometer, trait, issue, and emotion questions that were asked in the wave one telephone survey. In order to simplify the task, the answer sheet was restricted to a single page. Conducting a field experiment such as this is fraught with problems. In the rapidly unfolding campaign, there is always the risk that real-world events will have a differential impact on the subjects depending on when they viewed the stimulus. Therefore, the video experiment was kept in the field as briefly as possible and each respondent’s answer sheet contained a record of the date on which the experiment was completed and the date on which the answer sheet was received at the University of Michigan. In fact, no obviously disruptive event seemed to have occurred while the experiment was in the field. Interestingly, the presence of a difficult task (the video experiment) early in the panel design did not appear to depress participation in later waves of the survey. The rate of return for the September mail survey wave was about 80 percent, which means that many people who did not participate in the video experiment were still willing to return the mailed questionnaire. The combination of mail and telephone surveys poses other kinds of analytical problems, especially the difficulty of maintaining a consistent sample. But it has the advantage of substantial cost saving to the researcher, while permitting flexibility in capturing responses to particular campaign events, such as
THREE FACES OF NEGATIVE CAMPAIGNING
163
conventions and debates. To augment the panel and experimental data, we have conducted an extensive content analysis of candidate messages and news coverage of the presidential campaign. These data are available from the International Consortium for Political and Social Research at the University of Michigan. 13. Figures represent spots aired weighted by size of market. Dole’s advertising was also more negative than Clinton’s was, in terms of ads made (47 versus 23 percent), ad buys (55 versus 18 percent), and spots run (70 versus 28 percent). Source: Goldstein 1997. 14. These models were also estimated using retrospective and prospective measurements of respondents’ economic well-being in lieu of satisfaction with Clinton’s handling of the economy. No significant difference was obtained.
This page intentionally left blank
CHAPTER 9 RACIAL CUES IN CAMPAIGN NEWS: THE EFFECTS OF CANDIDATE STRATEGIES ON GROUP ACTIVATION AND POLITICAL ATTENTIVENESS AMONG AFRICAN AMERICANS* Vincent L. Hutchings, Nicholas A. Valentino, Tasha S. Philpot, and Ismail K. White
ecades of scholarship in the field of electoral behavior have established that the chief influence of campaigns is to reinforce or activate latent predispositions so that voters behave consistently with their underlying interests (Berelson, Lazarsfeld, and McPhee 1954; Campbell et al. 1960; Finkel 1993; Gelman and King 1993; Klapper 1960; Lazarsfeld, Berelson, and Gaudet 1944; Patterson and McClure 1976; Petrocik 1996; Zaller 1992). Little attention, however, has been devoted to identifying the specific process by which activation occurs. In particular, we have yet to identify specific catalysts, in the swirling confusion of campaign communication, for activating latent preferences. The question of how campaigns remind citizens where they stand in relation to the parties and candidates vying for political power may seem trivial at first blush: the candidates clearly convey their partisan affiliations and the individual’s own predispositions take over from there. The problem with this simple description of the activation process is that it brings us back to the point that the authors of the American Voter struggled with when they originally proposed the psychological model of voting behavior: The activation process does not seem to work equally well across elections
D
166
HUTCHINGS, VALENTINO, PHILPOT, AND WHITE
(Campbell et al. 1960). Sometimes Democratic voters are more effectively mobilized by the campaign, whereas at other times Republicans are more effectively activated. And this variation seems determinative in many elections (Finkel 1993). In other words, of the many messages articulated during the course of the campaign, what are the particular cues that effectively catalyze the activation process? In this chapter, we attempt to fill this explanatory void by synthesizing older ideas about the role of group cues in campaign communication with new theories about the impact of emotional reactions in the realm of politics. We argue that candidates can activate latent support among large social groupings by emphasizing the distance between them and their opposition in terms of their support for group interests. The reason group cues are so effective in the activation process is that they trigger various emotional reactions among group members, reactions that in turn affect candidate evaluations and stimulate campaign learning (Marcus, Neuman, and MacKuen 2000). In sum, we hope to illuminate and precisely test a mechanism by which different campaigns produce fairly consistent, yet never identical, electoral outcomes. Candidates, Campaigns, and Political Perceptions How do campaigns activate voters and stimulate campaign learning? Less attention has been devoted to this question than the logically prior issue of whether campaign messages influence political preferences.1 The implicit assumption of this literature is that campaign stimuli are influential because they raise the profile of group-relevant issues such that voters recognize the relationship between their political predispositions and their vote choice.2 For example, Berelson and his colleagues found that Harry Truman’s emphasis on working-class issues during the latter stages of the 1948 presidential campaign had the effect of activating support among voters interested in these issues (Berelson, Lazarsfeld, and McPhee 1954, pp. 262–270). Subsequent work focused on party identification as the most directly activated attachment during campaigns (Campbell et al. 1960). Since partisanship was theorized to lie “downstream” in the funnel of causality of factors ultimately producing the vote choice, it was reasonable to assume that campaign communication would be most effective if it activated this dimension. Still, none of these studies attempted to specify the types of campaign cues that would activate predispositions. Recently, scholars have begun to explore the possibility that emotional reactions play a vital role in political mobilization and persuasion (Brader 2005; MacKuen et al. 2001a, 2001b; Marcus and MacKuen 1993; Marcus, Neuman, and MacKuen 2000). The theory of “affective intelligence” argues that political stimuli generate emotional reactions such as enthusiasm
RACIAL CUES IN CAMPAIGN NEWS
167
or anxiety that then serve as powerful motivators of candidate evaluations and political attentiveness. Marcus and his colleagues draw important distinctions between the political consequences of enthusiasm and anxiety (Marcus, Neuman, and Mackuen 2000). The former is associated with what they refer to as the “dispositional system” that promotes the maintenance of previously learned behaviors, encouraging individuals to apply their traditional party allegiances. The latter emotion derives from the “surveillance system” and prompts individuals to respond to threatening information with increased attentiveness. Although the theory of affective intelligence is provocative and holds significant promise, no studies to date have explored the particular message factors in campaign communication that produce emotional reactions such as enthusiasm or anxiety. Many of the claims are based on cross-sectional survey data, which might provide good estimates of the extent of emotional reactions in a population and their relationship to political perceptions, but are very imprecise as means for determining the timing or environmental triggers for these emotions. For example, although Marcus and MacKuen (1993) show that anxiety is associated with learning, they cannot determine whether this association occurs because of campaigninduced anxiety. However, other studies by these researchers have utilized experimental designs to manipulate anxiety levels via cues about the aggressiveness of unpopular groups (Marcus et al. 1995; Marcus, Wood, and Theiss-Morse 1998). Additionally, Marcus, Neuman, and MacKuen (2000) report on a second set of experiments that explore emotional reactions to political advertisements. Still, these experiments rely on actual campaign ads and therefore cannot determine which particular aspects of the message produce the emotional reaction. Moreover, they do not demonstrate that these emotional reactions subsequently influenced candidate evaluations or led to increased learning effects. In other words, a general theoretical framework for linking campaign communication, emotional reactions, and campaign learning and persuasion has yet to be fully tested. We suggest that candidate distinctiveness, particularly with regard to competing social-group interests, represents a powerful class of emotionproducing stimuli that candidates employ in their campaigns. In politics, feelings of enthusiasm or anxiety may be produced when candidates diverge on group-relevant issues. Some previous work has also argued that emphasizing relevant issues is a necessary, but not sufficient, condition for activating political engagement and candidate support (Campbell et al. 1960; Key 1966; Page and Brody 1972). Specifically, this research argues that, in addition to raising salient issues, candidates must also clearly distinguish their position from their opponent’s. In the absence of such differentiation and conflict, the voters have no clear basis on which to bring
168
HUTCHINGS, VALENTINO, PHILPOT, AND WHITE
their political predispositions to bear on their political judgments, even if the issues are prominently discussed in the media.3 The social psychological literature on adaptation level theory and the “contrast effect” is also important, because varying distinctiveness itself can alter the evaluation of a given object, even when that object is unchanged (Helson 1964; Manis and Moore 1978; Petty and Cacioppo 1981). In brief, “. . .when an object or issue is evaluated in the context of very positive stimuli, it will be rated less favorably along the same dimension of judgment than if it is evaluated in the context of very negative stimuli” (Petty and Cacioppo 1981). Starting with these insights, we argue that political communication emphasizing candidate distance on issues salient to relevant social groups will trigger emotional responses, which will subsequently affect political evaluations and political learning. Our focus on groups is motivated by the assumption that the American party system is organized around various group cleavages in society such as race, gender, class, religion, ethnicity, and region (Axelrod 1972; Berelson, Lazarsfeld, and McPhee 1954; Carmines and Stimson 1989; Stanley and Niemi 1991). Additionally, decades of public opinion research suggests that group-based political thinking is pervasive and thus candidate appeals will be most effective when they resonate with group considerations (Converse 1964; Conover 1984, 1988; Conover and Feldman 1981; Dawson 1994; Nelson and Kinder 1996). Consequently, we hypothesize that an effective way for candidates to mobilize support among core constituencies and heighten campaign learning is to highlight the implications of the election for salient group interests by emphasizing candidate distinctiveness. Since group interests are frequently in conflict, campaign appeals that generate support among one group may well diminish support among another group. A group dimension commonly considered to produce zero-sum political conflict in America is race (Bobo 2000; Bobo and Hutchings 1996). Studies of non-campaign news content have indicated that the typical frames used to depict blacks and whites may consistently reinforce and exacerbate racial conflicts in society (Entman 1992; Entman and Rojecki 2000). We expect, therefore, that appeals highlighting candidate distinctiveness on issues relevant to African Americans may, on average, drive white support away from the Democratic candidate as it generates greater support from blacks. In addition to affecting political preferences, campaign messages emphasizing candidate distinctiveness should also lead to increased levels of campaign learning. This is because when candidates diverge on group-relevant policy dimensions, at least one candidate necessarily adopts positions antithetical to one or more social groups. For this reason, we do not expect all citizens exposed to this message to respond with increased anxiety and hence attentiveness. Instead,
RACIAL CUES IN CAMPAIGN NEWS
169
only individuals whose group interests are directly implicated should respond in this way (Boniniger, Krosnick, and Berent 1995). This expectation draws us into the long-standing debate in the field of political communication as to the most effective determinants of political news reception (Delli Carpini and Keeter 1996; Hutchings 2001; Iyengar 1990; Price and Zaller 1993). In short, our examination of campaign distinctiveness should also provide us with information as to whether general interest in politics or group-specific interests best facilitates campaign learning. In summary, we predict that media cues depicting candidates as distinctive on racial issues should effectively activate political support. If the candidates adopt the political parties’ traditional stand on racial issues, African Americans should be drawn toward the Democratic candidate whereas whites should move closer to the Republican candidate. Second, when candidates do not appear distinctive on racial issues, campaign communications should not activate group-specific support in this way. Third, emotions act as the mechanism that mediates campaign activation. That is, candidate distinctiveness activates support because it heightens feelings of enthusiasm for candidates whose policy positions are consistent with citizen’s underlying interests. Fourth, candidate distinctiveness should also produce greater political attentiveness, particularly among individuals most threatened by the issue positions of the candidate furthest from their ideal point. Fifth, this learning effect should also occur directly because of the heightened sense of anxiety engendered by the threat cues. Methods and Procedures In order to understand the process by which campaigns evoke feelings of enthusiasm and anxiety and thereby increase attentiveness and promote opinion change, one must manipulate the strategies adopted by major-party candidates and then measure the emotional and political reactions of the citizens exposed to those strategies. An experimental design is best suited to examine these issues. The chief virtue of experiments is that they allow the researcher to isolate and manipulate the factors that might produce changes in attitudes or behavior (Kinder and Palfrey 1993). This method has become increasingly popular in the study of political communications (Gilliam and Iyengar 2000; Iyengar and Kinder 1987; Nelson, Clawson, and Oxley 1997; Valentino, Hutchings, and White 2002). While surveys are vital for estimating population means and trends, they are weaker at determining the causal impact of specific media content. The data used to test our hypotheses are drawn from an experiment conducted in July of 2000 in our Media Lab at the University of Michigan in Ann Arbor. The study utilized a convenience sample of 198 adult, non-student
170
HUTCHINGS, VALENTINO, PHILPOT, AND WHITE
residents from the area surrounding the university. Blacks were oversampled in this study and constituted 29 percent of our sample whereas whites represented 56 percent and the remainder (15 percent) identified as Asian American, Native American, Hispanic, or “other.”4 Subjects were recruited individually with flyers distributed to local businesses, university office buildings and in a downtown area adjacent to the university. Each was told she or he would receive $15 for answering questions about “current events.” As subjects entered the lab they were randomly assigned to one of three experimental conditions (two treatment conditions and one control), and then escorted to a computer terminal. In order to minimize interviewer biases, subjects interacted solely with the computer throughout the session. After completing a pretest questionnaire about the type of radio and television programs they preferred, the computer instructed subjects to read a series of short newspaper articles. Each subject in the treatment groups viewed two different nonpolitical articles and one political story.5 Those assigned to the control group read only the nonpolitical articles.6 Following the articles, subjects responded to a number of questions involving their political views, knowledge, and participatory intentions. Our analyses focus on three dependent variables. The first are the emotions of enthusiasm and anxiety. Drawing on the work of Marcus, Neuman, and MacKuen (2000) we measure reactions of enthusiasm to the candidates with a single question asking whether Al Gore has ever made our subjects feel hopeful.7 Anxiety is measured with a similarly worded item asking if George W. Bush has ever made subjects feel afraid. Candidate evaluations are measured by subtracting the feeling thermometer score for Bush from the thermometer score for Gore. The feeling thermometer, which ranges from 0 to 100, is regarded as a good measure of general evaluations of the candidates, in part because it captures both the direction and intensity of support (Abramson, Aldrich, and Rhode 1994). The difference variable has been re-coded so that it ranges from ⫺1 to 1 with positive values indicating more favorable impressions of Gore, relative to Bush, and negative values indicating the reverse. Campaign learning is measured with perceptions of candidate issue differences on affirmative action, federal funding for public schools, and women’s equality in the work force. Perceptions of each candidate’s stand on these policies are derived from standard seven-point issue-scales.8 Difference measures are then computed by subtracting subjects’ perceptions of Bush’s position from their perception of Gore’s position. These variables have been re-coded to range from ⫺1 to 1, with positive values indicating the perception that Gore is the more liberal candidate, relative to Bush. The political stories used in the experiment were written by the researchers and were designed to highlight either the Democratic or
RACIAL CUES IN CAMPAIGN NEWS
171
Republican perspective on George W. Bush’s efforts to court African American voters in the 2000 presidential contest. One version of the story (referred to as the Similar Condition) begins with the headline “Gore, Bush Similar on Black Issues” and is accompanied by color photographs of each candidate interacting amicably with NAACP Chairman Julian Bond. The article focused on appearances by Gore and Bush before the annual convention of the civil rights organization. This version of the story begins with the reporter observing that though the candidates have tried to distinguish themselves from one another, “. . .on issues affecting African Americans the candidates are taking surprisingly similar positions.” The article goes on to note that Gore and Bush both support stronger enforcement of civil rights laws, racially diverse administrations, improvements in public education, and expanded access to health insurance. The alternate version (referred to as the Difference Condition) carries identical issue content but depicts the candidates as sharply diverging with regards to the interests of African Americans. For example, the headline for this version reads, “Gore, Bush Differ on Black Issues” and is accompanied by a photo of Bush amidst a crowd of white supporters at Bob Jones University.9 Gore is again shown interacting comfortably with Julian Bond. The text of this article begins much like the previous version, except that the writer concludes, “. . .on issues affecting African Americans the candidates have adopted dramatically different positions.” Their speeches before the NAACP are also characterized in starkly different fashions. In a colorful turn of phrase, Chairman Bond describes the vice-president’s speech as substantive whereas Governor Bush “. . .had a drive-by photo opportunity.” It is worth noting that Gore’s position on the issues is the same as in the previous version, whereas Bush is now portrayed as far less sympathetic than Gore to black interests. Perhaps the most noteworthy example involves civil rights issues. The article notes “. . .Gore has a solid record on civil rights issues and is a strong supporter of affirmative action policies. Conversely,. . . [Bush] strongly opposes affirmative action policies.”10
Results Candidate Distinctiveness and Emotions Our first set of analyses involves the impact of candidate issue difference on emotions. Our expectation is that the story emphasizing Gore’s support for black interests, relative to Bush’s opposition, will generate heightened enthusiasm for Gore among African Americans and weaker countervailing reactions among whites and other nonblack subjects.11 When the candidates appear similar on black issues, these effects should be muted. These
172
HUTCHINGS, VALENTINO, PHILPOT, AND WHITE
results are presented in figure 9.1.12 In the control group, we find that enthusiasm for Al Gore reaches only moderate levels for both blacks and nonblacks, in spite of the Democratic bias in our sample. When the candidates are depicted as similarly interested in serving black interests, enthusiasm for Gore among blacks rises only slightly. Interestingly, in the similar condition, enthusiasm for Gore among nonblacks begins to decline, although this result is not statistically significant. However, when Bush adopts a staunchly conservative position on issues relevant to African Americans, black enthusiasm for Gore increases dramatically. Indeed, the predicted probability more than doubles, going from .29 to .61. Levels of enthusiasm among nonblacks remain largely unaffected. Surprisingly, enthusiasm among whites and other nonblacks does not decline any further in this condition perhaps owing to a “floor effect” within this generally liberal sample. In general, the manipulation of candidate distance on racialized issues produced the pattern of results we expected at the outset. When the candidates are described as different on “black issues” African Americans are far more likely than whites to associate positive emotions with the Gore candidacy. However, this response is not preordained. When our subjects are not exposed to the story about the candidates or when Bush is described as a moderate, black enthusiasm for Gore remains muted. To the extent that whites and others are affected at all, they generally move in the opposite direction as a result of these same cues, becoming less enthusiastic about 0.61 Probability that gore has made you feel hopeful
0.6 0.5 0.39
0.4
0.34 0.3
0.29
0.32 0.28
0.2 0.1 0 Control
Same
Different
Experimental conditions Black
Nonblack
Figure 9.1 Probability that Something about Al Gore Makes Subject Feel Hopeful.
RACIAL CUES IN CAMPAIGN NEWS
173
Gore when he stakes out traditional, liberal positions on issues of race, regardless of his position relative to Bush. Candidate Distinctiveness, Enthusiasm, and Candidate Evaluations In order to be politically consequential, campaign strategies must do more than produce an emotional response in various groups. They must also influence candidate evaluations. In table 9.1, we explore, on the one hand, whether candidate distinctiveness can activate candidate support, and on the other hand, whether apparent candidate similarities can effectively prevent this activation. The independent variables are located on the left-hand side of the table. Dummy variables are entered into the analysis for each treatment condition, with the control group as the excluded category.13 The principal variables of interest in these analyses are the race-by-experimental condition interactions, located near the bottom of the table. As anticipated, only the race-by-difference condition interaction is both positive and statistically significant. The net effect of the difference condition among blacks Table 9.1 Predicting Effects of Manipulating Candidate Distance on Support for Presidential Candidates by Race Predictors
Candidate Evaluations
Intercept
⫺.44*** (.08) ⫺.14* (.07) ⫺.18** (.07) ⫺.07 (.09) .15 (.14) .31** (.13) 198 .36
Similar condition Different condition Black Similar * Black Different * Black N Adjusted R2
Note: The dependant variable is the difference between the Bush and Gore feeling thermometers. Higher values indicate greater support for Gore. Controls, not shown here, include party identification, ideology, and gender. Cell sizes are as follows: Control ⫽ 80; Similar Cell ⫽ 58; Different Cell ⫽ 60. * p ⱕ.05, ** p ⱕ.01, *** p ⱕ.001 for one-tailed test, except constant.
174
HUTCHINGS, VALENTINO, PHILPOT, AND WHITE
(⫺.18 ⫹.31) is a 13-point increase in support for Gore, relative to Bush.14 Interestingly, although Gore remains committed to black concerns in both experimental conditions, when Bush appears more moderate black support for Gore does not change relative to the control group.15 Blacks are not the only subjects affected by the political stories. The negative, and significant coefficients on both experimental variables suggests that support among nonblacks for Gore, relative to Bush, declines after reading either political story. As anticipated, this effect is somewhat larger when the candidates’ positions sharply diverge (⫺.18 versus ⫺.14). However, this difference is not dramatic. Thus, highlighting candidate distinctiveness is an effective way to mobilize one’s supporters. However, these results suggest that one risk in pursuing this strategy is that candidates may also succeed in mobilizing their opponents. Having now established that candidate distinctiveness produces heightened enthusiasm and activates candidate support among African Americans, we turn to the question of whether the latter occurs because of the former. That is, does candidate distinctiveness lead to increased support for Gore because it primes feelings of hopefulness? We test this hypothesis in table 9.2. Here, the variable of interest is the three-way interaction of the difference condition by feelings of hopefulness by race, located at the bottom of the table.16 If our hypothesis is correct, then the coefficient on this variable should be both positive and statistically significant. This expectation is confirmed. The substantive magnitude of the triple interaction is quite large. For example, turning first to those black subjects who feel hopeful about Gore, the vice-president enjoys no relative advantage over Bush on the feeling thermometers in the control group. Holding all other variables in the model to their mean or median produces a value of “( )” indicating that black subjects make no distinction between the two candidates. However, among blacks who feel hopeful about Gore and who are exposed to information suggesting that the candidates take starkly different positions on black issues, Gore’s relative advantage over Bush is significantly higher. These subjects give Gore a 36-point advantage over Bush on the feeling thermometer. The interaction of hopefulness and candidate distinctiveness has a different effect among nonblacks. In the control group, these subjects give Gore a 35-point advantage over Bush on the feeling thermometer. Interestingly, when the candidates are depicted as divergent on black issues, this advantage declines to 22 points. Among subjects who do not feel hopeful toward Gore, the effect of exposure to the difference condition results in less support for the Democrat. Moreover, this is true for both blacks and nonblacks. Thus, as anticipated, enthusiasm for Gore does act as a mediator of campaign activation but only for African Americans.17
RACIAL CUES IN CAMPAIGN NEWS
175
Table 9.2 Predicting Effects of Manipulating Candidate Distance on Support for Presidential Candidates by Feelings of Hopefulness toward Gore and Race Predictors Intercept Different condition Gore hopeful Different * Hopeful Black Different * Black Black * Gore hopeful Different * Hopeful * Black N Adjusted R2
Candidate Evaluations ⫺.44*** (.09) ⫺.19* (.09) .32*** (.09) .05 (.14) .13 (.11) .01 (.18) ⫺.49** (.18) .49* (.26) 140 .42
Note: The dependant variable is the difference between the Bush and Gore feeling thermometers. Higher values indicate greater support for Gore. Controls, not shown here, include party identification, ideology, and gender. Cell sizes are as follows: Control ⫽ 80; Different Cell ⫽ 60. * p ⱕ.05, ** p ⱕ.01, *** p ⱕ.001 for one-tailed test, except constant.
Threat Cues and Campaign Learning Our political stories manipulating candidate distinctiveness should have done more than just influence the relationship between enthusiasm and candidate evaluations. They should also have raised the specter of threat, especially for African Americans. This is because the prospect of a Bush presidency had much more dire implications for black interests in the difference condition than in the similar condition. In figure 9.2, we test whether the difference condition generates feelings of fear toward Bush. As anticipated, blacks are more likely to indicate that something about Bush makes them afraid when the candidates are depicted as different on black issues. This result is statistically significant, relative to the control, at the .05 level for a one-tailed test (results for the logistic regression analysis not shown). In the control group, neither blacks nor nonblacks have a high probability of indicating that Bush makes them fearful. Whites and other
176
HUTCHINGS, VALENTINO, PHILPOT, AND WHITE
Probability that bush has made you feel afraid
0.3
0.28
0.25
0.23
0.2 0.15
0.18 0.18 0.15 0.13
0.1 0.05 0 Control
Same Experimental conditions Black
Different
Nonblack
Figure 9.2 Probability that Something about George W. Bush Makes Subject Feel Afraid.
nonblacks are somewhat more likely to express this emotion, but this difference is not statistically significant. When the candidates’ positions on racialized issues appear similar, there is a modest increase in feelings of fear among blacks and an equally modest decline among other subjects. Again, these effects fall short of statistical significance. However, when Bush adopts a more threatening posture toward African American interests, black anxiety toward Bush rises considerably. Even with the Democratic bias of our sample, we find that whites move in the opposite direction. Clearly, Bush’s conservative position on black issues was perceived as threatening, but only for blacks. In table 9.3, we examine the impact of group threat cues on campaign learning. As indicated earlier, information on the candidates’ position on affirmative action and public school funding was included in our political articles. The articles made no mention of the candidates’ stand on women’s equality. Therefore, if the group threat cues were effective, they should increase perceptions of candidate difference on the two race-relevant issues and not on women’s equality. Further, and consistent with the issue salience hypothesis, African Americans should be especially sensitive to this information. When the candidates differ on issues they consider important, blacks should be far more likely to recognize this than nonblacks. In addition to the controls described earlier, the analyses in table 9.3 also
RACIAL CUES IN CAMPAIGN NEWS
177
Table 9.3 Predicting Effects of Manipulating Group Threat Cues on Campaign Learning by Race and Levels of Political Information Predictors Intercept Similar condition Different condition Black Similar * Black Different * Black Political Information Similar * Information Different * Information N Adjusted R2
Affirmative Action ⫺.01 (.13) ⫺.34* (.19) ⫺.32* (.18) ⫺.34** (.11) .39** (.16) .61*** (.15) .23 (.16) .51* (.26) .47* (.23) 198 .23
School Funding ⫺.17 (.10) ⫺.28* (.14) ⫺.15 (.13) ⫺.16* (.08) .09 (.12) .33** (.11) .20* (.12) .32* (.20) .30* (.17) 198 .27
Women’s Equality ⫺.31** (.11) ⫺.12 (.16) ⫺.10 (.15) ⫺.04 (.09) .16 (.13) .16 (.13) .27* (.13) .21 (.21) .14 (.19) 198 .18
Note: The dependant variables are the candidate issue positions, ranging from ⫺1 to 1, described in note 8. Higher values indicate perceptions that Gore is the more liberal candidate. Controls, not shown here, include party identification, ideology, and gender. Cell sizes are as follows: Control ⫽ 80; Similar Cell ⫽ 58; Different Cell ⫽ 60. * p ⱕ .05, ** p ⱕ .01, *** p ⱕ .001 for one-tailed test, except constant.
include controls for subjects’ level of political information, and the interaction of this variable with the experimental conditions.18 If, as some argue, citizens are primarily “information generalists” then the more politically informed should recognize when the candidates issue positions diverge, regardless of their race. The first column presents the results for affirmative action policies. Interpreting these results at first glance is somewhat difficult in light of the multiple interaction terms. However, by manipulating the values on the experimental conditions while holding all other variables constant at their mean, we can generate predicted scores for a number of different groups. For example, nonblacks in the control condition, on average, view Gore as .34 points more liberal than Bush on this dimension. Non, blacks exposed to either the similar or difference conditions are only marginally different (i.e. .299 and .301, respectively). Thus, on balance the experimental conditions provide no information gain with respect to the candidates’ position on affirmative action for the average nonblack subject. This conclusion is
178
HUTCHINGS, VALENTINO, PHILPOT, AND WHITE
somewhat different for nonblack subjects who score particularly high on the political information scale, which runs from 0 to 1. We find that Gore is perceived as significantly more liberal than Bush (i.e., a 15-to-17-point increase), after exposure to either of the manipulations, relative to the control group. Consequently, we conclude that citizens who are especially informed about politics in general are also more susceptible to learning about new political information. Are citizens with a more group-oriented view of politics also more likely to learn about group-relevant political information? To answer this question, we examine how blacks’ perception of the presidential candidates’ stance on various public policies are affected by exposure to our news stories. As anticipated, we find that blacks are considerably more receptive to information about the candidates’ position on affirmative action than are nonblacks. Interestingly, however, this effect is not a chronic one. Indeed, in the control group, blacks are significantly less likely than whites (i.e., 34 points) to view Gore as more supportive of affirmative action. It is unclear whether this is due to blacks’ relative lack of attention to this issue, or their displeasure that Gore was not more forceful in defense of this policy. In any case, black attentiveness to, and accuracy about, the candidates’ stands on affirmative action improves dramatically when exposed to the campaign news stories. Compared to the control condition, blacks recognize Gore as more liberal even when the candidates are depicted as similar on civil rights issues. Still, the net effect of this story merely eliminates the black-white gap observed in the control condition. When the candidates are characterized as different on affirmative action, blacks are far more likely to absorb and retain this information than are whites. In fact, the counterintuitive black-white gap in perceptions described in the control condition reverses. Instead of being substantially less likely than whites to view Gore as more liberal on affirmative action, blacks are substantially more likely to reach this conclusion. This difference translates into a 26-point black advantage in perceptions that Gore is the more liberal candidate.19 We uncover similar results, although more modest in size, for perceptions of candidate difference on public school funding (see column 2). Once again, very little campaign learning appears to occur among the typical nonblack in our sample. However, nonblack subjects particularly interested in, and informed about, politics are more apt to recognize candidate differences after exposure to the experimental conditions. As with perceptions of the candidates’ positions on affirmative action, blacks in the control condition are somewhat less likely than nonblacks to view Gore as the more liberal candidate on education. This counterintuitive difference is unchanged in the similar condition. However, when the candidates are described as distinctive, the racial gap between blacks and the comparison group is
179
RACIAL CUES IN CAMPAIGN NEWS
essentially eliminated. Finally, when asked about a nonracial issue that was not mentioned in our campaign coverage, the accuracy of neither blacks nor whites is affected. In summary, these results suggest that under some circumstances, more narrow group-oriented interests in politics are at least as effective a predictor of news acquisition as are more general interests. Our last tests involve whether feelings of anxiety act as mediators of campaign learning just as enthusiasm influenced candidate support. Table 9.4 presents the triple interaction of race-by-difference condition-by-anxiety for both affirmative action and public school funding.20 The three-way interaction is both positive and statistically significant for affirmative action but not school funding. In retrospect, this finding makes sense. Although education policies have indirect implications for black interests, affirmative action speaks directly to the mainstream black agenda. Efforts to dismantle these programs should provoke more anxiety among African Americans and this anxiety should induce more political attentiveness. The magnitude of the race-by-difference condition-by-anxiety effect is not trivial. We find, for example, that a typical nonblack subject in the Table 9.4 Predicting Effects of Manipulating Group Threat Cues on Campaign Learning by Race Predictors
Affirmative Action
School Funding
Intercept
.08 (.11) .04 (.11) .10 (.12) ⫺.15 (.19) ⫺.28** (.12) .36* (.19) ⫺.53* (.26) .67* (.34) 140 .15
⫺.03 (.07) .02 (.07) .08 (.09) .14 (.13) ⫺.13 (.09) .25* (.13) ⫺.30* (.18) .14 (.13) 140 .20
Different condition Afraid of Bush Different * Afraid Black Different * Black Black * Afraid Different * Afraid * Black N Adjusted R2
Note: The dependant variables are the candidate issue positions described in appendix 1. Higher values indicate perceptions that Gore is the more liberal candidate. Controls, not shown here, include party identification, ideology, and gender. Cell sizes are as follows: Control ⫽ 80; Different Cell ⫽ 60. * p ⱕ .05, ** p ⱕ .01.
180
HUTCHINGS, VALENTINO, PHILPOT, AND WHITE
control condition, who indicates that something about Bush makes them feel afraid, is more than capable of recognizing Gore’s relative liberalism on this issue and consequently places him about .48 points higher on this scale than Bush. Nevertheless, these perceptions are largely unaffected by the experimental condition, as these subjects place Gore about .37 points higher than Bush. Blacks in the control group, who express some anxiety about Bush, on the other hand, are more likely to view him as the more liberal candidate (i.e., ⫺.33) on affirmative action. This perception may reflect Bush’s largely successful efforts to portray himself as a more “compassionate” Republican during this phase of the campaign. In any case, blacks who indicate that something about Bush makes them afraid and who are also exposed to the difference condition are significantly more likely to recognize Gore’s more liberal position on affirmative action. The effect of exposure to the experimental condition is quite dramatic—on average, blacks in this condition view Gore as 60 points more liberal on this issue. Indeed, these blacks almost place the candidates as far apart as they can, given the range of this variable. Among subjects who do not express fear toward Bush, blacks are again more cognizant of Gore’s relative liberalism in the difference condition. However, the discrepancy between the control and difference conditions is, much less stark here (i.e., a 40-point difference relative to a 93-point difference). Additionally, nonblacks are again largely unaffected by the stimulus. Conclusion The aim of this chapter was to illuminate the process by which campaigns activate latent support in the electorate. We hypothesized that manipulating candidate distinctiveness with regard to a particular social group dimension would provide a powerful cue to group members. These cues were expected to produce emotional reactions that would in turn stimulate candidate support and boost issue attentiveness. In short, we argued that these processes underlie the real impact of the campaign: to activate latent group support for candidates, bringing evaluations and perceptions into line with preexisting interests. These hypotheses all found support in our results. When campaign coverage presents the candidates as distinctive, African Americans’ support for Al Gore increased. Whites, on the other hand, expressed greater support for George W. Bush. Consistent with the affective intelligence theory, the effects among blacks were mediated through emotions of enthusiasm. Just as important, when Bush appeared more moderate, Gore’s relative advantage among blacks did not materialize. Given that blacks represent arguably the most Democratic group in America, this result is significant. It suggests that even groups with a well-established
RACIAL CUES IN CAMPAIGN NEWS
181
reputation for support for one political party can be dislodged, to a degree, when the right campaign strategy is adopted. We also found that, when candidates highlight issues relevant to African American voters, blacks will be far more likely to absorb this information than nonblacks. Heightened emotional reactions, such as in this case anxiety, were also found to play a significant role in campaign learning. This was not the only mechanism at work here, however, as blacks who did not report such reactions nevertheless demonstrated evidence of increased attentiveness relative to nonblacks after exposure to the experimental conditions. This one study does not, of course, resolve the debate between proponents of the information generalist argument and the information specialist theorists. It does suggest, however, that some previous work may have been too quick to dismiss the role of domain-specific interests in encouraging political attentiveness. These results advance our knowledge about American elections in several ways. First, although previous work has demonstrated the linkage between emotional states and political judgments, no one has explored one of the particular triggers that can produce the effect: candidate distinctiveness. Previous work has identified the importance of threat cues in the political environment, but without fully specifying the contours and limitations of this strategy (Brader 2005; Hutchings 2001; Marcus and MacKuen 1993). This chapter has shown that highlighting candidate difference most effectively heightens candidate enthusiasm and anxiety. Political opponents, however, can defuse these appeals by appearing to adopt more moderate positions. This finding runs counter to the issue ownership hypothesis, which states that candidates will be most successful when they stick to issues on which their party’s reputations is strong (Petrocik 1996). However, our results fit nicely with the campaign strategies of recent presidential candidates. That is, we believe that neither George W. Bush’s “compassionate conservative” platform, nor Bill Clinton’s decision to run as a “new Democrat” were primarily ploys to gain supporters from the opposing camp. If that had been their goal, they would have to be considered failures. On the contrary, our results suggest that they were very effective ways of undercutting enthusiasm among their opponents’ core supporters. Second, our results help to explain why candidates highlight issue priorities even though most voters are inattentive to political matters most of the time. As Converse (1964) speculated almost 40 years ago, we found that voters are especially attentive when issues they regard as important are raised in political campaigns. In fact, we found that exposure to campaign news resulted in heightened learning effects among African Americans, in particular, even though this group traditionally scores lower than others on measures of general political knowledge. Thus, at least in some instances,
182
HUTCHINGS, VALENTINO, PHILPOT, AND WHITE
specialized interests are at least as effective a predictor of campaign learning as are more generalized measures of political interests. In short, issue appeals help to inform and mobilize core constituencies even if those constituencies are traditionally unengaged in politics. Third, our results indicate that, contrary to some accounts racial group concerns may not be chronically salient to African Americans (Dawson 1994; Lau 1989a). We found that, relative to the comparison group, blacks were not monolithic in their enthusiasm for the Democratic candidate nor immutably critical of the Republican candidate. Further, African Americans did not necessarily view the presidential candidates as divergent on “black issues.” These results suggest that, as with other voters, blacks’ views of the political landscape are heavily shaped by the communication strategies adopted by political candidates. Fourth, this chapter provides support for the notion that group-centric frames powerfully influence voters. Although our news articles were ostensibly directed toward African Americans they also influenced the views of whites and other nonblacks. For example, when Gore’s positions as a raceliberal are juxtaposed to Bush’s more conservative views, he gains black support but at the expense of support among other groups. These results suggest that adopting a position toward a high-profile group is an effective way to convey information about ones policy positions, but such a strategy does not come without costs. Finally, this study also contributes to our knowledge about the manner in which voters evaluate candidates. Although the debate between the relative merits of the spatial model of electoral choice versus the directional model remains unresolved, neither argument addresses the role that comparative judgments play in candidate evaluations (Downs 1957; MacDonald, Rabinowitz, and Listhaug 1998; Rabinowitz and MacDonald 1989; Westholm 1997). As we have shown, voters do not simply assess candidates based on how “close” they are to them on some uni-dimensional scale or, alternatively, which “side” of the issue the candidates come down on. They also consider the position of the candidates vis-à-vis each other. We hope these results cast additional light on the process by which campaigns activate latent support among groups in the electorate. By adopting divergent positions on group-relevant issues, and utilizing media strategies that highlight those distinctions, campaigns and news media outlets that cover them provoke emotional responses from large segments of the citizenry. These negative emotions are then translated into increased attention to politics, and support for the candidates who would best serve the interests of the group. The implications of this particular process of activation are important. The news media, with its preference for conflict and dramatic narrative, inadvertently feeds into this process. The public may or may not be well
RACIAL CUES IN CAMPAIGN NEWS
183
served, because the dimensions made salient by the candidates are chosen strategically and may not be the ones that would maximize the public good in some larger sense. If, for example, it is a fairly straightforward matter to activate latent support along the dimension of race, then depending on the racial breakdown of a constituency, one or the other party will always have an incentive to do so. The religious right may also receive a fairly steady stream of political cues during campaigns because they represent a politically homogeneous group that will likely respond quite consistently. Alternatively, women are likely more difficult to activate as a group in any election, because of their more heterogeneous political views. Although this study only focused on one group cleavage, we think a larger theory of the campaign activation process begins to emerge from these findings. However, more conclusive support for this theory must await future studies and an examination of additional groups. At this stage, our results suggest that campaigns will be best at activating those groups whose interests are relatively homogeneous with regard to the political system in which they are embedded. Notes * This research would not have been possible without the efforts of several graduate research assistants including Dmitri Williams, Lara Rusch, Matthew Beckmann, and Dara Faris. 1. The work of Charles Franklin represents an important exception to this trend (Franklin 1991). This scholar does examine the impact of campaign strategies on political perceptions, paying particular attention to the issue themes of incumbents and challengers. However, even Franklin’s work is more concerned with the number of issue themes raised rather than the specific content of those themes or the ways in which the issues emphasized by each candidate might influence voter perceptions. 2. Some recent work (Bartels 1997; Gelman and King 1993) arrives at a similar conclusion, although for somewhat different reasons. This work finds that although the presence of campaigns are important in activating political predispositions, this has less to do with the specific content of the campaign than with the presence of a vigorous and competitive partisan contest. 3. In news coverage of campaigns, it is unlikely that a given dimension will even become prominent in the political arena unless there is substantial elite conflict surrounding it, since such conflict is one of the news media’s most important triggers (Gamson 1992; Price 1989). 4. In spite of our reliance on adult subjects rather than students, our sample is not representative of the national population (although it compares quite well to the local population). For example, a disproportionate number of our subjects are Democrats (61%), and self-identify as ideologically liberal (50%).
184
HUTCHINGS, VALENTINO, PHILPOT, AND WHITE
5. All of the stories were presented so that they would appear authentic to the subjects. Thus, the masthead of the newspaper was superimposed over each article, and the byline was clearly visible at the beginning of each story. The entire stories are available upon request from the authors. 6. The nonpolitical stories dealt with wildfires throughout the United States, the new subway line in Los Angeles, and recent medical breakthroughs in restoring vision. The latter story was only viewed in the control group. 7. The specific question tapping this emotion read as follows: “Now we would like to know something about the feelings you have toward the candidates running for president. Has [Gore] ever made you feel [hopeful]?” This question (coded 0, 1), and all other emotion items used in this paper, is derived from the American National Election Study (ANES). 8. The specific wording for each question is as follows: “Some people feel that affirmative action policies should be ended. Others think that affirmative action should be continued.” “Some people believe that the federal government should spend much less money for public schools. Others feel that federal spending on public schools should be greatly increased.” “Some people think that women should have an equal role with men in running business, industry, and government. Others feel that women’s place is in the home.” Subjects were then presented with a 7-point scale and asked, “Where would you place [GORE or BUSH] on this scale?” 9. At the time of Bush’s visit, this southern university had a much-publicized policy of banning interracial dating. 10. To maximize the realism of the manipulation, great effort was expended in order to ensure that the information conveyed in each version would be credible to readers. For example, the candidates’ websites were carefully reviewed so as to accurately summarize their policy positions. Additionally, our stories drew upon actual news accounts of each candidate’s speech before the NACCP. 11. We group whites and other nonblacks together largely for convenience sake. Of the 30 subjects who indicate that they are neither black nor white, the vast majority classify themselves as “Other.” Moreover, the political profile of these subjects is closer to whites than blacks. For example, 74% of blacks identify as Democrats whereas this is true of only 62% of whites and 59% of the remaining subjects. Finally, our results in figure 9.1, and throughout the paper, are not substantively altered if we focus solely on black and white subjects. 12. The results in figures 9.1 and 9.2 are derived from logistic regression analysis. The predicted probabilities are estimated by manipulating a hypothetical subject’s race and the article that he or she read, while holding all other variables in the model at the population mean or median. The difference between blacks and nonblacks are statistically significant (p ⬍ .05; onetailed test), relative to the control group, only in the Difference Condition. Results from the full model are available upon request. 13. Although subjects were randomly assigned to conditions, we found that the experimental groups differed on a few key demographic variables: partisanship,
RACIAL CUES IN CAMPAIGN NEWS
14.
15.
16.
17.
18.
185
ideology, and gender. Consequently, all analyses include controls for these variables. We also examined specifications that included interactions for the experimental conditions and partisanship and ideology. In both instances we found that these interactions fell well short of statistical significance and that the race by difference condition was unaffected. An alternative explanation for these results might be attributed to an artifact of our experimental manipulations. Although we attempted to keep our experimental conditions as similar as possible it is nevertheless true that the “difference” condition specifically mentions the candidates’ positions on affirmation action whereas the “similar” condition only makes oblique reference to their positions on “civil rights.” This decision was a conscious one as depicting Bush as a moderate on affirmative action would not likely ring true with our more politically informed subjects. At any rate, if this minor difference is what really accounts for our results then we should find that attitudes on affirmative action, rather than race, are actually responsible for the findings presented in figure 9.1 and table 9.1. To examine this possibility, we re-ran the analyses described earlier with interactions for attitudes on affirmative action and the experimental conditions. In the case of both feelings of hopefulness for Gore, and candidate evaluations, the race-bydifference condition interaction remained robust whereas the affirmative action interactions were substantively and statistically inconsequential. Subjects in the “similar” condition are omitted here as figure 9.1 and table 9.1 have established that African Americans in this group do not view Gore more favorably, relative to the control condition. The fact that our experimental manipulations also activate candidate support among nonblacks suggests that candidate enthusiasm may also work as a mediator for this group. To explore this possibility, we analyzed the effects of the three-way interaction of experimental conditions-by-race-by hopefulness for George W. Bush on candidate evaluations. Consistent with our expectations, we found that the results were statistically significant for nonblacks, but only with the interaction containing the difference condition. However, the difference condition does not significantly increase feelings of hopefulness for Bush among nonblacks (although the results are in the anticipated direction). It is possible that the Democratic bias in our sample makes it difficult to uncover this relationship. The content of these scales vary somewhat from study to study. Still, they generally ask respondents to identify the political office of several national political figures or to provide basic information about the operation of the federal government. We adopted a combination of both strategies in our study. First, subjects were asked to identify the political office of Dennis Hastert. They were then asked whose responsibility it is to decide if a law is constitutional, how much of a congressional majority is required to override a presidential veto, which party has a majority in the House of Representatives, and which party is more conservative at the national level. Roughly 50% of our sample was able to answer 3 of the 5 questions
186
HUTCHINGS, VALENTINO, PHILPOT, AND WHITE
accurately. Finally, as in most representative samples, blacks scored considerable lower on this scale than other subjects (.49 versus .66, when the scale is re-coded 0–1). 19. Given the nature of our experimental manipulations, we are principally interested in subjects’ perceptions of candidate difference. Still, we also examined alternative ways of measuring campaign learning. For example, does exposure to our experimental conditions also encourage subjects to place the candidates on the right side (i.e., liberal or conservative) of an issue? In the case of affirmative action, this would entail recognizing that Bush is generally opposed to such policies whereas Gore typically favors them. Recoding the 7-point scales accordingly (e.g., all answers indicating the perception that Bush opposes this policy are coded “1” all else “0”), we find that both the similar and difference conditions significantly increase objectively accurate perceptions of both candidates. . .but again, only for African Americans. 20. We also examined the effects of the triple interaction for race*anxiety* similar condition. These results fell well short of statistical significance and were excluded on this basis.
CHAPTER 10 I LIKE HIM, BUT. . .: VOTE CHOICE WHEN CANDIDATE LIKEABILITY AND CLOSENESS ON ISSUES CLASH* David P. Redlawsk and Richard R. Lau
he importance of likeability, broadly defined, has been understood by political candidates and their handlers probably for as long as campaigns have existed. Common phrases such as “clothes make the man [woman]” and “looks can be deceiving” tell us much about how human beings are impacted by the visual (Dion, Berscheid, and Walster 1972). But candidate likeability goes beyond the physical, including personality traits that may be explicitly or implicitly applied to a candidate, sometimes simply on the basis of physical appearance (Riggle et al. 1992). Social psychologists have long documented a “beauty is good” stereotype (Berscheid and Walster 1974) where more physically attractive people are assumed to possess a range of more positive personality traits and to generate a more positive emotional response. Candidate physical attractiveness and personality clearly play some role, perhaps even the leading role, for many voters (Ottati 1990). It may be that in 2000, Vice-President Al Gore was at a distinct disadvantage to Texas Governor George Bush, given the general consensus that although quite knowledgeable on the issues—sort of a “policy wonk”—Gore appeared wooden and often less than likeable, while Bush despite his apparent limited grasp of many issue details, came across as warm and approachable. While campaign consultants know in their gut that likeability matters, a surprisingly limited amount of political science research has directly addressed the question of the role likeability plays when compared to other information voters may acquire.
T
188
REDLAWSK AND LAU
In an effort to fill some of this gap, we report a unique experiment where voters were exposed to presidential candidates in a simulated campaign. Some candidates were relatively good-looking with pleasant personalities—that is, they were likeable. At the same time, however, they were distant from the voter on the issues. Others candidates were ideologically close to the voter but sported less than attractive appearances and personalities. Throughout the course of the campaign, voters chose the information they wanted to learn about the candidates, and ultimately had to decide whether to support the more attractive candidate or the one who might better represent their policy interests. We find significant differences between political novices, who are more likely to be attracted to likeable candidates over issue-congruent ones, and experts, who seem more clearly focused on the issues. Theoretical Background In American presidential elections, voters can select from a wide variety of available information in making their choices. While not equally salient to all voters, issue positions, group endorsements, visual image, personality, and experience are all clearly important factors in distinguishing candidates (Lau and Redlawsk 1997, 2001a, 2001b). If all of these factors align themselves consistently in an election, the choice may be relatively easy. But we doubt this happens very frequently outside of the movies, where the hero is always smart, trustworthy, holds all the right issue positions, and looks like Robert Redford. More often, one candidate might have “good” positions on issues, but appear to have limited relevant experience. Another candidate might appear quite attractive physically and have a warm and likeable persona, but not be supported by the groups with which a voter identifies. What if voters have to choose between a candidate who is physically attractive and generally likeable (thus making the voter feel good) but on the wrong side of important issues, and one who takes issue positions more favorable to the voter but who is noticeably less attractive and sporting quite unlikeable traits that might generate a rather negative impression? What then? Do the cues inherent in physical image override the information provided by issue stands? And what if personality traits reinforce the physical image? We believe that while candidate-centered information such as pictures or personality traits may seem less valuable than issue positions, this does not mean they carry no useful content. It is reasonable to think that a candidate’s personality is relevant information for what it signals about the potential behavior of the candidate if elected; a politician’s personality is highly likely to have some influence on how he or she chooses from the myriad of actions available in every situation (Greenstein 1969).
I LIKE HIM, BUT. . .
189
Candidate likeability goes beyond the physical, including not just image, but the personality traits that people find more or less endearing. As Sigelman, Sigelman, and Fowler (1987) show, physical attractiveness, rather than directly affecting the vote, may actually condition assessments of personality, while it is the personality assessments that have more direct effects on the vote. In some elections, candidate pictures and the impressions they generate may be just about all voters have. Banducci et al. (2003) report a study of a low information election environment where voters seemed to favor the more attractive candidates. They suggest that when there is no other information, voters will extrapolate fit for office based at least in part on how a candidate looks. Riggle et al. (1992), find that when no other information is available, people will assign more positive traits, such as competence, to more attractive candidates. Rosenberg and McCafferty (1987; Rosenberg et al. 1986), argue that a candidate’s physical appearance “communicates a clear and politically relevant image of that person’s character” (1987, p. 114). Controlling for candidate party and issue positions, they find that physical appearance independently influences voter decisions, with attractive candidates more likely to garner support than unattractive ones. These studies focus primarily on physical appearance, while providing voters with relatively little additional information about the candidates under consideration. Budesheim and DePaola (1994) expand this line of work by giving their subjects candidate personality traits and issue positions, allowing examination of the individual and joint effects of likeability (defined as both visual image and personality) and issues on candidate evaluation. Their results are suggestive of the importance of likeability broadly defined, finding that issues have substantially less impact on evaluations than do the combination of a candidate’s picture and personality. While their focus is on evaluation rather than the vote choice, it seems that likeability is more important than issues when it comes to thinking about candidates. Others, however, are not so sure. When Riggle and her colleagues go beyond candidate appearance and provide their subjects with information about candidate partisanship and ideology, they find that the effects of physical attractiveness are attenuated. They conclude that while appearance might matter when little else is available, voters will use more substantive information—such as ideology—instead of relying on likeability when they are able to do so. And of course, there is a long history of research that suggests that issues matter to voters (e.g., Nie, Verba, and Petrocik 1975; Pomper 1972). Spatial models of the vote depend upon this (Hinnich and Munger 1994; Rabinowitz and MacDonald 1989), though as far back as 1960 the American Voter authors were not so sure issues mattered to very many voters (Campbell et al. 1960.) Recently, Bartels (2000) argues that the electoral relevance of partisanship has increased over the past few
190
REDLAWSK AND LAU
election cycles, suggesting that parties continue their important role in voter choice. All of these perspectives might suggest that likeability—image and personality—play a smaller role than many studies have found. It may be, then, that the evidence showing the importance of appearance simply reflects the less than comprehensive nature of many prior studies. Certainly, the Riggle et al. (1992) finding suggests that simply looking at appearance is not enough. At the same time, studies focused on the role of issues rarely, if ever, directly consider candidate attractiveness and personality. And none have examined whether in fact individual differences in voters condition the relative importance of likeability versus issues. Given what is known about candidate pictures as heuristic devices it seems likely that for some voters appearance matters more than for others (Lau and Redlawsk 2001b). Perhaps, more importantly, research to date has not explored the roles of likeability and issues in the type of environment that voters experience during a typical election campaign. Previous experiments have provided only very limited information about candidates and in a static format such that subjects have little choice but to process all of it, failing to consider the role a competitive campaign plays over time, when voters are choosing between two or more alternative candidates and where voters can more easily shape much of their own information environment. Process Tracing A comprehensive experimental study of the relative importance of likeability and issues to the vote requires a reasonable analog of a political campaign, where a wide range of information can be either processed or ignored about each of several candidates over a period of time. Using a computer-based dynamic process tracing methodology that allows us to track information search during a mock election campaign (Lau 1995; Lau and Redlawsk 1997, 2001a, 2001b, 2006; Redlawsk 2002, 2004), we can directly test how information such as pictures and personality compares with issue-oriented information in influencing voters’ choices. For this experiment the campaign environment was designed to require voters to choose between a visually appealing candidate with likeable personality traits but who was distant on the issues, and a less physically appealing, less likeable candidate who agreed with the voter on whatever issues the voter examined. While all voters saw the pictures of all of the candidates in the election, the amount of personality and issue information that any one voter learned depended on that own voter’s interests, and on what information the voter decided to examine. Thus, while all voters faced the same difficult choice, some voters were more aware than others of just how difficult the decision really was. It is important to understand the nature of the dynamic process tracing environment; however since we have described it in much detail elsewhere
I LIKE HIM, BUT. . .
191
(Lau 1995; Lau and Redlawsk 1997, 2001a, 2001b, 2006; Redlawsk 2001, 2002, 2004), we focus here only on key differences with traditional experimental approaches to studying campaigns. The most important difference is that in our dynamic environment candidate information comes and goes over time; subjects must choose which information they wish to learn about the candidates and in making any choice forgo choosing other pieces of information. Further, a very large database of candidate information is used, including dozens of issues, personality traits, group endorsements, and polls, where most studies have just a few items for each candidate. At any given time, though, a subject can only select from a small subset of this database, with the type of information that predominates (i.e., image, issues, polls, etc.) varying depending on the stage of the campaign. Early on, for example, specific issue positions may be scarce, as candidates try to introduce themselves to the voters via basic personal information. Later in the campaign issues become more available, mimicking the typical ebb and flow of a real-world presidential election (Lau 1995). The relative availability of information is also varied, so that generally available information (such as candidate’s party identification) is much easier to get (i.e., appears much more often) than other types of information (such as an obscure policy position). In addition to the information subjects can choose to examine, the campaign environment includes 20-second political advertisements, which periodically take over the computer screen, as they do the television at home. Overall, the dynamic process tracing environment provides a much closer real-world analog of a political campaign compared to traditional experimental methods. Procedure A non-probability sample of 196 nonstudent subjects were recruited from the central New Jersey area.1 The study began with the completion of a questionnaire about political knowledge, interests, and political preferences. After practicing on the computer, subjects participated in a simulated presidential primary election where two male Democrats and two male Republicans were competing for their respective party’s nomination, followed by a general election campaign with one of the Democrats facing one of the Republicans.2 The candidates in the elections, while fictitious, represented a realistic range of ideologies across both major political parties. Before the primary election, subjects “registered” with a political party, and were subsequently constrained to vote only for candidates from that party in the primary, although information was available about candidates from both parties during the campaign. As information “tags” appeared on the screen (e.g., “Fisher’s policy toward Russia”) subjects could choose which ones interested them enough to examine the details. At any given time there were six possible items from which voters could choose; these items
192
REDLAWSK AND LAU
appeared and disappeared from the screen over time, drawn from the larger underlying database. While the campaign commercials appeared without subject action, everything else subjects learned about the candidates was due to their own decisions about what information to examine. After the completion of the primary campaign (which lasted about 22 minutes) subjects voted and then rated all of the candidates on a 101-point feeling thermometer. Next they answered questions about the difficulty of their decision, learned which candidates were running in the general election during a short break, and then began that campaign (which lasted about 12 minutes). Following the general election campaign, subjects again voted, evaluated candidates, and answered questions about the difficulty of their decision. An unexpected memory test was then given about the candidates in the general election. Following this, subjects were debriefed, paid, and dismissed. Study Design We designed an election environment pitting a relatively unlikeable candidate who the voter agreed with on the issues, against a more likeable candidate who held many positions with which the voter disagreed. We created two such pairings, one in the Democratic primary and one in the Republican primary. Likeability was a combination of both the candidate’s physical appearance and his personality traits. For each candidate two pictures were created; one that was rated as relatively attractive in pretesting and another image of the same person rated as unattractive.3 These pictures were captured from the campaign ads of actual—but little known—candidates for Congress. Both the attractive and unattractive pictures were of the same person. Thus the variation in attractiveness is subtle and easily within the realm of what actually happens in many real campaigns. Some of us are just a little more attractive and a little more charming than others—and we all have our good days and our bad days—but we did not contrast beauty with the beast. The pictures were paired with a set of personality trait statements, also rated as either relatively positive or relatively negative. The unattractive picture was paired with the more negative trait descriptions, while the attractive picture was paired with more positive trait descriptions.4 Figure 10.1 displays the attractive and unattractive pictures of our four candidates, along with the likeable and less likeable personality descriptions associated with the different pictures. Again, we varied personality attractiveness by a matter of degree, not kind. We then took each candidate and assigned issue positions that were opposite of the candidate’s likeability. Unlikeable candidates took issue positions that were as close as possible to those espoused by the subject in response to the preexperiment political attitudes questionnaire, while likeable candidates
I LIKE HIM, BUT. . .
193
were assigned issue positions distant from the subject. But again, the variation in issue agreement was within the plausible range for American politics. We had prepared 8 different policy positions on every one of 23 different issues, 4 covering the range of stands typically found within the Democratic party (from the extreme left to slightly right of center), 4 covering the range of stands typically found within the Republican party (from slightly left of center to the extreme right). Each of our candidates was affiliated with one of the two major parties, and in choosing issue positions that were as close or
194
REDLAWSK AND LAU
Figure 10.1 Variations of Candidate Likeability.
as far as possible from the voter’s own positions, we selected from among the four that were available from the voter’s party. To make the general election choice as difficult as possible, we “fixed” the results of the primary elections so that the candidate the subject voted for in the primary election lost the nomination, and the candidate the subject had rejected in the primary was the party’s nominee. The out-party’s candidate in the general election, however, was the same type of candidate the voter had
I LIKE HIM, BUT. . .
195
preferred in the primary—likeable but relatively distant on the issues, or relatively unlikeable but as close as possible on the issues. Thus if voters were to choose the same type of candidate in the general election they had preferred in the primary, they would have to defect from their party affiliation to do so. The result of this process was a clear but difficult choice in the primary and again in the general election between two “flawed” candidates.5 Because our procedure for the most part allows voters to select the information they wished to learn about the candidates, subjects were differentially aware of the trade-offs between choosing either candidate. Every subject was exposed to two ads from each candidate during the primary, and three ads from each candidate during the general election. The last five seconds of every ad was the candidate’s (attractive or unattractive) picture with a voice-over offering the candidates campaign theme (e.g., “Vote Pat Thomas for a new tomorrow”). Thus all subjects saw each candidate’s picture at least twice, and this was the minimum level of “treatment” to which all subjects were exposed. However, all voters had many opportunities to choose to examine any candidate’s picture again, to choose to consider brief personality descriptions of the candidates from different people who know them, and to choose to learn about a variety of different issue stands from every candidate. But no voter had to look at any of these things, and very few subjects looked at every personality description of every candidate, or every issue stand that every candidate took. So the “strength” of our manipulation varies across voters. This variation allows us to consider how different amounts of issue information might contest against different amounts of likeability, rather than simply considering an all or nothing proposition. Results We have two different elections to consider in our analysis. First, subjects experienced a primary election where they were constrained to vote for one of the two candidates within their own party. Party is thus a constant in a primary and does not provide differentiating information to voters. Following the primary our subjects then participated in a general election, where partisanship clearly plays a role. Thus, in our analyses to follow we examine the primary and general elections independently, recognizing they are in many ways quite different from each other. Voting Issues or Voting Likeability To provide some context, we begin by looking at the vote decision. The top panel of table 10.1 reports the vote choice during the primary and general elections, for all voters.6 Slightly more subjects voted for the close and unlikeable candidate than for the distant/likeable alternative in both the
196
REDLAWSK AND LAU
Table 10.1 Voting in a Primary Election Choosing between “Close” and “Likeable” Candidates All Voters
Voted for: Close/Unlikeable Distant/Likeable Total
Primary
General All
General In-Party Close
General In-Party Likeable
53.1% (104) 46.9% (92) 100.0% (196)
56.1% (110) 43.9% (86) 100.0% (196)
76.1%a (70) 23.9% (22) 100.0% (92)
38.5% (44) 61.5%a (60) 100.0% (104)
Controlling on Political Sophistication General Election Primary Experts Voted for: Close/Unlikeable
Novices
64.3% 41.8% (63) (41) Distant/Likeable 35.7% 58.2% (35) (57) Total 100.0% 100.0% (98) (98) 2 ⫽ 9.915, 1 df, p ⬍ .01
In-Party Close Experts
Novices
71.4%a 78.9%a (25) (45) 28.6% 21.1% (10) (12) 100.0% 100.0% (35) (57) 2 ⫽ .674, 1 df, ns.
In-Party Likeable Experts
Novices
39.7% 36.6% (25) (15) 60.3%a 63.4%a (38) (26) 100.0% 100.0% (63) (41) 2 ⫽ .101, 1 df, ns.
Note: a Voted for in-party candidate.
primary and general elections. However, these results mask an important effect of partisanship in the general election. As the last two columns show, whether the close candidate is in the voter’s own party or in the other party, matters a great deal. A full 61.5 percent of subjects whose own party’s candidate was further away on the issues than the other candidate, remained with their party, while 38.5 percent defected to the opposition. On the other hand, when the close candidate was also the in-party candidate, defection to the more likeable out-party candidate was minimal, at only 23.9 percent. The pull of partisanship appears to outweigh either issues or likeability, though issues seem to trump likeability in their ability to draw defectors.7 At first glance this would seem to suggest that issue positions were generally seen to be more important than “likeability”—that is, physical attractiveness and a good personality. Still, a substantial set of subjects chose the more distant candidate. These voters either did not learn or chose to ignore the fact that there was a candidate closer to them on the issues. Of course, they may simply have considered policy agreement to be less important than likeability.
I LIKE HIM, BUT. . .
197
One possible explanation for this difference is expertise. Political experts may care more about issues and may be more interested in the candidates’ positions on those issues. If so, experts might be more likely to vote for the close candidate, regardless of the likeability of that candidate. During the preelection questionnaire, subjects were asked a battery of questions about political interest, knowledge, and activity. These variables were combined into a single comprehensive measure of political expertise, which was divided at its median, with political experts coded high and novices coded low. The bottom half of table 10.1 reports the vote by political expertise. The results for the primary election are as expected. Experts are more likely to vote for the close/unlikeable candidate than are novices (2 ⫽ 9.915, p ⬍ .01.) It appears that in primary elections political expertise is an important factor in determining whether voters will choose the candidate they agree with on the issues or the most likeable one. But again partisanship outweighs other considerations in the general election, overriding any expertise effect. When the in-party candidate is close on issues but less likeable, both experts and novices overwhelmingly vote for their party’s candidate. Likewise, when the in-party candidate is distant on the issues but attractive, most still stay within their party, although a larger number defect to the out-party candidate who is closer on the issues. Novices and experts alike are more likely to move toward an opposition candidate close to them on issues than toward an out-party candidate who is more likeable. The Difficulty of Voting for Likeable but Distant Candidates Our findings so far suggest that in a primary election expertise conditions whether voters choose the candidate closer to them or the more likeable candidate, while in the general election partisanship often seems to override both issue agreement and likeability. But some voters choose likeability over issues in either election. To the extent that voters know they are doing this—that is, they have learned enough about the candidates to recognize that the unlikeable candidate agrees with them on the issues, it might be instructive to see how difficult voters find it to ignore issues and vote likeability. As part of the postelection procedure, after voting and then evaluating the candidates, subjects were asked how difficult they found their decision, on a one to five scale with one labeled as extremely easy, and five as extremely difficult. For the primary election we conducted an ANOVA contrasting the two measures of interest here, political expertise and the type of chosen candidate (close or likeable). This analysis included controls for age, education, partisanship, gender, political party, and differential
198
REDLAWSK AND LAU Primary election 2.8
2.6 Novices 2.4
Experts 2.2
2 Voted close
Voted likeable
General election Voting for in-party candidate
Voting for out-party candidate
3.5
3.5
Novices
Experts
3
3 Novices
2.5
2.5
Experts 2
2 Voted close
Voted likeable
Voted close
Voted likeable
Figure 10.2 Difficulty of Vote Choice by Type of Candidate Chosen.
reading ability (since most of the information learned was text based). The same approach is used to model the general election, with the addition of a dummy variable indicating whether the voter chose an in-party or outparty candidate. The results are presented graphically in figure 10.2. In the primary, no significant effects are found for expertise, or for the interaction of expertise and the candidate type. But the main effects for candidate type are substantial (F ⫽ 6.592, p ⬍ .02). Voters choosing the candidate closer on the issues rated their choice a significantly easier one than those who chose the more likeable candidate who was distant on issues (2.27 versus 2.74). Thus while many voters were attracted to the candidate who looked better and had a better personality yet disagreed with
I LIKE HIM, BUT. . .
199
them on the issues, these same voters found it relatively difficult to actually choose that candidate. This is an important finding, because it is clear evidence that issue positions in general are more important to voters than likeability, and that overriding them is difficult. Turning to the general election it becomes clear that partisanship is a critical factor when we look at the relative difficulty of the vote choice for in-party and out-party candidates. The difference between the reported difficulty for an in-party choice (left chart) and an out-party choice (right chart) is both significant and substantive (F ⫽ 11.145, p ⬍ .001). Choosing the outparty candidate, whether close on issues or attractive, is just hard for voters to do. But choosing a likeable out-party candidate who is not close on issues is especially difficult. Yet some (nontrivial) number of voters does it anyway. Interestingly, the pattern of differences between experts and novices remains the same whether the vote is for an in-party or out-party candidate. Novices report that a vote for an unlikeable candidate close on issues is harder than for a likeable candidate further away from them. For experts, the opposite is the case. This interaction effect is statistically significant (F ⫽ 5.108, p ⬍ .05, controlling for partisanship of the candidate). Novices in a general election simply find it easier to vote for the more likeable candidate, while experts show the pattern we would expect in reporting a vote for a distant candidate to be more difficult. This may reflect differing informational bases between novices and experts, where novices may simply not know that a candidate is more or less close on issues.8 The Wellsprings of Preference for a Likeable Candidate Perhaps we can get some purchase on just who it is that prefers a likeable candidate over an issue-congruent candidate. As a first cut, we used a series of demographic variables to predict the likelihood of a vote for the attractive candidate. We built a simple logistic regression model regressing the vote for a likeable but distant candidate on the age, education, expertise, gender, ideology, and strength of partisanship or our subjects. The first and second data columns of table 10.2 presents this analysis. Two factors stand out, expertise and education. The bivariate findings about expertise reported earlier continue to hold in this multivariate analysis. Experts are significantly less likely to vote for the likeable candidate in the primary election (b ⫽ ⫺.894, p ⬍ .05). Calculating the effects of the logistic regression coefficient, holding all other predictors to their means, a female political novice is about 54 percent likely to vote for the likeable candidate, while an expert is only about 33 percent likely to do so. In other words, experts generally preferred that candidate they agreed with on the issues. However, controlling for political expertise, general education levels also predict the
200
REDLAWSK AND LAU
direction of the vote, with similar effects—less educated subjects are more likely to vote for the likeable candidate (b ⫽ ⫺.187, p ⬍ .10). A female novice with no college education is about 77 percent likely to vote for the distant, likeable candidate, while a political novice with a post graduate degree is only about 43 percent likely to cast the same vote. These effects are quite substantive. While the overall model is not terribly strong, the results suggest that those more likely to be interested in and knowledgeable about political issues are also more likely to avoid voting for a likeable candidate with whom they disagree on the issues. A significant strength of the process tracing methodology is that we know what types of information voters examine during the election. This seems particularly useful here since it is likely that examining different kinds of information might lead to different vote decisions. The third and fourth data column of table 10.2 reports an analysis for the primary election in which we add measures of the types of information examined during the campaign to the basic demographic analysis. We include measures of the share of pictures, personality items, issues, and ideology statements that our subjects accessed during the primary.9 We expect that voters who spend
Table 10.2 Vote for a Likeable but Distant Candidate Predictors
Primary B
Age in years Education Political expert Female Conservative Partisan strength Likeable/Distant In-party candidate
.001 ⫺.187* ⫺.894** ⫺.361 ⫺.152 ⫺.011
S.E. (.009) (.107) (.351) (.325) (.117) (.155)
Information examined Pictures Personality Issues Candidate ideology Constant 2 ⫺2LL Pseudo R2 Classified
General Election B
⫺.005 ⫺.204* ⫺.836** ⫺.567* ⫺.222* ⫺.020
S.E. (.011) (.113) (.365) (.342) (.124) (.162)
B .004 .049 ⫺.012 ⫺.808** .120 .037 1.723***
S.E. (.009) (.118) (.011) (.348) (.125) (.170) (.341)
.952* (.525) 1.723** (.865) ⫺2.901** (1.276) ⫺.990 (.643) 1.813** 14.870 6 df, p ⬍ .05 249.278 .100 60.7%
(.745)
2.835*** (.953) ⫺.862 29.008 35.867 10 df, p ⬍ .01 7 df, p ⬍ .001 235.141 226.601 .188 .229 63.4% 71.2%
B .001 .055 ⫺.009 ⫺.846** .095 .057 1.771***
.692 .013 ⫺1.130 ⫺.136 (.867)
⫺.985 38.552 11 df, p ⬍ .001 223.917 .245 70.7%
S.E. (.011) (.122) (.012) (.354) (.128) (.173) (.348)
(.546) (.966) (1.464) (.482) (1.009)
Note: Table entries are logistic regression coefficients, standard errors in parentheses. Dependent variable is a vote for the likeable but distant candidate, N ⫽ 191. * p ⬍ .1, ** p ⬍ .05, *** p ⬍ .01.
I LIKE HIM, BUT. . .
201
more time focusing on issues and ideology would be more inclined to support a candidate based on his issue positions. On the other hand, focusing more on pictures and the personality of candidates might generate a different response, and might point to a tendency to be swayed by candidate likeability. We can test this hypothesis by inserting into the model our measures for how much of each type of information voters examined. Adding these information-search variables does not change the significance of either education or expertise. Two additional demographic variables, conservative ideology and gender become marginally significant, with conservatives and women less likely to vote personality and attractiveness over issues.10 More importantly, however, we get strong results from the information-search variables, and in the expected directions. Examining a greater share of the available pictures and personality statements leads to a greater likelihood of a vote for the likeable candidate, ceterus paribus, while accessing a greater share of the issues substantially decreases the likelihood of such a vote. Candidate ideology, while not reaching statistical significance, is also in the expected direction. Clearly, the nature of the information that is examined, combined with key individual differences, has substantial impact on the primary vote. Given the difficulty of interpreting logistic regression coefficients, figure 10.3 examines graphically the relative effects of these different types of information for a female political expert. In this analysis, the effects of pictures and personality are combined into a single measure, as are those of issues and ideology. As our voter learns more of the former compared to the latter she becomes more likely to vote for the likeable candidate. But as she learns more of the latter compared to the former, she becomes less likely to vote for the likeable candidate. The relative impact of both kinds of information is clear— the increase in the probability of voting for the likeable candidate as more picture/personality information is learned is less steep than the increase as more issue/ideology information is learned. All else equal, issues and ideology have a greater impact on the vote choice than do the pictures and personality. What happens when we move to the general election, while partisanship plays such an important role? The general election analysis parallels that of the primary, with an important addition. To account for the effects of partisanship a dummy variable coded “1” when the in-party candidate is likeable but distant on issues is included in the analysis.11 This is a control variable, accounting for both the voter’s own partisan preference and the nature of the candidate choice the voter faced. However, our goal here is to understand the role played by the demographic and information factors apart from partisanship. The fifth one sixth data columns of table 10.2 examine just the demographics of general election voters. As expected, the nature of the in-party candidate is a
202
REDLAWSK AND LAU
Probability of vote for likeable candidate
1
0.8
0.6
0.4
0.2 1 0 0 0.25 0.5 Issue s and 0.75 ideolo gy
0 1
0.75 lity a on 0.5 rs e p 0.25 and s re ctu Pi
Figure 10.3 Probability of Voting for a Likeable Candidate, Primary Election by Proportion of Issues and Ideology Versus Pictures and Personality Examined Female, High Political Expertise.
significant positive predictor, simply indicating a strong tendency to vote for the party’s candidate in the general election, irrespective of whether that candidate was chosen during the primary. More interestingly, the only significant demographic variable is voter gender, with women less likely to vote for an attractive candidate, other things being equal, just as in the primary. But unlike in the primary, neither education nor expertise predicts the direction of the vote in the general election. Turning to the final two columns of table 10.2, even after adding the information-search measures, gender is the only significant predictor (other than the control for party). None of the added information variables is significant, although the signs are in the expected direction. Clearly, the effects of party, which are not present in the primary, change the dynamic of the general election.12
I LIKE HIM, BUT. . .
203
Likeability and Correct Voting Finally we turn to an important “so what” question. That is, do those who vote for the closer candidate end up making a better decision than those who vote for the more likeable one? The initial answer seems obvious: of course it is “better” to vote for the candidate who is closer on the issues. Presumably that is the rational thing to do, since an election is about determining who will get to carry out policy. By such a measure, we can simply declare that the 46.9 percent of our primary voters who chose the likeable but distant candidate failed to vote correctly. But this is too simple. After all, if as we suggested earlier, knowing a candidate’s personality traits means knowing something valuable about how that candidate will act in office, it is certainly possible that for some voters choosing the candidate who has the better personality over the one with better issue congruence is the right thing to do. Using a correct voting standard (Lau and Redlawsk 1997) we assess the extent to which voters make what for them is the right choice. Measurement of correct voting is guided by the subject’s own interests during the election. Rather than play the role of omnipotent seer determining a priori what is correct and what is not, we allow the nature of the information search to determine what would be correct for each subject. Our standard is actually quite simple. We assert that for each piece of information examined for one candidate during the election, the voter should have examined the same information for the other relevant candidate. Further, if a voter chooses not to examine certain information at all, we take that to mean that particular information is unimportant to the voter, and we do not include it in our assessment. Thus, our measure of a correct vote is a “normative-intuitive” measure, in that we normatively require information that is examined to be examined across the board, but only count that which the subject is actually interested in.13 For each of our voters, then, we know how they voted and, by the above criteria of “correctness,” we know how they should have voted. The top half of table 10.3 presents a simple analysis of correct voting in the primary election, examined separately for experts and novices. Of those who voted for the close but unattractive candidate, 81.7 percent overall voted correctly.14 But significantly fewer of those who chose the likeable candidate, by their own standards voted correctly (67.4 percent, t ⫽ 5.353, p ⬍ .05) with the difference especially clear for experts. Political experts who chose the likeable candidate were substantially led astray, with only about two-thirds as many voting correctly as those who chose the close candidate (t ⫽ 9.633, p ⬍ .01). But turning to novices, the story is quite difference. Novices who chose the likeable candidate were no less likely to be correct than those who voted for the close candidate. In fact, political novices who voted for the likeable candidate were actually more likely to be making the correct choice (for them) than were experts who went for likeability over issues.
204
REDLAWSK AND LAU
Table 10.3 Correct Voting and Candidate Likeability Overall Primary Election Voted Close Voted Likeable
General Election Voted Close Voted Likeable
% Voting Correctly Experts
Novices
81.7% (85) 67.4% (62) t ⫽ 5.353 p ⬍ .05
87.3% (55) 60.0% (21) t ⫽ 9.633 p ⬍ .01
73.2% (30) 71.9% (41) t ⫽ .018 n.s
68.2% (75) 55.8% (48) t ⫽ 3.159 p ⬍ .1
72.0% (36) 52.1% (25) t ⫽ 4.134 p ⬍ .05
65.0% (39) 60.5% (23) t ⫽ .200 n.s.
Turning once more to the general election, we find that overall correct voting is lower than in the primary, probably due to the more complicated environment we created where partisanship was balanced against either likeability or issue congruence. But as with the primary, those voting for the close candidate were more likely to vote correctly than those voting for the likeable candidate, all other things equal. And again we see experts who vote for the likeable candidate doing significantly worse than those voting for the close candidate, while there is no significant difference for novices. Discussion Our findings provide a more nuanced look than has heretofore been available at the question of whether likeability trumps issue positions. Earlier research, using limited pools of information in non-campaign experimental environments, showed a bias toward candidate physical attractiveness and personality over issues. While there is good psychological evidence that people do find attractive people more appealing, choosing political leaders is a much different task than simply evaluating someone as a friend or coworker. While voters must undoubtedly evaluate political candidates, they must consider a wide range of factors, including the extent to which the candidate not only has the personality to be president but also whether the candidate, if elected, will carry out preferred policies. While some previous studies have attempted to provide policy information along with personality and pictures, none has been able to do it in as comprehensive a way as we can with our dynamic process tracing methodology.
I LIKE HIM, BUT. . .
205
This new study presents voters with the full range of information that would be available in a typical presidential election. And, more importantly, our subjects generally choose the information they wish to learn about the various candidates. This is a crucial difference. Choosing information in an environment where time and other resources are limited means foregoing (other) information as well. Our voters could never learn everything about every candidate, and thus had to be guided by whatever considerations they thought most important. Of course, we provided our subjects with a very clear but difficult choice that probably does not occur very often in real presidential politics. Actual candidates are often more contradictory than this, so that the choice between likeability and issue agreement is rarely so stark. Even so, we do not find that likeability necessarily outweighs issues and ideological information. Preference for likeability compared to issues varies by individual factors, at least in a primary election where party cannot be a deciding factor. Some voters certainly are attracted to likeability. At least in the primary election, these tend to be less educated voters with less political expertise. Women, we find, are somewhat less likely to vote for a likeable candidate who is distant on the issues. And, importantly, the kind of information that voters actively seek out has substantial implications for their vote. Those voters who focus more on pictures and personality statements (and thus less on issues and ideology) are much more likely in a primary election to vote for the likeable candidate even though he is distant on issues. Is this wrong? That is, should we establish some normative standard that says these particular voters fail to choose the right candidate? We have mixed feelings about this. At one level, the argument that voters should choose candidates more likely to implement the policies they desire seems to comport with basic democratic theory, where representatives should be held accountable to citizens’ desires. On the other hand, for political novices in our study, the empirical evidence is that those who vote for the attractive candidate are just as likely to vote correctly as those who vote for the closer candidate. Why? Because the vote is guided by the information that voters examine, and our novices seem to process what they learn in ways that make sense to them. Having said that, it might be argued that they look at the “wrong” information; that these political novices should not place so much importance on image and personality. Yet whether a candidate has the “right” personality to be president may be nearly as important as whether he is likely to pursue certain policies. In fact, one might argue it is more important, since policy success may be based at least in part on a president’s personality traits (Neustadt 1991 [1960]; Greenstein 1969). It is worth noting, however, that for political experts who tend to focus a greater part of their search on issue-oriented information, choosing a
206
REDLAWSK AND LAU
likeable candidate over one close on the issues generally means lowering the chances of voting correctly, at least as we have defined it. Thus, those experts who were more attracted to the candidate image and personality were perversely more likely to have made the wrong choice based on the information they themselves chose to examine about the candidates. Notes * Support for data collection and analyses came from National Science Foundation grants SBR-9411162 to the first author and SBR-9321236 to the second author. Redlawsk gratefully acknowledges support from the Obermann Center for Advanced Studies at the University of Iowa. Earlier versions of this paper were presented at the annual meeting of the Midwest Political Science Association, Chicago, April 3–6, 2003 and at the Shambaugh Conference on Affect and Cognition in Political Action, University of Iowa, Iowa City, IA, March 6–9, 2003. 1. All subjects were required to be eligible U.S. voters, although they did not have to actually be registered to vote. Subjects could not currently be attending college. Just over half (51%) of subjects were male, 79.7% were white, and 13.5% black, with an average age of 40.5 (ranging from 18 to 82). Subjects were well educated, with 37.0% holding at least a BA degree, though 26.2% had not attended college at all. Finally, subjects were evenly distributed through the range of family incomes from under $15,000 to over $75,000. 2. See Lau and Redlawsk (1997, Figure 2, 588) for a detailed summary of the typical procedure. 3. In order to establish the attractiveness of both the pictures and the personality traits, we had a separate set of 63 pretest subjects (college students) respond to a series of questions about each picture and/or personality description as they were randomly presented on a computer. The questions included “How attractive is the person?” and “How much would you like this person for a friend?” 4. The trait descriptions included a range of statements about the candidate’s personality, including descriptions from a friend, a staff worker, a political opponent, and a telling anecdote from the candidate’s mother. 5. It is important to note here that because we varied candidates’ issue stands according to the voter’s own preferences, there was no such thing as “fixed” candidates in this experiment, and in fact it was probably the case that no two subjects ever faced a campaign involving exactly the same candidates—at least if by “candidate” we are referring to more than just their name and appearance. Each of our four candidates had a name and an attractive and unattractive picture and set of personality descriptions, although which candidate was assigned to the likeable and unlikeable conditions was randomly assigned within each party. Or at least that was our intention. However, a computer programming error that we only caught at the very end of our data gathering led to the same Republican candidates
I LIKE HIM, BUT. . .
6.
7.
8.
9.
10.
11.
207
being assigned to the likeable and unlikeable conditions in virtually every primary election. That the Republican candidates’ names are confounded with condition is probably irrelevant, but our efforts to obtain attractive and unattractive images of the same person went to naught in the Republican primary. One person’s attractive picture was almost always used for the likeable but distant candidate in the Republican primary, while a different person’s unattractive picture was almost always used in the unlikeable but close on the issues condition. Fortunately, this bug had no effect on assignment to the various candidate combinations in the general election. While a computer bug meant that virtually all Republican voters faced the same candidate-condition pairing, because we are not concerned about partisanship we feel comfortable in combining Republicans and Democrats in later primary analyses. Examination of the data for both primaries and the general election do not show any substantial differences between Republican and Democratic voters based solely on partisanship. Examining Republicans and Democrats separately, we find that partisans do not differ in the primary election, being equally likely to vote for close candidates over likeable ones. In the general election, Republicans are less likely to defect from their party overall, regardless of whether the Democrat was closer to them on the issues or more likeable. The process tracing nature of our experiment means we know specifically what kind of information our voters examine as they go through the campaign. We find that controlling for individual differences such as age, education, and the overall amount of information examined, experts examined more of the available issue information than did novices (Primary, F ⫽ 4.152, p ⬍ .05; General Election, F ⫽ 4.687, p ⬍ .05). Interestingly there are no significant differences for other types of information: personality, pictures, or ideology. Subjects could also look at endorsements from various interest groups, poll results and other campaign hoopla, and background information specific to the different candidates, e.g., their age, education, religion, military and political experience, family, and so on. The measures we use are calculated as the amount of each type of information acquired as a percentage of overall search. Thus providing some small evidence if George H. W. Bush tapped Dan Quayle as his vice-presidential running mate, believing that Quayle would appeal specifically to women, he was perhaps misguided. A subject who has a likeable in-party candidate in the general election by necessity voted for the unlikable but close on issues candidate in her primary election, since whatever candidate was chosen in the primary “lost” that election and his opponent made it to the general election. This measure is used in the analysis instead of the simple partisanship of the voter because it captures both the effects of partisanship and the nature of the choice facing the voter. Since the dependent variable in the analysis is a vote for the likeable candidate (regardless of party) a positive coefficient on this measure would suggest a tendency to vote for the in-party candidate, which is what is expected.
208
REDLAWSK AND LAU
12. These analyses raise a subtle but important point influencing our interpretation of these results. We suggested above that voters have chronic proclivities to seek out different types of information about candidates, presumably because they find that type of information particularly useful in forming their impressions of candidates. We want to interpret the results of the second equation in table 10.2 as illustrating that information search matters, because it determines, at least in part, the vote choice. But if the presumed proclivities to seek out different types of information are strong, then we may have illustrated nothing more than that there are different types of voters, and these voters are attracted to different types of candidates (see Lau 1986, 1989a). In that case our information search variables are merely serving as instruments to represent those different types of voters, rather than standing in for something that is important in its own right. To sort out these possibilities, we looked first to see if any individual difference variables could predict differential search in our experiment. Fortunately we have measures of what types of information our voters generally used to make their vote decisions from our pre-experiment questionnaire. We examined our measures of the relative importance of policy and likeability-based information, finding that the importance of policy-based information is positively related to the amount of issue-based search, and negatively related to the amount of likeability-related search, in both the primary and general election campaigns. The rated importance of candidate likeability is not as good a predictor, however. It has the wrong sign for the primary election, although not at all significant, but has the expected sign in the general election, just missing conventional significance levels (p ⬍ .07, one-tailed). Overall, though, we have reasonable individual-difference measures predicting differential information search. The question now becomes, what happens when we replace the information search variables added to the second equation of table 10.2, with these two indicators of a proclivity to seek out different types of information? The short answer is nothing. Whether these items replace the actual search variables, or are added to the equation along with actual search, they are never close to statistically significant, while the actual search measures remain significant. So it is actual information search—and not a proclivity to seek out different types of information, i.e., helping to determine the vote choice in the primary election. 13. For complete details on how a correct vote is calculated, see Lau and Redlawsk (1997, Appendix A). 14 There is some difference between experts and novices, as might be expected, with 87.3% of experts making a correct choice, while only 73.2% of novices did so.
CHAPTER 11 THE EMOTIONAL CALCULUS OF FOREIGN POLICY DECISIONS: GETTING EMOTIONS OUT OF THE CLOSET Nehemia Geva and J. Mark Skorick
or the past five decades, students of foreign policy analysis and foreign policy decision making have labored to develop theories that explain international behavior. Focusing on the individual as the unit of analysis, these studies have proceeded under the assumption that the political decision maker is the driving force behind the perception, interpretation, evaluation, and response to international political events (Snyder, Bruck, and Sapin 2002). However, within this growing field of inquiry, little empirical research has contributed to a scientific understanding of how emotions influence foreign policy decision making. This is surprising not only given our intuitive understanding that emotions are an important part of the human experience, but also given a history of research devoted to understanding the impact of emotions on social behavior. It has only been within the past decade that any serious attempt has been made to understand how emotions such as anger, hate, compassion, and fear impact politics. To date, no program of scientific inquiry has sought to empirically investigate the effect of emotions on foreign policy perceptions, interpretations, and choices. One need not search long to find emotions manifest within international relations. Recent headlines in world press outlets have used words like, “hatred,” “anger,” “fear,” and “rage” to describe events in Israel, United States, United Kingdom, and Macedonia. Syria’s president is accused of using anti-Semitic rhetoric to spark hatred against Jews. Ethnic resentments
F
210
GEVA AND SKORICK
against Albanians rise in the beleaguered Balkan nation of Macedonia, leading to clashes between local military officials and armed guerrillas. Italian opposition leaders suspend an election in response to threats and what they deem a “campaign of hatred.” Angry and rankled crowds gather outside the Indonesian parliament protesting the impeachment of their embattled president. Undercurrents of emotional rancor and acrimony flow in varying degrees within the domestic and international affairs of India and Pakistan, Jews and Palestinians, Serbia and Albania, French-Canadian Quebec, the Sudan, Northern Ireland, and China and the United States, to name only a few. Why has the study of international relations historically remained silent on the link between emotions and foreign policy? The failure of international relations theories to address the influence of emotions on foreign policy behavior can be broadly attributed to three causes. First, systemic and state level approaches have held sway among scholars of international relations for a number of years. As a result, individual level approaches to foreign policy analysis have not been as prominent in the field. Second, among those studies addressing foreign policy behavior at the individual level, the primary debate has been concerned with the contention between rational and cognitive schools of thought. Third, the concept of emotion has proven itself remarkably difficult to define and measure. Part of this difficulty lies in the fact that the study of emotions has followed two divergent theoretical tracks, leading to confusion in the literature over the nature, sources, location, and effects of emotion on political and social interaction. This chapter is structured in three main sections. First, it provides a review of some of the issues in the study of affect and emotions as related to international politics. Second, this chapter proposes a conceptual framework for studying the effects of emotion and affect on perception and interpretation of international events. Finally, we provide results of a preliminary experimental test of specific hypotheses derived from that conceptual framework. It should be noted that for the sake of this investigation, two important restrictions are necessarily placed on its content and scope. First, the current study focuses on the effect of negative emotions. Since the study of international relations has been concerned with the nature of conflict between states, a more direct relationship is assumed to exist between negative emotions (e.g., anger, hate, fear, etc.) and potentially conflictual foreign policy choices. This is, of course, not to suggest that other emotions do not impact the foreign policy decision-making process.1 However, in focusing on the effect of negative emotions, this analysis will (1) limit itself to a single albeit important source of international conflict and (2) reduce the number of measures necessary to test appropriate aspects of the theory.
EMOTIONAL CALCULUS OF FOREIGN POLICY
211
Second, this research program focuses on the impact emotions have on individual perceptions of international events. The generalizability of any findings is assumed to extend to the general public or similar political observers rather than political elites or foreign policy leaders. The reason for this restriction is the understanding that, in most cases, cultural and structural factors such as accountability to constituencies or power bases constrain the degree to which foreign policy leaders can express negative emotions. Furthermore, these same leaders are often adroit in the use of emotional stimuli to influence popular support for foreign policies (and thus apparently perceive the use of such cues to be potent measures). Since, therefore, the current study of emotions in politics has tended to stress its effects on the public rather than on elites,2 an important question in this vein asks how the individual interpretation and evaluation of important foreign policy events is influenced by negative emotions such as hate, anger, or fear. This emphasis on the individual’s perceptions stems from the voluminous literature linking public opinion and foreign policy decision making (e.g., Holsti 1996; Sobel 2001). Emotions in International Relations A review of the literature on the effects of emotions in international relations supports Crawford’s (2000) argument “Systematic analysis of emotion may have important implications for international relations theory and the practices of diplomacy, negotiation, and postconflict peacebuilding” (116). One could suggest that the first indirect studies of emotion in international relations were those dealing with the effects of stress on decision making and studies of attitudes or beliefs. Since the 1960s, attention has focused on the effects of stress in political decision making and crisis (Holsti 1972, 1979; Hermann 1972; Lamb 1984; Mor 1993; Roberts 1988). The majority of these studies defined stress in such a way as to emphasize the relationship between time pressures, goal commitments, and decisional performance, rather than emphasize any emotional component of the crisis situation. It can be argued, however, that these studies implied an emotional component of international decision-making behavior. Janis and Mann’s (1977) discussion of “hot cognitions” (c.f. Abelson 1963) acknowledged that emotional arousal associated with critical and stressful decisions compelled decision makers to undertake suboptimal problem-solving measures. Emotions or feelings have also been either directly or indirectly addressed in the research focusing on political attitudes and beliefs in international relations. These include studies of belief systems and operational codes (George 1969; Holsti 1976; Walker 1977, 1983) and images (Herrmann 1988; Herrmann et al. 1997; Cottam 1986, 1994; Hudson
212
GEVA AND SKORICK
1995). Rather than suggesting a direct effect of emotions on foreign policy behavior, these models imply that actor-oriented beliefs and attitudes in international relations include affective evaluations of and orientations toward other nations or leaders. Thus, Dixon (1983), Hermann, Hermann and Hutchins (1982), and Hudson, Hermann and Singer (1989) conceptualize affect as a reflection of governmental attitudes toward another international actor. The role of emotion has also been a subject of some focus within the broader study of political science. The affective elements of political attitudes (Brady and Sniderman 1985), emotional responses to political issues (Conover and Feldman 1986); Kinder and Sanders 1996), and the emotional character of national patriotism have received attention in the literature (Rahn et al. 1996). Gibson (1992), Gibson and Bingham (1982), and Marcus, Wood, and Theiss-Morse (1998) have also addressed political intolerance as a form of emotional reaction. Lodge (1995), and Lodge and Taber (2000) note that the affective tag is an inherent feature of political candidate evaluation and biases voters’ judgments toward or away from a candidate. Departing from the focus on attitudes and emotions, Marcus and colleagues (Marcus 1998; Marcus, Wood, and Theiss-Morse 1998; Marcus and MacKuen 1993; Marcus 2000; Marcus, Neuman, and MacKuen 2000) have recently developed a theory emphasizing the physiology of emotion and have stressed the need to understand the interplay of emotion and politics in light of neural and biological research.3 Introducing Emotions to the Cognitive Calculus Model The Cognitive Calculus (CC) model (Geva and Skorick 1999, 2000a, 2000b; Geva, Clare, and Mosher 2002; Geva, Mayhar, and Redd 1997; Geva, Mayhar, and Skorick 2000; Geva, Mosher, and Clare 2003) is an attempt to bridge the gap between the process and outcome validity research orientations in the international relations literature and will serve as the basic framework for this chapter. “This theory models the mental calculus of decision making while taking as a fundamental premise that an individual is the engine that conducts the decision-making process, and that the model should therefore represent his/her capabilities” (Geva, Mayhar, and Skorick 2000, p. 448). It argues that the way in which information is evaluated as it enters into the decision-making process will influence the ultimate outcome of this process (i.e., the decision). While a complete review of the model is not outlined here, a brief summary is provided below.4 The CC model suggests that information characteristics, such as the reliability, relevance, redundancy, and valence of individual information items
EMOTIONAL CALCULUS OF FOREIGN POLICY
213
moves the individual’s choice propensity (i.e., the likelihood of choosing one alternative over another) toward or away from any available alternative. In line with other online models of information processing (Lodge 1995; Lodge and Taber 2000), the model assumes that the cumulative contribution of all items reviewed will direct the decision maker toward one choice over another. The CC model departs from other such theories in that it specifies a decision threshold which indicates the point at which a decision maker stops attending to information and finally makes a choice (see figure 11.1). The critical components of the model are the cumulative choice propensity counter (CCP), which moves as a function of each item’s information utility and its valence. In turn an item’s information utility is a function of its reliability, relevance, and redundancy (with previous information). Finally, the decisional threshold (TH) indicates the point at which a decision is made and no further information is accessed by the decision maker. The CC model proposes that incoming information is translated into the information utility parameters via an individual’s implicit theory of international relations (ITIR), which represents the stored knowledge, beliefs, and attitudes held regarding international events. An important process question concerns what happens when the translation of information via the ITIR is usurped by the inclusion of emotion? In other words, what are the implications of introducing affect or emotion as part of the ingredients of the cognitive calculus? The definition of emotions proposed here takes into account the total interplay of these factors: Emotions are responses to external stimuli that are perceived to influence an individual’s well-being (as defined by the self or
TH (alternative A)
CCP
TH (alternative B) 1 2 3 4 5 Number of items of information
6
7
8
9
10
Figure 11.1 The Cognitive Calculus Model of Decision Making.
11
12
214
GEVA AND SKORICK
primary group).5 These responses are characterized by phenomenological, physiological, and behavioral properties and in turn are associated with structures within the cognitive system. The aggregate properties (phenomenological, physiological, behavioral dispositions) of emotions become labeled experiences (anger, joy, hate, fear) connected to event-specific objects (e.g., other individuals, groups, states, organizations, or institutions). Thus, emotions participate in the experience of social and political phenomena as expressions of a cognitive and physiological interaction. Our new theorizing is founded upon a basic proposition about the interaction of incoming information, cognitions, and emotions. Proposition 1: Generally, negative emotions limit both the process capacity and content of items to be processed.6 Derivative 1: The rise of negative emotion and its association toward a target results in (1) fewer items being drawn in from the environment and/or the cognitive system into the process; (2) these few items will be primarily relevant and schema-consistent with the emotion: and (3) these constraining effects occur only at high intensity of the negative emotional state.
At low emotional levels such as liking or disliking some nation, group, or individual, new incoming information can change or reinforce the emotional state. However, feelings of hate or anger toward some nation, group, or individual while being triggered by cognitions, (e.g., reading news about the target nation, hearing the name of a hated country, etc.) works independently of such cognitions. Furthermore, such feelings constrict the ability of the cognitive system to incorporate new information about the object toward which negative emotions are felt. New information that is allowed into the process will consist of few and relatively homogeneous clusters of information. This in turn results in biased interpretive processes that are consistent with appropriate schema or images (e.g., the use of stereotypes). Since individuals tend to process information in an online manner, these effects of intense emotions result in fewer items of information accessed, less time spent reviewing information, greater attention to consistent items of information, and higher recall of such items. This proposition allows us to understand how emotions impact foreign policy behavior by reflecting findings which suggest that negative emotional stimuli receive more attention (Marcus and MacKuen 1993; Derryberry 1991; Pratto and John 1991), provide critical information to perceivers (Schwarz and Clore 1988), and result in suboptimal information processing strategies (Janis and Mann 1977).
EMOTIONAL CALCULUS OF FOREIGN POLICY
215
How do Negative Emotions Influence Event Interpretation? In line with the current research in international relations which stresses the importance of problem representation (Billings and Hermann 1998), and the definition of the situation (Sylvan and Voss 1998), the interpretation of an event is of critical importance to the understanding of foreign policy decision making. Event interpretation primarily revolves around the concept of making sense of some occurrence. While interpretation of any event is fundamentally cognitive, our model suggests that emotions may impact cognitions relevant to stimuli. An event can be interpreted, or made sense of, in a number of ways. One way in which event interpretation can be represented is by analyzing the attributions, evaluations, and inferences individuals make about an object. Subjects who hate or are distressed by the actions of another country will make inferences about that actor’s motives, intentions, and qualities in a far different manner than one who does not feel any particular emotion toward that nation. The presence of a negative emotion toward a target dramatically influences the types of cognitions brought to bear about the object. We postulate that at a certain point, a dramatic “step” exists between low level negative affective reactions and the existence or occurrence of intense negative emotion. The step from dislike to hate implies a differentiation in processing at the level of the emotional system versus that of the cognitive system. Two general hypotheses about event interpretation can be derived from the proposition and derivation noted earlier. First, while low level negative emotion—in comparison with a neutral psychological state—may increase attention to new incoming information (and thus may be affected by it), we expect high level of negative emotions to decrease attention to new information. Second, we expect that negative emotions will lead to more biased (emotionally congruent) processing of the information.
Method General To test our hypotheses, an experiment was designed around a hypothetical international crisis concerning the fictitious island nation of Manova. The crisis involved the taking of American and foreign hostages at the U.S. embassy by armed local rioters, an attempt by U.S. soldiers to rescue the hostages, and the subsequent execution of some American hostages. We employed a computerized process tracing instrument similar to that used in
216
GEVA AND SKORICK
previous studies (Geva, Mayhar, and Skorick 2000; Geva and Skorick 2000a, 2000b). The program presents written, audio, and visual information in a controlled setting while recording subjects’ behavior. One hundred upper division college students from Texas A&M University participated in the main phase of the experiment. Subjects were randomly assigned to one of nine experimental conditions. The experiment was structured as a 3 ⫻ 3 between groups factorial design. The manipulated factors were: (1) affective/emotional manipulation (video clip, written text, or none), and (2) the valence of the information describing the target nation (positive, negative, and mixed). The dependent variables addressed processing parameters (processing time and information recall), and the interpretation of the event.
Instrument and Procedure The Manova Case All subjects were informed that they would be exposed to events ongoing between the United States and Manova and would be asked to make sense out of what is going on in Manova. Following this, all subjects were introduced to the Manovan crisis via the following description: Manova is a country on an island in the Gulf of Guinea. Since the end of World War 2, Manova has been a member of the Organization of African States. It has historically had a good relationship with the United States and, since 1973, has been an important port for US Naval ships. Because of its strategic location, Manova has allowed US vessels to use its ports for refueling, repair, and crew leisure. An agreement reached in 1995 with the Foreign Affairs Office of Manova allowed some US and foreign personnel to establish temporary residence in Manova. However, a recent measure in the Manovan elections called for the removal of US military housing from Manova. Though the ballot measure lost, a number of Manovans protested against the US presence, some of them demanding that the Manovan government no longer allow access to US Naval ships in its ports. Other Manovan officials have publicly stated that these protesters are a radical minority. This morning you have learned that civil unrest has broken out in the capital of Manova. Initial details are sketchy but it appears that some US military and embassy personnel, as well as a small number of foreign diplomats, are being held hostage at an unknown location in the capital of Manova.
EMOTIONAL CALCULUS OF FOREIGN POLICY
217
The Affective/Emotive Manipulation Three levels of this factor were introduced in the experiment. (1) Intense Negative Emotions manipulation was introduced via the use of an audio-visual account of the events unfolding in Manova, and was fashioned after a typical “on the scene” news report.7 The short video (approx. 1 minute in length) depicted riots by Manovans, described the execution of American hostages, and showed American bodies being dragged through the streets by reveling Manovans.8 (2) Low Intensity Negative Emotions: The manipulation employed a written account of the same event, i.e. the audio portion in writing, without the accompanying video. The text read as follows: Events in Manova took a dramatic turn today as protests over the US and Western presence in Manova turned violent. As early as yesterday, bands of armed men were seen roaming through the streets of the capital, calling for an uprising against the current government and all Western nations. Then, without warning, the city seemed to explode this morning as militia groups began cruising through the city destroying American, British, and French business offices and shooting at suspected foreigners. However, the most dramatic event occurred hours ago as armed groups attempted to storm the US embassy and take hostages. In fact, it appears that they were initially successful as a small group of US and foreign diplomats was captured at the outset. A US marine unit was dispatched to rescue the Americans and was able to do so only after a fierce gun battle with armed Manovans. However, several other Americans and foreigners were not rescued and were executed by Manovan crowds. Afterward, their bodies were dragged through the streets by cheering Manovans. At the moment the State Department has no comment on this situation. (3) A control condition was included wherein the subjects received neither a written or audio/visual story, but instead skipped directly to the information series.
Nineteen undergraduate students (recruited from the same pool of subjects of the experiment) were involved in the pretest of the emotional/ affective manipulation. In the neutral condition the subjects were exposed merely to the basic scenario of the international crisis, in the Low Intensity treatment the subjects read the written version of the news report, and in the Intense Emotional induction subjects viewed the full audio/video clip. Following the exposure to 1 of the 3 treatments the students responded on a scale ranging from 0 to 10 whether they have experienced one of the following emotional reactions: Anger; Hate; Sympathy; Fear toward the Manovans.
218
GEVA AND SKORICK
Table 11.1 Pretest Results for Experimental Manipulations Treatment Neutral Written Text Video Clip F
Angry
Hate
Sympathetic
Fear
1.17 6.00 9.43 206.97 p ⬍ .0001
0.17 2.00 7.29 88.63 p ⬍ .0001
1.17 1.67 0.29 4.42 p ⬍ .03
1.33 1.50 0.86 0.57 ns
Table 11.1 illustrates the means of subjects’ responses as a function of the treatment. The responses suggest the effectiveness of the manipulation along three of the four emotions. The only case where the treatment did not yield a statistically significant effect is along the fear dimension.9 The Valence of the Information Following the emotional/affective manipulation subjects were informed that, “As a context in which to gain better insight into what is happening in Manova, you can now view additional information about Manova gathered from news and governmental sources.” Subjects were randomly assigned to review one of three possible information sets, a positive, negative, or mixed set of information. Each set consisted of 22 items of information. The positive set consisted of 18 positive and 4 negative items of information. Items were considered positive if they suggested that Manova or Manovans were similar to the US or sympathetic to US interests. Examples included, “Recently unclassified US intelligence documents indicate that the Manovan diplomatic corps helped US agents during times of peak Cold War hostilities,” and “Manova has long stood in support of the US role in the Organization of South Aegean States and has regularly voted in support of US interests in the region.” The negative set consisted of 18 negative and 4 positive items of information. Items were considered negative if they suggested that Manova or Manovans were dissimilar to the US or antagonistic to US interests. Examples included, “In recent years, the Manovan military has been accused by opposition and Western human rights agencies of corruption and brutality against its citizens,” and “The Manovan government has recently, and without explanation, expelled all foreign western missionaries and has seized all of their property and belongings, giving the westerners only 24 hours to leave the country or face detention and arrest.” The mixed set consisted of 11 positive and 11 negative items of information. To test the effectiveness of the valence manipulation we asked 25 other students (taken from the same subjects pool we used in the experiment) to read the items of a particular set and to form an impression of that nation on a scale ranging from ⫺5 to ⫹5. The effect of the manipulation is statistically
EMOTIONAL CALCULUS OF FOREIGN POLICY
219
significant, F(2,22) ⫽ 39.05 p ⬍ .0001. The positive set of items led to a more positive impression of Manova (M ⫽ 3.33) than the mixed set (M ⫽ .25) and the “worst impression” was in the negative set (M ⫽ ⫺4.13).10 The Procedure Subjects were seated at individual computer terminals in the computer lab of the political science department and were instructed in the use of the computer application before being asked to begin reviewing information. In accordance with our goal of testing hypotheses related to the interpretation of information, subjects were informed at the start of the experiment that their task was to make sense of events occurring in Manova. The computerized software guided subjects through each section of the experiment. Following the affective/emotional manipulation, subjects were presented with the additional information as described earlier. The program allowed subjects to review an item of information and then, by clicking on the button labeled “next item,” review a new item. Subjects were required to review all 22 items of information prior to moving to the final section of the experiment wherein the dependent variable measures were introduced. Upon completion of the experiment, all subjects were debriefed about the theoretical and empirical context of the experiment. Results The findings we report divided into two. The first set of results pertains to effects emotions have on inferences and judgments the participants made on the target nation, that is, their interpretation of the information. The second set addresses the processing parameters that are assumed to mediate these effects. Inferences and Judgment of the Target Nation as a Function of Emotive States We expected variations in the valence of the sequential information sets to influence the interpretation of the crisis and particularly the perception of Manova and its people. Moreover, we hypothesized that intense negative emotions (i.e., hate and anger) will suppress the effects of the different contents compared to the influence of a low intensity emotive induction and the control conditions. The responses of the subjects addressed several dimensions: How democratic is Manova? How similar is that country to the United States of America? How trustworthy are they? How likely is that country to negotiate in a context of crisis?
220
GEVA AND SKORICK
Table 11.2 Inferences and Judgments of the Target Nation as a Function of Information Valence Dimension
Similar Democratic Trustworthy Negotiate
Valence of Information Set Negative
Mixed
Positive
F(2,91)
1.91 0.88 2.19 1.84
3.58 4.15 3.52 4.82
5.57 6.29 5.17 5.77
39.12 p ⬍ .0001 111.98 p ⬍ .0001 26.95 p ⬍ .0001 50.16 p ⬍ .0001
The findings suggest that along these dimensions the information sets had a significant effect. The positively valence sets led to more favorable impressions than the mixed sets, while the least favorable impressions were generated when the valence was mainly negative. These results are shown in table 11.2. More importantly for our conceptualization, the “suppressive” effect of emotion—which should be expressed as an interaction between the emotive manipulation and the valence of the information sets—was obtained along all these dimensions. Specifically, our proposition suggests that when we compare the interpretation of the three sets of information that have different valences, the emotive manipulation would lead to a small gap between the impressions based on positive information than those based on negative information as compared to such gaps that are generated by control or low emotion conditions. Thus, the intense emotion carries an extra amount of “weight” that overrides in part the information conveyed by the items describing the target nation. Table 11.3 illustrates the gap between the impression based on positive and negative information sets generated by the subjects exposed to the video clip (intense negative emotion) and those in the other two conditions (low intensity emotion and control) along each of the earlier mentioned dimensions (the mixed sets were always an in between value). The smaller gaps in the emotive condition are the source for the statistically significant interaction between the experimental factors, and specifically support our contention. It is not that the valence of information is totally ignored. The main effect of the information sets validates the idea that the information matters. Yet the pattern of the interaction implies that intense negative emotions decreases the reliance on the information and affects the interpretation. These reported findings are interesting because when we explore the direct effects of the emotive manipulation on the judgments of the target nation we obtain only two significant main effects. The subjects in the negative emotion condition attributed less of a likelihood that Manova would
Table 11.3 The Gap in Inferences and Judgments of the Target Nation between Positive and Negatively Valenced Information Sets in the Different Emotive States Dimension
High Intensity Emotions
Low Intensity Emotions
Control
F(4,91) Interaction
Positive Items
Negative Items
The Gap
Positive Items
Negative Items
The Gap
Positive Items
Negative Items
The Gap
Similar
4.36
2.39
1.97
5.92
1.22
4.70
7.00
1.90
5.10
Democratic
5.24
1.15
4.09
6.75
.56
6.19
7.33
.80
6.53
Trustworthy
4.07
1.92
2.15
5.50
1.67
3.83
6.44
3.00
3.44
Negotiate
5.14
1.92
3.22
6.50
2.44
4.06
5.78
1.20
4.58
4.23 p ⬍ .005 5.78 p ⬍ .0005 3.24 p ⬍ .02 4.90 p ⬍ .005
221
222
GEVA AND SKORICK
resort to negotiation (M ⫽ 3.33) than did subjects in the condition of low intensity emotion (M ⫽ 5.64) or control [(M ⫽ 3.71), F(2,91) ⫽ 15.09 p ⬍ .0001]. In addition, intense negative emotions led subjects to perceive Manova as less similar to the United States (M ⫽ 3.13) than under conditions of low intensity emotion (M ⫽ 4.21) and control [(M ⫽ 4.04), F(2,91) ⫽ 3.38 p ⬍ .05]. Along the other dimensions the emotive state lowered evaluations but did not do so in a significantly different manner from the other experimental conditions. Affect and Emotion in the Processing of International Crisis Information Our basic premise implies that intense negative emotions (hate and anger) will lower the processing threshold and that this effect will be expressed in less time participants spend acquiring/processing information. The 3 ⫻ 3 ANOVA yielded the following results. First, we found a weak trend (on the verge of statistical significance) in which the intense emotion condition (video clip) generated faster processing time (M ⫽ 175.88 seconds) than the low intensity emotive treatment (text) (M ⫽ 187.04 seconds), and the control conditions [(M ⫽ 191.33), F(2,91) ⫽ 2.89 p ⬍ .06]. This trend is compatible with our expectation and corresponds with the findings on the effects of negative emotion on the interpretation reported earlier. Second, we
230 214.0
210 190
Time
170
202.1
199.8
201.6 186.9
176.4
173.7 171.4 157.0
150 130 110 90 70 50 Mixed
Negative Control
Text
Positive Video
Figure 11.2 Processing Time as a Function of Experimental Condition.
EMOTIONAL CALCULUS OF FOREIGN POLICY
223
obtained a two-way interaction between the emotive state and the type of information subjects processed [F(4,91) ⫽ 3.72 p ⬍ .005]. The pattern of the interaction is illustrated in figure 11.2. Within the experimental conditions of mixed information and positive information, intense negative emotions, as expected, decreased the time spent on acquiring information. The results highlight the distinction between levels of emotion, particularly in the condition of mixed information. In this condition, where the information is complex (mixed), low intensity emotion increased processing time while the intense emotion treatment reduced it. In other words, intense negative emotions dampened information processing when in actuality the situation called for an increase in information processing. Conversely, subjects in the low intensity emotive condition were more in tune to the increased informational demand. However, in the face of predominantly negative or positive information about the target nation (i.e., consistent sets), we find that actually the effect of emotion is quite similar irrespective of the intensity: in both cases there is a reduction in processing time as compared to control (neutral) conditions. The second processing parameter is unforewarned free recall which is used as an indirect measure of how much attention subjects paid to the information they have seen (see Geva, Mayhar, and Skorick 2000). A lowered processing threshold would imply that the subjects who were required to go over all the items in the information set did that ritualistically rather than paying close attention to each item. Shifts in attention would be expressed in the accuracy of their recall.11 The overall recall was not high (about 7 of 22 items). Generally, the low intensity emotive treatment led to more recall (M ⫽ 7.61) than did the high intensity emotion treatment (M ⫽ 6.21) or the control [(M ⫽ 6.50), F(2,91) ⫽ 4.19 p ⬍ .05]. Once more this trend coincides with our proposition that an emotional state has the potential to suppress attention. This finding also coincides with previous studies finding that negative affect increases vigilance to information (Marcus, Wood, and Theiss-Morse 1998; Marcus and MacKuen 1993). However, if attention to incoming information is thematically mediated, then it is important to examine whether the emotive states influenced differently the retrieval of positive versus negative items from the information sets. Since each of the three sets contained a different distribution of positive and negative items we analyzed the proportion of accurate recall of positive and negative items in relation to their number in a specific set. Thus, the proportion of recall of a particular group of valence of items served as the repeated measure in a 3 ⫻ 3 ⫻ 2 ANOVA. The main finding we report of this analysis pertains to the interaction of the emotion/affective manipulation with the items’ subcategory (positive
224
GEVA AND SKORICK 1.00 0.90 Proportion recall
0.80 0.70 0.60 0.50 0.40 0.30
0.46 0.40
0.40 0.29
0.27
0.23
0.20 0.10 0.00 Control
Text
Video
Emotive state Prop-positive
Prop-negative
Figure 11.3 Proportion of Items Recalled as a Function of Emotional/Affective Manipulation.
versus negative) [F(2,91) ⫽ 12.83 p ⬍ .0001]. Figure 11.3 illustrates the pattern. In the control condition, subjects recalled a higher proportion of positive items (.40) than negative items (.23). This result may represent a salience effect whereas the positive items are actually inconsistent with general story line of the initial scenario (cf. Hastie and Kumar 1979). However, in both affective/emotive conditions the trend was reversed, that is, a higher recall level of negative (consistent items with both the story and with the affect/emotion inducement) than positive information. It seems that the negative emotion at both levels led the participants to pay more attention to “confirming” negative information than to thematically “disconfirming” information. Conclusion In this chapter we argued that emotions influence the processing of information and, more specifically, that different levels of their intensity may produce different effects. Our findings support this contention as our experimental results indicate specific effects of high levels of anger and hate on attention to incoming information as well as the impact of such emotions on judgments and inferences made on the basis of new information. In general the experiments illustrate that anger and hate decrease processing time and suppress recall. Furthermore these emotions “flatten” the effects of the differences between positive and negative information.
EMOTIONAL CALCULUS OF FOREIGN POLICY
225
This research represents an attempt to inter-relate affect, emotion, and cognition in the study of international relations and particularly in foreign policy decision making. This task is difficult given the understanding that even the study of cognition is little well received within the field, and that the utility of experimental methods is often questioned (see McDermott 2002). However, recent events such as September 11 and the war on terror have made political scientists more cognizant of symbolic aspects of international relations as well as the need to include emotions in the equation. Our approach and experiment are unique in that we examine the “gradient” of the intensity of negative emotions and suggest different effects for the different levels. Hence in agreement with Marcus and his colleagues, we perceive that low levels of certain negative emotions (anger and hate) have the potential to increase vigilant attention to the environment—yet we refer to ancient Greek logic (or should we say “pathos”) when we support the notion that at more intense levels hate and anger “blind” the individual. Notes 1. Furthermore, we do not claim that all negative emotions have the same implication on the underlying information processes of foreign policy choices. Cf. work by Mosher (2005) that suggests that sadness and anger have different implications for foreign policy decisions. 2. Marcus and colleagues have advanced a model of threat detection which focuses on the public’s perception and evaluation of candidates (Marcus 1988, 2000; Marcus and MacKuen 1993). 3. It should be noted that the theoretical distance between studies which have emphasized emotion as an attitudinal component, and those which have stressed emotion as physiological and neurobiological function mirrors the long-existing debate in the social psychology in relations to the study of emotion and affect, and especially with regard to the status of the emotional system as separate or integrated with the cognitive system (Leventhal 1980; Zajonc 1980, 1984, 2000). 4. See Geva, Mayhar, and Skorick (2000) for a more detailed explication of the model. 5. See also Damasio (1999), Ekman (1994), Kleinginna and Kleinginna (1981), Lewis and Granic (1999), and Lyons (1999) for further discussion of defining emotion. 6. It should be noted that not all negative emotions have such an effect. Boedenhausan et al. (1994) illustrates that sadness is not different from neural conditions in information processing. 7. The use of video clips is a common manipulation of emotions (Pillard, McNair, and Fisher 1974; Brown, Corriveau, and Monti 1977; Engel et al. 1984; Marston et al. 1984; Gross and Levenson 1995; Palomba, Angrilli, and
226
GEVA AND SKORICK
8.
9.
10. 11.
Mini 1997; Hagemann et al. 1999; Niedenthal, Halberstadt, and Setterlund 1997). The validity of these manipulations has been confirmed primarily through physiological observation and self-reports—for which several scales have been developed (Watson, Clark, and Tellegen 1988; Izard et al. 1974). The news clip consist of actual news footage taken from the U.S. involvement in Somalia, and the subsequent dragging of a U.S. soldier’s body through the streets of Mogadishu by Somalians in October of 1993. In additional research we find that when the emotion triggering treatment does not imply “direct” consequences to the subjects (e.g., similarity to the target casualties or physical proximity) it is quite difficult to observe variation along the fear factor (Geva, Redd, and Mosher 2004; Mosher 2005). All the posteriori contrasts (using Scheffe) are significant. Our recall coding of an item addresses accuracy as a retrieval of the main gist of the information item.
REFERENCES
Abele, A. E., and G. H. E. Gendolla. 1999. Satisfaction Judgments in Positive and Negative Moods: Effects of Concurrent Assimilation and Contrast Producing Processes. Personality and Social Psychology Bulletin 25:883–895. Abele, A., G. H. E. Gendolla, and P. Petzold. 1998. Positive Mood and In-GroupOut-Group Differentiation in a Minimal Intergroup Situation. Personality and Social Psychology Bulletin 24:1343–1357. Abelson, Robert P. 1963. Computer Simulation of “Hot” Cognition. In S. Tomkins and S. Messick (Eds.), Computer Simulation of Personality. New York: Wiley. Abelson, Robert P., Donald R. Kinder, Mark D. Peters, and Susan T. Fiske. 1982. Affective and Semantic Components in Political Personal Perception. Journal of Personality and Social Psychology 42(4):619–630. Abramson, Paul, John Aldrich, and David Rhode. 1994. Change and Continuity in the 1992 Elections. Washington, DC: Congressional Quarterly Press. Adolphs, R. 1999. Social Cognition and the Human Brain. Trends in Cognitive Sciences 3:469–479. Adolphs, R., and A. R. Damasio. 2001. The Interaction of Affect and Cognition: A Neurobiological Perspective. In J. P. Forgas (Ed.), Handbook of Affect and Social Cognition. Mahwah, NJ: Lawrence Erlbaum Associates. Anderson, J. R. 1983. A Spreading Activation Theory of Memory. Journal of Verbal Learning and Verbal Behavior 22:261–295. Anderson, J. R., and L. M. Reder. 1999. The Fan Effect: New Results and New Theories. Journal of Experimental Psychology: General 128:186–197. Ansolabehere, Stephen, and Shanto Iyengar. 1995. Going Negative: How Political Advertisements Shrink and Polarize the Electorate. New York: Free Press. Ansolabehere, Stephen, Shanto Iyengar, and Adam Simon. 1999. Replicating Experiments Using Aggregate and Survey Data: The Case of Negative Advertising and Turnout. American Political Science Review 93(4):901–909. Ansolabehere, Stephen, Shanto Iyengar, Adam Simon, and Nicholas Valentino. 1994. Does Attack Advertising Demobilize the Electorate? American Political Science Review 88(4):829–838. Averill, J. R. 1980. A Constructionist View of Emotion. In R. Plutchik and H. Kellerman (Eds.), Emotions: Theory, Research, and Experience, Vol. 1. New York: Academic Press, 305–339. ———. 1982. Anger and Aggression: An Essay on Emotion. New York: Springer Verlag.
228
REFERENCES
Averill, J. R. 1983. Studies on Anger and Aggression. American Psychologist 38:1145–1160. ———. 1988. Disorders of Emotion. Journal of Social and Clinical Psychology 6(3S4):247–268. Axelrod, Robert. 1972. Where the Votes Come From: An Analysis of Electoral Coalitions. American Political Science Review 66:11–20. ———. 1984. The Evolution of Cooperation. New York: Basic Books. Ajzen, I., and M. Fishbein. 1980. Understanding Attitudes and Predicting Social Behavior. New Jersey: Prentice Hall. Banducci, Susan, Michael Thrasher, Colin Rallings, and Jeffrey A. Karp. 2003. Candidate Appearance Cues in Low-Information Elections. Paper presented at the annual meeting of the American Political Science Association, Philadelphia, PA. Barber, James David. 1985. The Presidential Character: Predicting Performance in the White House. 3rd edition. Englewood Cliffs: Prentice-Hall. Bargh, John A. 1994. The Four Horseman of Automaticity: Awareness, Intention, Efficiency, and Control in Social Cognition. In Robert Wyer and Thomas Srull (Eds.), Handbook of Social Cognition: Basic Processes, Vol. 1. Hillsdale, NJ: Erlbaum. ———. 1997. The Automaticity of Everyday Life. In Robert Wyer (Ed.), Advances in Social Cognition, Vol 10. Mahwah, NJ: Erlbaum. ———. 1999. The Cognitive Monster: The Case Against Controllability of Automatic Stereotype Effects. In S. Chaiken and Y. Trope (Eds.), Dual Process Theories in Social Psychology. New York: Guilford. Bargh, John A., and K. Barndollar. 1996. Automaticity in Action: The Unconscious as Repository of Chronic Goals and Motives. In P. M. Gollwitzer and J. A. Bargh (Eds.), The Psychology of Action. New York: Guilford. Bargh, John A., Shelly Chaiken, Rajen Govender, and Felicia Pratto. 1992. The Generality of the Automatic Attitude Activation Effect. Journal of Personality and Social Psychology 62(6):893–912. Bargh, John A., M. Chen, and L. Burrows. 1996. Automaticity of Social Behavior: Direct Effects of Trait Construct and Stereotype Activation on Action. Journal of Personality and Social Psychology 71:230–244. Barrett, L. F., and J. A. Russell. 1998. Independence and Bipolarity in the Structure of Current Affect. Journal of Personality and Social Psychology 74(4):967–984. Barsalou, Lawrence. 1992. Cognitive Psychology: An Overview for Cognitive Scientists. Hillsdale: Erlbaum. Bartels, Larry M. 1996. Going Negative: How Political Advertisements Shrink and Polarize the Electorate (Book review). Public Opinion Quarterly 60(3):456–461. ———. 1997. How Campaigns Matter. Unpublished manuscript. Princeton University. ———. 2000. Partisanship and Voting Behavior, 1952–1996. American Journal of Political Science 44:35–50. Basil, Michael, Caroline Schooler, and Byron Reeves. 1991. Positive and Negative Political Advertising: Effectiveness of Advertisements and Perceptions of
REFERENCES
229
Candidates. In Frank Biocca (Ed.), Television and Political Advertising, Volume 1: Psychological Processes. Hillsdale, NJ: Lawrence Erlbaum. Berelson, Bernard, Paul Lazarsfeld, and William McPhee. 1954. Voting: A Study of Opinion Formation in a Presidential Campaign. Chicago: Chicago University Press. Berkowitz, Leonard. 1990. On the Formation and Regulation of Anger and Aggression: A Cognitive-Neoassociationistic Analysis. American Psychologist 45:494–503. ———. 1993. Towards a General Theory of Anger and Emotional Aggression. In T. K. Srull and R. S. Wyer (Eds.), Advances in Social Cognition, Vol. 6. Hillsdale, NJ: Erlbaum. ———. 2000. Causes and Consequences of Feelings. In K. Oatley and A. Manstead (Eds.), Studies in Emotion and Social Interaction. Cambridge: Cambridge University Press. Berscheid, E., and E. H. Walster. 1974. Physical Attractiveness. In Leonard Berkowitz (Ed.), Advances in Experimental Social Psychology, vol. 7. New York: Academic Press. Betsch, Tilman, Henning Plessner, Christiane Schwieren, and Robert Gutig. 2001. I Like It But I Don’t Know Why: A Value Account Approach to Implied Attitude Formation. Personality and Social Psychology Bulletin 27:242–253. Billings, Robert S., and Charles F. Hermann. 1998. Problem Identification in Sequential Policy Decision Making: The Re-representation of Problems. In James F. Voss and Donald Sylvan (Eds.), Problem Representation in Foreign Policy Decision Making. Cambridge: Cambridge University Press. Bless, H. 2000. The Interplay of Affect and Cognition: The Mediating Role of General Knowledge Structures. In J. P. Forgas (Ed.), Feeling and Thinking: The Role of Affect in Social Cognition. New York: Cambridge University Press. ———. Mood and the Use of General Knowledge Structures. In L. L. Martin and G. L. Clore (Eds.), Theories of Mood and Cognition: A User’s Guidebook. Mahwah, NJ: Erlbaum. Bless, H., G. Bohner, N. Schwarz, and F. Strack. 1990. Mood and Persuasion: A Cognitive Response Analysis. Personality and Social Psychology Bulletin 16:331–345. Bless, H., G. L. Clore, N. Schwarz, V. Golisano, C. Rabe, and M. Wolk. 1996. Mood and the Use of Scripts: Does a Happy Mood Really Lead to Mindlessness? Journal of Personality and Social Psychology 71:665–679. Bless, H., and N. Schwarz. 1999. Sufficient and Necessary Conditions in Dual Process Models: The Case of Mood and Information Processing. In S. Chaiken and Y. Trope (Eds.), Dual Process Theories in Social Psychology. New York: Guilford. Bless, H., N. Schwarz, and R. Wieland. 1996. Mood and the Impact of Category Membership and Individuating Information. European Journal of Social Psychology 26:935–959. Bobo, Lawrence. 2000. Race and Beliefs about Affirmative Action: Assessing the Effects of Interests, Group threat, Ideology, and Racism. In David O. Sears, Jim Sidanius, and Lawrence Bobo (Eds.), Racialized Politics: The Debate about Racism in America. Chicago: University of Chicago Press.
230
REFERENCES
Bobo, Lawrence, and Vincent L. Hutchings. 1996. Perceptions of Racial Group Competition: Extending Blumer’s Theory of Group Position to a Multiracial Social Context. American Sociological Review 61:951–972. Bodenhausen, G. V., G. P. Kramer, and K. Susser. 1994. Happiness and Stereotypic Thinking in Social Judgment. Journal of Personality and Social Psychology 66:621–632. Bodenhausen, G. V., L. A. Sheppard, and G. P. Kramer. 1994. Negative Affect and Social Judgment: The Differential Impact of Anger and Sadness. European Journal of Social Psychology 24:45–62. Bodenhasen, G. V., and R. S. Wyer, Jr. 1985. Effects of Stereotypes on Decision Making and Information-Processing Strategies. Journal of Personality and Social Psychology 48:267–282. Bohner, G., and T. Weinerth. 2001. Negative Affect Can Increase or Decrease Message Scrutiny: The Affect Interpretation Hypothesis. Personality and Social Psychology Bulletin 27:1417–1428. Bohner, G., K. Crow, H. P. Erb, and N. Schwarz. 1992. Affect and Persuasion: Mood Effects on the Processing of Message Content and Context Cues on Subsequent Behavior. European Journal of Social Psychology 22:511–530. Boninger, David S., Jon A. Krosnick, and Matthew K. Berent. 1995. Origins of Attitude Importance: Self-Interest, Social Identification, and Value Relevance. Journal of Personality and Social Psychology 68:61–80. Bower, G. H. 1981. Mood and Memory. American Psychologist 6:129–148. ———. 1983. Affect and Cognition. Transactions of the Royal Philosophical Society of London, Series B 302:387–402. ———. 1991. Mood Congruity of Social Judgments. In J. P. Forgas (Ed.), Emotion and Social Judgments. Oxford: Pergamon. Bower, G. H., and J. P. Forgas. 2000. Affect, Memory, and Social Cognition. In E. Eich, J. F. Kihlstrom, G. Bower, J. P. Forgas, and P. M. Niedenthal (Eds.), Cognition and Emotion. Oxford: Oxford University Press. ———. 2001. Affect and Social Memory. In J. P. Forgas (Ed.), Handbook of Affect and Social Cognition. Mahwah, NJ: Erlbaum. Bower, G. H., S. G. Gilligan, and K. P. Montiero. 1981. Selectivity of Learning Caused by Affective States. Journal of Experimental Psychology: General 100:451–473. Bowlby, John Y. 1973. Attachment and Loss, Vol. 2, Separation: Anxiety and Anger. New York: Basic Books. Boynton, G. R., and John Nelson. 2003. Sound Arguments. Paper presented at the Shambaugh Conference of Affect and Cognition in Political Action, University of Iowa. Brader, Ted. 1999. Campaigning for Hearts and Minds: How Campaign Ads Use Emotion and Information to Sway the Electorate. Ph.D. Thesis. Department of Political Science, Harvard University, Cambridge. ———. 2005. Campaigning for Hearts and Minds: How Political Ads Use Emotion to Sway the Electorate. Chicago: University of Chicago Press. Brady, Henry E., and Richard Johnston. 1987. What was the Primary Message: Horse Race or Issue Journalism? In Gary R. Orren and Nelson W. Polsby
REFERENCES
231
(Eds.), Media and Momentum: The New Hampshire Primary and Momentum Politics. Chatham, NH: Chatham House. Brady, Henry E., and Paul M. Sniderman. 1985. Attitude Attribution: A Group Basis for Political Reasoning. American Political Science Review 79:1061–1078. Branscombe, N. 1988. Conscious and Unconscious Processing of Affective and Cognitive Information. In K. Fiedler and J. P. Forgas (Eds.), Affect, Cognition, and Social Behavior. Toronto: Hogrefe International. Brehm, J. W. 1956. Postdecision Changes in the Desirability of Alternatives. Journal of Abnormal and Social Psychology 52:384–389. Brians, Craig L., and Martin P. Wattenberg. 1996. Campaign Issue Knowledge and Salience: Comparing Reception from TV Commercials, TV News and Newspapers. American Journal of Political Science 40(1):172–193. Broadbent, D., and M. Broadbent. 1988. Anxiety and Attentional Bias: State and Trait. Cognition and Emotion 2:165–183. Brody, Richard A., and Paul M. Sniderman. 1977. From Life Space to Polling Place: The Relevance of Personal Concerns for Voting Behavior. British Journal of Political Science 7(3):337–360. Brown, W. A., D. P. Corriveau, and P. M. Monti. 1977. Anger Arousal by a Motion Picture: A Methodological Note. American Journal of Psychiatry 134:930–931. Bruce, John M. 1994. The Question of Emotional Response. Paper presented at the annual meeting of the Midwest Political Science Association, Chicago, IL, April 14–16. Budesheim, T. L., and S. J. DePaola. 1994. Beauty or the Beast? The Effects of Appearance, Personality, and Issue Information on Evaluations of Political Candidates. Personality and Social Psychology Bulletin 20(4):339–348. Burke, M., and A. M. Mathews. 1992. Autobiographical Memory and Clinical Anxiety. Cognition and Emotion 6:23–35. Butler, G., and A. M. Mathews. 1983. Cognitive Processes in Anxiety. Advanced Behavior Research Therapy 5:51–62. ———. Anticipatory Anxiety and Risk Perception. Cognitive Therapy and Research 11:551–565. Butler, J. L., and R. F. Baumeister. 1998. The Trouble with Friendly Faces: Skilled Performance with a Supportive Audience. Journal of Personality and Social Psychology 75:1213–1230. Cacioppo, John T., and Gary G. Berntson. 1999. The Affect System: Architecture and Operating Characteristics. Current Directions in Psychological Science 8(5):133–137. Cacioppo, John T., Wendi L. Gardner, and Gary G. Berntson. 1997. Beyond Bipolar Conceptualizations and Measures of Attitudes and Evaluative Space. Personality and Social Psychological Review 1:3–25. ———. 1999. The Affect System has Parallel and Integrative Processing Components: Form Follows Function. Journal of Personality and Social Psychology 76(5):839–855. Cacioppo, John T., David J. Klein, Gary G. Berntson, and Elaine Hatfield. 1993. The Psychophysiology of Emotion. In M. Lewis and J. M. Haviland (Eds.), Handbook of Emotions. New York: Guilford Press.
232
REFERENCES
Calder, B. J., and C. L. Gruder. 1988. A Network Activation Theory of Attitudinal Affect. Unpublished manuscript. Kellogg School of Business, Northwestern University. Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald Stokes. 1960. The American Voter. New York: John Wiley and Sons. Cappella, Joseph N., and Kathleen Hall Jamieson. 1997. Spiral of Cynicism: The Press and the Public Good. New York: Oxford University Press. Cappella, Louis, and Ronald D. Taylor. 1992. An Analysis of the Effectiveness of Negative Political Campaigning. Business and Public Affairs 18(2):10–17. Carmines, Edward G., and James A. Stimson. 1989. Issue Evolution: Race and the Transformation of American Politics. Princeton, NJ: Princeton University Press. Chaiken, S., and J. A. Bargh. 1993. Occurrence versus Moderation of the Automatic Attitude Activation Effect: Reply to Fazio. Journal of Personality and Social Psychology 64:759–765. Chaiken, S., and D. Maheswaran. 1994. Heuristic Processing Can Bias Systematic Processing: Effects of Source Credibility, Argument Ambiguity, and Task Importance on Attitude Judgment. Journal of Personality and Social Psychology 66:460–473. Cialdini, R. B. 2003. Crafting Normative Messages to Protect the Environment. Current Directions in Psychological Science 12:105–109. Ciarrochi, J. V., and J. P. Forgas. 1999. On Being Tense Yet Tolerant: The Paradoxical Effects of Trait Anxiety and Aversive Mood on Intergroup Judgments. Group Dynamics: Theory, Research and Practice 3:227–238. Clark, A. M., and B. A. Waddell. 1983. Effects of Moods on Thoughts about Helping, Attraction and Information Acquisition. Social Psychology Quarterly 46:31–35. Clore, G. L. 1992. Cognitive Phenomenology: Feelings and the Construction of Judgment. In L. L. Martin and A. Tesser (Eds.), The Construction of Social Judgments. Hillsdale, NJ: Erlbaum. Clore, G. L., and S. Colcombe. 2001. The Role of Context in Suboptimal Affective Priming. Unpublished manuscript. University of Illinois. Clore, G. L., and L. M. Isbell. 2001. Emotion as Virtue and Vice. In J. H. Kuklinski (Ed.), Citizens and Politics: Perspectives from Political Psychology. New York: Cambridge University Press. Clore, G. L., N. Schwarz, and M. Conway. 1994. Affective Causes and Consequences of Social Information Processing. In R. S. Wyer and T. K. Srull (Eds.), Handbook of Social Cognition. 2nd edition. Hillsdale, NJ: Erlbaum. Clore, G. L., and A. Ortony. 2000. Cognition in Emotion: Always, Sometimes, or Never? In R. D. Lane and L. Nadel (Eds.), Cognitive Neuroscience of Emotion. New York: Oxford University Press. Clore, G. L., and W. G. Parrot. 1991. Moods and Their Vicissitudes: Thoughts and Feelings as Information. In J. Forgas (Ed.), Emotion and Social Judgment. Oxford: Pergamon. ———. 1994. Cognitive Feelings and Metacognitive Judgments. European Journal of Social Psychology 24:101–116.
REFERENCES
233
Clore, G. L., R. S. Wyer, B. Dienes, K. Gasper, C. Gohm, and L. M. Isbell. 2001. Affective Feelings as Feedback: Some Cognitive Consequences. In L. L. Martin and G. L. Clore (Eds.), Theories of Mood and Cognition: A User’s Guidebook. Mahwah, NJ: Erlbaum. Colcombe, S., L. M. Isbell, L., and G. L. Clore. 2001. The Effects of Suboptimal Exposure to Smiles and Frowns on Information Processing. Unpublished manuscript. University of Illinois at Urbana-Champaign. Collins, A. M., and Loftus, E. F. 1975. A Spreading-Activation Theory of Semantic Processing. Psychological Review 82:407–428. Committee of Concerned Journalists. Project for Excellence in Journalism. URL: www.journalism.org/statementl Conover, Pamela J. 1984. The Influence of Group Identifications on Political Perception and Evaluation. The Journal of Politics 46:760–784. ———. The Role of Social Groups in Political Thinking. British Journal of Political Science 18:51–76. Conover, Pamela J., and Stanley Feldman. 1981. The Origin and Meaning of Liberal / Conservative Self-identification. American Journal of Political Science 25:617–645. ———. 1986. Emotional Reactions to the Economy: I’m Mad as Hell and I’m not Going to Take it Anymore. American Journal of Political Science 30:30–78. Converse, Philip E. 1964. The Nature of Belief Systems in Mass Publics. In David Apter (Ed.), Ideology and Discontent. New York: Free Press. Cottam, Martha L. 1986. Foreign Policy Decision Making: The Influence of Cognition. Boulder: Westview Press. ———. 1994. Images and Intervention: U.S. Policies in Latin America. Pittsburgh: Pittsburgh University Press. Crawford, Neta C. 2000. The Passion of World Politics. International Security 24:116–156. Damasio, Antonio R. 1994. Descartes’ Error: Emotion, Reason and the Human Brain. New York: G. P. Putnam’s Sons. ———. 1996. The Somatic Marker Hypothesis and the Possible Functions of the Prefrontal Cortex. Philosophical Transactions of the Royal Society of London. Series B-Biological Sciences 351:1413–1420. ———. 1999. The Feeling of What Happens. New York: Harcourt, Brace and Co. Dawson, Michael C. 1994. Behind the Mule: Race and Class in African-American Politics. Princeton: Princeton University Press. DeRivera, Joseph H. 1981. The Structure of Anger. In J. H. De Rivera (Ed.), Conceptual Encounter: A Method for the Exploration of Human Experience. Washington, DC: University Press of America. Delli Carpini, Michael X., and Scott Keeter. 1996. What Americans Know about Politics and Why it Matters. New Haven: Yale University Press. Derryberry, John 1991. The Immediate Effects of Positive and Negative Feedback Signals. Journal of Personality and Social Psychology 61:267–278. Devine, Patricia. 1989. Stereotypes and Prejudice: Their Automatic and Controlled Components. Journal of Personality and Social Psychology 56:680–690.
234
REFERENCES
Devine, P. G., and M. J. Monteith. 1993. The Role of Discrepancy-Associated Affect in Prejudice Reduction. In D. M. Mackie and D. L. Hamilton (Eds.), Affect, Cognition, and Stereotyping: Interactive Processes in Group Perception. San Diego: Academic Press. Devine, P. G., M. J. Monteith, J. R. Zuwerink, and A. J. Elliot. 1991. Prejudice With and Without Compunction. Journal of Personality and Social Psychology 60:817–830. Dillard, James P., Courtney A. Plotnick, Linda C. Godbold, Vicki S. Freimuth, and Timothy Edgar. 1996. The Multiple Affective Outcomes of AIDS PSAs: Fear Appeals Do More than Scare People. Communication Research 23(1):44–72. Dion, K. K., E. Berscheid, and E. H. Walster. 1972. What is Beautiful is Good. Journal of Personality and Social Psychology 24:285–290. Ditto, Peter H., and David F. Lopez. 1992. Motivated Skepticism: Use of Differential Decision Criteria for Preferred and Nonpreferred Conclusions. Journal of Personality and Social Psychology 63:568–584. Dixon, William J. 1983. Measuring Interstate Affect. American Journal of Political Science 27:828–851. Dovidio, J. F., S. L. Gaertner, and S. Loux. 2000. Subjective Experiences and Intergroup Relations: The Role of Positive Affect. In H. Bless and J. P. Forgas (Eds.), The Message Within: The Role of Subjective Experience in Social Cognition and Behavior. Philadelphia: Psychology Press. Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row. Eagly, Alice H., and Shelly Chaiken. 1993. The Psychology of Attitudes. Fort Worth, TX: Harcourt Brace Jovanovich. Edwards, John A., Gifford Weary, William von Hippel, and Jill A. Jacobson. 2000. The Effects of Depression on Impression Formation: The Role of Trait and Category Diagnosticity. Personality and Social Psychology Bulletin 26(4):462–473. Edwards, Kari, and Edward Smith. 1996. A Disconfirmation Bias in the Evaluation of Arguments. Journal of Personality and Social Psychology 71(1):5–24. Ekman, Paul. 1994. All Emotions are Basic. In P. Ekman and R. J. Davidson (Eds.), The Nature of Emotion. Oxford: Oxford University Press. Ellsworth, P. C., and C. A. Smith. 1988. Shades of Joy: Patterns of Appraisal Differentiating Pleasant Emotions. Cognition and Emotion 2:301–331. Engel, G. L., M. Frader, C. Barry, and G. R. Morrow. 1984. Sadness Evoked by a Film on Grief: An Experimental Study. International Journal of Psychiatry in Medicine 14:1–30. Entman, Robert M. 1992. Blacks in the News; Television, Modern Racism, and Cultural Change. Journalism Quarterly 69:341–362. Entman, Robert M., and Andrew Rojecki. 2000. The Black Image in the White Mind: Media and Race in America. Chicago: University of Chicago Press. Epstein, Seymour. 1972. The Nature of Anxiety with Emphasis on Its Relationship to Expectancy. In C. D. Spielberger (Ed.), Anxiety: Current Trends in Theory and Research, Vol. 2. New York: Academic Press. Erber, R. 1991. Affective and Semantic Priming: Effects of Mood on Category Accessibility and Inference. Journal of Experimental Social Psychology 27:480–498.
REFERENCES
235
Evans, J. St. B. T., and D. E. Over. 1996. Rationality and Reasoning. Hove, UK: Psychology Press. Erber, R., and M. W. Erber. 1994. Beyond Mood and Social Judgment: Mood Incongruent Recall and Mood Regulation. European Journal of Social Psychology 24:79–88. Eysenck, Michael, and Mark Keane. 1995. Cognitive Psychology. Sussex, UK: Erlbaum Fabrigar, Leandre R., Penny S. Visser, and Michael W. Browne. 1997. Conceptual and Methodological Issues in Testing the Circumplex Structure of Data in Personality and Social Psychology. Personality and Social Psychology Review 1(3):184–203. Fazio, Russell. 1986. How do Attitudes Guide Behavior? In R. M. Sorrento and E. T. Higgins (Eds.), The Handbook of Motivation and Cognition: Foundation of Social Behavior. New York: Guilford Press. ———. 1993. Variability in the Likelihood of Automatic Activation: Data Reanalysis and Commentary on Bargh, Chaiken, Govender and Pratto. Journal of Personality and Social Psychology 64(5):753–758. ———. 1995. Attitudes as Object-Evaluation Associations: Determinants, Consequences, and Correlates of Attitude Accessibility. In Richard Petty and Jon Krosnick (Eds.), Attitude Strength: Antecedents and Consequences. Hillsdale: Erlbaum. ———. 2001. On the Automatic Activation of Associated Evaluations: An Overview. Cognition and Emotion 15(2):115–141. Fazio, Russell, H. Russel, J. R. Jackson, B. C. Dunton, and C. J. Williams. 1995. Variability in Automatic Activation as an Unobtrusive Measure of Racial Attitudes: A Bona Fide Pipeline? Journal of Personality and Social Psychology 69:1013–1027. Fazio, Russell, David Sanbonmatsu, Martha Powell, and Frank Kardes. 1986. On the Automatic Activation of Attitudes. Journal of Personality and Social Psychology 50:229–238. Festinger, L. 1957. A Theory of Cognitive Dissonance. Stanford, CA: Stanford University Press. Finkel, Steven E. 1993. Reexamining the Minimal Effects Model in Recent Presidential Campaigns. The Journal of Politics 55:1–21. Finkel, Steven E., and John Geer. 1998. A Spot Check: Casting Doubt on the Demobilizing Effect of Attack Advertising. American Journal of Political Science 42(2):573–595. Fiorina, M. P. 1981. Retrospective Voting in American National Elections. New Haven: Yale University Press. Fiske, S. T., and S. L. Neuberg. 1990. A Continuum of Impression Formation, from Category Based to Individuating Processes: Influences of Information and Motivation on Attention and Interpretation. In M. P. Zanna (Ed.), Advances in Experimental Social Psychology, Vol. 23. New York: Academic Press. Fiske, S. T., and S. E. Taylor. 1991. Social Cognition. New York: McGraw-Hill. Foa, Edna B., and Michael J. Kozak. 1993. Pathological Anxiety: The Structure and Meaning of Fear. In N. Birbaumer and A. Öhman (Eds.), The Structure of Emotion: Psychophysiological, Cognitive and Clinical Aspects. Seattle: Hogrefe and Huber.
236
REFERENCES
Forgas, J. P. 1992. Affect in Social Judgments and Decisions: A Multi-Process Model. In M. P. Zanna (Ed.), Advances in Experimental Social Psychology, Vol. 25. San Diego: Academic Press. ———. 1995. Mood and Judgment: The Affect Infusion Model (AIM). Psychological Review 117:39–66. Forgas, J. P., and G. H. Bower. 1987. Mood Effects on Person-Perception Judgments. Journal of Personality and Social Psychology 20:497–513. ———. 1988a. Affect in Social and Personal Judgments. In K. Fiedler and J. Forgas (Eds.), Affect, Cognition, and Social Behavior. Toronto: Hogrefe International. ———. 1988b. Affect in Social Judgments. Australian Journal of Psychology 40:125–145. Forgas, J. P., J. V. Ciarrochi, and S. Moylan. 2000. Affective Influences on the Production of Persuasive Messages. Unpublished Manuscript. University of New South Wales, Sydney, Australia. Franklin, Charles. 1991. Eschewing Obfuscation? Campaigns and the Perception of U.S. Senate Incumbents. American Political Science Review 85:1193–1214. Freedman, Paul, and Ken Goldstein. 1999. Measuring Media Exposure and the Effects of Negative Campaign Ads. American Journal of Political Science 43(4):1189–1208. Friedman, R. S., and J. Foerster. 2000. The Effects of Approach and Avoidance Motor Actions on the Elements of Creative Insight. Journal of Personality and Social Psychology 79:477–492. Frijda, Nico H. 1986. The Emotions. Cambridge: Cambridge University Press. Gal, Reuven, and Richard S. Lazarus. 1975. The Role of Activity in Anticipating and Confronting Stressful Situations. Journal of Human Stress 1(4):4–20. Gamson, W. A. 1992. Talking Politics. New York: Cambridge University Press. Garramone, Gina M. 1984. Voter Response to Negative Political Ads. Journalism Quarterly 61(2):250–259. Garramone, Gina M., Charles T. Atkin, Bruce E. Pinkleton, and Richard T. Cole. 1990. Effects of Negative Political Advertising on the Political Process. Journal of Broadcasting and Electronic Media 34(3):299–311. Gasper, K., and G. L. Clore. 1998. The Persistent Use of Negative Affect by Anxious Individuals to Estimate Risk. Journal of Personality and Social Psychology 74:1350–1363. Gehm, Theodor L., and Klaus R. Scherer. 1988. Relating Situation Evaluation to Emotional Differentiation: Nonmetric Analysis of Cross-Cultural Questionnaire Data. In K. R. Scherer (Ed.), Facets of Emotion. Hillsdale: Erlbaum. Gelman, Andrew, and Gary King. 1993. Why are American Presidential Election Campaign Polls So Variable When Votes are So Predictable? British Journal of Political Science 23:409–451. George, Alexander L. 1969. The “Operational Code”: A Neglected Approach to the Study of Political Leaders and Decision-Making. International Studies Quarterly 13:190–222. Geva, Nehemia, Joe Clare, and Katrina N. Mosher. 2002. Information Reliability and the Cognitive Calculus of Foreign Policy Decision Making. Paper
REFERENCES
237
presented at the twenty-fifth Annual Scientific Meeting of the International Society of Political Psychology, Berlin, Germany, July 16–19. Geva, Nehemia, James Mayhar, and Steven B. Redd. 1997. The Three R’s in Foreign Policy Decision Making: A Computational Model of the Role of Relevance, Reliability and Redundancy of Information. Paper presented at the annual meeting of the International Society of Political Psychology, Toronto, March 18–22. Geva, Nehemia, James Mayhar, and J. Mark Skorick. 2000. The Cognitive Calculus of Foreign Policy Decision Making: An Experimental Assessment. Journal of Conflict Resolution 44:447–471. Geva, Nehemia, Katrina N. Mosher, and Joe Clare. 2003. Quality vs. Quantity: Information Reliability and the Cognitive Calculus of Foreign Policy Decision Making. Paper presented at the annual meeting of the Midwest Political Science Association, Chicago, IL, April 3–6. Geva, Nehemia, Steven B. Redd, and Katrina N. Mosher. 2004. International Terror, Emotions, and Foreign Policy Decision Making. Paper presented at the International Studies Association 2004 Conference, Montreal, Canada. Geva, Nehemia, and J. Mark Skorick. 1999. Information Inconsistency and the Cognitive Algebra of Foreign Policy Decision Making. International Interactions 25:333–362. ———. 2000a. Process and Outcome Consequences of Simultaneous Foreign Policy Decisions. Paper presented at the annual meeting of the International Society of Political Psychology, Seattle, WA. ———. 2000b. The Cognitive Algebra of Sequential Foreign Policy Decision Making: Selective Exposure in Post-Decisional Information Processing. Paper presented at the annual meeting of the Midwest Political Science Association, Chicago, IL. Gibson, James L. 1992. The Political Consequences of Intolerance: Cultural Conformity and Political Freedom. American Political Science Review 86:338–356. Gibson, James L., and Richard D. Bingham. 1982. On the Conceptualization and Measurement of Political Tolerance. American Political Science Review 76:603–620. Gilliam, Jr., Franklin D., and Shanto Iyengar. 2000. Prime Suspects: The Influence of Local Television News on the Viewing Public. American Journal of Political Science 44:560–573. Gilligan, S. G. 1982. Mood Intensity and Learning of Congruous Material. Ph.D. Dissertation. Department of Psychology, Stanford University. Goffman, Erving. 1959. The Presentation of Self in Everyday Life. New York: Doubleday Books. ———. 1963. Behavior in Public Places: Notes on the Social Organization of Gatherings. New York: Free Press. Goldstein, Kenneth. 1997. Political Advertising and Political Persuasion in the 1996 Election. Paper Presented at the annual meeting of the American Political Science Association, Washington, DC. Gollob, Harry F. 1968. Impression Formation and Word Combination in Sentences. Journal of Personality and Social Psychology 10:341–353.
238
REFERENCES
Gordon, Steven L. 1990. The Sociology of Sentiments and Emotion. In Morris Rosenberg and Ralph H. Turner (Eds.), Social Psychology: Sociological Perspectives. New Brunswick, NJ: Transaction Publishers. Gorsuch, Richard L. 1983. Factor Analysis. Hillsdale: Lawrence Erlbaum Associates. Graber, Doris A. 1984. Processing the News: Taming the Information Tide. New York: Longman. Gray, Jeffrey A. 1973. Causal Theories of Personality and How to Test Them. In J. R. Joyce (Ed.), Multivariate Analysis and Psychological Theory. New York: Academic Press. ———. 1981. The Psychophysiology of Anxiety. In Richard Lynn (Ed.), Dimensions of Personality: Papers in Honour of H. J. Eysenck. New York: Pergamon. ———. 1982. The Neuropsychology of Anxiety: An Enquiry into the Functions of the Septo-Hippocampal System. Oxford: Clarendon Press. ———. 1987a. The Neuropsychology of Emotion and Personality. In S. M. Stahl, S. D. Iversen, and E. C. Goodman (Eds.), Cognitive Neurochemistry. Oxford, England: Oxford University Press. ———. 1987b. The Psychology of Fear and Stress. 2nd edition. Cambridge: Cambridge University Press. ———. 1990. Brain Systems That Mediate both Emotion and Cognition. Cognition and Emotion 4(3):269–288. ———. 2004. Consciousness: Creeping Up on the Hard Problem. New York: Oxford University Press. Green, Donald P., and Jack Citrin. 1994. Measurement Error and the Structure of Attitudes: Are Positive and Negative Judgments Opposites? American Journal of Political Science 38(1):256–281. Green, Donald P., and Alan Gerber. 1999. Misperceptions about Perceptual Bias. Annual Review of Political Science 2:189–210. Green, Donald P., Susan Lee Goldman, and Peter Salovey. 1993. Measurement Error Masks Bipolarity in Affect Ratings. Journal of Personality and Social Psychology 64(6):1029–1041. Green, Donald P., and Ian Shapiro. 1994. Pathologies of Rational Choice Theory: A Critique of Applications in Political Science. New Haven: Yale University Press. Greene, Steven. 2002. The Social-Psychological Measurement of Partisanship. Political Behavior 24(3):171–197. Greenstein, Fred. 1969. Personality and Politics: Problems of Evidence, Inference, and Conceptualization. Chicago: Markham Publishing Company. Greenwald, A., and M. Banaji. 1995. Implicit Social Cognition: Attitudes, Selfesteem, and Stereotypes. Psychological Review 102:4–27. Greenwald, A., M. Banaji, L. A. Rudman, S. D. Farnham, B. A. Nosek, and D. S. Mellott. 2002. A Unified Theory of Implicit Attitudes, Stereotypes, Selfesteem, and Self Concept. Psychological Review 109:3–25. Greenwald, A., J. E. Pickrell, and S. D. Farnham. 2002. Implicit Partisanship: Taking Sides for No Reason. Journal of Personality and Social Psychology 83:367–379. Gross, James J., and Robert W. Levenson. 1995. Emotion Elicitation Using Films. Cognition and Emotion 9:87–108.
REFERENCES
239
Hagemann, Dirk, Ewald Naumann, Stefanie Maier, Gabriele Becker, Alexander Lurken, and Dieter Bartussek. 1999. The Assessment of Affective Reactivity Using Films: Validity, Reliability and Sex Differences. Personality and Individual Differences 26:627–639. Hampton, Peter J. 1978. The Many Faces of Anger. Psychology: A Journal of Human Behavior 15(1):35–44. Hastie, R. 1988. A Computer Simulation Model of Person Memory. Journal of Experimental Social Psychology 24:423–447. Hastie, R., and P. A. Kumar. 1979. Person Memory: Personality Traits as Organizing Principles in Memory for Behaviors. Journal of Personality and Social Psychology 37:25–38. Hastie, R., and B. Park. 1986. The Relationship between Memory and Judgment Depends on Whether the Task is Memory-Based or On-Line. Psychological Review 93:258–268. Hebb, D. O. 1949. The Organization of Behavior: A Neuropsychological Theory. New York: Wiley. ———. 1955. Drives and the C.N.S. (Conceptual Nervous System). Psychological Review 62:243–254. Heise, David R. 1968. Norms and Individual Patterns in Student Deviancy. Social Problems 16:78–92. ———. 1969. Affective dynamics in simple sentences. Journal of Personality and Social Psychology 11:204–213. ———. 1977. Social Action as the Control of Affect. Behavioral Science 22:163–177. ———. 1979. Understanding Events: Affect and the Construction of Social Action. New York: Cambridge University Press. ———. 1987. Affect Control Theory: Concepts and Model. Journal of Mathematical Sociology 13:1–33. ———. 2001. Project Magellan: Collecting Cross-Cultural Affective Meanings via the Internet. Electronic Journal of Sociology 5(3). Heise, David R., and Neil J. MacKinnon. 1987. Affective Bases of Likelihood Judgments. Journal of Mathematical Sociology 13:133–151. Helson, H. 1964. Adaptation-Level Theory: An Experimental and Systematic Approach to Behavior. New York: Harper and Row. Hepworth, J. T., and S. G. West. 1988. Lynchings and the Economy: A TimeSeries Reanalysis of Hovland and Sears, 1940. Journal of Personality and Social Psychology 55:239–247. Hermann, Charles F. 1972. International Crises: Insights from Behavioral Research. New York: Free Press. Hermann, Charles F., Margaret G. Hermann, and Gerald L. Hutchins. 1982. Affect. In Patrick Callahan, Lindy P. Brady, and Margaret G. Hermann (Eds.), Describing Foreign Policy Behavior. Beverly Hills, CA: Sage. Herrmann, Richard K. 1988. The Empirical Challenge of the Cognitive Revolution: A Strategy for Drawing Inferences about Perceptions. International Studies Quarterly 32:175–203.
240
REFERENCES
Herrmann, Richard K., James F. Voss, Tonya Y. E. Schooler, and Joseph Ciarrochi. 1997. Images and International Relations: An Experimental Test of Cognitive Schemata. International Studies Quarterly 41:403–433. Higgins, E. T., J. A. Bargh, and W. Lombardi. 1985. The Nature of Priming Effects on Categorization. Journal of Experimental Psychology: Learning, Memory and Cognition 11:59–69. Hildebrand-Saints, L., and G. Weary. 1989. Depression and social information gathering. Personality and Social Psychology Bulletin 15:150–160. Hinnich, Melvin J., and Michael C. Munger. 1994. Ideology and the Theory of Political Choice. Ann Arbor: University of Michigan Press. Hitchon, Jacqueline C., and Chingching Chang. 1995. Effects of Gender Schematic Processing on the Reception of Political Commercials for Men and Women Candidates. Communication Research 22(3):430–458. Hitchon, Jacqueline C., Chingching Chang, and Rhonda Harris. 1997. Should Women Emote? Perceptual Bias and Opinion Change in Response to Political Ads for Candidates of Different Genders. Political Communication 14(1):49–69. Hochschild, Arlie R. 1979. Emotion Work, Feeling Rules, and Social Structure. American Journal of Sociology 85:551–575. Holbrook, Thomas. 1996. Do Campaigns Matter? Thousand Oaks, CA: Sage. Holsti, Ole R. 1972. Crisis, Escalation, War. Montreal: McGill-Queen’s University Press. ———. 1976. Foreign Policy Formation Viewed Cognitively. In Robert Axelrod (Ed.), Structure of Decision: The Cognitive Maps of Political Elites. Princeton, NJ: Princeton University Press. ———. 1979. Theories of Crisis Decision Making. In Paul Gordon Lauren (Ed.), Diplomacy: New Approaches in History, Theory, and Policy. New York: Free Press. ———. 1996. Public Opinion and American Foreign Policy. Ann Arbor: University of Michigan Press. Hovland, C. I., and R. R. Sears. 1940. Minor Studies in Aggression: VI. Correlation of Lynchings with Economic Indices. Journal of Psychology 9:301–310. Huckfeldt, R., J. Levine, W. Morgan, and J. Sprague. 1997. Accessibility and the Political Utility of Partisan and Ideological Orientations. American Journal of Political Science 3:888–911. Hudson, Valerie M. 1995. Culture and Foreign Policy. Paper presented at the annual meeting of the International Studies Association, Chicago, IL. Hudson, Valerie M., Charles F. Hermann, and Eric Singer. 1989. The Situational Imperative: A Predictive Model of Foreign Policy Behavior. Cooperation and Conflict 24:117–139. Hutchings, Vincent L. 2001. Political Context, Issue Salience, and Selective Attentiveness: Constituent Knowledge of the Clarence Thomas Confirmation Vote. Journal of Politics 63:846–868. Isbell, L. M. 2004. Not All Happy People Are Lazy or Stupid: Evidence of Systematic Processing in Happy Moods. Journal of Experimental Social Psychology 40:341–349.
REFERENCES
241
Isbell, L. M., K. C. Burns, and T. Haar. In Press. The Influence of Affect on Information Selection in Impression Formation. Social Cognition. Isbell, L. M., K. C. Burns, and L. James. 2004. The Influence of Affect on Question Selection during an Interview. Unpublished Manuscript. Isbell, L. M., and V. Ottati. 2002. The Emotional Voter: Effects of Episodic Affective Reactions on Candidate Evaluation. In V. Ottati, R. S. Tindale, J. Edwards, F. B. Bryant, L. Heath, D. C. O’Connell, Y. Suarez-Balcazar, and E. J. Posavac (Eds.), The Social Psychology of Politics. Social Psychological Application to Social Issues, Vol. 5. New York: Kluwer Academic-Plenum Publishers. Isbell, L. M., and R. S. Wyer. 1999. Correcting For Mood-Induced Bias in the Evaluation of Political Candidates: The Roles of Intrinsic and Extrinsic Motivation. Personality and Social Psychology Bulletin 25:237–249. Isen, A. M. 1984. Toward Understanding the Role of Affect in Cognition. In R. S. Wyer, Jr., and T. K. Srull (Eds.), Handbook of social cognition Vol. 3. Hillsdale, NJ: Erlbaum. ———. 1987. Positive Affect, Cognitive Processes, and Social Behavior. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology Vol. 20. San Diego, CA: Academic Press. Isen, A. M., T. E. Shalker, M. Clark, and L. Karp. 1978. Affect, Accessibility of Materials in Memory, and Behavior: A Cognitive Loop? Journal of Personality and Social Psychology 36:1–12. Iyengar, Shanto. 1990. Shortcuts to Political Knowledge: The Role of Selective Attention and Accessibility. In J. Ferejohn and J. Kuklinski (Eds.), Information and Democratic Processes. Urbana and Chicago: University of Illinois Press. Iyengar, Shanto, and Donald R. Kinder. 1987. News That Matters: Television and American Opinion. Chicago: University of Chicago Press. Izard, Caroll E. 1977. Human Emotions. New York: Plenum Press. Izard, Caroll E., and Edmund S. Bartlett. 1972. Patterns of Emotions: A New Analysis of Anxiety and Depression. Oxford: Academic Press. Izard, Caroll E., F. E. Dougherty, B. M. Bloxom, and N. E. Kotsch. 1974. The Differential Emotions Scale: A Method of Measuring the Meaning of Subjective Experience of Discrete Emotions. Nashville: Vanderbilt University. James, William. 1890. The Principles of Psychology. Cambridge, MA: Harvard University Press. Jamieson, Kathleen Hall. 1992. Dirty Politics: Deception, Distraction and Democracy. New York: Oxford University Press. Jamieson, Kathleen Hall, and Paul A. Waldman. 2000. Watching the Adwatches. In Larry Bartels and Lynn Vavreck (Eds.), Campaign Reform: Insights and Evidence. Ann Arbor, MI: University of Michigan Press. Janis, Irving L., and Leon Mann. 1977. Decision Making: A Psychological Analysis of Conflict, Choice, and Commitment. New York: Free Press. Johnston, L. C. 1996. Resisting Change: Information-Seeking and Stereotype Change. European Journal of Social Psychology 26:799–825. Johnston, L. C., and C. N. Macrae. 1994. Changing Social Stereotypes: The Case of the Information Seeker. European Journal of Social Psychology 24:581–592.
242
REFERENCES
Jones, Edward E. 1971. Attribution: Perceiving the Causes of Behavior. Morristown: General Learning Press. Just, Marion, Ann Crigler, and Tami Buhr. 1999. Voice, Substance and Cynicism. Political Communication 16(1):25–43. Just, Marion, Ann Crigler, Dean Alger, Timothy Cook, Montague Kern, and Darrell West. 1996. Crosstalk: Citizens, Candidates, and the Media in a Presidential Campaign. Chicago: University of Chicago Press. Just, Marion, Ann Crigler, and Todd Belt. 2004. Don’t Give up Hope. Paper presented at the Affect Effect Conference, University of Michigan, Ann Arbor, MI, October 2004. Kahn, Kim Fridkin, and John G. Geer. 1994. Creating Impressions: An Experimental Investigation of Political Advertising on Television. Political Behavior 16(1):93–116. Kahn, Kim Fridkin, and Patrick J. Kenney. 1999. Do Negative Campaigns Mobilize or Suppress Turnout? Clarifying the Relationship between Negativity and Participation. American Political Science Review 93(4):877–889. Kaid, Lynda Lee. 1996. Presidential Ads as Nightly News: A Content Analysis of 1988 and 1992 Televised Ad Watches. Journal of Broadcasting and Electronic Media 40:297–308. ———. 1997. Effects of Television Spots on Images of Dole and Clinton. American Behavioral Scientist 40(3):1085–1094. Kaid, Lynda Lee, and Anne Johnston. 1991. Negative versus Positive Television Advertising in US Presidential Campaigns, 1960–1988. Journal of Communication 41(3):53–64. Karpinski, A., R. Steinman, and J. L. Hilton. 2005. Attitude Importance as a Moderator of the Relationship Between Implicit and Explicit Attitude Measures. Personality and Social Psychology Bulletin 31:949–962. Keltner, D., P. C. Ellsworth, and K. Edwards. 1993. Beyond Simple Pessimism: Effects of Sadness and Anger on Social Perception. Journal of Personality and Social Psychology 64:740–752. Keltner, D. P., K. D. Locke, and P. C. Audrain. 1993. The Influence of Attributions on the Relevance of Negative Feelings to Personal Satisfaction. Personality and Social Psychology Bulletin 19:21–29. Kemper, Theodore D. 1991. Predicting Emotions from Social Relations. Social Psychology Quarterly 54(4):330–342. Kerbel, Matthew. 1995. Edited for Television: ABC, CNN and the 1992 Presidential Campaign. Boulder, CO: Westview. Kern, Montague. 1989. 30 Second Politics: Political Advertising in the Eighties. New York: Praeger. Kessel, J. H. 2004. Views of Voters. In H F. Weisberg and C. Wilcox (Eds.), Models of Voting in Presidential Elections: The 2000 U.S. Election. Stanford, CA: Stanford University Press. Key, V. O. 1966. The Responsible Electorate: Rationality in Presidential Voting, 1936–1960. Cambridge: Harvard University Press. Kinder, Donald R. 1986. Presidential Character Revisited. In R. R. Lau and D. O. Sears (Eds.), Political Cognition: The 19th Annual Carnegie Symposium on Cognition. Hillsdale: Erlbaum.
REFERENCES
243
Kinder, Donald R. 1998. Opinion and Action in the Realm of Politics. In D. Gibert, S. Fiske, and S. Gardner (Eds.), Handbook of Social Psychology. 4th edition. New York: Oxford University Press. Kinder, Donald R., and Thomas R. Palfrey. 1993. On Behalf of an Experimental Political Science. In Donald R. Kinder and Thomas R. Palfrey (Eds.), Experimental Foundations of Political Science. Ann Arbor: University of Michigan Press. Kinder, Donald R., and Lynn M. Sanders. 1996. Divided by Color: Racial Politics and Democratic Ideals. Chicago: University of Chicago Press. King, Erika G., Robert W. Henderson, and Hong C. Chen. 1998. Viewer Response to Positive vs. Negative Ads in the 1996 Presidential Campaign. Paper presented at the annual meeting of the Midwest Political Science Association, Chicago, IL. Klapper, Joseph T. 1960. The Effects of Mass Communications. Glencoe, IL: Free Press. Kleinginna, Paul R., and Anne M. Kleinginna. 1981. A Categorized List of Emotion Definitions, with Suggestions for a Consensual Definition. Motivation and Emotion 5(4):345–359. Krauth-Gruber, S., and F. Ric. 2000. Affect and Stereotypic Thinking: A Test of the Mood-and-General-Knowledge Model. Personality and Social Psychology Bulletin 26:1587–1597. Krosnick, John. 2002a. The Challenges of Political Psychology: Lessons to be Learned from Research on Attitude Perception. In James Kuklinski (Ed.), Thinking about Political Psychology. New York: Cambridge University Press. ———. 2002b. The Causes of No-Opinion Responses to Attitude Measures in Surveys: They are Rarely What They Appear to Be. In R. M. Groves, D. A. Dillman, J. L. Eltinge, and R. J. A. Little (Eds.), Survey Nonresponse. New York: Wiley. Kuklinski, James H. (Ed.) 2001. Citizens and Politics: Perspectives from Political Psychology. New York: Cambridge University Press. Kunda, Ziva. 1987. Motivated Inference: Self-Serving Generation and Evaluation of Evidence. Journal of Personality and Social Psychology 53(4):636–647. ———. 1990. The Case for Motivated Political Reasoning. Psychological Bulletin 108(3):480–498. Lachman, R., J. L. Lachman, and E. C. Butterfield. Cognitive Psychology and Information Processing. Hillsdale, NJ: Erlbaum. Lakoff, George. 1997. Moral Politics: What Conservatives Know That Liberals Don’t. Chicago: University of Chicago Press. Lamb, Christopher J. 1984. Belief Systems and Decision Making in the Mayaguez Crisis. Political Science Quarterly 99:681–702. Lambert, A. J., S. R. Khan, B. A. Lickel, and K. Fricke. 1997. Mood and the Correction of Positive versus Negative Stereotypes. Journal of Personality and Social Psychology 72:1002–1016. Lang, Peter J., Mark K. Greenwald, Margaret M. Bradley, and Alfons O. Hamm. 1993. Looking at Pictures: Affective, Facial, Visceral and Behavioral Reactions. Psychophysiology 30:261–273. Lanzetta, John T., Denis G. Sullivan, Roger D. Masters, and Gregory J. McHugo. 1985. Emotional and Cognitive Responses to Televised Images of Political
244
REFERENCES
Leaders. In Sidney Kraus and Richard M. Perloff (Eds.), Mass Media and Political Thought. Beverly Hills, CA: Sage. Larsen, Randy J., and Edward Diener. 1992. Promises and Problems with the Circumplex Model of Emotion. In M. S. Clark (Ed.), Emotion. Newbury Park, CA: Sage. Lau, Richard R. 1985. Two Explanations for Negativity Effects in Political Behavior. American Journal of Political Science 29(1):119–138. ———. 1986. Political Schemata, Candidate Evaluations, and Voting Behavior. In Richard R. Lau and David O. Sears (Eds.), Political Cognition: The 19th Annual Carnegie Symposium on Cognition. Hillsdale, NJ: Lawrence Erlbaum. ———. 1989a. Construct Accessibility and Electoral Choice. Political Behavior 11:5–32. ———. 1989b. Individual and Contextual Influences on Group Identification. Social Psychology Quarterly 52:220–231. ———. 1995. Information Search during an Election Campaign: Introducing a Process Tracing Methodology for Political Scientists. In M. Lodge and K. Mcgraw (Eds.), Political Judgment: Structure and Process. Ann Arbor, MI: University of Michigan Press. Lau, Richard R., and Gerald Pomper. 2001. Effects of Negative Campaigning on Turnout in US Senate Elections, 1988–1998. Journal of Politics 63(3):804–819. ———. 2002. Effectiveness of Negative Campaigning in US Senate Elections. American Journal of Political Science 46(1):47–66. ———. 2004. Negative Campaigning: An Analysis of US Senate Elections (Campaigning American Style). Lanham, MD: Rowman and Littlefield. Lau, Richard R., Gerald Pomper, and Grace A. Mumoli. 1998. Effects of Negative Campaigning on Senate Election Outcomes: 1988, 1990, 1994, and 1996. Paper presented at the annual meeting of the Midwest Political Science Association, Chicago, IL. Lau, Richard R., and David P. Redlawsk. 1997. Voting Correctly. American Political Science Review 91:585–598. ———. 2001a. An Experimental Study of Information Search, Memory, and Decision-Making During a Political Campaign. In J. Kuklinski (Ed.), Citizens and Politics: Perspectives from Political Psychology. New York: Cambridge University Press. ———. 2001b. Advantages and Disadvantages of Cognitive Heuristics in Political Decision Making. American Journal of Political Science 45(4):951–971. ———. 2006. How Voters Decide: Information Processing during an Election Campaign. New York: Cambridge University Press. Lau, Richard R., and David O. Sears. 1981. Cognitive Links between Economic Grievances and Political Responses. Political Behavior 3(4):279–302. ———. 1986. Political Cognition: The 19th Annual Carnegie Symposium on Cognition. Hillsdale, NJ: Earlbaum. Lau, Richard R., Lee Sigelman, Caroline Heldman, and Paul Babbitt. 1999. The Effects of Negative Political Advertisements: A Meta-Analytic Assessment. American Political Science Review 93(4):851–875.
REFERENCES
245
Lazarsfeld, Paul, Benard Berelson, and Helen Gaudet. 1944. The People’s Choice. New York: Duell, Sloane, and Pearce. Lazarus, Richard S. 1991. Emotion and Adaptation. New York: Oxford University Press. LeDoux, Joseph. 1996. The Emotional Brain: The Mysterious Underpinnings of Emotional Life. New York: Simon and Schuster. ———. 2003. Synaptic Self: How Our Brains Become Who We Are. London: Penguin Books. Lerner, J. S., J. H. Goldberg, and P. E. Tetlock. 1998. Sober Second Thought: The Effects of Accountability, Anger, and Authoritarianism on Attributions of Responsibility. Personality and Social Psychology Bulletin 24:563–574. Lerner, J. S., R. M. Gonzalez, D. A. Small, and B. Fischhoff. 2003. Effects of Fear and Anger on Perceived Risks of Terrorism: A National Field Experiment. Psychological Science 14: 144–150. Lerner, J. S., and D. Keltner. 2000. Beyond Valence: Toward a Model of EmotionSpecific Influences on Judgment and Choice. Cognition and Emotion 14:473–493. ———. 2001. Fear, Anger, and Risk. Journal of Personality and Social Psychology 81:146–159. Lewis, Marc D., and Isabela Granic. 1999. Self-organization of Cognition-Emotion Interactions. In Tim Dalgleish and Mick Power (Eds.), Handbook of Cognition and Emotion. New York: John Wiley and Sons. Lewis-Beck, Michael S. 1988. Economics and Elections: The Major Western Democracies. Ann Arbor: University of Michigan Press. Leventhal, Howard. 1980. Toward a Comprehensive Theory of Emotion. Advances in Experimental Social Psychology 13:139–207. Lichtenstein, M., and T. K. Srull. 1987. Processing Objectives as a Determinant of the Relationship Between Recall and Judgment. Journal of Experimental Social Psychology 23:93–118. Lindsay, Peter, and Donald Norman. 1977. Human Information Processing: An Introduction to Psychology. New York: Academic Press. Lloyd, G. G., and W. A. Lishman. 1975. Effect of Depression on the Speed of Recall of Pleasant and Unpleasant Experiences. Psychological Medicine 5:173–180. Lodge, Milton. 1995. Toward a Procedural Model of Candidate Evaluation. In Milton Lodge and Kathleen M. McGraw (Eds.), Political Judgment: Structure and Process. Ann Arbor: University of Michigan Press. Lodge, Milton, Kathleen McGraw, and Patrick Stroh. 1989. An Impression-driven Model of Candidate Evaluation. American Political Science Review 83:399–419. Lodge, Milton, Marco Steenbergen, and Shawn Brau. 1995. The Responsive Voter: Campaign Information and the Dynamics of Candidate Evaluation. American Political Science Review 89:309–326. Lodge, Milton, and Patrick Stroh. 1993. Inside the Mental Voting Booth: An Impression-driven Model. In Shanto Iyengar and William McGuire (Eds.), Explorations in Political Psychology. Durham, NC: Duke University Press.
246
REFERENCES
Lodge, Milton, and Charles Taber. 1996. Three Steps toward a Theory of Motivated Political Reasoning. Paper presented at the International Society of Political Psychology, Vancouver, Canada. Lodge, Milton, and Charles Taber. 2000. Three Steps toward a Theory of Motivated Political Reasoning. In Arthur Lupia, Matthew McCubbins, and Samuel Popkin (Eds.), Elements of Political Reason: Understanding and Expanding the Limits of Rationality. London: Cambridge University Press. ———. 2002. The Primacy of Affect for Political Candidates. Groups, and Issues: An Experimental Test of the Hot Cognition Hypothesis. Unpublished manuscript. ———. 2005. The Primacy of Affect for Political Candidates, Groups, and Issues: An Experimental Test of the Hot Cognition Hypothesis. Political Psychology 26:455–482. Lord, Charles, Michael Ross, and Mark Lepper. 1979. Biased Assimilation and Attitude Polarization: The Effects of Prior Theories on Subsequently Considered Evidence. Journal of Personality and Social Psychology 27:2098–2109. Lyons, William. 1999. The Philosophy of Cognition and Emotion. In Tim Dalgleish and Mick Power (Eds.), Handbook of Cognition and Emotion. New York: John Wiley and Sons. MacDonald, Stuart Elaine, George Rabinowitz, and Ola Listhaug. 1998. On Attempting to Rehabilitate the Proximity Model: Sometimes the Patient Just Can’t Be Helped. Journal of Politics 60:653–690. Mackie, D. M., and L. T. Worth. 1989. Processing Deficits and the Mediation of Positive Affect in Persuasion. Journal of Personality and Social Psychology 57:27–40. MacKuen, Michael B., George E. Marcus, W. Russell Neuman, Luke Keele, and Jennifer Wolak. 2001a. Emotional Framing, Information Search, and the Operation of Affective Intelligence in Matters of Public Policy. Unpublished manuscript. University of North Carolina, Chapel Hill. ———. 2001b. Emotions, Information, and Political Cooperation. Paper presented at annual meeting of the American Political Science Association, at San Francisco. MacLeod, C., and I. L. Cohen. 1993. Anxiety and the Interpretation of Ambiguity: A Text Comprehension Study. Journal of Abnormal Psychology 102:238–247. Manis, M., and J. C. Moore. 1978. Summarizing Controversial Messages: Retroactive Effects due to Subsequent Information. Social Psychology Quarterly 41:62–68. Marcus, George E. 1988. The Structure of Emotional Response: 1984 Presidential Candidates. American Political Science Review 82(3):727–761. ———. 2000. Emotions in Politics. In N. W. Polsby (Ed.), Annual Review of Political Science. Palo Alto, CA: Annual Reviews. ———. 2002. The Sentimental Citizen: Emotion in Democratic Politics. University Park, PA: Pennsylvania State University Press. ———. 2003. The Psychology of Emotion and Politics. In D. O. Sears, L. Huddy, and R. Jervis (Eds.), Handbook of Political Psychology. Oxford: Oxford University Press.
REFERENCES
247
Marcus, George E., and Michael B. MacKuen. 1993. Anxiety, Enthusiasm, and the Vote: The Emotional Underpinnings of Learning and Involvement during Presidential Campaigns. American Political Science Review 87(4):672–685. Marcus, George E., Michael B. MacKuen, Jenny Wolak, and Luke Keele. 2005. The Measure and Mismeasure of Emotion. This Volume. Marcus, George E., Russell Neuman, and Michael B. Mackuen. 2000. Affective Intelligence and Political Judgment. Chicago: University of Chicago Press. Marcus, George E., and Wendy Rahn. 1990. Emotions and Democratic Politics. In S. Long (Ed.), Research in Micropolitics. Greenwich, CT: JAI Press. Marcus, George E., John L. Sullivan, Elizabeth Theiss-Morse, and Sandra Wood. 1995. With Malice Toward Some: How People Make Civil Liberties Judgments. New York: Cambridge University Press. Marcus, George E., Sandra L. Wood, and Elizabeth Theiss-Morse. 1998. Linking Neuroscience to Political Intolerance and Political Judgment. Politics and the Life Sciences 17(2):165–178. Marcus-Newall, A., W. C. Pederson, M. Carlson, and N. Miller. 2000. Displace Aggression is Alive and Well: A Meta-Analytic Review. Journal of Personality and Social Psychology 78:670–689. Markus, Gregory B. 1982. Political Attitudes during an Election Year: A Report on the 1980 NES Panel Study. American Political Science Review 76(3):538–560. Marston, A., J. Hart, C. Hileman, and W. V. Faunce. 1984. Toward the Laboratory Study of Sadness and Crying. American Journal of Psychology 97:127–131. Martin, L. L. 1986. Set/Reset: Use and Disuse of Concepts in Impression Formation. Journal of Personality and Social Psychology 51:493–504. Martin, L. L., J. J. Seta, and R. A. Crelia. 1990. Assimilation and Contrast as a Function of People’s Willingness and Ability to Expend Effort in Forming an Impression. Journal of Personality and Social Psychology 59:27–37. Martinez, Michael D., and Tad Delegal. 1990. The Irrelevance of Negative Campaigns to Political Trust: Experimental and Survey Results. Political Communication and Persuasion 7(1):25–40. Marx, Karl. 1978 [1844]. Estranged Labor. In Robert C. Tucker (Ed.), The MarxEngels Reader. 2nd edition. New York: W. W. Norton and Company. Mathews, A. M., and C. MacLeod. 1985. Selective Processing of Threat Cues in Anxiety States. Behavior Research and Therapy 23:563–569. ———. 1994. Cognitive Approaches to Emotion and Emotional Disorders. Annual Review of Psychology 45:25–50. Mayer, J. E., M. Gayle, M. E. Meehan, and A. K. Harman. 1990. Toward a Better Specification of the Mood-Congruency Effect in Recall. Journal of Experimental Social Psychology 26:465–480. Mayer, William G. 1996. In Defense of Negative Campaigning. Political Science Quarterly 111(3):437–455. McDermott, Rose. Experimental Methods in Political Science. Annual Review of Political Science 5:31–64. McKinnon, Lori Melton, and Lynda Lee Kaid. 1999. Exposing Negative Campaigning or Enhancing Advertising Effects: An Experimental Study of Ad
248
REFERENCES
Watch Effects on Voters’ Evaluations of Candidates and their Ads. Journal of Applied Communication Research 27:217–236. McNally, R. J. 1994. Cognitive Bias in Panic Disorders. Current Directions in Psychological Science 3:129–132. Mehrabian, Albert. 1980. Basic Dimensions for a General Psychological Theory. Cambridge: Oelgeschlager, Gunn, and Hain. Merritt, Sharyne. 1984. Negative Political Advertising: Some Empirical Findings. Journal of Advertising 13(3):27–38. Miller, G. A. 1957. The Magic Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information. Psychological Review 63:81–97. Mogg, K., A. M. Mathews, and J. Weinman. 1987. Memory Bias in Clinical Psychology. Journal of Abnormal Psychology 96:94–98. Monro, Geoffrey, Peter Ditto, Lisa Lockhart, Angele Fagerlin, Mitchell Gready, and Elizabeth Peterson. 2002. Biased Assimilation of Sociopolitical Arguments: Evaluating the 1996 U.S. Presidential Debate. Basic and Applied Social Psychology 24(1):15–26. Monteith, M. J. 1993. Self-Regulation of Prejudiced Responses: Implications for Progress in Prejudiced Reduction Efforts. Journal of Personality and Social Psychology 65:469–485. Mor, Ben D. 1993. Decision and Interaction in Crisis. Westport, CT: Praeger. Morgan, Rick L., and David Heise. 1988. Structure of Emotions. Social Psychology Quarterly 51(1):19–31. Mosher, Katrina N. 2005. Public Interpretations and Reactions to Terror Related Casualties: The Effects of Numbers and Identities. Paper presented at the annual meeting of the International Society of Political Psychology, Honolulu, Hawaii, March 1–5. Murphy, S. T., and Robert Zajonc. 1993. Affect, Cognition and Awareness: Affective Priming with Optimal and Suboptimal Stimulus Exposures. Journal of Personality and Social Psychology 64:723–729. Nabi, Robin L. 1999. A Cognitive-Functional Model for the Effects of Discrete Negative Emotions on Information Processing, Attitude Change, and Recall. Communication Theory 9(3):292–320. ———. 2002. Anger, Fear, Uncertainty, and Attitudes: A Test of the CognitiveFunctional Model. Communication Monographs 69(3):204–216. Nadeau, R., R. G. Niemi, and T. Amato. 1995. Emotions, Issue Importance, and Political Learning. American Journal of Political Science 39:558–574. Neely, J. 1977. Semantic Priming and the Retrieval of Lexical Memory: Roles of Inhibitionless Spreading Activation and Limited-Capacity Attention. Journal of Experimental Psychology: General 106:226–254. Nelson, John S., and G. Robert Boynton. 1997. Video Rhetorics: Televised Advertising in American Polticis. Urbana: University of Illinois Press. Nelson, Thomas E., and Donald R. Kinder. 1996. Issue Frames and GroupCentrism in American Public Opinion. The Journal of Politics 58:1055–1078. Nelson, Thomas, Rosalee Clawson, and Zoe Oxley. 1997. Media Framing of a Civil Liberties Conflict and Its Effect on Tolerance. American Political Science Review 91:567–583.
REFERENCES
249
Neuman W. Russell, George E. Marcus, and Michael B. MacKuen. 1996. Emotion and Reason in Mass Political Behavior. Paper presented at the International Society of Political Psychology, Vancouver, Canada. Neuman, W. Russell, Marion Just, and Ann Crigler. 1992. Common Knowledge: News and the Construction of Political Meaning. Chicago: University of Chicago Press. Neustadt, Richard. 1991 [1960]. Presidential Power and the Modern Presidents. Revised edition. New York: Free Press. Nie, Norman H., Sidney Verba, and John R. Petrocik. 1975. The Changing American Voter. Cambridge, MA: Harvard University Press. Niedenthal, P. M. 1990. Implicit Perception of Affective Information. Journal of Experimental Social Psychology 26:505–527. Niedenthal, P. M., J. B. Halberstadt, and M. B. Setterlund. 1997. Being Happy and Seeing Happy: Emotional State Mediates Visual Word Recognition. Cognition and Emotion 11:403–432. Nisbett, R., and Wilson, T. 1977. Telling More Than We Can Know: Verbal Reports on Mental Processes. Psychological Review 84:231–259. Nussbaum, Martha Craven. 2001. Upheavals of Thought: The Intelligence of Emotions. Cambridge, England: Cambridge University Press. Oatley, Keith. 2000. The Sentiments and Beliefs of Distributed Cognition. In N. H. Frijda, A. S. R. Manstead, and S. Bem (Eds.), Emotions and Belief: How Feelings Influence Thoughts. New York: Cambridge University Press. Ohman, Arne. 1993. Fear and Anxiety as Emotional Phenomena: Clinical Phenomenology, Evolutionary Perspectives, and Information-Processing Mechanisms. In M. Lewis, and J. M. Haviland (Eds.), Handbook of Emotion. New York: Guilford Press. Ortony, Andrew, Gerald L. Clore, and Allan Collins. 1988. The Cognitive Structure of Emotions. Cambridge: Cambridge University Press. Osgood, Charles E., George C. Suci, and Perry H. Tannenbaum. 1957. The Measurement of Meaning. Urbana, IL: University of Illinois Press. Ottati, V. C. 1990. Determinants of Political Judgments: The Joint Influence of Native and Heuristic Rules of Inference. Political Behavior 12:159–179. ———. 1997. When the Survey Question Directs Retrieval: Implications for Assessing the Cognitive and Affective Predictors of Global Evaluation. European Journal of Social Psychology 27:1–21. Ottati, V. C., and L. M. Isbell. 1996. Effects of Mood during Exposure to Target Information on Subsequently Reported Judgments: An On-Line Model of Misattribution and Correction. Journal of Personality and Social Psychology 71:39–53. Ottati, V. C., E. Riggle, R. S. Wyer, N. Schwarz, and J. Kuklinski. 1989. Cognitive and Affective Bases of Opinion Survey Responses. Journal of Personality and Social Psychology 57:404–415. Ottati, V. C., M. Steenbergen, and E. Riggle. 1992. The Cognitive and Affective Components of Political Attitudes: Measuring the Determinants of Candidate Evaluations. Political Behavior 14:423–442. Ottati, V. C., N. Terkildsen, and C. Hubbard. 1997. Happy Faces Elicit Heuristic Processing in a Televised Impression Formation Task: A Cognitive Tuning Account. Personality and Social Psychology Bulletin 23:1144–1156.
250
REFERENCES
Ottati, V. C., and R. S. Wyer. 1993. Affect and Political Judgment. In S. Iyengar, and W. J. McGuire (Eds.), Explorations in Political Psychology. Durham, NC: Duke University Press. Page, Benjamin I. 1976. A Theory of Political Ambiguity. American Political Science Review 70:742–752. Page, Benjamin I., and Richard A. Brody. 1972. Policy Voting and the Electoral Process: The Vietnam War Issue. American Political Science Review 66:979–995. Palomba, Daniela, Alessandro Angrilli, and Alessio Mini. 1997. Visual Evoked Potentials, Heart Rate Responses and Memory to Emotional Pictorial Stimuli. International Journal of Psychophysiology 27:55–67. Park, J., and M. R. Banaji. 2000. Mood and Heuristics: The Influence of Happy and Sad States on Sensitivity and Bias in Stereotyping. Journal of Personality and Social Personality 78:1005–1023. Parrot, W. G., and J. Sabini. 1990. Mood and Memory under Natural Conditions: Evidence for Mood Incongruent Recall. Journal of Personality and Social Psychology 59:321–336. Parsons, Talcott, and Edwared A. Shils. 1962. Toward a General Theory of Social Action. New York: Harper and Row. Patterson, Thomas E. 1980. The Mass Media Election: How Americans Choose their President. New York: Praeger. ———. 1993. Out of Order. New York: Knopf. Patterson, Thomas E., and Robert McClure. 1976. The Unseeing Eye: The Myth of Television Power in National Elections. New York: Paragon Books. Pederson, W. C., C. Gonzales, and N. Miller. 2000. The Moderating Effect of Trivial Triggering Provocation on Aggression. Journal of Personality and Social Psychology 78:913–927. Pentony, Joseph F. 1998. Effects of Negative Campaigning on Vote, Semantic Differential, and Thought Listing. Journal of Applied Social Psychology 28(23):2131–2149. Petrocik, John R. 1996. Issue Ownership in Presidential Elections with a 1980 Case Study. American Journal of Political Science 40:825–850. Petty, R. E., and J. T. Cacioppo. 1981. Attitudes and Persuasion: Classic and Contemporary Approaches. Dubuque: Brown. ———. 1986. The Elaboration Likelihood Model of Persuasion. Advances in Experimental Social Psychology 19:123–205. Petty, R. E., D. Schumann, S. Richman, and A. Strathman. 1993. Positive Mood and Persuasion: Different Roles for Affect Under High and Low Elaboration Conditions. Journal of Personality and Social Psychology 64:5–20. Petty, R. E., and D. T. Wegener. 1993. Flexible Correction Processes in Social Judgment: Correcting for Context-Induced Contrast. Journal of Experimental Social Psychology 29:137–165. Pillard, R. C., D. M. McNair, and S. Fisher. 1974. Does Marijuana Enhance Experimentally Induced Anxiety? Psychopharmacologia 40:205–210. Pinkleton, Bruce. 1997. The Effects of Negative Comparative Political Advertising on Candidate Evaluations and Advertising Evaluations: An Exploration. Journal of Advertising 16(1):19–29.
REFERENCES
251
Plutchik, Robert, and Hope R. Conte. 1997. Circumplex Models of Personality and Emotions. Washington, DC: American Psychological Association. Pomper, Gerald M. 1972. From Confusion to Clarity: Issues and American voters, 1956–1968. American Political Science Review 66(2):415–428. Posner M. I., and C. R. R. Snyder. 1975. Attention and Cognitive Control. In R. L. Solso (Ed.), Information Processing and Cognition. Hillsdale, NJ: Erlbaum. Pratto, Felicia, and Oliver P. John. 1991. Automatic Vigilance: The Attentiongrabbing Power of Negative Social Information. Journal of Personality and Social Psychology 61:380–391. Price, Vincent. 1989. Social Identification and Public Opinion: Effects of Communicating Group Conflict. Public Opinion Quarterly 53:197–224. Price, Vincent, and John Zaller. 1993. Who Gets the News? Alternative Measures of News Reception and their Implications for Research. Public Opinion Quarterly 57:133–164. Rabinowitz, George, and Stuart Elaine Macdonald. 1989. A Directional Theory of Issue Voting. American Political Science Review 83:93–121. Ragsdale, L. 1991. Strong Feelings: Emotional Responses to Presidents. Political Behavior 13:33–65. Rahn, Wendy M. 2000. Affect as Information: The Role of Public Mood in Political Reasoning. In A. Lupia, M. McCubbins, and S. Popkin (Eds.), Elements of Reason: Cognition, Choice, and the Bounds of Rationality. New York: Cambridge University Press. Rahn, Wendy M., John H. Aldrich, Eugene Borgida, and John L. Sullivan. 1990. A Social-Cognitive Model of Candidate Appraisal. In J. Ferejohn, and J. Kuklinski (Eds.), Information and Democratic Processes. Urbana-Champaign: University of Illinois Press. Rahn, Wendy M., B. Kroeger, and C. M. Kite. 1996. A Framework for the Study of Public Mood. Political Psychology 17:29–58. Rapoport, Ronald B., Kelly L. Metcalf, and Jon A. Hartman. 1989. Candidate Traits and Voter Inferences: An Experimental Study. Journal of Politics 51(4):917–932. Redlawsk, David P. 2001. You Must Remember This: A Test of the On-Line Model of Voting. Journal of Politics 63:29–58. ———. 2002. Hot Cognition or Cool Consideration? Testing the Effects of Motivated Reasoning on Political Decision Making. Journal of Politics 64:1021–1044. ———. 2004. What Voters Do: Information Search during Election Campaigns. Political Psychology 25(4):595–610. Remington, Nancy A., Leandre R. Fabrigrar, and Penny S. Visser. 2000. Reexamining the Circumplex Model of Affect. Journal of Social Psychology 79(2):286–300. Richardson, Glenn. 2001. Looking for Meaning in All the Wrong Places. Journal of Communication 51:775–880. Riggle, E. D. B., V. C. Ottati, R. S. Wyer, R. S., J. Kuklinski, and N. Schwartz. 1992. Bases of Political Judgments: The Role of Stereotypic and Nonstereotypic Information. Political Behavior 14:67–87.
252
REFERENCES
Roberts, Jonathan M. 1988. Decision-Making during International Crises. New York: St. Martin’s Press. Robinson, Michael J., and Margaret A. Sheehan. 1983. Over the Wire and on TV: CBS and UPI in Campaign ’80. New York: Russell Sage Foundation. Roddy, B. L., and Gina M. Garramone. 1988. Appeals and Strategies of Negative Political Advertising. Journal of Broadcasting and Electronic Media 32(4):415–427. Roseman, I. J. 1984. Cognitive Determinants of Emotions: A Structural Theory. In P. Shaver (Ed.), Review of Personality and Social Psychology. Beverly Hills, CA: Sage. ———. 1991. Appraisal Determinants of Discrete Emotions. Cognition and Emotion 5:161–200. Rosenberg, S. W., L. Bohan, P. McCafferty, and K. Harris. 1986. The Image and the Vote: The Effect of Candidate Presentation on Voter Preference. American Journal of Political Science 30(1):108–127. Rosenberg, S. W., and P. McCafferty. 1987. The Image and the Vote: Manipulating Voters’ Preferences. Public Opinion Quarterly 51(Spring):31–47. Rumelhart, David, and Anthony Ortony. 1977. The Representation of Knowledge in Memory. In Richard C. Anderson, R. J. Spiro, and W. E. Montague (Eds.), Schooling and the Acquisition of Knowledge. Hillsdale, NJ: Erlbaum. Russell, James A. 1980. A Circumplex Model of Affect. Journal of Personality and Social Psychology 39:1161–1178. Russell, James A., and Lisa Feldmen Barrett. 1999. Core Affect, Prototypical Emotional Episodes, and Other Things Called Emotion: Dissecting the Elephant. Journal of Personality and Social Psychology 76(5):805–819. Russell, James A., and James M. Carroll. 1999a. On the Bipolarity of Positive and Negative Affect. Psychological Bulletin 125(1):3–30. ———. 1999b. The Phoenix of Bipolarity: Reply to Watson and Tellegen. Psychological Bulletin 125(5):611–617. Sabato, Larry J. (Ed.) 1997. Toward the Millennium: The Elections of 1996. Boston, MA: Allyn and Bacon. Sanford, Anthony. 1986. The Mind of Man: Models of Human Understanding. New Haven: Yale University Press. Schank, R., and R. Abelson. 1977. Scripts, Plans, Goals, and Understanding: An Inquiry into Human Knowledge Structures. Hillsdale, NJ: Lawrence Erlbaum Associates. Scherer, Klaus R. 1984. On the Nature and Function of Emotion: A Component Process Approach. In K. R. Scherer, and P. Ekman (Eds.), Approaches to Emotion. Hillsdale: Erlbaum. Schiffenbauer, A. I. 1974. Effect of Observer’s Emotional State on Judgments of the Emotional State of Others. Journal of Personality and Social Psychology 30(1):31–35. Schwarz, N. 1990. Feelings as Information: Informational and Motivational Functions of Affective States. In R. M. Sorrentino, and E. T. Higgins (Eds.), Handbook of Motivation and Cognition: Foundations of Social Behavior, Vol. 2. New York: Guilford Press. ———. 2002. Situation Cognition and the Wisdom of Feelings: Cognitive Tuning. In L. Feldman-Barrett, and P. Saolvey (Eds.), Wisdom in Feeling: Psychological Processes in Emotional Intelligence. New York: Guilford Press.
REFERENCES
253
Schwarz, N., H. Bless, and G. Bohner. 1991. Mood and Persuasion: Affective States Influence the Processing of Persuasive Communications. Advances in Experimental Social Psychology 24:161–195. Schwarz, N., and G. L. Clore. 1983. Mood, Misattribution, and Judgments of Well-Being: Informative and Directive Functions of Affective States. Journal of Personality and Social Psychology 45:513–523. ———. 1988. How do I feel about it? Informative Functions of Affective States. In K. Fiedler and J. Forgas (Eds.), Affect, Cognition, and Social Behavior. Toronto, Canada: Hofgrefe International. ———. 1996. Feelings and Phenomenal Experiences. In E. T. Higgins, and A. W. Kruglanski (Eds.), Social Psychology Handbook of Principles. New York: Guilford Press. Schwarz, N., F. Strack, D. Kommer, and D. Wagner. 1987. Soccer, Rooms and the Quality of Your Life: Mood Effect on Judgments of Satisfaction with Life in General and with Specific Life-Domains. European Journal of Social Psychology 17:69–79. Sedikides, C. 1994. Incongruent Effects of Sad Mood on Self-Conception Valence: It’s a Matter of Time. European Journal of Social Psychology 24:161–172. Shaver, Phillip, Judith Schwartz, Donald Kirson, and Cary O’Connor. 1987. Emotion Knowledge: Further Exploration of a Prototype Approach. Journal of Personality and Social Psychology 52(6):1061–1086. Shepsle, K. A. 1972. The Strategy of Ambiguity: Uncertainty and Electoral Competition. American Political Science Review 66:555–568. Sherman, D. K., and H. S. Kim. 2002. Affective Perseverance: The Resistance of Affect to Cognitive Invalidation. Personality and Social Psychology Bulletin 28:224–237. Shiffrin, R. M., and W. Schneider. 1977. Controlled and Automatic Information Processing: Perceptual Learning, Automatic Attending, and a General Theory. Psychological Review 84:127–190. Sigal, Leon V. 1973. Reporters and Officials: The Organization and Politics of Newsmaking. Lexington, MA: D. C. Heath and Company. Sigelman, Lee, Carol K. Sigelman, and Christopher Fowler. 1987. A Bird of a Different Feather? An Experimental Investigation of Physical Attractiveness and the Electability of Female Candidates. Social Psychology Quarterly 50(1):132–143. Simon, Herbert. 1969. Sciences of the Artificial. Cambridge, MA: MIT Press. Sinclair, R. C., M. M. Mark, and G. L. Clore. 1994. Mood-Related Persuasion Depends on (Mis)attributions. Social Cognition 12:309–326. Singer, J. A., and P. Salovey. 1988. Mood and Memory: Evaluating the Network Theory of Affect. Clinical Psychology Review 8:211–251. Skaperdas, Stergios, and Bernard Grofman. 1995. Modeling Negative Campaigning. American Political Science Review 89(1):49–61. Smith, Craig A., and Phoebe C. Ellsworth. 1985. Patterns of Cognitive Appraisal in Emotion. Journal of Personality and Social Psychology 48:813–838. Smith, Craig A., and Richard S. Lazarus. 1993. Appraisal Components, Core Relational Themes, and the Emotions. Cognition and Emotion 7(3–4): 233–269.
254
REFERENCES
Smith, E. R., R. H. Fazio, and M. A. Cejka. 1996. Accessible Attitudes Influence Categorization of Multiple Categorizable Objects. Journal of Personality and Social Psychology 71(5):888–898. Smith, E. R., and J. R. Kluegel. 1982. Cognitive and Social Bases of Emotional Experience: Outcome, Attribution, and Affect. Journal of Personality and Social Psychology 43(6):1129–1141. Smith-Lovin, Lynn. 1995. The Sociology of Affect and Emotion. In Karen S. Cook, Gary A. Fine, and James S. House (Eds.), Sociological Perspectives on Social Psychology. Needham Heights, MA: Allyn and Bacon. Snyder, Richard C., H. W. Bruck, and Burton Sapin. 2002. Decision-Making as an Approach to the Study of International Politics. In Richard C. Snyder, H. W. Bruck, and Burton Sapin (Eds.), Foreign Policy Decision Making. New York: Palgrave Macmillan. Sobel, Richard. 2001. The Impact of Public Opinion on U.S. Foreign Policy. Oxford: Oxford University Press. Soldat, A. S., and R. C. Sinclair. 2001. Colors, Smiles, and Frowns: External Affective Cues can Directly Affect Responses to Persuasive Communications in a Mood-Like Manner without Affecting Mood. Social Cognition 19:469–490. Soldat, A. S., R. C. Sinclair, and M. M. Mark. 1997. Color as an Environmental Processing Cue: External Affective Cues can Directly Affect Processing Strategy without Affecting Mood. Social Cognition 15:55–71. Srull, T. K. 1981. Person Memory: Some Tests of Associative Storage and Retrieval Models. Journal of Experimental Psychology: Human Learning and Memory 7:440–463. Srull, T. K., and R. S. Wyer, Jr. 1989. Person Memory and Judgement. Psychological Review 96:58–83. Stanley, Harold W., and Richard Niemi. 1991. Partisanship and Group Support, 1952–1988. American Politics Quarterly 19:189–210. Steenbergen, M. R. 2001. The Reverend Bayes Meets John Q. Public: Patterns of Political Belief Updating in Citizens. Paper presented at the annual meeting of the International Society of Political Psychology, Cuernavaca, Mexico. Steenbergen, M. R., and M. Lodge. 1998. Process Matters: Cognitive Models of Candidate Evaluation. Paper presented at the annual meeting of the American Political Science Association, Boston. Sylvan, Donald A., and James F. Voss. 1998. Problem Representation in Foreign Policy Decision Making. Cambridge: Cambridge University Press. Taber, Charles. 2003. Information Processing and Public Opinion. In David O. Sears, Leonie Huddy, and Robert Jervis (Eds.), Handbook of Political Psychology. London: Cambridge University Press. Taber, Charles S., Jill Glather, and Milton Lodge. 2001. The Motivated Construction of Political Judgements. In James Kuklinski (Ed.), Citizens and Politics: Perspectives from Political Psychology. New York: Cambridge University Press. Tangney, June P., Patricia Wagner, Carey Fletcher, and Richard Gramzow. 2001. Shamed into Anger? The Relation of Shame and Guilt to Anger and SelfReported Aggression. In W. G. Parrott (Ed.), Emotions in Social Psychology: Essential Readings. Philadelphia: Psychology Press.
REFERENCES
255
Teasdale, J. D., and M. L. Russell. 1983. Differential Effect of Induced Mood on the Recall of Positive, Negative, and Neutral Words. British Journal of Clinical Psychology 22:163–171. Tellegen, Auke. 1985. Structures of Mood and Personality and their Relevance to Assessing Anxiety, with an Emphasis on Self-Report. In Hussain Tuma and J. D. Maser (Eds.), Anxiety and the Anxiety Disorders. Hillsdale, NJ: Erlbaum. Tellegen, Auke, David Watson, and Lee Anna Clark. 1999a. Further Support for a Hierarchical Model of Affect. Psychological Science 10(4):307–309. ———. 1999b. On the Dimensional and Hierarchical Structure of Affect. Psychological Science 10(4):297–303. Thomas, David L., and Ed Diender. 1990. Memory Accuracy in Recall of Emotions. Journal of Personality and Social Psychology 59(2):291–297. Thorson, Esther, William G. Christ, and Clarke Caywood. 1991a. Selling Candidates Like Tubes of Toothpaste: Is the Comparison Apt? In F. Biocca (Ed.), Television and political advertising. Hillsdale, NJ: Erlbaum. ———. 1991b. Effects of Issue-Image Strategies, Attack and Support Appeals, Music and Visual Content in Political Commercials. Journal of Broadcasting and Electronic Media 35(4):465–486. Tiedens, L. Z., and S. Linton. 2001. Judgment Under Emotional Certainty and Uncertainty: The Effects of Specific Emotions on Information Processing. Journal of Personality and Social Psychology 81:973–988. Tomkins, Silvan S. 1962. Affect, Imagery, Consciousness. New York: Springer. Tversky, A., and D. Kahneman. 1974. Judgment Under Uncertainty: Heuristics and Biases. Science 185:1124–1131. Valentino, Nicholas A., Vincent L. Hutchings, and Ismail K. White. 2002. Cues That Matter: How Political Ads Prime Racial Attitudes During Campaigns. American Political Science Review 96:75–90. Walker, Stephen G. 1977. The Interface Between Beliefs and Behavior: Henry Kissinger’s Operational Code and the Vietnam War. Journal of Conflict Resolution 21:129–168. ———. 1983. The Motivational Foundations of Political Belief Systems: A Re-Analysis of the Operational Code Construct. International Studies Quarterly 27:179–201. Watkins, T., A. M. Mathews, D. A. Williamson, and R. Fuller. 1992. Mood Congruent Memory in Depression: Emotional Priming or Elaboration. Journal of Abnormal Psychology 101:581–586. Watson, David. 1988a. Intraindividual and Interindividual Analyses of Positive and Negative Affect: Their Relation to Health Complaints, Perceived Stress, and Daily Activities. Journal of Personality and Social Psychology 54(6):1020–1030. ———. 1988b. The Vicissitudes of Mood Measurement: Effects of Varying Descriptors, Time Frames, and Response Formats on Measures of Positive and Negative Affect. Journal of Personality and Social Psychology 55(1):128–141.
256
REFERENCES
Watson, David, and Lee Anna Clark. 1991. Self-Versus Peer Ratings of Specific Emotional Traits: Evidence of Convergent and Discriminant Validity. Journal of Personality and Social Psychology 60(6):927–940. ———. 1994. The PANAS-X: Manual for the Positive and Negative Affect Schedule-Expanded Form: University of Iowa. ———. 1997. Measurement and Mismeasurement of Mood: Recurrent and Emergent Issues. Journal of Personality Assessment 68:267–296. Watson, David, Lee Anna Clark, and Auke Tellegen. 1988. Development and Validation of a Brief Measure of Positive and Negative Affect. Journal of Personality and Social Psychology 51:1063–1070. Watson, David, and Auke Tellegen. 1985. Toward a Consensual Structure of Mood. Psychological Bulletin 98(2):219–235. ———. 1999. Issues in the Dimensional Structure of Affect: Effects of Descriptors, Measurement Error, and Response Formats: Comment on Russell and Carroll. 1999. Psychological Bulletin 125(5):601–610. Watson, David, David Wiese, Jatin Vaidya, and Auke Tellegen. 1999. The Two General Activation Systems of Affect: Structural Findings, Evolutionary Considerations, and Psychobiological Evidence. Journal of Personality and Social Psychology 76(5):820–838. Wattenberg, Martin P., and Craig Leonard Brians. 1999. Negative Campaign Advertising: Demobilizer or Mobilizer? American Political Science Review 93(4):891–899. Weary, Gifford, and Jill A. Jacobson. 1997. Causal Uncertainty Beliefs and Diagnostic Information. Seeking. Journal of Personality and Social Psychology 73:839–848. Weber, Christopher, Milton Lodge, and Charles Taber. 2005. Subliminal Priming and Political Campaigns: The Impact of Subliminally Presented Affective Primes on Campaign Ad Evaluations. Paper presented at the Annual Meeting of the American Political Science Association, Washington, DC. Wegener, D. T., R. E. Petty, and D. J. Klein. 1994. Effects of Mood on High Elaboration Attitude Change: The Mediating Role of Likelihood Judgments. European Journal of Social Psychology 24:25–44. Wegener, D. T., R. E. Petty, and S. M. Smith. 1995. Positive Mood Can Increase or Decrease Message Scrutiny: The Hedonic Contingency View of Mood and Message Processing. Journal of Personality and Social Psychology 69:5–15. Weiner, Bernard. 1982. The Emotional Consequences of Causal Attributions. In M. S. Clark, and S. T. Fiske (Eds.), Affect and Cognition: The Seventeenth Annual Carnegie Symposium on Cognition. Hillsdale: Erlbaum. West, Darrell M. 1997. Air Wars. 2nd edition. Washington, DC: Congressional Quarterly. ———. 2005. Air Wars. 4th edition. Washington, DC: Congressional Quarterly. Westholm, Anders. 1997. Distance versus Direction: The Illusory Defeat of the Proximity Theory of Electoral Choice. American Political Science Review 91(4):865–884. Wilson, Timothy D. 2002. Strangers to Ourselves: Discovering the Adaptive Unconscious. Cambridge, MA: Belknap Press of Harvard University Press.
REFERENCES
257
Wilson, Timothy D., and Jonathan W. Schooler. 1991. Thinking Too Much: Introspection Can Reduce the Quality of Preferences and Decisions. Journal of Personality and Social Psychology 60(2):181–192. Worth, L. T., and D. M. Mackie. 1987. Cognitive Mediation of Positive Affect in Persuasion. Social Cognition 5:76–94. Wyer, R. S., Jr., and T. K. Srull. 1980. The Processing of Social Stimulus Information: A Conceptual Integration. In R. Hastie, T. M. Ostrom, E. B. Ebbesen, R. S. Wyer, Jr., D. L. Hamilton, and D. E. Carlston (Eds.), Person Memory: The Cognitive Basis of Social Perception. Hillsdale, NJ: Erlbaum. ———. 1986. Human Cognition in its Social Context. Psychological Review 93:322–359. ———. 1989. Memory and Cognition in its Social Context. Hillsdale, NJ: Erlbaum. Wyer, R. S., Jr., G. V. Bodenhausen, and T. K. Srull. 1984. The Cognitive Representation of Persons and Groups and Its Effect on Recall and Recognition Memory. Journal of Experimental Social Psychology 20:445–469. Wyer, R. S., Jr., G. L. Clore, and L. M. Isbell. 1999. Affect and Information Processing. In M. P. Zanna (Ed.), Advances in Experimental Social Psychology, Vol. 31. San Diego, CA: Academic Press. Wyer, R. S., Jr., and L. L. Martin. 1986. Person Memory: The Role of Traits, Group Stereotypes, and Specific Behaviors in the Cognitive Representation of Persons. Journal of Personality and Social Psychology 50:661–675. Zajonc, Robert B. 1980. Feeling and Thinking: Preferences Need no Inferences. American Psychologist 35:151–175. ———. 1984. On the Primacy of Affect. American Psychologist 39:117–123. ———. 2000. Feeling and Thinking: Closing the Debate Over the Independence of Affect. In Joseph P. Forgas (Ed.), Feeling and Thinking: The Role of Affect in Social Cognition. Cambridge: Cambridge University Press. Zaller, John. 1992. The Nature and Origins of Mass Opinion. Cambridge: Cambridge University Press. Zaller, John, and Stanley Feldman. 1992. A Simple Theory of the Survey Response. American Journal of Political Science 36:579–616. Zelenski, J. M., and R. J. Larsen. 2002. Predicting the Future: How Affect-Related Personality Traits Influence Likelihood Judgments of Future Events. Personality and Social Psychology Bulletin 28:1000–1010. Zellner, Arnold. 1962. An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias. Journal of the American Statistical Association 57(298):348–368. Zevon, M. A., and A. Tellegen. 1982. Experimental Methods in Political Science. Annual Review of Political Science 5:31–64.
This page intentionally left blank
INDEX
AAI model see affect-as-information model Abelson, Robert P. 2, 34, 59, 125 ACT see affect control theory actions reparative 48, 52, 53 social 48, 55 activation 14, 81, 165, 182 automatic 15, 25 processes 165, 166 spreading 18–20 Adolphs, Ralph 10, 72 advertising attack (negative) 8, 53, 138–40, 144–6, 153, 155, 156 positive 139, 143, 147, 148, 151, 153, 156 affect infusion 62 primacy of 13 affect control theory 7, 48, 49, 51–6 affect-as-information model 62, 67, 71, 74–6, 78, 82, 83 affective cues 7, 58, 71, 72, 76, 78–81, 83, 85 experiences 74, 75, 86 feelings 71, 72, 78, 81, 104, 105 information 78, 81 intelligence 3, 4, 6, 7, 10, 36, 111, 166, 167, 180 priming 22, 28 reactions 3, 59, 68, 110, 133 responses 6, 22, 24, 26–9, 118, 122
states 7, 58, 59, 61, 62, 64, 65, 74, 78, 85 afraid 36–9, 109, 117, 118, 122, 142, 170, 175, 180 see also fear aggression, displaced 66, 69 anger 7, 8, 22, 35–9, 41, 57, 58, 64–6, 68–70, 74, 75, 81, 84, 85, 109–14, 117, 118, 209–11, 214, 224, 225 anxiety 35–9, 38, 41, 42, 68, 76, 109, 111, 117, 122, 124, 127, 167 aversion and 123 campaign-induced 167 effects of 66, 67 experience of 110, 113, 132 inducing 117 information-seeking hypothesis 150 items 119, 122, 123 measures of 41, 45, 118 operationalizing 117 appearance, physical 187, 189, 192 assimilation effects 58, 59, 62, 63, 65, 67, 68, 85 mood-induced 59 attention 4, 5, 10, 14, 18, 29, 31, 33, 53, 70, 111, 112, 142, 150, 153, 154, 214, 215, 223, 224 attentiveness 141, 168, 169 political 167, 169, 179, 181 attitude object 12, 20, 21, 24 attitudes 2, 4, 10, 12, 13, 16–18, 21, 26, 28, 30, 58, 59, 64, 75, 78, 185, 211–13 base 69
260
INDEX
attitudes––continued global 58, 59 positive 58 priming 20–2, 24 prior 27, 30 attractiveness 9, 73, 189, 192, 196, 201, 204, 206 attributions 113, 215 automatic processes 12, 14, 15 responses 15, 16, 20 automaticity 13–16, 20–2, 24 aversion 7, 8, 36–41, 43, 45, 109–28, 131–3 dimensions 111, 117, 119, 122 experience of 116, 125, 132 items 38, 39, 119, 123 measures of 41, 42, 45, 118 mobilization of 127, 128 predicting 121 Banaji, M. R. 70 Banducci, Susan 189 Bargh, John A. 6, 14, 21 Bartels, Larry 189 behavioral dispositions 11, 214 beliefs 2, 4, 7, 11, 12, 15–19, 22, 27–30, 112–14, 116, 125, 211, 213 automaticity of 20 core 110, 113, 126, 127 political 13, 17, 27, 29 Belt, Todd 8 Berntson, Gary G. 33, 125 biases 4, 5, 27, 29, 69, 72, 73, 127, 204 anger-induced 69 bitter 36–9, 42, 64, 109, 117, 122 Bless, H. 72 Bodenhausen, G. V. 35, 74 Bohner, G. 76, 77 Bower, G. H. 60 Burns, K. C. 7, 82–4 Butler, J. L. 67, 79 Cacioppo, J. T. 12, 33, 168 campaign anxiety-laden messages 149
anxiety-producing messages 157 attention 155 communications 142, 149, 150, 155, 157, 165–7, 169 coverage 179, 180 emotional commitment 8 environment 88, 102, 190, 191 events 144 information 8, 150 learning 166–70, 175, 176, 179, 181, 182 negative messages 137, 138 negativity 153, 154 news 138, 140, 150, 154, 181 strategies 173, 181, 183 candidate appearance 189, 190 assessment 137 attacks 139, 143, 148, 153 attractive 188, 189, 199, 202, 205 debates 154, 156 Democrat 168, 169, 182 difference, perceptions of 176, 178, 186 disliked 90, 94–6, 98, 105, 106 evaluations 8, 87, 88, 90, 104, 115, 125, 143, 166, 167, 182 ideology 201 image 206 in-party 92, 100, 196, 197, 201, 207 likeability 187, 189, 194, 201, 208 liked 8, 89, 90, 94–6, 98, 106 memories 98 out-party 9, 106, 196–9 personality 92 Republican 169, 182, 206, 207 traits 110, 126, 132 unlikeable 192, 195, 197, 199 category membership 69, 70, 72 Chaiken, S. 12 choice, making a 6, 17, 29, 102, 115, 205, 207, 209 Clore, G. L. 62, 67, 79, 80, 110 cognition 2–6, 11, 24, 59, 206, 214, 215, 225
INDEX
cognitive architecture 17 associations 14, 15, 24, 27, 29 calculus model, in foreign policy 212, 213 Colcombe, S. 79, 80 communications, political 155, 168, 169 compassion 209 conflict 167, 168, 182, 183, 210 congruent information 4, 90, 91, 95, 96, 98–104 memories 91, 101, 104 conscious awareness 6, 24, 28, 29 considerations, conscious 11–13, 17 contempt 36–9, 42, 117, 122 contrast effects 58, 63, 64, 168 correct voting 203, 204 Crigler, Ann 8 cues 58, 70, 79, 80, 166, 167, 172, 180, 188, 211 contextual 59, 62, 65 peripheral 69, 70, 75 threat 176 cynicism 138, 140, 142, 143, 148, 149, 151, 152, 156 Damasio, Antonio 3 deflection, in affect control theory 51–3 deliberative processes 13–15, 27, 30 Devine, P. G. 77 disgust 7, 21, 22, 37, 39, 109 displaced aggression hypothesis 66 Downs, Anthony 1 dual-process accessibility model 6, 12, 13 Edwards, John A. 76, 214 emotional appeals 8, 140, 149 appraisals 35, 41 arousal 137, 151, 157, 211 attachment 153, 154 commitments 141 experience 33–5, 41, 64
261
impact 151 rancor 210 reactions, distinct 114 response items 39 responses 4, 5, 7, 8, 31, 33, 35, 36 38, 41, 111, 137, 139, 168, 212, 217: best measure of 31; campaigns provoking 182; concurrent 35; dimensions of 31, 33–5, 38, 39; distinct 38, 137; measurement of 2, 31, 32, 42, 43 self-report 43, 44 states 33, 137, 181, 214, 223 system 10, 114, 215, 225 emotion-eliciting process 59 emotion-laden messages 140 emotion-producing stimuli 167 emotions see also specific emotion terms arousing 141 circumplex model of 32, 33, 35, 110, 111, 117, 125, 137 common 113, 126 definitions of 37, 225 discrete 110 distinct 8, 110, 111, 115, 116 effects of 59, 150, 209–11, 212, 222, 223 intense 215, 217, 219, 220, 222, 223 interplay of 2, 212 items 36, 41, 43, 184 low intensity 217, 220, 222, 223 measuring 3, 7, 31, 32, 35, 39, 41, 43 mobilizing 111 negative 7, 8, 10, 38, 39, 58, 85, 110, 112, 115, 126, 142, 150, 182, 210, 211, 214, 215, 224, 225 positive 58, 110, 142, 150, 154, 187 potency 49, 50, 111, 113 role of 1–3, 5, 7, 10, 43, 137, 212 structure of 32, 35, 38, 110, 117, 137
262
INDEX
emotions––continued three-dimensional model of 8, 110, 111 two-dimensional model of 109, 110 valence model of 33, 35–7 emotive states 219, 222, 223 encoding 60, 65, 66 selective 59, 60, 66, 68 enthusiasm 3, 35–7, 39, 41, 43, 84, 110, 111, 137, 147, 167, 169, 170, 175, 179 Entman, Robert M. 168 expectancy-disconfirming information 82 expertise 34, 70, 98, 197–9, 201, 202 experts, political 9, 63, 64, 69, 197, 199, 201, 203, 205 exposure to stimuli 2, 12, 17, 22, 24, 29, 66, 136, 140, 142, 144, 148, 154–6, 178, 181 Facial expression 79, 80 facilitation effects 25, 26, 30 Fazio, R. H. 6, 20 fear 7, 8, 57, 58, 64, 66, 74–7, 81, 84, 85, 111–14, 140, 152, 154, 157, 175, 176, 180, 209–11 see also afraid fear-arousing messages 137, 155 feeling rules 47, 48 states, discrete emotional 43 field experiment 8, 148, 162 foreign policy behavior 210, 212, 214 decisions 9, 209, 211, 215, 225 Franklin, Charles 183 functional emotion theories 110, 111 fundamental sentiments, in affect control theory 50–2 Gasper, K. 67 Gehm, Theodor 132 gender 168, 185, 197, 199, 201, 202
George, Alexander 211 Geva, Nehamia 9 goals 12, 15, 17, 20, 27, 30, 71, 112, 113 Green, Donald P. 4 groups 22–4, 26, 112, 168, 173, 180, 183, 188, 207 guilt 57, 77, 110 Haar, T. 82 happiness 70, 72–4, 78, 79 happy individuals 65, 70–3 moods 62, 63, 70–3, 81, 83 participants 71–4, 82–4 Hastie, R. 90 hate 9, 37, 38, 42, 66, 109, 209–11, 214, 215, 217, 219, 222, 224, 225 Heise, David R. 49–51, 53, 56, 111, 113 helplessness 112, 114 Hermann, Charles F. 212 Herrmann, Richard K. 211 heuristics 69, 70, 75, 84 hope 37, 39, 41, 174, 185 hot cognition 4, 6, 22, 25, 26, 89, 211 Hudson, Valerie M. 212 Hutchings, Vincent 8, 9 identity 18, 47–50, 52, 55, 115, 116 implicit processes 16, 29 impressions 16, 52, 58, 64, 69, 79, 81, 82, 84, 88, 90, 170, 189, 208, 218, 220 transient 50, 51 incongruent information 4, 7, 8, 21, 89–92, 94–104, 106 information board, dynamic 9, 92 information overload 83 information processing dampened 223 heuristic 76 human 18
INDEX
political 10, 15, 17, 58, 77, 80, 82, 86 requirements 102 social 66 strategies 161: heuristic 69; suboptimal 214; systematic 69 style 58, 69–71, 74–81, 84 information-seeking behavior 82, 83, 143, 149–54, 157 inhibition effects 22, 25, 26, 30 intensity 26, 41, 44, 45, 110, 170, 223–5 intentions 11–18, 20, 27–30, 148, 206, 215 behavioral 2, 11, 12 international relations, emotions in 209–13, 215, 225 Isbell, L. M. 7, 53, 63, 79, 80, 82, 84 Isen, A. M. 60 James, L. 83, 84 journalists 135, 156 judgments 6, 15, 17, 29, 53, 57–60, 62, 63, 65, 67, 69–73, 76, 80, 81, 85, 86, 219, 220 Kahn, Kim Fridkin 138 Kenney, Patrick J. 138 knowledge structures 72–4 Kramer, G. P. 35 Lambert, A. J. 73, 83 Lau, Richard R. 2, 9, 139 leadership 119–21, 123, 126, 132 learning, political 137, 168 Lerner, J. S. 68 likeability 187, 189, 190, 192, 195–7, 199, 203–5 Linton, S. 75 loathing 36, 109, 110 Lodge, Milton 4–7, 30, 87–90, 102, 103, 212 Mackie, D. M. 71 Mackuen, W. Russell
84, 167
263
Marcus, George E. 4, 6, 7, 76, 84, 111, 114, 137, 167, 212, 225 Mathews, A. M. 67 McGraw, Kathleen 87 memory 7, 8, 10, 15, 18, 25–8, 59, 60, 71, 79, 87–91, 94, 96, 98–107 effects 88, 89, 102, 104 long term (LTM) 12, 13, 17–19, 21, 27 nodes 18, 19 online processing and 101 processing 88, 90, 92, 93, 103, 104 recall 10, 45, 65, 70, 73, 90, 91, 94, 98, 106, 214, 223 role of 7, 102 working 18–20 Miller, G. A. 18 Monteith, M. J. 77 mood 32, 33, 36, 37, 44, 60–3, 65, 67, 68, 71, 72, 77, 80, 137 manipulated 59, 60 misattribution 59, 62 mood-congruent information 60 moral traditionalism 123, 125, 133 Mosher, Katrina N. 225 motivated reasoning, see reasoning motivation 2, 12, 15, 16, 28, 62, 63, 69, 71, 110 negative campaigning 8, 135–7, 141, 157 effects of 136, 156 negative events 61, 67, 113 negative feelings 50, 57, 141, 150, 154 negative information 60, 84, 96, 138, 220, 224 negativity 8, 13, 18, 39, 50, 61, 136, 138, 147, 155, 156 neutral information 94, 101 novel stimuli 111, 112, 114 novices, political 69, 95, 188, 199, 200, 203, 205
264
INDEX
online processing 8, 14, 87–90, 92, 102–4, 213 tally 20, 26, 87, 88, 94, 101, 102, 107 Ottati, V. C. 7, 59, 63, 79 Park, J. 70 partisanship 9, 45, 53, 69, 72, 73, 76, 83–5, 119, 123, 124, 127, 184, 185, 195–7, 199, 201, 207 party identification 87, 146, 149, 152, 155, 166, 191 Patterson, Thomas E. 138 perception, emotional effects on 209–11 personality 9, 92, 187–90, 196, 198, 201, 203–7 descriptions 195, 206 traits 9, 73, 187–92, 203, 205, 206 person-evaluation tasks 88 persuasion 62, 77, 80, 166, 167 Petty, R. E. 12, 168 Philpot, Tasha S. 8 policy-based information 208 political action 7, 48 attitudes 11, 22, 80, 211, 212 campaigns 34, 89, 154, 181, 190, 191 candidates 54, 58, 59, 63, 69, 79, 81, 89, 90, 105, 110, 121, 126, 135, 182, 187, 204 cynicism 148, 149, 155 events 54, 57, 61, 64, 75, 76, 80 figures 8, 57, 88, 115–17, 121, 126–8 identity 7, 48, 54 information 2, 7, 10, 63, 68, 76, 177, 178 issues 26, 84, 116, 200, 212 judgments 5, 7, 29, 58, 59, 63–5, 80, 85, 168, 181 leaders 22, 26, 37, 42, 58, 59, 126, 204
parties 2, 70, 76, 115, 169, 181, 191, 197 perceptions 166, 167, 183 sophistication 6, 26, 30, 95, 100, 106 stimuli 2, 5, 6, 34–6, 39, 166 positive affect-eliciting stimuli 58, 72, 73 predispositions 18, 127, 165, 166 presidential campaign 53, 80, 138, 141, 154, 166 candidates 37, 139, 142, 178, 181, 182, 188 priming 14, 18, 23, 19, 25–8, 61 process tracing, dynamic 190, 191 processing affect influencing 76, 80 deliberative 12, 14, 16, 28, 51 group-relevant information 77 incongruent information 90, 91, 96, 98, 103 issue-relevant information 78 memory-based 11 motivation 71, 72 online, see online processing systematic 68, 69, 71, 74–9 proud 39, 41 race
69, 83, 84, 119, 123, 168, 173, 174, 177, 183–6 rage 209 Rahn, Wendy M. 137 reaction time 22, 24 reasoning, motivated 3, 4, 6, 8, 10, 28, 89–92, 94, 98, 101–4, 127 Redlawsk, David P. 7, 9, 106 resentful 37, 38, 42 resources attentional 15, 16 cognitive 12, 16, 71 Riggle, E. D. B. 189, 190 risk 67, 68, 162, 174 Rosenberg, S. W. 189
INDEX
sad individuals 65, 70–5 82–4 moods 60, 62, 63, 70, 73, 75, 76, 81, 83 sadness 41, 57, 64, 65, 70–5, 78, 79, 82, 110, 225 salience 51, 76, 77 schema-inconsistent information 90 Scherer, K. R. 132 Schwarz, N. 62, 71, 72 Sheppard, L. A. 35 Sigelman, Lee 189 Sinclair, R. C. 78, 80 Skorick, J. Mark 9 social interaction 7, 10, 48, 49, 52, 210 issues 77, 78 judgment 66 life 47, 48, 53, 55 Soldat, A. S. 78, 80 sophisticates, political 23, 26, 95, 96, 115 Srull, T. K. 90 Steenbergen, Marco R. 4, 8, 89 stereotypes 72–7, 79, 82–4, 111, 214 stop-and-think process 89, 91, 96, 102, 103 Stroh, Patrick J. 87 sympathy 217 systematic processing 75 systems cognitive 28, 214, 215, 225 dispositional 36, 111, 114 surveillance 3, 36, 111–13, 167
265
Taber, Charles 4, 6, 7, 89, 90, 103, 212 Tellegen, Auke 44, 45 Thomas, David L. 106 threat 4, 68, 77, 112–14, 126, 132, 175, 210 cues 169, 175 Tiedens, L. Z. 75 traits, candidate 110, 115 trust 8, 54, 136, 140, 142, 153, 161 trustworthy 188, 219 uneasy 39, 109, 117, 118 utility theories, subjective 12 Valentino, Nicholas A. 8 vote choice 76, 81, 142, 166, 189, 195, 199, 201, 208 voter decisions 7, 89, 91, 93, 95, 99, 101, 103, 105, 107, 189 Watson, David 33, 41 Weinerth, T. 76, 77 well-being 8, 110, 113–15, 121, 126, 213 Wieland, R. 72 women 133, 170, 176, 183, 184, 201, 202, 205, 207 worry 37, 42, 109 Worth, L. T. 71 Wyer, R. S. 63, 79, 90 Zajonc, Robert B. 6, 22, 26, 30