Rumor

Psychology
Social and
Organizational
Approaches

Nicholas DiFonzo and Prashant Bordia

American Psychological Association • Washington, DC

Copyright © 2007 by the American Psychological Association. All rights reserved. Except as
permitted under the United States Copyright Act of 1976, no part of this publication may be
reproduced or distributed in any form or by any means, including, but not limited to, the process
of scanning and digitization, or stored in a database or retrieval system, without the prior written
permission of the publisher.

Published by
American Psychological Association
750 First Street, NE
Washington, DC 20002
www.apa.org

To order
APA Order Department
P.O. Box 92984
Washington, DC 20090-2984
Tel: (800) 374-2721; Direct: (202) 336-5510
Fax: (202) 336-5502; TDD/TTY: (202) 336-6123
Online: www.apa.org/books/
E-mail: [email protected]

In the U.K., Europe, Africa, and the Middle East, copies may be ordered from
American Psychological Association
3 Henrietta Street
Covent Garden, London
WC2E 8LU England

Typeset in Meridien by World Composition Services, Inc., Sterling, VA

Printer: Hamilton Printing, Castleton, NY

Cover Designer: Naylor Design, Washington, DC
Technical/Production Editor: Tiffany L. Klaff
The opinions and statements published are the responsibility of the authors, and such opinions
and statements do not necessarily represent the policies of the American Psychological
Association.

Library of Congress Cataloging-in-Publication Data

DiFonzo, Nicholas.
Rumor psychology : social and organizational approaches / Nicholas DiFonzo and Prashant
Bordia.—1st ed.
p. cm.
Includes bibliographical references and index.
ISBN-13: 978-1-59147-426-5
ISBN-10: 1-59147-426-4
1. Rumor. 2. Social psychology. 3. Organizational behavior. I. Bordia, Prashant. II. Title.
HM1241.D54 2007
302.2'4—dc22 2006009552

British Library Cataloguing-in-Publication Data

A CIP record is available from the British Library.

Printed in the United States of America

First Edition
Dedicated to Mary Josephine DiFonzo and Manjula Bordia.
Contents

ACKNOWLEDGMENTS IX
INTRODUCTION 3

1 Defining Rumor 11
2 Forms, Frequency, and Fallout of Rumors 35
3 Psychological Factors in Rumor Spread 69
4 Factors Associated With Belief in Rumor 89
5 Rumor as Sense Making 113
6 Rumor Accuracy: Patterns of Content Change,
Conceptualization, and Overall Accuracy 133
7 Mechanisms Facilitating Rumor Accuracy
and Inaccuracy 159
8 Trust and Organizational Rumor Transmission 185
9 Rumor Management 205
10 Summary, Model, and Research Agenda 229

REFERENCES 261
AUTHOR INDEX 277
SUBJECT INDEX 283
ABOUT THE A U T H O R S 291

VII
Acknowledgments

everal people have been especially helpful in the production

of this volume. We are particularly indebted to Ralph L.
Rosnow, our ever-helpful colleague and mentor at Temple
University, a true pioneer in the study of rumor transmis-
sion. We also thank Charles Walker of St. Bonaventure Uni-
versity for his perspicacious advice on chapter 1. We thank
the anonymous corporation, including workers, question-
naire administrators, and management, who participated in
the longitudinal survey described in chapters 2 and 8. Credit
with regard to this project is particularly due to Rob Winter-
korn for his determined—and successful—efforts in data
collection. We thank H. Taylor Buckner and Frederick
Koenig for their comments on chapters 6 and 7. We appreci-
ate comments on a draft of chapter 8 made by Kurt Dirks,
Chip Heath, and Ralph L. Rosnow. We thank Eric K. Foster,
Holly Horn, Frederick Koenig, Mark Pezzo, Charles Walker,
Sarah Wert, and John Yost for consistently stimulating our
thinking about rumor and gossip via the electronic discus-
sion group, [email protected]. The
first author (DiFonzo) acknowledges funding to assist in the
preparation of this volume from two sources at Rochester
Institute of Technology: the College of Liberal Arts Faculty
Education and Development Fund and the Provost's Faculty
Leave Fund. Research reported in chapter 9 was funded
by the Australian Research Council and the University of
Queensland Foundation. We are grateful to collaborators
past and present. We thank Simon Lloyd Restubog and

IX
x ACKNOWLEDGMENTS

Bernd Irmer for help at various stages in the preparation of this manu-
script. We thank Emily Leonard, development editor, and Tiffany Klaff,
production editor, in the Books department at the American Psycholog-
ical Association, and two anonymous peer reviewers for their helpful
comments on a draft of this volume.
Rumor
Psychology
 Introduction

We are swimming in rumors.

—Manager at a small company facing severe downsizing
Tropical Fantasy Fruit Punch contains a substance that
causes black men to become sterile.
—False rumor circulating in New York City that
caused sales to plummet 70% (Freedman, 1991)
Israel warned 4000 Jews not to report for work at the
World Trade Center on September llth, 2001.
—False rumor circulating among anti-Zionist groups
(Hari, 2002; U.S. Department of State, 2005)

umors are an enduring feature of social and organizational

R landscapes. They attract attention, evoke emotion, incite

involvement, affect attitudes and actions—and they are
ubiquitous. A small example includes the groundless rumors
that McDonald's uses worm meat in its hamburgers
grounded sales in Atlanta (Goggins, 1979). Sober reports
that Paul McCartney was dead were discussed with sadness
and snowballed, even after a photo and interview with a
very much alive McCartney was published in LIFE magazine
(Rosnow, 1991). Office scuttlebutt often eats away at trust—
and feeds on distrust—among organizational members
(DiFonzo, Bordia, & Rosnow, 1994). False rumors that a
Haitian coup leader was to be set free spurred angry riots
that killed 10 people ("10 Die in Haiti," 1991). Seven million
people heard the incorrect claim that Coca-Cola contains
4 RUMOR PSYCHOLOGY

carcinogens (Kapferer, 1989). Two bizarre and fallacious rumors, wide-

spread in Africa, were that the AIDS virus was developed in a western
laboratory, and that a World Health Organization team inoculated
100,000 Africans with an untested vaccine that caused the continent's
pandemic of AIDS (Lynch, 1989). Harmful or potentially harmful ru-
mors reach the ears of top corporate public relations personnel nearly
once per week on average (DiFonzo & Bordia, 2000). E-mailed
computer-related hoaxes, such as the "Good Times" virus that will
rewrite one's hard drive and the "teddy bear" icon that destroys your
whole system, regularly alarm novice Internet users (Bordia & DiFonzo,
2004; "JDBGMGR.EXE," 2002). The catalog continues in abundance;
rumors flourish, fascinate, and frustrate.
It is not surprising then that the record of scholarly interest in the
psychology of rumor is long and illustrious; for over 7 decades social and
organizational researchers in psychology and sociology have researched
rumor. Some brief highlights we note include the early and substantial
work of Jamuna Prasad (1935) who studied rumors circulating after a
cataclysmic Indian earthquake. Interest in the subject of rumor psychol-
ogy peaked during World War II and rumor researchers included such
well-known social psychologists as Floyd H. Allport, Kurt Back, Dorwin
Cartwright, Leon Festinger, Stanley Schachter, and John Thibaut (e.g.,
F. H. Allport & Lepkin, 1945; Back et al, 1950; Festinger et al, 1948;
Schachter & Burdick, 1955). The standard work during this period was
G. W. Allport and Leo J. Postman's The Psychology of Rumor published
in 1947. The eminent Tamotsu Shibutani published the landmark socio-
logical treatise Improvised News: A Sociological Study of Rumor in 1966.
Ralph L. Rosnow and his associates refined the conceptual understand-
ing of rumor and systematically investigated the dynamics of rumor
transmission in the latter decades of the 20th century (e.g., Jaeger,
Anthony, & Rosnow, 1980; Rosnow, 1974, 1980, 1988, 1991; Rosnow,
Esposito, & Gibney, 1988; Rosnow & Fine, 1976; Rosnow & Georgoudi,
1985; Rosnow, Yost, & Esposito, 1986). Other social and organizational
psychologists and sociologists contributed significantly to the body of
knowledge regarding rumor as well during this period (e.g., K. Davis,
1972; Fine, 1992; Kapferer, 1987/1990; Knopf, 1975; Koenig, 1985;
Morin, 1971; Pratkanis & Aronson, 1991; P. A. Turner, 1993; R. H.
Turner & Killian, 1972). And within the past decade, social and organi-
zational psychologists have paid increased attention to this topic (e.g.,
R. S. Baron, David, Brunsman, & Inman, 1997; Bordia & DiFonzo,
2002, 2004, 2005; Bordia, DiFonzo, & Schulz, 2000; Bordia & Rosnow,
1998; DiFonzo & Bordia, 1997, 2002b, 2006, in press; DiFonzo, Bordia,
& Winterkorn, 2003; DiFonzo et al., 1994; Fine, Heath, & Campion-
Vincent, 2005; Fiske, 2004; Heath, Bell, &• Sternberg, 2001; Houmanfar
& Johnson, 2003; Kimmel, 2004a, 2004b; Michelson & Mouly, 2004;
 Introduction 5

Pendleton, 1998; Pratkanis & Aronson, 2001; Rosnow, 2001; Rosnow

& Foster, 2005). Indeed, an electronic discussion group of scholars
interested in both rumor and gossip research was formed in 2003
(Rumor- [email protected]).'
Therefore, some of what is known about rumor is quite old and
some of it is quite new; and so rumor theory is due for synthesis into
an updated whole. That is our first purpose in writing this book. We
have made a concerted effort to integrate and update findings from all
phases of rumor's long research history. Such a work is very much
needed; during a recent electronic discussion among social psycholo-
gists, the topic of post-9/11 rumors surfaced. The conversation was
perspicacious and insightful; however, it was informed mostly by G.
W. Allport and Postman's seminal—but dated—The Psychology of Rumor
(1947b). Another kind of integration is also needed; the existing body
of scholarly and practitioner knowledge about rumor tends to be iso-
lated from recent social-psychological theory. Rumor has become an
eddy among mainstream currents. It is no longer included in popular
social-psychological textbooks (although this is changing; e.g., Fiske,
2004, p. 517). This omission is unfortunate because rumor—an intrinsi-
cally fascinating topic in social psychology—is closely entwined with a
host of social and organizational phenomena including social cognition,
attitude formation and maintenance, prejudice and stereotyping, group
dynamics, interpersonal and intergroup relations, social influence, and
organizational trust and communication. These connections have not
been clearly set forth; that is the second reason we have written this
book.
In this volume, we address a series of questions—some old, some
new—that surround rumor. We set these questions against a backdrop
of current social and organizational psychological theory, and our
approach has a strong empirical flavor. Following are the questions
we tackle.

What Is Rumor?

Despite rumor's long history of inquiry, the rumor construct continues

to elude clarification. In chapter 1, we attempt to bring conceptual
clarity to rumor by presenting a new definition that takes into account

1
Researchers wishing to subscribe to this list should send the following message in
the first line of an e-mail to [email protected]: SUBscribe Rumor-Gossip
Research your name (please insert your actual name [e.g., Gordon Allport] at your name).
6 R U M O R P S YC H O L O GY

the content, contexts, and functions of rumor, gossip, and urban legend.
We develop these dimensions and present empirical evidence to but-
tress our contention that rumor, gossip, and urban legend tend to
exhibit different information-dimension patterns.

What Are the

Categories of Rumor,
How Frequent Are
They, and What Effects
Do They Have?

Chapter 2 addresses these descriptive questions, overviews rumor cate-

gorization schemes, and presents evidence pertaining to how frequently
various forms of rumor occur. In addition, this chapter addresses a
basic question, rarely investigated in a systematic fashion: Do rumors
cause or contribute to various outcomes of interest? This chapter re-
views a series of published and unpublished studies exploring the corre-
lates and consequences of rumor activity. We pay special attention in
this chapter to rumor effects in an organizational context.

Why Do People Spread

Rumors?

Rumor transmission has been perhaps the most studied aspect of rumor
research, and several antecedents have been implicated (Rosnow,
1991). In chapter 3, we review these antecedents and conceptually
integrate them within a more recent social-psychological motivational
framework used by attitude formation and maintenance researchers.
The result is a better understanding of both the causes (external ante-
cedents) and reasons (internal motivations) for rumor transmission.
We present research investigating how the pattern of motivation differs
in different social contexts. This chapter integrates the work on rumor
 Introduction 7

transmission by Rosnow (1991) and colleagues within a motivational

reasoning framework familiar to social psychologists.

Why Do People Believe

Rumors?

A puzzling question indeed, especially for fantastic tales! In chapter 4

we view the question using Egon Brunswik's lens model of judgment
in which distal attributes are inferred from proximal cues. We review
and meta-analytically combine studies bearing on this question to iden-
tify cues that people use to infer the truthfulness of a rumor. We present
evidence from a field study of stockbrokers suggesting the existence
of these—and other—belief cues. To our knowledge, this topic has
heretofore not been systematically studied.

How Do Rumors Help

People Make Sense of
Uncertainty?

Rumors typically help people make sense of ambiguous situations.

Chapter 5 explores the psychological mechanisms involved in this pro-
cess at the individual and interpersonal levels. At the individual level,
we explicate rumor sense making using frameworks within social cogni-
tion: explanation theory, illusory correlation, causal attribution, and
antiregressive prediction. At the interpersonal level, we explore collec-
tive rumor sense making by examining the content, functions, and
flow of rumor statements uttered in rumor discussions on the Internet.
Rumor has often been referred to as a sense-making process; this chap-
ter materially advances the field's understanding of that process. It
represents a much-needed integration of rumor theory and social cogni-
tion and advances the seminal collective sense-making theories pro-
posed by sociologists Shibutani (1966) and Ralph H. Turner (Turner &
Killian, 1972).
RUMOR P S Y C H O L O G Y

How Accurate Are

Rumors and How Do
They Become More—or
Less—Accurate?

Rumors have a bad reputation as being false. Is this reputation de-

served? Chapters 6 and 7 address this question and several others
related to rumor content change with respect to veracity. In chapter 6
we first present a review of decades of research on rumor content
change and attempt to resolve two divergent findings: Some rumors
seem to expand (become more detailed), whereas others seem to level
(become less detailed) over their lifetimes. We then present a sorely
needed conceptualization of the term accuracy and discuss how accuracy
is measured. Next, we pose a seldom-asked, but epistemologically im-
portant, question: Overall, how accurate are rumors? We review rumor
accuracy studies and report three investigations into organizational
rumor accuracy. In chapter 7 we review literature concerning the pro-
cesses by which rumors become more accurate and less accurate: Cogni-
tive, motivational, situational, group, and network mechanisms are
proposed. We present data from our accuracy investigations to explore
these processes. This chapter materially advances theory about accuracy
and revives the lesser known work of the sociologist H. Taylor Buckner
(1965) on this topic.

How Does Trust Affect

Rumor Transmission?

In our investigations of rumor, we have frequently observed that where

distrust grows, rumor flows—yet rumor research has not systematically
investigated this connection. In chapter 8 we present a longitudinal
study of the relationship between rumor and trust in an organization
undergoing radical downsizing. We use a framework proposed by orga-
nizational trust researchers Kurt Dirks and Donald Ferrin (2001) in
which trust is proposed to have both direct and moderating effects.
This chapter is a new avenue for understanding the dynamics of rumor
transmission from the vantage point of interpersonal trust.
 Introduction 9

How Can Harmful

Rumors Be Managed?

Popular literature about how to handle rumor is fairly common—but

almost all of it is nonempirical; it relies on intuition and individual
experience. In chapter 9, we summarize and synthesize empirical re-
search on rumor management strategies. Factors that moderate the
effectiveness of rumor denials are examined. Along the way, recurring
questions are addressed: Should a no-comments strategy be used?
Should one repeat a rumor when rebutting it? We apply research on
persuasion and causal attribution in addressing these questions.

What's Next in Rumor

Research?

Finally, in chapter 10 we summarize the main elements of previous

chapters, propose an integrative model of rumor processes, and set
forth a comprehensive agenda for future research. We hope that our
efforts in this volume will encourage and enable a renewed interest in
the social and organizational psychology of rumor.
 Defining Rumor
I

Anyone living in the United States during the weeks

following the terror attacks of September 11, 2001,
remembers two things: trying to make sense of these
senseless acts, and feeling threatened. As people
huddled in engaged conversations, rumors circulated
widely:
Avoid Boston on September 22nd because drunken
Arabs at a bar let it slip that there would be a second
wave of attacks that day.
—Marks, 2001
A hijacked plane is headed for the Sears Tower in
Chicago!
—Deener, 2001
Osama bin Laden owns Snapple (Emory, n.d.) and
Citibank (Cantera, 2002) and was sighted in Utah.
—Mikkdson, 2001a
Federal Emergency Management Agency personnel
were dispatched to New York City the day before the
September 11 attacks, proving that the government
knew of them in advance.
— "Monday, Monday," 2002
All of these rumors were false.

irthful friends posed the following question to one of us

(DiFonzo) after revealing my research interest in rumor:
"So, what's the latest on Ben and JLo?" They were refer-
ring to what, at the time, was the much publicized breakup
of film stars Ben Affleck and Jennifer Lopez. My reply,

11
12 R U M O R P S Y C H O L O G Y

"Actually that's gossip—not rumor," was greeted with further merri-

ment. I understood how I had tickled their funny bone: They were
amused that such a shadowy and often humorous topic—rumor—
was studied in a serious academic fashion. The conversation took an
educational turn while maintaining its jovial spirit: Using Socrates's
methods, I helped them "recall" that rumor had more to do with making
sense whereas gossip had more to do with evaluative social talk. We
attempt to do this and more in this chapter (and likewise retain our
sense of humor!).
My friends' reactions were understandable: Rumor is often lumped
with other genres of informal communication such as gossip, innuendo,
urban legend, and idle chitchat. Among the general public, "Have you
heard any rumors lately?" might elicit a juicy bit of office tattle, any
statement thought to be false, a pejorative stereotype, a morality tale, or
an anxious prognostication of office downsizing. Failing to distinguish
rumor from gossip is also the case among the broader psychological
public; indeed, rumor in the American Psychological Association's Psyc-
INFO Thesaurus is currently subsumed under the term gossip. And
among academics studying these forms of communication there is some
disagreement about what constitutes rumor. This conceptual murkiness
is not new; the Indian psychologist A. B. Ojha once observed that
academic definitions of the term rumor varied greatly, thereby leaving
the reader "in a fix" (1973, p. 61). Progress has been made (e.g., Rosnow
& Georgoudi, 1985), but in some ways ambiguity persists: While dining
together at a recent conference of social psychologists, rumor and gossip
researchers disputed for an entire evening over whether rumor and
gossip really are two different forms. And at a recent interdisciplinary
conference on rumors and legends, rumor and legend researchers
contended—again during meals—over whether rumor and urban leg-
ends differed in their essential features. Digestion notwithstanding, is
this ambiguity a problem?
Yes, it is. Conceptual ambiguity has contributed to problems in
rumor theory, research methodology, and management. For example,
requesting participants to think of a rumor may actually produce a set
of gossip statements; study results would then not be generalizable to
rumor. A second example is telling an urban legend about tourists in
Australia dressing up a stunned kangaroo who then awakens and hops
away with one of their wallets is likely to serve different social functions
than would telling a rumor about impending layoffs. In a similar way,
remedies for common gossip about a friend's extramarital affair may
not be a proper prescription for a rampant rumor about contaminated
soda pop. Glossing over the differences between these genres of infor-
mal communication leads to inadequate conceptualizations of how
 Defining Rumor 13

they operate, how they are validly studied, and how they are effectively
prevented and managed.
Meaningful distinctions between rumor and other forms of infor-
mal communication do exist (Rosnow & Georgoudi, 1985; Rosnow &
Kimmel, 2000), although remarkably these have rarely been empiri-
cally investigated. In this chapter, we further sharpen the concept of
rumor by defining it, comparing and contrasting it with gossip and
urban legend, and presenting empirical evidence investigating the di-
mensions of these types of informal communication. In particular, we
focus on the contextual, content, and functional elements of each form
of communication. We begin with a definition of rumor.

Rumor

We define rumors as unverified and instrumentally relevant information

statements in circulation that arise in contexts of ambiguity, danger,
or potential threat and that function to help people make sense and
manage risk. In this definition, we focus on the contexts in which
rumors crop up, contents of rumor statements, and group functions
that rumor serves, all of which are summarized in Table 1.1 along with
the contexts, contents, and functions of gossip and urban legend.

RUMOR CONTEXTS AND FUNCTIONS

Rumors arise in situations that are ambiguous or threatening in some
way (G. W. Airport & Postman, 1947b; Rosnow, 1991; Shibutani, 1966).
Ambiguous situations are those in which the meaning or significance
of events is unclear, or in which the effects of events are not certain.
Ambiguity is problematic for people. Why? In any context, humans
have a core social motive to understand and to act effectively (Fiske,
2004). Culturally defined categories ordinarily help individuals do this
(R. H. Turner, 1964). Yet sometimes events do not fit well together or
fail to convey meaning. In these cases, individuals refer back to the
group to understand the situation and to act (Asch, 1955; Sherif, 1936).
This referring back to the group—or group thinking—is rumor discus-
sion. Thus, rumors occur when a group is attempting to make sense
of ambiguous, uncertain, or confusing situations. Sociologist Tamotsu
Shibutani (1966) proposed that when formal information is absent,
people compensate by informally interpreting the situation (cf., Bauer
& Gleicher, 1953; R. H. Turner & Killian, 1972). Rumor discussion is
14 RUMOR P S Y C H O L O G Y

a process of group interpretation; rumor is a product of that process

(Rosnow, 1974).
Rumor activity in ambiguous contexts therefore functions to make
sense; it is a response to the core human motivation to understand
(Fiske, 2004). The resolving of ambiguity is always a sense-making
explanation of events that are unclear for some part of a group or
community. As G. W. Allport and Postman put it in their seminal text,
The Psychology of Rumor, "in ordinary rumor we find a marked tendency
for the agent to attribute causes to events, motives to characters, a raison
d'etre to the episode in question" [italics in original] (1947b, p. 121).
Explanations primarily attempt to make reality perceptible and mean-
ingful (Antaki, 1988); rumors are a collective effort at such explanations
when a group—or subset of a group—is faced with uncertainty (Di-
Fonzo & Bordia, 1998). To some extent, these explanations must pass
some group norm of plausibility. When group standards of plausibility
are high, rumor discussions look very much like fact finding. When
the group standards are low, rumor discussions look very much like
contagion or panic. Rumor is thus part of "normal collective information
seeking" [italics in original] (R. H. Turner, 1994, p. 247) wherein the
group is trying to define an ambiguous situation with a "lower degree

Contexts, Content, and Functions of Rumor, Gossip, and Urban Legend

Context Content Group function
Rumor Ambiguous or Instrumentally relevant To make sense of
threatening events information ambiguity
or situations statements that are To manage threat or
unverified potential threat
Gossip Social network Evaluative statements To entertain
building, about individuals' To supply social
structuring, or private lives information
maintaining To establish, change, or
maintain group
membership, group
power structure, or
group norms
Urban Storytelling Entertaining narratives To entertain
legend To establish, maintain, or
impart cultural mores
or values

Note. Each genre of communication may exhibit all contexts, contents, and functions in this table (e.g., rumor
also functions to impart cultural mores and gossip also functions to help the group make sense of ambiguity),
though each genre's quintessential contexts, contents, and functions are listed here.
 Defining Rumor 15

ofNormalization of many of its component acts" [italics in original] (Shi-

butani, 1966, p. 23). Norms for verification, sources, and other aspects
of fact finding are typically relaxed, but it remains a group sense-
making activity.
Rumor also functions to manage threat; it is a response to the core
human motivations to control one's environment or to protect one's
self-image (Fiske, 2004). Threatening situations are those in which
people feel that their welfare or sense of self is endangered. Threatening
situations may indeed imperil one's welfare, as when a company faces
possible layoffs. They may also involve threats to one's health or life,
as with catastrophic situations such as earthquakes, floods, and nuclear
accidents. Prasad (1935) documented one such outbreak of catastrophic
rumors after the great Indian earthquake of 1934. Rumors also
abounded after the Chernobyl nuclear accident in 1986 (Peters, Al-
brecht, Hennen, & Stegelmann, 1990). In the face of such threats to
welfare, rumors help groups to make sense of the situation and thereby
prepare for or act effectively against the threat. The preparation for
threat may take the form of "secondary control" whereby simply inter-
preting events within a framework that makes sense to people helps
them gain a sense of control (Walker, 1996). Threatening situations
may also be those in which one feels emotionally endangered, in which
one's sense of self—or indeed anything that one cherishes—seems
threatened. "Eleanor Club" rumors (among Whites) that Black servants
were found using the "lady of the house's" combs occurred during
times of racial turbulence in American history (these rumors were so
named after Eleanor Roosevelt; G. W. Airport & Postman, 1947b). The
context of these rumors was a sense that one's identity as a White
person was under attack. In a similar way, at the heart of wedge-driving
rumors—rumors that derogate other groups or classes of people—is a
defensive sentiment: one feels threatened. To cope with such feelings,
the wedge-driving rumor monger enhances his or her sense of self by
putting others down. Thus, in the face of threats to one's sense of self,
rumors that portray other groups in a negative light help people think
of themselves in a more pleasing way.
Rumor may also fulfill other functions such as entertainment, wish
fulfillment, alliance making and maintenance, and enforcement of
communal norms, but these are secondary. For example, rumors specu-
lating about the identity of the person who murdered a babysitter
(Peterson & Gist, 1951) may be entertaining for some, but this rumor
function takes second place behind attempts to ascertain "whodunit,"
how it could have happened in one's community, what the implications
are for future community safety, and so on. In short, the essence of
rumor has to do with sense making and threat management amid
uncertainty.
16 R U M O R P S YC H O L O GY

RUMOR CONTENT
Rumor content refers to the substance of rumor—what type of state-
ment constitutes a rumor? Rumors are information statements that
circulate among people, are instrumentally relevant, and are unverified.

Imtrumentally Relevant and Communicated

Information Statements
Rumors are first of all information statements. They tell rather than
interrogate or command (although a rumor statement may indeed raise
questions or direct behavior). The kernel of a rumor is an idea conveyed
by a meaningful statement or set of statements. "Carter Hawley is
bankrupt" (Lev, 1991), "The Haitian coup leader will be released from
prison" ("10 Die in Haiti," 1991), and "Paul McCartney is dead"
(Rosnow, 1991) are fundamentally subjects and verbs. Second, rumors
are communicated information. A rumor is not seated at rest inside an
individual; it moves among and through a set of persons. A rumor is
never merely a private thought. Rumors are threads in a complex fabric
of social exchange, informational commodities exchanged between
traders (Rosnow, 2001). In this sense, rumors are a subclass of memes,
ideas that survive through processes similar to those of natural selection;
"unfit" rumors die—they stop circulating—whereas "fit" ones persist
(Heath, Bell, & Sternberg, 2001).
Third, rumor is information that is circulated because it is instru-
mentally relevant to rumor participants; the information relates to,
affects, interests, or threatens them in some way. We mean instrumental
in the sense that Dewey (1925) intended: having some weighty pur-
pose, rather than being entertaining, sociable, or aimless. That is, ru-
mors tend to be about topics of significance to participants (Rosnow &
Kimmel, 2000). This instrumental relevance may be of a recent nature,
that is, it may pertain to current events. Rumor is similar to news in
this way; it tends to be information that is new. Relevance of this sort
often stems from potential consequences to persons (Rosnow & Fine,
1976); Rosnow (1991) dubbed it outcome relevance. For example, organi-
zational rumors in a study by DiFonzo et al. (1994) arose from collective
concerns such as job duties and advancement opportunities (turnover
rumors), job security and compensation (layoff rumors), and deleteri-
ous health effects (contamination rumors). However, the relevance
may be of an established nature. Rumors may pertain to topics of long-
standing interest such as ritual murder and flying saucers (Rosnow &
Kimmel, 2000). To the extent that participants considered such topics
to relate to, affect, or threaten them in some way, such statements
would be considered rumors. Put another way, although they some-
 Defining Rumor 17

times make people laugh, rumors are not primarily jokes; although
they may make people more sociable, they are not primarily meant to
pass the time. Rumors are about topics that people consider relatively
more urgent, significant, purposeful, or important.

Unverified Statements
Fourth—and most central—rumor is important communicated infor-
mation that is unverified. To verify is "to prove to be true by demonstra-
tion, evidence, etc.; to confirm" (Agnes, 1996, p. 683); unverified state-
ments, therefore, are unproven, not demonstrated to be true, and
unaccompanied by "secure standards of evidence" (G. W. Airport &
Postman, 1947b, p. ix). As Rosnow put it, rumor is "constructed around
unauthenticated information" (1974, p. 27). Note that this is not to
say that rumors never have a basis; they often do. Rather, the basis is
simply weak or absent—it is not secure evidence. Secure or stable
evidence is usually empirical in nature or consists of testimony from
a credible source. This type of evidence coheres even under scrutiny,
testing, and questioning. The difference between news and rumor is
helpful here; news is always confirmed, but rumor is always uncon-
firmed (Shibutani, 1966).
To some people, statements appear to possess secure evidential
moorings but on further scrutiny do not. Case in point: In 1981, a false
rumor spread that the president of the Procter & Gamble Corporation
(P&G) appeared on a nationwide talk show and proclaimed that P&G
donated to the Church of Satan; the rumor was spread by means of a
paper flyer (Koenig, 1985). The flyer proclaimed there was evidence:
It directed people to call the talk show and obtain for $3.00 a transcript
of the alleged broadcast. Anyone who took the time to call, of course,
discovered that the event never happened; indeed, P&G's president
has never appeared on any talk show. Thus, rumor includes what is
sometimes called misinformation, or false statements thought to be true
by some people. The implications of this reflection are sobering and
(rightly) lead to a general sense of caution about what is heard: Evidence
that initially seems firm may in fact easily crumble.
This characteristic of being unverified highlights the fact that some
rumors are more vulnerable to reality testing than are others.1 The
eminent philosopher of science, Karl R. Popper (1962), proposed that
a scientific theory is vulnerable to falsification; nonscientific theories
are not. This is the principle of falsifiability. Like scientific theories,

1
We are indebted to Charles Walker for the ideas and first example contained in
this paragraph.
18 R U M O R P SYC H O L O GY

some rumors are observable or specific in character. For example, "Sam

Dickson at corporate headquarters in Atlanta has asked all the branch
offices in New York to cut clerical staff by 20%" can be refuted; it is a
risky rumor. Mr. Dickson can rebut the rumor, clerical staff records
can be combed, and human resource personnel can deny receiving
cutback directives. In contrast, "Management are aliens!" is the kind
of claim that is difficult to disprove. Less risky rumors—like some
nonscientific theories—might be more immune to falsification because
they assert nonobservable phenomena. This Popperian aspect of rumor
may help to explain why some rumors are quite long-lived.

Mode of Transmission and Structure

Attempts have been made to define rumor on the basis of transmission
mechanisms. G. W. Airport and Postman (1947b) stated that rumors
were "passed along from person to person, usually by word of mouth"
(p. ix). Rosnow (1980, 2001) pointed out that this definition was out-
dated. Today, print, electronic media, and the Internet may be added
to word-of-mouth channels of communication through which ru-
mors flow.
Attempts have also been made to define rumor by its structure.
For example, rumors are often prefaced with a cautionary statement
indicating lack of verification such as "I don't know if this is true . . . " or
"I heard that. . . ." Because these prefaces indicate a lack of verification,
however, they are attached only to those rumors for which the trans-
mitter perceives the statement to be a rumor. They are thus only a
subset of statements constituting rumor because they signal only those
statements about which people are unsure. In addition, some research-
ers propose that rumors are simply short expressions. Rumors are non-
narrative in structure, that is, they tend not to have a plot sequence.
They are one-liners. In contrast, longer tales are classified as legends,
tall tales, fables, or myths. There is some merit in this distinction,
although the storylike quality of rumors varies considerably. For exam-
ple, we have collected many rumors that can be encapsulated in a
single sentence (e.g., see Tables 4.3, 4.4, 6.3, 6.4), but Caplow (1947)
observed that military rumors were typically composed of three state-
ments rather than one. Also, Peterson and Gist (1951) reported rumors
surrounding the murder of a babysitter; these rumors had snowballed—
become very elaborate—and were quite storylike in nature. The distinc-
tion in structure stems from the differential functions of these state-
ments. Rumors tend to relate information that pertains to an ambiguous
situation or event of which people are already apprised and in which
they are involved in ferreting out the facts; legends and myths tend to
provide a setting, plot line, and interpretation all-in-one. Viewed in
 Defining Rumor 19

this way, rumors are like plot twists revealed serially over time rather
than prepackaged as a coherent story. We return to this distinction in
our comparison of rumors and legends later. We turn now to rumor's
cousin, gossip.

Gossip

Gossip is evaluative social talk about individuals, usually not present,

that arises in the context of social network formation, change, and
maintenance—that is, in the context of building group solidarity. Gossip
fulfills a variety of essential social network functions including enter-
tainment, maintaining group cohesiveness, and establishing, changing,
and maintaining group norms, group power structure, and group
membership.

CONTEXTS AND FUNCTIONS

Gossip arises in the context of social network formation, change, and
maintenance (L. C. Smith, Lucas, & Latkin, 1999), that is, in situations
concerned with building group solidarity (Foster, 2004). One core hu-
man motivation is to belong, to fit in and be part of a group (Fiske,
2004). Gossip is talk that helps people do that by informing persons
about the group and individuals within it, helping one to keep track
of people in one's social network, advertising oneself as a potential
friend or mate, influencing people to conform to group norms, and
providing mutual enjoyment of an entertaining tidbit together (Dunbar,
2004; Foster & Rosnow, 2006; Rosnow & Georgoudi, 1985). Like pri-
mate grooming, activity that promotes interpersonal bonding is essen-
tial to group cohesion; gossip is thus a very important activity—without
it, societies would not be sustainable (Dunbar, 1996, 2004).
Gossip performs several functions key to social network formation
and maintenance (Foster, 2004). Gossip first of all provides information
about complex social environments; it informs people about aspects of
the group (Levin & Arluke, 1987). An organizational rookie benefits
from gossip such as "By the way, the boss is a real fascist."2 Wert and
Salovey (2004) claimed that all gossip is information gained by social
comparison with other people. For example, gossip informs people

2
We thank Eric Foster for providing this illustrative example of the informative
function of gossip.
20 R U M O R P S Y C H O L O G Y

about what to do and what not to do in a given social setting (Don't

get drunk at the Christmas party!) and who is to be included or excluded
in one's social group (Johnny is "cool"; Jason is a nerd). In most general
terms, gossip has been proposed as information gained through observ-
ing the "adventures and misadventures of others"; gossip thus provides
cultural information in a second- (and third- and fourth- and fifth-)
hand fashion (Baumeister, Zhang, & Vohs, 2004, p. 112). Some exam-
ples of such cultural information include "'Just don't drink'; 'Don't
forget your true friends'; 'Infidelity will eventually catch up with you'"
(Baumeister et al., 2004, p. 119).
A second way that gossip builds group solidarity is by providing
social entertainment (Litman & Pezzo, 2005; Rosnow & Fine, 1976).
Gossip is a mutual mood enhancer—together people laugh at other
people's peccadilloes. They enjoy a bit of private information about
someone else (Rosnow & Georgoudi, 1985); sharing such information
helps to pass the time. Tales of the office Don Juan's weekend escapades
are amusing; gossip therefore fulfills an entertainment function (Gluck-
man, 1963).
Another primary function of gossip is to define intimacy boundaries
and group membership; gossiping with another makes members feel
closer (L. C. Smith et al., 1999) or bonded (Horn & Haidt, 2002). It has
been argued that gossip is an efficient means of social bonding, enabling
friendship groups to include many members (Dunbar, 2004). Gossip
helps people affiliate with others by helping them keep track of people
in the social network and by helping them to advertise their own
desirability as a friend or mate (Dunbar, 2004). One knows that he or
she is part of the social group when someone whispers some delightful
inside information; one at last becomes an insider. One does not gossip
with one's enemies, but with friends or people with whom one wishes
to be more strongly affiliated. Thus, through gossip people gain friend-
ships and alliances. Of course, the darker side of delineating intimacy
boundaries is exclusion: Through gossip, people ostracize (L. C. Smith
et al., 1999). Thus, gossip is evaluative talk behind someone's back
(Foster, 2004; Sabini & Silver, 1982). It often evaluates the behavior
of a person or persons known to the participants—that is, in their social
network—in a conversational context in which the "evaluative talk
[is] about a person who is not present" (Eder & Enke, 1991, p. 494).
Gossip is a key weapon in "relational aggression" (Crick et al., 2001).
These painful experiences of exclusion are often the most memorable
aspects of childhood gossip experiences. Consistent with gossip's useful-
ness in excluding others is the finding that people sometimes feel
remorse after spreading negative gossip (Horn & Haidt, 2002).
Fourth, gossip defines not only who is in and out of the group,
but also power relationships within a group (Kurland & Felled, 2000;
 Defining Rumor 21

L. C. Smith et al., 1999). Gossip preserves or enhances the gossiper's

social status by slyly deprecating others or by enhancing self. Horn
and Haidt (2002) factor analyzed items related to social "people talk"
episodes: Gossiping made participants feel more empowered and popu-
lar, and that their status was elevated when telling critical gossip.
A fifth and final function is to propagate and enforce group norms
essential to group functioning. Gossip is perhaps particularly useful in
monitoring and cracking down on free riders—those who receive the
benefits that society gives without adequately returning in kind; too
many free riders can severely limit group functioning and thus gossip
performs an invaluable adaptive function (Dunbar, 2004; Foster &
Rosnow, 2006). Gossip does this by "informally communicating value-
laden information about members of a social setting" (Noon & Del-
bridge, 1993, p. 24). Norms may be communicated by commenting
on people personally known (proximate gossip) or those probably not
personally known, such as celebrities (e.g., Ben and JLo) and political
figures (distal gossip). Gossip thus has a moral orientation; it is value-
laden. It forms, maintains, enforces, or disseminates group norms. In
this way gossip influences and controls attitudes and actions (Rosnow
& Georgoudi, 1985). In more broad terms, gossip educates people about
how to act effectively in complex social environments (Foster, 2004),
especially by specific comparisons with the behavior of real or imagined
others (Wert & Salovey, 2004).

CONTENT
Although gossip is important, gossip content is typically presented in
a noninstrumental way by participants. That is, gossip is typically done
with an apparently aimless or idle purpose or simply to pass the time
(Rosnow & Georgoudi, 1985). Although it may indeed have a consid-
ered objective (trying to persuade, affiliate, exclude), it is packaged in
tones of relative disinterest. In a similar way, gossip is talk about matters
that are typically considered not that urgent or weighty. The content
is "nonessential in the context of the exchange" (Rosnow & Georgoudi,
1985, p. 62; Michelson & Mouly, 2000). Chat about office romances,
classmate peccadilloes, and family members' personality traits consti-
tutes gossip if offered without apparent serious intent. The same topics,
explored by a social psychologist in a conference paper, would not be
gossip because their function would be more central to the purpose of
the exchange. Again, gossip and gossiping perform central and signifi-
cant functions in social life, but gossip content is typically considered
less central, relevant, or important by participants.
In addition, gossip is evaluative talk—sometimes positive and
sometimes negative (Foster, 2004)—although it is predominantly
22 RUMOR PSYCHOLOGY

viewed as negative, slanderous, and derogatory (Wert & Salovey, 2004).

For example, a content analysis of gossip heard by college students at
a northeastern college found that 61% of it was intended to shame
the gossip target; in contrast, only 2% was intended to venerate the
gossipee (Walker & Struzyk, 1998). Gossip is tattle about someone—
a praise or critique of their actions or attitudes (Michelson & Mouly,
2000). Tales about Joe's behavior at the office Christmas party make
a statement that such behavior is laudable or laughable. Third, gossip
is evaluative talk about the private and personal life of an individual
rather than about a group or event (Foster, 2004). For example, middle
school gossip content often includes current couple configurations
("Johnny and Jasmine are a thing"}, personality traits ("Jackson is really
nice"), sexual orientation ("Frank is homosexual"), family matters
("Brittany's mother is having an affair"), and personal histories
("George was busted for cocaine last year").

RUMOR VERSUS GOSSIP

In common usage, rumor and gossip are often used interchangeably
(Rosnow, 1974; Rosnow & Fine, 1976). As previously stated, the Psyc-
INFO Thesaurus currently lists rumor under the term gossip. Gossip, like
rumor, is a transaction in which the hearsay is exchanged for some
other commodity (Rosnow & Fine, 1976). Both have been referred to
as "unofficial communication" (Kapferer, 1987/1990), "informal com-
munication" (Michelson & Mouly, 2000), and "hearsay" (Fine, 1985).
However, rumor and gossip are not equivalent concepts; they differ
in function and content. Rumor's function is to make sense of an
ambiguous situation or to help people adapt to known or potential
threats; gossip serves social network formation and maintenance. Put
another way, rumor is intended as a hypothesis to help make sense of
an unclear situation whereas gossip entertains, bonds, and normatively
influences group members.
Rumor can be distinguished from gossip in three ways (Rosnow,
1974; Rosnow & Georgoudi, 1985; Rosnow & Kimmel, 2000). First,
rumor is not based on solid evidence (it is unverified), whereas gossip
may or may not be firmly substantiated. Second, rumor is typically
about a topic of importance or significance to its participants, whereas
gossip is perceived in a less urgent fashion. As Sabini and Silver (1982)
put it, "Calling a story gossip and calling it rumor are both dismissive,
but they dismiss in different ways. 'Rumor' attacks the speaker's claim;
'gossip' attacks its idleness" (p. 92). Third, rumor may or may not be
about the private lives of individuals, but gossip is always about this
topic. Ambrosini (1983) stated it this way:" Gossip focuses on the private
affairs of individuals; rumor focuses on the larger sphere of human
 Defining Rumor 23

events" (p. 70). Although gossip may venerate (Foster, 2004), in con-
versation, at least, it is predominantly derogatory and slanderous in
nature (Walker, 2003; Walker & Struzyk, 1998; Wert & Salovey, 2004;
however, see Dunbar, 2004 for a contradictory finding: Less than 5%
of gossip overheard in public places was derogatory).
Although rumor and gossip differ, there exist "nebulous forms"
that are hard to classify (Rosnow, 2001, p. 211). For example, hearsay
that the boss is embezzling funds to pay for his sizable gambling debts
is unverified, is instrumentally relevant, and arises in a situation of
potential threat: Embezzlement may affect company livelihood. How-
ever, such hearsay is also evaluative idle talk that might be spread by
someone attempting to satisfy status or ego needs in the context of a
social hierarchy. Like rumor, gossip may convey useful social informa-
tion (Rosnow & Georgoudi, 1985). For example, coworkers may help
one another understand the boss's motivation: "She has an absurd
need for power; approach her only with requests that will make her
look good to her superiors." Such a message is private information
about another individual, it is entertaining, and it serves important
social network formation functions, yet it is also helps make sense of
an ambiguous situation and manage potential threat.

Urban, Modern, or
Contemporary Legends

Urban legends are stories of unusual, humorous, or horrible events

that contain themes related to the modern world; are told as something
that did or may have happened, variations of which are found in
numerous places and times; and contain moral implications (Cornwell
& Hobbs, 1992; Fine, 1992; Kapferer, 1987/1990). The term urban
legend is a misnomer—urban legends often involve any location, not
just cities; they are more properly termed modern or contemporary leg-
ends (P. B. Mullen, 1972). The terms urban, modern, and contemporary
legends are therefore used interchangeably in this discussion.

CONTEXT AND FUNCTIONS

Contemporary legends are told in a storytelling context. They have a
setting, a plot, a climax, and an epilogue. For example, tourists traveling
in a van in Australia hit a kangaroo and apparently killed it. They got
out of the van and dressed it up in one of the tourist's jacket and took
a photograph to show back home—what a laugh it would bring! But
24 R U M O R P S Y C H O L O G Y

the kangaroo had the last laugh; he was merely stunned. He awoke
and hopped away—jacket (which contained a wallet and a passport)
and all! Moral of the story: Be kind to animals (adapted from a version
circulating on the Internet in 1997, as quoted in Mikkelson, 2004a).
Contemporary legends are appropriate for situations in which enter-
taining stories are recounted, such as in casual conversation, Internet
chat episodes, and social gatherings. They serve important functions:
to amuse and to propagate moral values within a culture.
First they entertain. Modern urban legends are like tall tales in
their exaggeratedness (Bennett, 1985). They are interesting to listen
to. Consider the story of the hitchhiker who vanished (Brunvand,
1981): Driving on a country road, a father and daughter picked up a
young girl hitchhiking. She got into the backseat and told them that
she lived in a house 5 miles farther on. When they arrived, the girl
had vanished! Knocking on the door of the house, they discovered
that a young girl, who looked like the person they had picked up, had
disappeared several years ago and had last been seen hitchhiking on
that very country road. And, that day was her birthday.
Second, urban legends propagate mores and values. All good stories
signify a theme or meaning; in other words, there is a moral to the
story. As Kapferer (1987/1990, p. 123) stated, they are "exemplary
stories . . . since, like fables, their function is to set forth examples from
which moral implications can be drawn." Wilkie posited that three
popular contemporary legends in the 1970s circulated among Ameri-
cans "to criticize and regulate the behaviors of other Americans" (1986,
p. 5): Jumping up and down after sex prevents pregnancy (Don't get
pregnant!); child actor Jerry Mathers—Beaver in the popular TV series
Leave it to Beaver—died in action in the Vietnam conflict (Get out of
Vietnam!); and six students high on LSD blinded themselves by staring
into the sun (Don't take drugs!). The tale entitled "The Hook" criticizes
teenage promiscuity. In this urban legend, a teenage couple in a parked
car in the dead of night stop necking after hearing scratching noises;
after they arrive home, the prosthetic hook of an escaped mental patient
is found hanging on the car door handle (Brunvand, 1981). Modern
legends are thus like fables that focus on "fears, warnings, threats, and
promises" (Bennett, 1985, p. 223). Urban legends often contain the
funny and the horrible—but the "horror often 'punishes' someone
who flouts society's conventions" (Van der Linden & Chan, 2003). The
story of the traveler who is lured to his apartment by a seductive
woman but wakes up to discover that his kidney has been removed
as part of an illegal organ-selling operation is a morality tale about
one-night stands (Mikkelson, 2002). Like traditional legends, modern
legends persist because they answer long-standing questions and make
sense of the world; they symbolize underlying truths and values. For
 Defining Rumor 25

example, the legend of George Washington and the cherry tree symbol-
izes and reinforces the virtue of honesty (G. W. Allport & Postman,
1947b).

CONTENT
As in the kangaroo tale, urban legend content is first of all a narrative
tale, usually complete with setting, plot, climax, and denouement.
Second, these stories are unusual, horrible, or funny. They are of "nota-
ble happenings of the kind that allege 'strange but true'" (Fine, 1992,
p. 2). It would be unusual, to say the least, for a Doberman to bite off
and choke on the fingers of a burglar, but it could happen (Brunvand,
1984). Third, contemporary legend content contains contemporary ma-
terial as opposed to traditional themes and events. The topics of the
contemporary legend are "events that happened in contemporary soci-
ety and depict persons, relations, organizations, and institutions, that
are recognized by narrator and audience to characterize the modern
world" (Fine, 1992, p. 2; although see Bennett, 1985, for a counter-
view). These topics include, for example, automobiles, hitchhikers,
carcinogens, necking, photography, dating, and organ removal.

RUMOR VERSUS URBAN LEGENDS

Both rumor and modern legends have been seen as propositions for
belief, collective processes and transactions, verbal expressions (P. B.
Mullen, 1972), and unofficial news (Kapferer, 1987/1990). In addition,
both consist, at their core, of beliefs, statements, or verbal expressions
(P. B. Mullen, 1972). Distortion has occurred in both—in the form of
inclusion of concrete details—so as to make a tale more plausible (G. W.
Allport & Postman, 1947b; P. B. Mullen, 1972).
However, rumors and urban legends differ in their primary con-
texts, function, and content, their typical structure, and the extent to
which they migrate. First, although legends may, broadly speaking,
help to make sense of the world (P. B. Mullen, 1972), they often do
not pertain to a particular situation. For example, they are not often
proposed during company downsizings because they are of limited
value in ferreting out the facts and preparing for the future. However,
rumors are not set forth primarily to entertain and promote mores;
rumors of low projected fourth-quarter corporate earnings are not
amusing nor do they possess a moral adage. Thus, rumors tend to be
about a current event or topic of discussion (Rosnow, 1974) and how
these events predict the future; legends typically consist of a storylike
series of events that has already occurred.
26 R U M O R P S Y C H O L O G Y

Second, rumor and contemporary legend tend to differ in structure;

legends tend to be longer than rumors and to have narrative elements
(P. B. Mullen, 1972). Rumors are "short, nonnarrative expressions of
belief" (P. A. Turner, 1993, p. 5; Bird, 1979; Caplow, 1947; Fine, 1985).
As previously noted, this distinction accrues from the differing primary
functions of the two. Legends are for storytelling and amusement and
therefore tend to contain a setting, plot, climax, and denouement.
Rumors are for ferreting out the facts, making sense, and managing
risk and thus the information tends to be received in shorter packets that
are relevant to a particular situation. A narrative cannot be presented
because the sense making is contemporaneous rather than post hoc.
Finally, modern legends are rumors that become migratory—they
are "brought up to date" and located locally (Kapferer, 1987/1990,
p. 29); thus legends are renowned for the plethora of variations in
detail in different versions. For example, the K-mart snake rumor—
the false story about a woman bitten by a snake when trying on clothing
at a K-mart—migrated from geographic locale to locale and from
K-mart to Wal-Mart. Urban legends are not anchored in time and
space as rumors are (Kapferer, 1987/1990). In fact, legends have been
proposed to be rumors that persisted over time, and rumors have arisen
out of legends also; thus both "feed off one another" (P. B. Mullen,
1972). "Some legends may reappear occasionally embodied in the form
of a rumor" (Rosnow & Fine, 1976, p. 11). Put another way, "A legend
may be regarded as a solidified rumor" (D. L. Miller, 1985, p. 162).
Legends, then, are rumors that, after some distortion, persist for genera-
tions (G. W. Allport & Postman, 1947b). These legends seem to touch
down as rumors in a particular locale for a particular time, then disap-
pear, only to reappear years later in another place and with the charac-
ters changed.

Empirical Evidence:
Information
Dimensions

So far we have sought to define and distinguish rumor, gossip, and

urban legend by exploring their contexts, functions, and contents. We
have in mind methodological and practical reasons for this exploration;
a sharpened conceptualization of rumor will lead to more valid investi-
gations of rumor and more valid prescriptions for handling rumor.
With these objectives and concerns in mind, we investigated the ques-
 Defining Rumor 27

tion, "Do people make distinctions between pure forms of rumor, gos-
sip, and urban legends? And if so, do they make the same distinctions
that we do?" Thus far, the answer to both questions has turned out to
be "yes." In the remainder of this chapter we describe a series of studies
we conducted that investigated these questions.
On the basis of the knowledge we discussed earlier in this chapter,
we hypothesized that classic forms of rumor, news, gossip, and urban
legend would be differentially regarded along six dimensions of infor-
mation: evidentiary basis, importance, extent to which content is about
individuals, extent to which content is slanderous, how entertaining
the information is, and how useful the information is. In specific terms,
rumor should be rated low on evidentiary basis and high on importance
and usefulness. News ratings should mirror these elements except for
evidentiary basis, which ought to be rated highly. In contrast, gossip
should be considered by participants to be low in importance and
usefulness, and high in slanderous content about individuals and enter-
tainment value. Finally, urban legends should be low in evidentiary
basis, importance, and usefulness, but high in entertainment. These
hypotheses are summarized in Table 1.2.
To explore these hypotheses, we generated the Information Dimen-
sions Scale (IDS) to measure perceived dimensions of information. We
presented prototypical examples of each information type to partici-
pants and they rated each on bipolar 9-point scales. We rated four
information dimensions related to content: We measured evidentiary
basis by rating the extent to which the information is "information
that has been verified" versus "information that has not been verified,"
"information that you are absolutely [vs. not at all] sure is true," and
"information that is [vs. is not] based on strong evidence." We measured
importance of content by rating the extent to which the information
is important, is significant, and will be talked about seriously. We
measured content about individuals by rating the extent to which the
information is about a person's private life, about individuals, and not

Hypothesized Information Dimensions of Rumor, News, Gossip, and Urban Legend

Perceived Content Perceived
Evidentiary importance by about Content usefulness by
basis participants individuals slanderous Entertaining participants

Rumor Low High L/M/H L/M/H L/M/H High

News High High L/M/H L/M/H L/M/H High
Gossip L/M/H Low High High High Low
Urban legend Low Low L/M/H L/M/H High Low

Note. L/M/H = low, medium, or high.

28 RUMOR PSYCHOLOGY

about matters pertaining to the group or organization. We measured

slanderous content by rating the extent to which the information dis-
credits someone, is slanderous, and is derogatory. Participants rated
one exemplar of rumor, gossip, news, or urban legend; we created
two versions for each information type (see Exhibit 1.1). Fifty-nine
Rochester Institute of Technology undergraduates were asked to imag-
ine that they were at work and heard the statement from a coworker;

• E X - H I BIT 1

Rumor, Gossip, News, and Urban Legend Statements

Rumor 1: "/ heard that our department is about to be downsized." Your supervisor has not
heard anything about this but you know that the economy has not been doing that well lately.
Rumor 2: "/ heard that our department is about to be moved to another building and will be
merged with another department." Your supervisor has not heard anything about this but you
know that another department was recently moved and merged because of reorganization.
Gossip 1: "/ heard that one of the managers is having an extramarital affair with his secretary."
Your supervisor has not heard anything about this but you rarely see that manager with his
wife lately.
Gossip 2: "I heard that Sally is a wild and crazy kind of girl." (Sally works in the building but
you don't know her that well.) Your supervisor has not heard anything about this but you have
noticed that Sally is moderately attractive.
News 1: "Our company must respond to a tough economy in order to survive. We will be
downsizing the development department." (You work for the development department.) Your
supervisor confirms this and you know that the economy has not been doing that well lately.
News 2: "Jim Jones, head of development, will be promoted to senior vice president of consumer
relations." (You work for the development department and Jim Jones is your supervisor.) Your
supervisor confirms this and you know that he has had a long string of successes in the past
2 years.
Urban Legend 1: "/ heard this from a friend of a friend of mine: This guy was driving with a
group of tourists through the Australian bush when they hit a large kangaroo. He thought,
'What a great photo opportunity! The animal stood about 6 feet tall and would really impress
the pals back home.' So they propped the kangaroo up, and to add that little bit of humor,
one of them put his jacket on the kangaroo. However, the kangaroo was not dead! It was only
stunned and promptly hopped off into the distance complete with jacket, wallet, and passport."
Urban Legend 2: "/ heard this from a friend of a friend of mine: A Swiss couple fled home
from Hong Kong after their pet poodle, Rosa, was cooked and served to them garnished with
pepper sauce and bamboo snoots at a Chinese restaurant. The couple said they took Rosa with
them to the restaurant and asked a waiter to give her something to eat. The waiter had trouble
understanding the couple but eventually picked up the dog and carried her to the kitchen
where they thought she would be fed. Eventually the waiter returned carrying a dish. When
the couple removed the silver lid they found Rosa."

Note. Urban Legend 1 is adapted from a version circulating on the Internet in 1997, as quoted in Mikkelson
(2004a); Urban Legend 2 is from a Reuters news story that circulated in August 1971, as quoted in Brunvand
(1984, p. 95).
 Defining Rumor 29

in news conditions, the statement was read from a memo from the
company president. Participants then rated the statement on eviden-
tiary basis, importance, content about individuals, and slanderous
content.3
Content information dimension means for rumor, news, gossip,
and urban legend are presented in Figure 1.1. Across the two versions,
information dimension means for each type of statement were similar
and were therefore collapsed. Consistent with our hypotheses, rumor
and news differed only with regard to evidentiary basis, and were both
rated as important, not about individuals, and not slanderous. Rumor
and gossip, however, differed on every dimension except evidentiary
basis (both were low). Gossip exemplars were rated as not important,
about individuals, and slanderous. In addition, urban legend exemplars
were rated low on evidentiary basis and importance. Therefore, our
exemplars of rumor, gossip, news, and urban legend were meaningfully
distinguished by content along hypothesized information dimensions.
Participants also rated functional information dimensions: the ex-
tent to which the information is entertaining and useful. The primary
function of rumor and news is to make sense of an ambiguous or
threatening situation. Rumor and news should therefore be useful
information that may or may not be entertaining. Such information
would be more likely to be discussed in a serious conversation with a
boss or coworkers than at a lighthearted party. In contrast, gossip is
about social-network configuration, entertainment, and communicat-
ing social norms. In a similar way, urban legends are stories told for
entertainment and to convey mores. Gossip and urban legends should
therefore be rated high on entertainment and low on usefulness. Gossip
would be more likely to be discussed at a lighthearted party than in a
serious conversation with one's boss.
To measure these information dimensions, we constructed an addi-
tional set of bipolar IDS items. We measured entertaining by rating
the extent to which the statement was entertaining, amusing, and
enjoyable; we measured useful by rating the extent to which the state-
ment was "useful to you," beneficial to know, and helpful to know. A
second set of Rochester Institute of Technology undergraduate partici-
pants rated one of the same eight statements on the extent to which
the statement was entertaining and useful.4 Participants also rated how
likely they were to mention the statement "in a serious conversation

3
Standardized alpha coefficients were as follows: evidentiary basis, .91; importance,
.87; content about individuals, .70; and slanderous content, .79. This study used a
between-groups design («s = 7 or 8).
4
Standardized alpha coefficient for each was .83; N = 50 in a between-groups design
(m = 5, 6, or 8).
30 RUMOR P S Y C H O L O G Y

Information Type
Gossip D News H Urban Legend

Evidentiary Basis Importance About Individuals Slanderous

Information Dimension

Mean information dimension ratings for exemplars of rumor, gossip, news, and urban
legend: evidentiary basis, importance, about individuals, and slanderous.

with your boss," "at a party with friends where you are having a good
time," and "to other coworkers within an hour after hearing it."
Mean ratings for entertaining and useful are presented in
Figure 1.2; means were again similar across versions and therefore
collapsed within information type. Consistent with our hypotheses,
rumor and news exemplars served similar functions: Both were rated
as highly useful and low on entertainment. However, gossip and urban
legend were entertaining but not very useful. Thus, rumor differed
from gossip and urban legend on these dimensions. Our exemplars of
rumor, gossip, news, and urban legend were meaningfully distin-
guished by function along hypothesized information dimensions.
 Defining Rumor 31

Information Type
0 Gossip D News H Urban Legend

Entertaining Useful
Information Dimension

Mean information dimension ratings for exemplars of rumor, gossip, news, and urban
legend: entertaining and useful.

Transmission likelihood means for rumor, gossip, and news exem-

plars also reflected hypothesized functions. Figure 1.3 presents the
mean likelihood of transmitting each type of information within various
settings. Rumor and news transmission likelihoods did not differ across
situation, and both were more likely to be transmitted in a serious
conversation with the boss or to coworkers within 1 hour than to
friends having a good time at a party. These results indicate a sense-
making function. In contrast, gossip and urban legend were much more
likely to be shared with coworkers and within 1 hour of hearing it
than in a serious conversation with the boss, which indicates the enter-
tainment function. Once again, participants meaningfully distinguished
by function between exemplars of rumor, gossip, news, and urban
legend.
32 RUMOR P S Y C H O L O G Y

Information Type
H Gossip D News S Urban Legend

Boss Friends at Party Coworkers 1 Hour

Setting

Mean likelihood of transmitting rumor, gossip, news, and urban legend in different
settings. Scale anchored at 1 (very unlikely) to 9 (very likely). With boss = "in a serious
conversation with your boss," friends at party = "at a party with friends where you
are having a good time," and coworkers 1 hour = "to other coworkers within an hour
after hearing it."

Conclusions,
Implications, and
Future Research

In this chapter we have defined rumor as unverified and instrumentally

relevant information statements in circulation that arise in contexts of
ambiguity, danger, or potential threat and that function to help people
make sense and manage risk. We differentiated rumor from gossip
(social chat that entertains and serves valuable social network func-
tions) and urban legend (narrative that entertains and reinforces val-
 Defining Rumor 33

ues). We have taken a position that these genres of communication,

although sharing some similarities, exhibit meaningful distinctions. For
each form of communication, we explored context, function, and con-
tent differences. In addition, we created IDSs to investigate perceptions
of exemplars of rumor, gossip, news, and urban legend. Our exemplars
of these information types were appropriately distinguished: We ob-
served both content and functional differences.
What are the implications of the distinctions between rumor, gos-
sip, and urban legend? First, it is clear that in this volume we are
discussing rumors. Although the dynamics discussed in this book may
apply to gossip and urban legend, they are primarily intended for rumor.
Second, for rumor researchers, providing rumor exemplars (as in Ex-
hibit 1.1) or scenarios is likely to produce responses that are pertinent
to rumor rather than to gossip or urban legend. This point is important
especially for rumor and gossip. In unpublished data, we have uni-
formly found that people do not distinguish between the terms rumor
and gossip. That is, they rate them both as they would rate gossip: low
in importance, slanderous, about individuals, and so on. The rumor
researcher cannot simply request participants to recall rumors; rather,
rumors must be denned or exemplified in such a way as to distinguish
them from gossip. For example, participants could be given several
exemplars of rumor and then asked to provide additional examples.5
Profiles from the IDS may also be effectively used in content analysis
to distinguish rumor from gossip and legend. These steps will help
ensure that the phenomenon under consideration is truly rumor.
We note that the study presented in this chapter involved the use
of pure statement forms and that these pure forms exhibited different
information-dimension patterns with continuous scales. The use of
continuous scales represents an advance over attempts to place a cate-
gorical label on a statement because it allows for the comparative em-
phasis of the statement's content, context, and function. For example,
rumor may now be differentiated from gossip with comparative (e.g.,
rumor tends to score higher on the importance dimension than does
gossip) rather than dichotomous language (e.g., rumor is considered
important whereas gossip is not). Indeed, our approach easily incorpo-
rates the concept of nebulous forms—those statements that are difficult
to classify as rumor, gossip, or urban legend. Future research could use
IDSs to measure the information dimensions of nebulous forms and
the extent to which they match prototypical rumor, gossip, or urban
legend profiles. For example, the unverified statement about the boss's
embezzlement of funds might turn out to score highly on prototypical

5
We thank Charles Walker for this suggestion.
34 R U M O R P S Y C H O L O G Y

rumor and moderately on prototypical gossip. Such an approach would

prove more fruitful than would attempting to categorize the statement
in one or the other form, or the equally undesirable alternative of
minimizing distinctions between rumor, gossip, and urban legend.
Armed with a clearer conceptualization of rumor, we turn first to
a description of the phenomena. What types of rumors typically present
themselves during times of uncertainty and threat? How frequently
does each type of rumor occur? Do rumors cause or contribute to
human attitudes and behaviors? We address these questions in the
next chapter.
 Forms, Frequency, and
Fallout of Rumors

In late August 2005, 950 people died because of a

rumor. Hundreds of thousands of Shiite pilgrims had
converged in Baghdad to commemorate the
anniversary of Imam Musa al-Kazim's death. Pilgrims
were crossing a bridge in northern Baghdad when
rumors of a suicide bomber caused them to panic: "We
were all chanting slogans about Imam Musa, and then
people started shouting about a suicide bomber," an
injured eyewitness stated. "They started crashing into
each other; no one would look back or give a hand to
help the ones who had fallen. People started running
on top of each other, and everyone was trying to save
himself."
—Worth, 2005

e denned rumor as collective sense making in response to

uncertainty or threat. Before we approach the social and
organizational processes associated with rumor, we will
more fully describe the phenomena. Rumor researchers
have attempted to address three descriptive questions. First,
in what ways does rumor sense making manifest itself? It is
obvious that rumors differ in many ways, including content,
style, periodicity, and motivational goal; it will be helpful
to explore what categories or forms rumor takes. Second,
how often does this collective sense making occur? We once
spoke with the senior vice president of PR at a large corpora-
tion who said, "I deal with rumors all the time." It is valuable

35
36 R U M O R P S Y C H O L O G Y

to know how prevalent—how frequent—rumors are in human dis-

course. Third, does it matter? That is, do rumors cause or contribute
to human action or attitudes in a significant way? It is surprising that
this question has not often been addressed. It is frequently taken for
granted—through anecdotes and intuition—that rumors affect people.
Yet do they really, and if so, how strongly and to what end? Again, it
is beneficial to systematically assess what the effect—the fallout—of
rumor is. In the previous chapter we defined rumor; in this one we
describe it.

Forms of Rumors

G. W. Allport and Postman (1947b) noted that there are many ways
to classify rumors depending on the interest of the analyst: "The rumor
pie may be sliced in many ways" (p. 170). Rumors may be divided
according to temporal aspects such as periodicity: Some rumors (as we
saw in chap. 1, this volume) are versions of long-standing urban legends
that touch down from time to time and whose details adapt to the
current time and locale. For example, the story that a thief, lying in
wait under a woman's car that was parked in a shopping mall, slashed
her ankles and stole her car surfaces periodically as a rumor about
one's local shopping mall (Mikkelson, 1999). Rumors may also be
segmented according to subject matter; D. L. Miller (1985) surveyed
examples of product rumors (a leper had been discovered working in
the Chesterfield cigarette factory), disaster rumors (a Swedish nuclear
power plant leaked radiation), and atrocity rumors (a prisoner of war
communicated his torture via a postage stamp). In a similar way, rumors
have often been categorized by their content or theme. In this way
Knopf (1975) and P. A. Turner (1993) classified rumors related to race,
whereas Fine (1992) and Koenig (1985) grouped commercial rumors
(see also Bird, 1979). Rumors may also be differentiated by the pattern
of collective discussion surrounding them: Shibutani (1966) posited
rumors arising out of low-anxiety deliberative sense-making discus-
sions in contrast to those coming out of high-anxiety extemporaneous
discussions. Extemporaneous rumor discussions are similar to Wilke's
(1986) crisis rumors, which are endemic in situations in which there
is a dearth of or ambiguity about information about an important topic.
Kapferer (1987/1990) presented an etiological matrix of rumor based
on the origin of the rumor (an ambiguous event, a hitherto-unnoticed
detail, or no event) and the birth process of the rumor (did it arise
spontaneously or was it deliberately set forth?).
 Forms, Frequency, and Fallout of Rumors 37

Rumors have been cataloged according to the motivational tension

that characterizes the rumor: R. H. Knapp (1944) categorized over
1,000 wartime rumors as either dread rumors (fearful of a negative
event), wish rumors (hopeful of a positive event), or wedge-driving ru-
mors (expressive of hostility toward a people-group). To this motiva-
tional scheme, G. W. Allport and Postman (1947b) added curiosity (intel-
lectually puzzling) rumors. Knapp's classification scheme is perhaps
the most popular and has been adopted or extended in a variety of
contexts (Bordia & DiFonzo, 2004; Hellweg, 1987; Hershey, 1956).
For example, U.S. Air Force Captain Stephanie R. Kelley categorized
wartime rumors surfacing in and around Baghdad according to Knapp's
scheme (S. R. Kelley, 2004). Among other interesting findings, Kelley
observed a smaller percentage of wedge-driving and a greater percent-
age of fear rumors than did Knapp. Kelley then categorized these
wedge-driving rumors according to the target of their hostility, includ-
ing U.S.-Coalition forces, the Iraqi interim government, Saddam
Hussein, Kurds, Shi'a, and Sunni. Finally, Kelley categorized rumors
according to collective concerns of the Iraqis. A quarter of the rumors
reflected apprehension regarding the emerging government and politi-
cal process such as the transfer of authority to Iraq, alleged international
(usually Zionist-United States) plots to rule over Iraq, and possible civil
war because of sectarian strife; other concerns included quality of life,
the insurgency, and security.
With Ralph Rosnow (DiFonzo, Bordia, & Rosnow, 1994), we first
categorized organizational rumors recalled by company managers ac-
cording to content and object of collective interest or concern, which
reflected our emphasis on the sense-making function and ambiguous
or threatening contexts in which rumors often arise (see chap. 1, this
volume). Turnover rumors were about people leaving the organization
and reflected employee interest in how turnover might affect job duties,
advancement opportunities, and working conditions. Pecking order
rumors were about changes in management hierarchy and reflected
interest in how these changes would affect job stability and stock prices.
Rumors about job security and job quality reflected concerns about
employment longevity, duties, and compensation. Costly error rumors
were about mistakes and reflected concerns about stock prices, corpo-
rate reputation, and job security. Finally, consumer concern rumors were
about customer apprehensions and reflected concerns about sales, envi-
ronmental risks, and health effects.
We also classified organizational rumors according to their rumor
public: the people among which the rumor circulates. Some rumors are
primarily internal: "those of primary interest to company personnel,
suppliers, or vendors (i.e., people who are associated with the produc-
tion or distribution or sale of the organization's products or services)"
38 R U M O R P S Y C H O L O G Y

(DiFonzo & Bordia, 2000, p. 176). Turnover, pecking order, job security,
and job quality rumors are most likely to circulate among internal
rumor publics. Some rumors are primarily external: "Those of primary
interest to customers, press, stockholders, or the general public (i.e.,
people who purchase and use or potentially purchase and use the
organization's products, services, or stocks)" (p. 176). Costly error,
consumer-concern, and stock-market rumors are most likely to be
external rumors. This typology reflects the likely distinctions in both
content and object of collective concern in rumors circulating among
different organizational constituencies.
Organizational rumors are often most populous—and
troublesome—during periods of change. During a downsizing at a large
hospital, we categorized internal rumors according to content and ob-
ject of collective concern with regard to change (Bordia, Jones, Gallois,
Callan, & DiFonzo, in press). Our organizational change rumor typology
consisted of four types of rumors concerned with change: Rumors
about changes to job and working conditions were about job loss,
work practice changes, impacts on careers, loss of facilities, and staff
reductions (e.g., "operational officers to be downsized from 300 to
100"). Rumors about the nature of the organizational change reflected
concerns about changes to the structure and nature of the organization
(e.g., "mental health will be collocated with geriatrics"). Rumors about
poor change management were concerned with how badly change was
being accomplished, such as this rumor about waste: "they are paying
an extra $ 1 million to put an 'aesthetically pleasing' bend in the build-
ing!" Finally, some rumors were about the consequences of change for
organizational performance (e.g., "there will be no oxygen available
in bathrooms, only portable oxygen"). This typology again reflects the
sense making and threat management functions of rumor along several
different aspects of the organizational change: how well the change is
being managed, and its impact on jobs, organization structure, and
organizational performance.
What is to be gained from these varied attempts to parse rumors?
Classifying rumors highlights the contours of the collective sense
making and threat management functions of rumor: They tell us what
people are concerned about. Organizational members are obviously
apprehensive about changes that may affect their jobs, working condi-
tions, and financial security; when uncertain about such issues, they
will participate in the rumor mill. In addition, classification often reveals
underlying attitudes and beliefs (G. W. Allport & Postman, 1947b; R. H.
Knapp, 1944); dread rumors reveal an underlying fear, for example,
whereas hostility rumors indicate group conflict. S. R. Kelley's (2004)
observation of an entire family of rumors about alleged United States-
 Forms, Frequency, and Fallout of Rumors 39

Zionist conspiracies to subdue Iraq revealed deep distrust and antipathy

toward Israel. On a practical note, these categorization schemes prepare
managers, PR officers, and opinion leaders for the types and nature of
rumors they can expect.

Frequency

How often are rumors encountered? Our investigations concerning

organizational rumors over a dozen years have included interviews,
surveys, and lab experiments. According to anecdotal evidence, manag-
ers and communications officials often relate that they deal with rumors
frequently. One manager of a software development firm undergoing
downsizing quipped: "We are swimming in rumors" (DiFonzo & Bordia,
2006). Another complained, "I deal with rumors all the time!" Research
into this question is scant, but what little there is characterizes rumor—
contrary to our anecdotes—as less frequent than other forms of com-
munication. Horn and Haidt found that rumors arose much less fre-
quently than did gossip in the "social talk" of college students (Holly
Horn, 2003, personal communication). Hellweg's (1987) review of
grapevine research concluded that a small portion of informal network
information is rumor. During 2 years of service in a regimental military
unit of over 1,700 men in the Pacific during World War II, Caplow
(1947) noted that rumor frequency was low; Caplow's highest count
was 17 rumors in 1 month. Rumors are, of course, typically episodic
in nature. In our longitudinal study of rumors in an organization under-
going radical downsizing (see the rest of this chapter and chap. 8,
this volume), the average number of different rumors heard by each
employee each month peaked at seven just prior to layoff
announcements—a time of great uncertainty and anxiety—then re-
ceded to less than two in succeeding months. Some situations (e.g.,
those filled with uncertainty and anxiety) and conditions (e.g., close
networks) increase rumor frequency.
Managers are sensitive to the frequency of rumors, particularly
those rumors that are harmful. We investigated the frequency of harm-
ful organizational rumors in a sample of 74 very experienced corporate
PR officers (DiFonzo & Bordia, 2000). Participants were members of
the Arthur W. Page Society (a prestigious group of PR professionals)
or associated with the Institute for Public Relations in Gainesville,
Florida. The sample consisted of senior vice presidents of PR from
Fortune 500 corporations and well-known PR agency consultants. This
40 R U M O R P S Y C H O L O G Y

cream-of-the-PR-crop sample averaged over 26 years of PR experience

in a wide variety of industries. Participants reported that, on average,
rumors that were of concern or potential concern reached their ears
almost once per week.1 Almost 90% of participants reported hearing
such a rumor at least once per month. Thus, harmful or potentially
harmful corporate rumors appear to be encountered frequently by
PR professionals.
We also asked participants to estimate the overall proportion of
harmful rumors according to our internal-external classification
scheme discussed earlier (DiFonzo & Bordia, 2000). The median esti-
mate of the percentage of primarily internal rumors encountered was
50%; for primarily external rumors, the median estimate was 30%.
Thus, it appears that most harmful corporate rumors heard by PR
professionals circulate among internal constituencies. Of internal ru-
mors, the median estimates of the percentage of rumors about person-
nel changes, job security, and job satisfaction were 30%, 20%, and
10%, respectively. In comparison, the median estimate of the percent-
age of gossip rumors (i.e., gossip) was 15%.2 We asked participants for
examples of internal rumors that had reached their ears; what was
most evident was that nearly all of them related to episodes of significant
organizational change such as layoffs, mergers, outsourcing, staff
changes, restructurings, and plant closings. Downsizing and restructur-
ing, for example, led to rumors about benefit reductions and the transfer
of job duties. Of the external rumors, the median estimates of the
percentage of rumors about organizational reputation, stock market,
and product service or quality were 30%, 18%, and 10%, respectively.
As with internal rumors, perusal of the content of examples of these
rumors showed that the preponderance of them related to organiza-
tional change. Impending mergers, for example, spawned stock-market
rumors and rumors about the discontinuance of certain products. In
organizational contexts at least, both internal and external rumors
clearly become more frequent during times of change.
More recent work has focused on the frequency and flavor of
internal organizational change rumors. We assessed the relative fre-
quencies of internal organizational change rumors in the study of the
hospital undergoing change mentioned earlier (Bordia et al., in press).
The hospital was a hotbed of major and multiple changes taking place
over several years: new hospital construction, new patient care technol-

1
"M = 5.68, SD = 1.15, « = 74, where 5 = monthly, 6 = weekly, and 7 = daily"
(DiFonzo & Bordia, 2000, p. 177; see Appendix 2.1 for the full survey instrument).
2
Median estimates sum to less than 100% if distributions are positively skewed, as
many were.
 Forms, Frequency, and Fallout of Rumors 41

ogies, organizational restructuring, partial privatization, patient bed

reduction, new use of multidisciplinary teams, and relocation. All 3,200
hospital staff members were mailed our questionnaire and 1,610
(50.3%) responded. We asked respondents, "Please describe the last
rumor you heard about the changes going on at [this organization]."
Of 776 responses 368 (47%) were about changes to job and working
conditions; 147 (19%) concerned the nature of the organizational
change; 89 (11%) concerned the management of the change; and 53
(7%) concerned how change would affect organizational performance;
10 (1%) were gossip statements; and 109 (14%) were uncodable. In
addition, the overwhelming motivational tension of these rumors was
fear: Of the 510 reported rumors that could be classified as either
negative or positive, 479 were dread rumors and 31 were wish rumors.
Again, change was a catalyst for rumors; in organizations, the nature of
these rumors is sense making (amid anxiety) about how organizational
change may adversely affect job and working conditions. Furthermore,
the most frequent types of rumors are those concerning negative conse-
quences to employees (e.g., loss of job). Rumor participants seem to
attempt to gain a sense of control over their own situation by becoming
aware of the bad things that might happen to them because of the
change.

Rumor Fallout

Do rumors matter? That is, do they cause or contribute to behavior and

mental processes? The answer to this question clearly is yes. Popular,
business, and scientific literature is replete with instances in which
rumors exerted or exacerbated powerful effects. For example, rumors
during natural calamities have resulted in a number of outcomes (Pra-
sad, 1935; Shibutani, 1966). For example, Chinese earthquake rumors
spawned "panicky and fatalistic [behaviors] such as killing and eating
livestock and spending savings, stockpiling food supplies, trying to leave
the area, not going to work and postponing essential agricultural activi-
ties" as well as "active information seeking" (R. H. Turner, 1994, p. 252)
and a revival of mystic protection rituals in traditional areas. Economic
fallout of such rumors included food supply depletion and short-term
inflation. Rumors have also long been implicated in precipitating ethnic
riots (Horowitz, 2001; Knopf, 1975) and exacerbating racial tensions
(G. W. Allport & Postman, 1947b; Fine & Turner, 2001; P. A. Turner,
1993). For example, the Report of the American National Advisory
Commission on Civil Disorders cited rumor as responsible for inflaming
42 R U M O R P S Y C H O L O G Y

racial tensions in "more than 65% of the disorders studied by the

Commission" (Kerner et al., 1968, p. 173). Medical rumors are known
to affect health behaviors as well (Suls & Goodkin, 1994). For example,
a rumor that contraceptive use causes weakness was associated with
reduced use among Egyptian women (DeClerque, Tsui, Abul-Ata, &
Barcelona, 1986). A second example: False rumors that Hong Kong had
been declared an area infected by severe acute respiratory syndrome
(SARS) caused widespread panic there ("Teenager Arrested," 2003).
Telephone networks became jammed with people spreading the rumor,
which resulted in bank and supermarket runs. Rumors may even affect
population rates; in one economic model, rumors that government-
funded social security systems can fail because of aging populations
reduce expectations about the future and lead citizens to bear fewer
children (Van Groezen, Leers, & Meijdam, 2002).
Rumor effects occur even if rumors themselves are not believed.
Prasad (1935) noticed that people may have disbelieved rumors circu-
lating after the calamitous Indian earthquake of 1934, but they still
acted on them. For example, trips to the Indian city of Patna were
canceled on the day that a rumor predicted that Patna would cease to
exist. People apparently wanted to be safe rather than sorry; that is,
even though the outcomes predicted by the rumor were very unlikely,
they were extremely negative. We found the same sort of inference
making at work in a rumor episode we observed in 1993: the headlights
hoax. This rumor was transmitted by means of a flyer resembling a
safety memo; the memo spread like a contagious virus via fax machines
throughout the United States (Mikkelson, 2004b). Exhibit 2.1 presents
a reproduction of the fax we received in our graduate school mailboxes
at Temple University (a well-meaning employee had been alarmed
by the contents of the fax and had distributed it to everyone in the
psychology building).
The fax urged people to not flash their car lights for anyone because
of the deadly consequences that might occur. It alleged that nighttime
motorists blinked their headlights at oncoming automobiles that were
traveling with their headlights off as a kindly reminder to turn them
on. This neighborly act resulted in the opposing motorist circling about,
following the blinking motorist to his or her destination, and killing
him or her—all as part of a grisly new gang initiation rite. The fax
allegedly originated from the Illinois state police; when we contacted
them to verify the rumor, a tired trooper stated: "We've been getting
phone calls about this for 2 weeks. It never happened; it's a hoax!"
After interviewing 36 people in the psychology building, we found that
virtually all of them planned to abstain from headlight blinking of any
sort, even after being informed of the false nature of the fax. Why?
One likely explanation of these effects is prospect theory. Prospect
 Forms, Frequency, and Fallout of Rumors 43

Headlights-Hoax Flyer

SAFETY NEWS
Date: October 15, 1993
From: Pat Duffy, Manager, Safety Department
To: All Employees and Their Families
We were made aware of the following bulletin from the Norfolk Southern police department
(Virginia) and have confirmed through the New Castle County and Wilmington police depart-
ments that similar events have occurred in Los Angeles, Chicago, and Baltimore. Please take
the time to read the remainder of this memo and inform your family members and friends.
This awareness and precaution is important for both drivers and passengers whether at home
or traveling on business or pleasure.
BULLETIN
MITHERE IS A NEW GANG INITIATION!!!
This new initiation of "MURDER" is brought about by gang members driving around with their
car lights off. When you flash your car lights to signal them that their lights are out, the gang
members take it literally as "LIGHTS OUT", so they follow you to your destination and kill
you!!! That's their initiation.
Two families have already fallen victim to this initiation ritual in the St. Louis and Chicago areas.
This information should be given widespread distribution on our respective territories and
posted on all bulletin boards. Beware and inform your families and friends.
DO NOT FLASH YOUR CAR LIGHTS FOR ANYONE
The above information was furnished by the Illinois State police department.

theory says that people are loss-averse, that is, they tend to feel losses
more intensely than they feel equivalent gains (Kahneman & Tversky,
1979). People tend to overestimate the probability of improbable nega-
tive events, perhaps because negative information is processed more
thoroughly than is positive information (Baumeister, Bratslavsky,
Finkenauer, & Vohs, 2001), and therefore take steps to avoid them.
The consequences of blinking one's headlights, although improbable,
were vividly and catastrophically negative. The net effect of this rumor:
a neighborly cultural practice diminished.
Rumor effects may be classified as behavioral or attitudinal. In
business settings, behavioral effects of rumor include those that affect
purchase behaviors. The false rumor that Tropical Fantasy, a soft drink,
was owned by the Ku Klux Klan and made Black men sterile reportedly
caused sales to drop by 70% and incited attacks on delivery trucks
(Freedman, 1991). Unger (1979) reported similar losses in sales result-
ing from false product rumors: Bubble Yum bubble gum is contaminated
44 R U M O R P S Y C H O L O G Y

with spider eggs, and Pop Rocks candy, when ingested with soda pop,
explodes in one's stomach. Rumors have also affected stock purchase
behaviors and thus stock values (Lazar, 1973; Rose, 1951). For example,
prior to publication of takeover rumors in the "Heard on the Street"
column of The Wall Street Journal, price runups occurred, indicating
that the takeover rumors pushed prices up as they diffused through
the financial community (Pound & Zeckhauser, 1990); investors often
"buy on the rumor." Workplace productivity has also been affected—
usually negatively—by internal rumors (DiFonzo & Bordia, 2000). Tan-
gible effects such as these are often mediated through rumor's impact
on attitudes. One such attitude is reputation; clearly, rumors can wreak
havoc on a company's public standing (Zingales, 1998). Koenig (1985)
documented the case of Continental Bank whose reputation was be-
smirched by rumors of impending bankruptcy. Rumors during organi-
zational change episodes such as restructuring and layoffs may also
have damaging effects on organizational attitudes such as morale and
trust (DiFonzo & Bordia, 1998; DiFonzo et al., 1994; Smeltzer & Zener,
1992). These effects are but a small sampling of the many outcomes
that rumors can result in or contribute to. The remainder of this chapter
summarizes systematic research from several studies conducted to in-
vestigate the scope and nature of organizational rumor effects.

EFFECTS OF RUMORS ON STOCK

TRADING BEHAVIOR
Several experiments we conducted (DiFonzo & Bordia, 1997, 2002b)
suggested that individual trading behavior is dramatically affected by
the presence of rumors. In these studies, "investors" played a computer-
ized stock market game in which they could buy or sell (or do neither)
Goodyear stock. This microworld simulation (DiFonzo, Hantula, &
Bordia, 1998) generally lasted for 60 "days" of trading; each "day"
lasted 20 seconds. Stock prices started and ended at $35 per share, and
each day's price change was unrelated to the next day's price change;
prices rose during 30 of the 60 days and fell during the remainder. The
direction of the next day's price change was thus objectively unpredict-
able, as it is in the actual stock market (Fama, Fisher, Jensen, & Roll,
1969; Malkiel, 1985). During each day of trading, participants were
presented with the stock price for that day, the price change from the
previous day, and the value of their stock and cash holdings. A sample
display is reprinted in Figure 2.1. Some participants were presented
with eight stock-related rumors and others received no such informa-
tion. The rumors were either published in the "Heard on the Street"
column of The Wall Street Journal or heard from an unpublished source,
brother-in-law Harry (e.g., "You hear from Harry that Goodyear profits
 Forms, Frequency, and Fallout of Rumors 45

Day: 13 1 Seconds Left

The 'Heard On The Street' column reported this niMor today:
,Goodyear loses large government contract,

YOUR HOLDINGS TOD

CftSH =""==">$ 245,00 [enough to ky 6 «ore shares) PRICE PRICE CHflNGE
YOU OWN :==:==> 7 SHARES
WORTH :i::::i:>$ 277,55 539,65 $ +4,49
TOTfiL flSSETS==>$ 522,55 [you began uith $490,00]

Example of a display presented on a training "day." From "Rumors and Stable Cause
Attribution in Prediction and Behavior," by N. DiFonzo and P. Bordia, 2002b, Organiza-
tional Behavior and Human Decision Processes, 88, p, 787. Copyright 2002 by Elsevier.
Reprinted with permission.

are up"). The conditions under which these rumors were heard were
carefully varied: In one study, the rumor valence (whether it was a
positive or negative rumor) agreed with the direction of that day's price
change (up or down) 0%, 25%, 50%, 75%, or 100% of the time. The
predictive validity of the rumor for tomorrow's price change was also
varied. In one study, today's rumor predicted the direction of tomor-
row's price change (e.g., a positive rumor would be followed by a price
increase the next day); in most of the investigations, it was rigor-
ously unrelated.
In all of these studies, participants' trading patterns were systemati-
cally affected by the presence of rumors: They tended to depart from
a buy-low-sell-high trading strategy. Buying stock when the price is
relatively low and selling it when it is relatively high has been dubbed
a tracking strategy (Andreassen, 1987) and is, of course, essential to
making a profit. Departing from this strategy means, for example, that
participants continued to buy when the price was on the rise, or sell
when the price was declining; such patterns are not very economically
wise! Why did this happen? We describe the social cognitive processes
involved in these studies in more detail in chapter 5 (this volume), but
46 R U M O R P S Y C H O L O G Y

suffice to say that rumors led investors to attribute the causes of price
changes to stable forces, which are called stable-cause attributions. Thus
participants exposed to rumors thought that recent price trends would
continue despite the fact that participants rated published and unpub-
lished rumors as not credible, untrustworthy, and risky. The result was
financially deleterious: Participants exposed to rumors departed from
tracking and made significantly less profit during the simulations than
did control participants.
In our most recent microworld study (DiFonzo & Bordia, 2002b)
we were able to counteract the rumor-based formation of stable-cause
attributions through training. In the unstable-cause training condition,
participants were taught that stock-market price changes are random
and unpredictable. These participants were taught to perceive that the
information contained in rumors had already been incorporated into
the stock price for that day—that the price efficiently reflected the
aggregate opinion of the market. This efficient market theory (Fama et
al., 1969) remains the prevailing understanding of the stock market.
In the stable-cause training condition, participants were taught that
stock price changes were affected for 2 or 3 days after a rumor surfaced;
thus tomorrow's price change was somewhat predictable from today's
rumor. Control participants received no training. Results: Investors
taught to see price changes as caused by unstable forces departed less
from tracking (buy low, sell high) trading strategies than did control
participants or those trained to perceive stable causes. In other words,
trained to perceive random variation, investors were less prone to
the stable-cause attributional effects of rumor. In sum, these studies
implicated stable-cause attributional mechanisms behind the systematic
effects that rumors have on predictions of sequential events and on
behaviors based on those predictions; these mechanisms are especially
relevant to effects of rumors felt on the stock market.

EFFECTS OF CORPORATE RUMORS

In our study on the effects of corporate rumors (DiFonzo & Bordia,
2000) we presented our sample of 74 experienced PR officers with a
list of 17 rumor effects (see Appendix 2.1) that we had developed from
the literature and from our own previous studies (DiFonzo et al., 1994;
DiFonzo & Bordia, 1998). Participants rated whether or not they had
ever witnessed each effect as the result of a rumor they had personally
dealt with. They also rated each effect for its average severity on a scale
of small, medium, and large (we later assigned numeric values of 1, 2,
and 3, respectively, to these ratings). Finally, we asked participants to
give us a recent example of how a rumor caused the most serious
effect(s) to occur.
 Forms, Frequency, and Fallout of Rumors 47

Average severity rating, as well as the percentage of the sample

that had ever observed the effects, is presented in Figure 2.2. Our
experienced sample had observed a large majority of the effects during
their long tenures; each of the top 11 effects had been witnessed by
at least 78% of the sample. The most commonly experienced—
90% or greater—rumor effects included detrimental consequences for
employee morale, press reports, productivity, stress levels, and trust
held by employees and customers. Effects were rated, overall, as
moderately severe: 13 of the 17 effects were given an average severity
rating between 1.50 and 2.50 (indicating medium severity). The most
severe of these—those with an average rating at or above 1.75—

Lowered morale: 97.3%A

Bad press: 95.9% E
Loss of trust between management & workers/staff: 95.8% A
Increased employee stress at work: 95.8% A
Loss of trust among coworkers: 94.4% A
o>
Loss of trust by customers: 91.7% E
Decreased productivity: 90.3% B
UJ
— Sullied reputation of an individual: 88.7% E
O
E Sullied reputation of corporation: 88.7% E
DC Loss of trust between management & stockholders: 78.3% E
Unnecessary turnover: 78.1% B
Securities price/earnings dropped: 71.2% E
Decreased sales: 66.2% E
Fueled union action or strike: 55.1% E
Increased lateness/absenteeism: 54.3% B
Punitive action against management: 49.3% E
Sabotage-type behavior: 48.6%

1.00 1.25 1.50 1.75 2.00 2.25

Mean Severity

Rumor effects and mean severity ratings. Effects are in decreasing order by the percent-
age of respondents (n ranged from 66-73) who had ever observed the effect in their
overall experience. Mean severity ratings are on a scale in which 1, 2, and 3 indicate
small, medium, and large average effects, respectively. E indicates an external ramifica-
tion, A indicates effects related to internal attitudes, and B indicates effects associated
with internal behaviors (see text). From "How Top PR Professionals Handle Hearsay:
Corporate Rumors, Their Effects, and Strategies to Manage Them," by N. DiFonzo and
P. Bordia, 2000, Public Relations Review, 26, p. 180. Copyright 2000 by Elsevier.
Reprinted with permission.
48 RUMOR PSYCHOLOGY

duplicated the list of most often experienced, and included a "sullied

reputation of corporation" as well. Examples of how rumors caused
such effects were illustrative: A rumor that the company was "up for
sale" resulted in "lowered morale, resulting in loss of direction, focus,
which reverberated to sales force and caused job security concerns."
A rumor that "two plants in distant cities were to be closed and 4,000
people were to be terminated" had the effect of lowering morale in
these plants and reducing productivity by 10%. A rumor that the
corporation "would release 1998 first quarter earnings early because
of losses stemming from its Asian business" caused an "intraday [stock]
price dip of 2 points (price about 135) in company's common stock,
which was reversed by end of day." These instances provide a glimpse
of the many varieties of rumor effects that have been experienced
by corporate PR officers.
Are there broader categories of rumor effects? We were interested
in reducing the 17 effects to a more manageable set. We therefore
performed a principal components analysis on the severity ratings for
each effect. Principal components analysis, commonly used as a data
reduction technique, is similar to factor analysis in which correlated
items are put together as components. A three-component solution,
accounting for 58% of the total variance in effect severity ratings, was
selected as most meaningful. The effects and their component loadings
are presented in Table 2.1; component effects are also indicated in
Figure 2.2.
The three components were labeled external ramifications (e.g., bad
press, sullied reputation of corporation, decreased sales), internal atti-
tudes (e.g., loss of trust among coworkers, loss of trust between manage-
ment and workers or staff, lowered morale), and internal behaviors
(e.g., decreased productivity, increased lateness or absenteeism). These
components correspond to the internal-external categorization (dis-
cussed earlier), indicating that the nature of the effect results in part
from the rumor public through which the rumor recirculates.3 We then
created three component scores for each participant by averaging the
severity ratings for the items within each component; thus, each partici-
pant had an external ramifications component score, an internal atti-
tudes component score, and an internal behaviors component score.
The mean of these component scores showed that external ramifica-
tions were considered to be somewhat severe, internal attitude effects
were moderately severe, and internal behaviors were also somewhat

3
We performed item analysis for each set of items in the three components; each
was reliable (alpha coefficient for nine external ramification items was .89; for four
internal attitude items, .78; and for three internal behavior items, .69).
 Forms, Frequency, and Fallout of Rumors 49

Rotated Factor Pattern and Final Communality Estimates From Principal Component Analysis
of Rumor Effect Item Severity Ratings
Component
External Internal Internal Communality
ramifications attitudes behaviors estimates Rumor effect
77* 18 32 72 Punitive action against management
76* 1 11 59 Loss of trust by customers
73* 33 3 64 Sullied reputation of corporation
73* 15 6 55 Securities price/earnings ratio dropped
72* 21 -9 57 Loss of trust between management
and stockholders
72* -8 15 55 Bad press
69* -9 48 71 Decreased sales
61* 22 -4 42 Fueled union action or strike
57* 29 23 46 Sullied reputation of individual
8 81* 34 78 Lowered morale
2 76* 9 59 Loss of trust among coworkers
23 66* 7 49 Loss of trust between management
and workers or staff
19 65* 18 48 Increased employee stress at work
41 44 8 37 Sabotage-type behaviors
17 11 85* 76 Unnecessary turnover
8 34 74* 67 Increased lateness or absenteeism
5 44 54* 49 Decreased productivity

Note. N = 63. Printed values are multiplied by 100 and rounded to the nearest integer. Component loadings
greater than 0.50 have been flagged by an asterisk (*). From "How Top PR Professionals Handle Hearsay: Corpo-
rate Rumors, Their Effects, and Strategies to Manage Them," by N. DiFonzo and P. Bordia, 2000, Public Relations
Review, 26, p. 181. Copyright 2000 by Elsevier. Reprinted with permission.

severe.4 In sum, rumors that experienced PR officers have encountered

are perceived to have somewhat severe external ramifications, moder-
ately severe effects on internal attitudes, and somewhat severe effects
on internal behaviors.

EFFECTS OF NEGATIVE CHANGE RUMORS

ON EMPLOYEE STRESS
It is clear that our survey of corporate PR officers found that they
perceived the most serious consequences of rumor to be negative im-
pacts on employee attitudes. We wished to further investigate this idea

"External ramifications M = 1.58, SD = .54, n = 73 (where 1, 2, and 3 indicated

the effect was small, medium, and large in average severity, respectively); internal atti-
tudes M = 1.90, SD = .58, n = 73; internal behaviors M = 1.45, SD = .52, « = 70.
50 R U M O R P S Y C H O L O G Y

by comparing the stress levels of people hearing rumors with the levels
of those not hearing rumors. In the hospital study of change-related
rumors introduced earlier (Bordia et al., in press), we gained a large-
scale opportunity to do this.
This study was part of a broader investigation of the dynamics and
effects of organizational change, and therefore change-related stress
was measured. Each respondent rated how stressful the organizational
changes were, using the following four dimensions taken from Terry,
Tonge, and Callan (1995): (a) not at all stressful to extremely stressful;
(b) not at all disruptive to extremely disruptive; (c) not at all upsetting
to extremely upsetting; and (d) not at all difficult to extremely difficult.
These ratings were aggregated into a single change-related stress score.5
Now, recall that in this study, 776 of the 1,610 respondents reported
the most recent change-related rumor they had heard; 834 did not
report hearing a rumor. Of the 776 who heard a rumor, 479 heard
negative (dread) rumors and 31 heard positive (wish) rumors. The
remaining rumors (e.g., "mental health will be collocated with geriat-
rics") could not be classified either way and were therefore dropped
from this analysis. To assess the relation between rumor and stress, we
compared change-related stress scores between the negative-rumor
group (n = 479), the positive-rumor group (n = 31), and the group
that did not report any rumor (n = 834). Change-related stress was
higher in the negative-rumor group than in the positive-rumor group—
which is to be expected because the anticipation of a dreaded event is
more stressful than that of a wished-for event—but it was also higher
than in the no-rumor group. That is, employees who reported recently
hearing a negative rumor were more stressed than those who didn't
report hearing a rumor; ignorance may indeed be more blissful—or at
least less stressful. The finding is, of course, only correlational; those
who are more stressed may be more likely to hear or remember a
negative rumor. Indeed, in light of the fact that uncertainty and anxiety
have been linked to rumor transmission (see chap. 3, this volume),
this explanation seems likely. At the very least, however, this result
suggests that negative organizational rumors are associated with greater
employee stress.

EFFECTS OF NEGATIVE RUMORS ON

EMPLOYEE ATTITUDES, PRODUCTIVITY,
AND INTENTION TO STAY
Again, our survey of corporate PR officers found that they perceived
the most serious consequences of rumor to be negative impacts on

5
Alpha = .92.
 Forms, Frequency, and Fallout of Rumors 51

employee attitudes. And the cross-sectional organizational survey

discussed earlier points to an association between hearing rumors and
stress. We investigated these ideas in a more comprehensive and
longitudinal way by measuring several well-known employee attitudes
and behaviors over time in a change situation. For example, do
employees who hear negative rumors over time tend to become less
satisfied with their jobs? In this next study, we (and Rob Winterkorn)
assessed the relationship over time between hearing negative rumors
and uncertainty, anxiety, perceptions of quality of formal organiza-
tional communication, job satisfaction, organizational commitment,
trust in the company, productivity, and intention to stay. We did this
in a longitudinal investigation of a division of a company in Rochester,
New York undergoing radical downsizing. In chapter 8 (this volume)
we present the methodology of this study in greater detail and
investigate the direct and moderating effects of trust on the likelihood
of rumor transmission—defined as the likelihood of passing rumors;
here, we focus on how these attitudes and behaviors are related to
hearing rumors.
How might hearing rumors affect employee uncertainty, anxiety,
attitudes, intentions, and productivity? It is important to note that in
this study, as well as the organizational change rumor studies discussed
earlier, rumors are most often dread (the division will be downsized)
or wedge-driving (management has made budgetary errors) in nature.
Let's begin with uncertainty, anxiety, and formal communication. As
we discuss in chapter 3 (this volume), many investigations have linked
anxiety and uncertainty to rumor activity; we therefore posited that
uncertainty and anxiety would be associated with hearing rumors. The
causal direction could go both ways, however: People who are anxious
and uncertain may be more receptive to rumors (as we previously
speculated), or hearing dread and wedge-driving rumors of (usually)
uncontrollable events might also sustain or even heighten anxiety and
uncertainty. In a similar way, formal communication quality should
be negatively related to hearing rumors. Here also, causality may be
bidirectional and, in addition, may be mediated by uncertainty. Poor
formal communication would lead to higher uncertainty and hence
greater receptivity to hearing rumors (DiFonzo & Bordia, 1998), but
hearing such rumors—especially repeatedly—might lead one to con-
clude that formal communication efforts were poor.
Positive employee attitudes (job satisfaction, commitment to the
organization, and trust in the company) and intentions to remain with
the company are likely to be negatively associated with hearing negative
rumors. This negative association could happen in at least three ways:
Social learning (Lott & Lott, 1985) and social information processing
(Salancik & Pfeffer, 1977) through negative rumors may teach employ-
ees that others feel negatively toward the company or that it is not a
52 R U M O R P S Y C H O L O G Y

nice place to work. Indeed, rumors may be the quintessential vehicle

by which social learning occurs. Second, equity theory (Adams, 1965)
posits that employees keep account of the ratio between what they put
into a job and what they get out of it, and that they compare this ratio
with that of others in the company. Wedge-driving rumors seem likely
to result in perceptions of inequity, and thus reduce job satisfaction,
commitment, and trust. As a result, employees become more likely to
think about leaving the organization. Indeed, employees are more likely
to spread negative rumor about the organization when they feel that
the organization has treated them unfairly (Skarlicki & Folger, 1997).
Third, the law of cognitive structure activation contends that ambiguous
stimuli will be encoded as consistent with the most salient schema and
thereby affect relevant judgments and behaviors (Sedikides & Skow-
ronski, 1991; also see chap. 5, this volume). Negative rumors almost
certainly prime negative interpretation of ambiguous events by making
negative schema salient, resulting in lower satisfaction, commitment,
and trust. Again, these relations may be causally bidirectional: For
example, negative rumors might reduce trust in management, but low
trust in management might predispose an employee to hear—and
recall—more negative rumors.
The positive employee behavior of productivity (increasing work
output) might be increased or decreased by negative rumors. In our
previous research, rumors of plant layoffs allegedly motivated workers
to increase output (DiFonzo et al., 1994). K. Davis (1975) similarly
found an average 8% increase in productivity as a result of grapevine
rumors. However, anecdotal evidence suggests that engaging in rumor
discussions is time consuming and distracting and thus reduces produc-
tivity (Weiss, 1982, chap. 16).
On the basis of these ideas, we hypothesized that hearing negative
organizational rumors would be associated with higher levels of uncer-
tainty and anxiety and lower levels of satisfaction with formal commu-
nication quality, job satisfaction, organizational commitment, trust in
management, and intention to stay with the company; we made no
prediction regarding productivity. The results of this study were consis-
tent with these ideas.
The survey was administered once per month for 4 consecutive
months to all 75 (initial) employees; response rates were high during
each wave. Prior to the first and second waves, the division experienced
work slowdowns and restructuring. During these periods, negative
rumors—typically pertaining to dreaded layoff outcomes—abounded.
After the second wave, layoff announcements occurred; 50% of the
division was laid off between the third and fourth wave of the survey.
Uncertainty and anxiety were measured, along with these employee
attitudes: perceptions of the quality of formal organizational communi-
 Forms, Frequency, and Fallout of Rumors 53

cation, job satisfaction, organizational commitment, and perceptions of

the management as caring and trustworthy. Also, employees provided a
self-rating of their recent productivity, and of their intention to stay
with the company. All items were measured on a 7-point Likert-type
scale and are presented in Exhibit 2.2. In addition, participants recorded
the number of different rumors they had heard during the past month
and rated overall how positive versus negative these rumors were. As
one might expect during a time of radical downsizing, the rumors were
overwhelmingly negative.6
Let's first examine the overall pattern of means over time for each
variable. Mean numbers of rumors heard, along with uncertainty, anxi-
ety, communication quality, job satisfaction, organizational commit-
ment, trust, productivity, and intention to stay, are presented in Figures
2.3 and 2.4 for each wave of the survey. The number of different
rumors heard spiked from the first to the second wave, then receded
dramatically at Time 3 (T3) and Time 4 (T4). As the number of rumors
heard increased, uncertainty and anxiety also increased, whereas job
satisfaction, organizational commitment, and trust decreased. Self-
rated productivity and intention to stay also decreased. These trends
reversed when the number of rumors heard decreased; in addition,
communication quality ratings increased. In the aggregate, then, in-
creased numbers of negative rumors were accompanied by increased
uncertainty and anxiety and more negative employee attitudes, behav-
iors, and intentions.
Would these same patterns obtain at the level of the individual?
That is, how strongly is hearing rumors associated with each outcome?
Tables 2.2 through 2.5 present correlations between the number of
rumors heard during various time periods and Tl, T2, T3, and T4
outcome variables. To obtain an estimate of the overall strength of
these associations, we calculated average correlations.7 A clear pattern
emerged in line with predictions. First, number of rumors heard during
any specified or cumulative period was strongly correlated with uncer-
tainty (ravg = .42 for 20 correlations) and moderately correlated with
anxiety (ravg = .35). Hearing negative rumors was thus associated with
greater levels of uncertainty and anxiety. Second, moderate-sized nega-
tive correlations were observed between the number of rumors heard

6
Mean rating = 2.02, SD= 1.19, n = 169 (1 = extremely negative, 7 = extremely positive).
1
More sophisticated confidence intervals and meta-analytic combinations were not
appropriate here as correlations were not independent; we simply wished to calculate
a descriptive central tendency of each population of 20 correlations. Except for productiv-
ity (as discussed earlier) each set appeared to be homogeneous. In this discussion, designa-
tions of weak ( 0 < r < . 2 0 ) , moderate (.20 < r< .40), and strong (.40 < r< .60) correlations
were guided by J. Cohen (1988, pp. 79-81).
54 RUMOR PSYCHOLOGY

Variables Measured in a Four-Wave Longitudinal Study of an Organization Undergoing

Downsizing

Communication Quality (.66, .91, .88, .84)a

On the whole over the past month,
1. How well informed has the company kept you about upcoming changes? 7 = Completely
Informed, 1 = Completely Uninformed.
2. How do you feel about the adequacy of communication that you receive from the company?
7 = Completely Adequate, 1 = Completely Inadequate.

Uncertainty (.77, .84, .81, .83)

On the average over the past month.
1. was filled with questions about what current events in my company meant,
2. was uncertain about whether friends and colleagues will lose their jobs,
3. was uncertain about whether the company will be a good place to work,
4. was uncertain about whether the overall quality of my job will change,
5. was uncertain about whether I will be laid off.
6. was filled with uncertainty related to my job and/or my company.

Anxiety (.87, .96, .88, .84)

On the average over the past month,
1. I felt anxious about possible changes that will occur in this company.
2. The thought of upcoming changes in this company worried me.

Rumors Heard
In the past month, how many different rumors have you heard related to this organization?
(Write approximate number.)

Rumors Passed
Of the above number of rumors, how many did you pass on to someone else within the
organization? (Write approximate number.)

Job Satisfaction (.86, .84, .88, .89)

1. In general over the past month, how well would you say your job measures up to the sort
of job you hoped it would be when you took it? 7 = Very Much Like, 4 = Somewhat Like,
1 = Not Very Much Like.
2. All in all, how satisfied would you say you were with your job over the past month?
7 = Extremely, 4 = Moderately, 1 = Not at All.
3. If a good friend of yours told you he/she was interested in working in a job like yours for
your employer, on average over the past month what would you have told him/her?
7 = Definitely Recommend, 1 = Definitely Not Recommend.
4. Knowing what you know now, if you had to decide all over again whether to take the job
you now have, what would you have decided over the past month? 7 = Definitely Take
Same Job, 1 = Definitely Not Take Same Job.

continued
 Forms, Frequency, and Fallout of Rumors 55

'"'ILKHIBIT 2::,;2C|C:ej: rt:t( .fl:

Variables Measured in a Four-Wave Longitudinal Study of an Organization
Undergoing Downsizing

Organizational Commitment (.82, .82, .84, .93)

On average over the past month,

1. I did not feel a strong sense of belonging to my organization, (r)
2. I felt that this organization had a great deal of personal meaning for me.
3. I felt that I would be very happy to spend the rest of my career with this organization.
4. I felt that I would enjoy discussing my organization with people outside it.
5. I did not feel like "part of the family" at this organization, (r)
6. I did not feel "emotionally attached" to this organization, (r)

Trust (.86, .87, .94, .93)

On the average over the past month,
1. I felt that the company takes advantage of its employees, (r)
2. I felt that management is concerned about employee potential and development.
3. I felt that the company is honest in its dealings with the employees.
4. I felt that the company cares about what happens to its employees.
5. I felt that the company listens to its employees.

Intention to Stay (.79, .86, .91, .96)

On the average over the past month,
1. I often thought seriously about quitting my job. (r)
2. I often had serious intentions of leaving the company, (r)

Productivity (.96, .97, .95, .95)

Over the past month,
1. How productive were you in comparison to your usual level of productivity? 7 = Much More
Than Usual, 4 = Same as Usual, 1 = Much Less Than Usual.
2. How much work did you accomplish in comparison to what you ordinarily accomplish?
7 = Much More Work, 4 = About the Same, 1 = Much Less Work.

Note. Unless otherwise indicated, participants were asked to rate their level of agreement or disagreement with
each statement on a 7-point scale (1 = strongly disagree; 4 = neither agree nor disagree; 7 = strongly agree).
Measures of uncertainty, job satisfaction, and intention to stay are from Schweiger and DeNisi (1991) and previ-
ous conceptualizations of uncertainty. Communication quality, anxiety, rumors heard, rumors passed, and produc-
tivity items were generated. Trust was assessed with five items from Meglino, DeNisi, Youngblood, and Williams
(as cited in Schweiger & DeNisi, 1991). Organizational commitment from the Affective Commitment Scale
(McGee & Ford, 1987).
"Standardized alphas for time periods 1 through 4 are listed sequentially within parentheses after variable title (T1,
T2, T3, T4). N = 61, 48, 40, and 29 for each successive survey administration. Items with (r) were reverse-scored.

and the following employee attitudes: trust in the company (ravg = -.29),
job satisfaction (ravg = -.26), ratings of formal communication quality
('"avg = -.26), and—although weaker—organizational commitment
(7"avg = -.18). As predicted, hearing rumors was negatively associated
with key employee attitudes. Third, hearing rumors was moderately
56 RUMOR P S Y C H O L O G Y

8- Rumors Heard
Uncertainty
• • Anxiety
7-
— Intention to Stay
— Productivity
6-

5-

4-

3-

2-

1-

~r
4
Wave

Mean number of rumors heard, employee uncertainty, anxiety, self-rated productivity,

and intention to stay during an organizational downsizing.

negatively associated with the intention to stay (ravg = -.30). Again as

predicted, hearing negative rumors was associated with thoughts of
leaving the company. Finally, hearing negative rumors was moderately
negatively associated with productivity (r avg = -.22), although this cor-
relation varied widely from month to month (ravg = -.32 for T2 and T4
outcomes, ravg = -.03forTl and T3 outcomes). Hearing negative rumors
was sometimes moderately negatively associated with productivity, and
sometimes it was not associated with productivity.
It is important to note that these correlations obtained when the
rumors were heard both during the same time period as the outcome
variable and across time periods. That is, the number of rumors heard
during a certain month was associated with future months' outcomes.
We therefore wondered whether hearing negative rumors month after
month might exert even stronger effects, and indeed there does appear
to be a cumulative effect of hearing negative rumors. In general, the
 Forms, Frequency, and Fallout of Rumors 57

8- ^ Rumors Heard
• in Comm. Quality
• • Trust
7- — Job Satisfaction
— Org. Committment
6-

5-
ro
I
4-

lit!!

3-

2-

1-
—r
4
Wave

Mean number of rumors heard, perceptions of communication quality, perceptions of

management as caring and trustworthy, job satisfaction, and organizational commit-
ment during an organizational downsizing.

cumulative number of rumors heard in any given period was more

(positively) associated with uncertainty and anxiety—and more (nega-
tively) associated with job satisfaction, organizational commitment,
trust, productivity, and intention to stay—than was the number of
rumors heard in the most recent period only. Consider the correlations
presented in Table 2.4 for T3 outcome variables. The cumulative num-
ber of rumors heard by the end of the third month was more strongly
associated with every variable than was the number of rumors heard
during the third month alone (correlations in the Tl +12 + T3 row are
stronger than those in the T3 row in every column). Furthermore, the
pattern obtains even when one starts counting at T2: The cumulative
number of rumors heard from the second through the third month
was more strongly associated with every variable than was the number
of rumors heard during the third month alone (correlations in the
T2 + T3 row are stronger than those in the T3 row in every column).
58 RUMOR P S Y C H O L O G Y

Jx c

H
|s r
10
t * *
ro m
r f
<
(N
f
IS

£
f > * *
(N
cl inr "d-i' inr
^ u ^
H 0
fNI

l
p

a
1
a

•K #
<
«- 3 on
H C <N pi PO PO
m m
00
IN
f
" l' f l'

in in
Oi 15 *• _«
.a c c
re
•c
re <-;
.2 ll * <
to vo
m IN

I-
"1! 1

P
'E |
8
r i' r
•o 0 8
•o I
c C
CO rg
in in
T3
|
<D
1 Ol
_O
'w
c
o
'£ <00
0. ZI "o ™ o
*J
a. 00 o
(N •-
r
N -g S
t
01 "^ *!n 01 '~ M I* i' 1*
E
1 E 'i
in in
3 3
O _ O
•c
,o « .0
>
Ol S 5* ^"• O O 00 in
c
•c
3
^ "c ~
H
C
g "
g.
i •c
3
pff C gi
fM

1
T-

1 1
O g O |
•5re 8 ^
re
8

0)
I
umors 1

O
E
3
clc 00

"I
* <
IN IN
IN
IN
te. ra (C
"

&. q
1 X
E E jQ.
3 ^ o 3 1 * *
Z •"1 IN Z C" IB r~ *
* in on
* ^
on »a; iri
2
01

01 c p
01
01 3 01 3 ..

10)
DO
1
00
^0.

CD
o
relation n)
•relations

VI
"c
e period

V
.0 li V f
o.
Time peri
(correlati
Correlal

PP
§ S + I
S o
u I
i= £•§P H
IN
1-
S
s
 Forms, Frequency, and Fallout of Rumors 59

**
CM in r-. no
rfl fN no oo rf> no fN

I- 1 1 1 1 1 f

> 00 1C LD «- fM
m tJ O O O O O 0
^ 3
1 1 o
^
1 1 1
r
1

m S < »—
cn * *cri vo
* *in T—

P g r\i 't m no m fN
1 1 1 1 1 r

i/i -^
o>
.0 .2 p *ffl

<
CO oo in
•c 2 S •-c (N
H ,S
l~- O
m r\i m i- rvi fN

>
g C
1 1 1 1 1 r
p |i
XI
c
(0

1
VI
|
(N 1^ T- O 00
dj H O i|S r\l fN no r\i rvi fN
Q. ^ .£ 1 1 1 1 | f
0) 1
M

o
•c
1
s pif O *
00 *
O *
00 *
0
PO *t
r-.
r\i
_c f 3
E (T 1 I 1 1 1 r
3 g
0 "
S
X < *
*J * r-
52 H 'x
r\i * oo
* O no LD
Q PO ro no no m fSI

3
CC
"S ^.
OJ .£ 0
jQ * * *
in ID o O P^ 1C
E P
8 no ^ in no
3 Q.
-tt
3
C *
V L/i
q
V
10 a
(Q £-•• *
rrelations 1

0
1C P CD
1.1 a. + no V

M PO _i_
%-- + _i+ PO I §
&
3 II P P P pp
P
60 R U M O R P S Y C H O L O G Y

Is- 3 O O ID o o t t IN
O
~ PO ^" PO m *t m rM PM o »-
1 1 1 1 1 1 1 1 Q.
8
52.
TZn
< * < 0)
M ID 00 T-
M m ^f t
1 1 1
oo m LH
IN ^r n
1 1 1
ID IN
PM PM
1 1
I
S2
o si »- o

ce
01

1 O PM 'tf Ol »- IN ID in
- m mm IN m m O PM _
1 1 3
o «-
1 1 1 1 1 1

2 =
0
•• O ^" 00 ^f CM r^« in oo
all
H- .™ f
r r f
O i— »—
r r f
o «—
f oo

a.

•> m PO o 00 00 LO r^ in IN (0 <
- PO «a- *r rvi m m O PM (V) Q)
r r r r r r r r LT1

ce
in 01
O
'C n ID o PO O ID [^ PM <7l
as "" PO ^f 00 po m PM «- O
r r f r r r i' r I 2
O PO
O PM
Oi (D P
E cr
c 01
•c

* *
OS
01 ..f 1 0 «- 10
- 10 in <tf" ^D LH ^ £
X
0) 00
S2
o
E
cz 2!
•s o ID m T- 00
- t in in ^
T- 00
LTl ^~
•* o
Q.
01
OJ q 00 00
O PM
E V
I/I
S o LT>
IN O

V PO V
SK a. Ol Q.
01 ^^ _l_ rM
GO N—*• T
_
c S?
II
12
Ci + +
^ (N rM IN
IN i
^^

—^ m m
T
IN
£ (S
3I 01
D)
Q.
<

I!
n 1- 1- h- LD t- 1- m H (D

•t
Is
2 + + + o. + + Ci. +
PO PO
- P PP H HH 1- 1-
 Forms, Frequency, and Fallout of Rumors 61

Similar patterns were generally observed for T2 and T4 correlational

outcomes.
We also performed hierarchical regressions on these data as a
stronger test of the cumulative effect for rumor; these results generally
mirrored the correlational comparisons. In these analyses, we com-
puted the variance (R 2 ) for each outcome variable during a time period:
In step 1, we used the number of rumors heard during that same time
period, and in step 2 we used the cumulative rumors heard. In each
regression, the cumulative number of rumors heard almost always
accounted for additional variance, and this additional variance was
sometimes significant. For example, Table 2.6 presents results of hierar-
chical regressions for T4 outcomes; step 1 used the number of rumors
heard during T4 whereas step 2 used the cumulative number of rumor
heard during all 4 months. In this set of regressions, the cumulative
number of rumors heard is a significantly better predictor than is the
number of rumors heard in the same month for uncertainty, anxiety,
communication quality, job satisfaction, and productivity. Similar, al-
though weaker, patterns occurred for T2 and T3 outcomes, and included
effects on organizational commitment, trust, and intention to stay.
These results suggest the existence of cumulative effects for rumors
heard. In other words, the total number of rumors heard tends to be
more strongly (positively) associated with one's current uncertainty
and anxiety than does the number of rumors one has heard in the past
month. Because current uncertainty and anxiety cannot affect the
number of rumors heard in past months, this result suggests that nega-
tive rumors do increase uncertainty and anxiety; it is unsettling to hear
a layoff rumor about one's department. Furthermore, the cumulative
number of rumors heard tends to be more strongly (negatively) associ-
ated with one's current view of formal communication quality, job
satisfaction, organizational commitment, trust, productivity, and inten-
tion to stay than does the number of rumors one has heard in the past
month alone. Again, this result suggests that negative rumors do indeed
negatively affect these job attitudes, intentions, and behaviors.

Summary

In this chapter we have examined the forms, frequency, and fallout of

rumor. The way in which rumors can be categorized varies consider-
ably; we presented recent typologies of rumor based primarily on the
central function of rumor: collective sense making. Our recent typolo-
gies were therefore parsed according to thematic content, object of
62 R U M O R P S Y C H O L O G Y

collective concern and locus of rumor public (internal vs. external).

We also presented a typology of organizational change rumors. Al-
though the literature presents rumor as relatively infrequent, our recent
research points toward rumor as—of course—episodic in nature, and
closely related to organizational change. Negative rumors seem to pre-
dominate, and harmful or potentially harmful rumors reach the ear of
corporate PR officers almost once per week on average.
Rumors matter. They cause or contribute to a variety of important
outcomes. Furthermore, there is evidence that these effects do not
depend on belief in the rumor. Rumor effects are both intangible (in-
cluding attitude and reputation) and tangible (including sales, produc-
tivity, and stock prices). We summarized a number of recent experi-
mental studies showing that rumors systematically draw investors away
from tracking trading strategies. Furthermore, we presented a number
of field studies, both cross sectional and longitudinal, suggesting that
rumors negatively affect a number of important organizational attitudes
and behaviors: stress, job satisfaction, organizational commitment,
trust, productivity, and intention to stay.
As stated in chapter 1 (this volume), rumors are never merely a
private thought; rather, they are communicated. They are not simply
heard, but often passed along as well. Some rumors are more often
communicated than others. Why? In the next chapter we examine the
psychological factors involved in rumor transmission.
 Forms, Frequency, and Fallout of Rumors 63

Appendix 2.1
Managing Internal and External
Rumors: A Survey of Experienced
Communications Professionals

This survey1 is being administered to a select group of communications

and public relations professionals. It will assess the prevalence of differ-
ent types of rumors that exist inside and outside of organizations, the
effects of these rumors, and strategies used to prevent and neutralize
rumors. Please answer all questions. If you wish to comment on any
question or qualify your answers, please feel free to use the space in
the margins or the back page. Your comments will be read and taken
into account. Thank you for your help.
A rumor is an unverified bit of information about something of
importance to a group. It is like news in every way except that it is
not verified. It may or may not be true. It may be spread by word-of-
mouth, fax, electronic mail, or any other communication channel. It
is often introduced by the phrase "I heard that. . . . "
1. On average, about how frequently does a rumor reach your ear
(i.e., you hear it or hear that it is circulating) that is of concern or
potential concern to you? (Circle one number): 1 LESS THAN ONCE PER
YEAR, 2 YEARLY, 3 SEMI-ANNUALLY, 4 QUARTERLY, 5 MONTHLY,
6 WEEKLY, 7 DAILY, 8 MORE THAN ONCE PER DAY.

1
Survey instrument from DiFonzo and Bordia (2002b).
64 RUMOR P S Y C H O L O G Y

In the next three questions, you will be asked about how prevalent
different types of rumors are.
Internal rumors are of primary interest to company personnel or
suppliers or vendors, that is, people who are associated with the produc-
tion or distribution or sale of the organization's products or services.
External rumors are of primary interest to customers or press or stock-
holders or the general public, that is, people who purchase or use
or potentially purchase or use the organization's products, services,
or stocks.
2. Out of all the rumors that have reached your ear, about what
percentage were primarily internal rumors, primarily external rumors,
or primarily both types? (Please give approximate percentages for each type.
Please keep in mind that these percentages should add up to 100%): A. %
WERE PRIMARILY INTERNAL, B % WERE PRIMARILY EXTER-
NAL, C % WERE PRIMARILY BOTH, D % WERE PRIMARILY
OTHER (specify), TOTAL = 100%.
There are many types of internal rumors. Some are primarily about
changes that may threaten job security (the loss or potential loss of jobs)
or about changes that may threaten;^ satisfaction. These would typically
include rumors of layoffs, downsizing, reorganization, reduced pay,
or increased job responsibilities. Other internal rumors may also be
primarily about personnel changes (changes in staffing as a result of
turnover or promotion or demotion) or may be gossip (slander or innu-
endo primarily about private or personal matters).
3. Out of all the internal rumors that have reached your ear, about
what percentage were primarily about job security, primarily about job
satisfaction, primarily about personnel changes, primarily gossip, or
primarily about some other topic? (Please give approximate percentages
for each type. Please keep in mind that these percentages should add up to
100%): A % WERE PRIMARILY JOB SECURITY, B % WERE
PRIMARILY JOB SATISFACTION, C % WERE PRIMARILY PER-
SONNEL CHANGES, D % WERE PRIMARILY GOSSIP, E %
WERE PRIMARILY OTHER (specify), TOTAL = 100%.
The next question (and some others in this survey) requests an example.
We have included these requests because we wish to ensure that we accurately
understand your responses and we wish to obtain some sense of the rich detail
that is often involved in rumor situations. Please feel free to mask any details
so as to preserve the anonymity of this questionnaire.
4. Please give an example of a recent internal rumor that reached
your ear and that was of concern or potential concern to you. The
rumor stated:
In a similar way, there are many types of external rumors. Some
are primarily about changes that would affect stock prices or earnings
(e.g., as with potential mergers, forthcoming earnings reports, or costly
 Forms, Frequency, and Fallout of Rumors 65

management errors). Others are primarily about the quality of the

product or service that the organization produces (e.g., that it may be
unsafe). Still others are primarily about the organization's reputation
(e.g., that the organization is in some way untrustworthy or bad).
5. Out of all the external rumors that have reached your ear,
about what percentage were primarily stock market rumors, primarily
product or service quality rumors, primarily organizational reputation
rumors, or primarily about some other topic? (Please give approximate
percentages for each type. Please keep in mind that these percentages should add
up to 100%): A % WERE PRIMARILY STOCK MARKET, B %
WERE PRIMARILY PRODUCT OR SERVICE QUALITY, C % WERE
PRIMARILY ORGANIZATIONAL REPUTATION, D % WERE PRI-
MARILY OTHER (specify), TOTAL = 100%.
6. Please give an example of a recent external rumor that reached
your ear and that was of concern or potential concern to you. The
rumor stated:
Another important purpose of this study is to assess what the effects
of rumors are and how severe these effects typically are.
7. On the basis of your overall experience with rumors, how severe,
on average, were each of the following rumor effects? NEVER means
the effect never occurred, SMALL means the effect was small, MEDIUM
means the effect was moderate, and LARGE means the effect was large.
Circle one word to indicate severity (if effect has never occurred, circle
"NEVER"): 7.1 loss of trust between management and workers or staff,
7.2 loss of trust between management and stockholders, 7.3 loss of
trust among coworkers, 7.4 loss of trust by customers, 7.5 decreased
productivity, 7.6 unnecessary turnover, 7.7 decreased sales, 7.8 in-
creased lateness or absenteeism, 7.9 lowered morale, 7.10 sullied repu-
tation of an individual, 7.11 sullied reputation of corporation, 7.12
sabotage-type behaviors, 7.13 bad press, 7.14 securities price or earn-
ings dropped, 7.15 punitive action against management, 7.16 increased
employee stress at work, 7.17 fueled union action or strike, 7.18
other (specify).
8. Of the most serious effects that have occurred, please give a
recent example of how a rumor caused the effect(s) to occur (if possible,
please quantify the effect in terms of % change, such as % change in
productivity or information requests):
A further purpose of this study is to assess what strategies are in
use to prevent or neutralize rumors and how effective these strategies
are perceived to be.
9. On the basis of your overall experience with rumors, how effec-
tive, on average, were each of the following specific strategies in pre-
venting and/or neutralizing rumors? NEVER means strategy was never
used, LOW means low average effectiveness, MEDIUM means medium
66 R U M O R P S Y C H O L O G Y

average effectiveness, and HIGH means high average effectiveness. Cir-

cle one word to indicate effectiveness (if strategy was never used, circle
"NEVER"): 9.1 ignore the rumor, 9.2 state "no comment," 9.3 explain
why you cannot comment or give full information, 9.4 confirm the
rumor, 9.5 deny rumor by company official, 9.6 deny rumor by trusted
outside source, 9.7 spread counterrumors, 9.8 search for and/or seek
to punish people who planted the rumor, 9.9 specifically attempt to
increase trust, 9.10 encourage potentially affected parties to prepare
for the worst, 9.11 set time line for official message that will contain
full information, 9.12 establish committees to explore options for people
affected by upcoming change, 9.13 state values that will be used to guide
the upcoming changes, 9.14 state procedures by which the upcoming
changes will be decided, 9.15 establish a rumor hotline, 9.16 explain
how upcoming change decisions will be made, 9.17 have an employee
inform you of rumors that are circulating, 9.18 other (specify).
10. Please give a recent example of how one or more of these
strategies was effective in preventing or neutralizing a rumor:
This next section asks questions related to the psychological and
situational variables affecting the rumor. When you answer these ques-
tions, it is helpful to have a particular rumor in mind. Please think of
an instance when a rumor (but not a gossip-type rumor) reached your
ear that you considered harmful or potentially harmful. Please choose
a rumor that has since been proven true [false] beyond a reasonable
doubt (other participants are being asked for rumors that proved to be
false [true]; we wish to obtain an even sampling of both kinds).
11. Please state the true [false] rumor here (it may be one you
have already mentioned):
12. Please describe the situation out of which the rumor arose:
Rumors may spread through different groups of people. For the
remaining questions in this section, answer with respect to one group
of people who heard and/or passed the rumor. Also, rumors may change
over time. For the remaining questions in this section, please answer
for the point in time just prior to when the rumor was proven true.
The first question deals with how accurate or distorted the ru-
mor became.
13. For the rumor above, how accurate or true (as opposed to
distorted or false) did the rumor prove to be? 1 COMPLETELY FALSE,
2 MOSTLY FALSE, 3 HALF FALSE AND HALF TRUE, 4 MOSTLY TRUE,
5 COMPLETELY TRUE.
The next question deals with how the rumor changed over time.
14. From the time when the rumor started until when the rumor
was proven true, to what extent did the rumor tend to become more
accurate (as opposed to more distorted)? 1 BECAME MUCH MORE
DISTORTED, 2 BECAME SOMEWHAT MORE DISTORTED, 3 STAYED
 Forms, Frequency, and Fallout of Rumors 67

ABOUT THE SAME, 4 BECAME SOMEWHAT MORE ACCURATE,

5 BECAME MUCH MORE ACCURATE.
The next several questions deal with psychological and situational
variables. As with all the questions in this section remember to answer
with respect to a particular group of people who heard or passed the
rumor or both.
15. Overall, how uncertain (filled with questions about what cur-
rent events meant or what future events were likely to occur) were
people? 1 NOT AT ALL UNCERTAIN, 2 A LITTLE UNCERTAIN,
3 SOMEWHAT UNCERTAIN, 4 MOSTLY UNCERTAIN, 5 EXTREME-
LY UNCERTAIN.
16. Overall, how confident were people that this rumor was true?
1 NOT AT ALL CONFIDENT, 2 A LITTLE CONFIDENT, 3 SOMEWHAT
CONFIDENT, 4 MOSTLY CONFIDENT, 5 EXTREMELY CONFIDENT.
17. Overall, to what extent did people have a reduced sense of control
or power over events affecting their lives? 1 NOT AT ALL REDUCED
SENSE OF CONTROL, 2 A LITTLE REDUCED SENSE OF CONTROL,
3 SOMEWHAT REDUCED SENSE OF CONTROL, 4 MOSTLY REDUCED
SENSE OF CONTROL, 5 EXTREMELY REDUCED SENSE OF CONTROL.
18. Overall, how important (significant, consequential) was the ru-
mor to people? 1 NOT AT ALL IMPORTANT, 2 A LITTLE IMPORTANT,
3 SOMEWHAT IMPORTANT, 4 MOSTLY IMPORTANT, 5 EXTREME-
LY IMPORTANT.
19. Overall, how anxious (worried or concerned) were people?
1 NOT AT ALL ANXIOUS, 2 A LITTLE ANXIOUS, 3 SOMEWHAT ANX-
IOUS, 4 MOSTLY ANXIOUS, 5 EXTREMELY ANXIOUS.
These next questions continue with psychological variables. Please
continue to answer with respect to a particular group of people who
heard and/or passed the rumor.
20. Overall, to what extent did people desire or want the rumor to
be true? 1 GREATLY DESIRED RUMOR TO BE FALSE, 2 SOMEWHAT
DESIRED RUMOR TO BE FALSE, 3 NEUTRAL, 4 SOMEWHAT DE-
SIRED RUMOR TO BE TRUE, 5 GREATLY DESIRED RUMOR TO BE
TRUE.
21. Overall, to what extent did the rumor agree with or was con-
sistent with peoples' biases (preconceived notions, attitudes, and/or
prejudices)? 1 GREATLY DISAGREED WITH PEOPLES' BIASES, 2
SOMEWHAT DISAGREED WITH PEOPLES' BIASES, 3 NEITHER
AGREED NOR DISAGREED, 4 SOMEWHAT AGREED WITH PEOPLES'
BIASES, 5 GREATLY AGREED WITH PEOPLES' BIASES.
22. Overall, to what extent would you characterize the people as
gullible (as opposed to skeptical)? 1 HIGHLY SKEPTICAL, 2 SOME-
WHAT SKEPTICAL, 3 NEITHER SKEPTICAL NOR GULLIBLE, 4 SOME-
WHAT GULLIBLE, 5 HIGHLY GULLIBLE.
68 R U M O R P S Y C H O L O G Y

The next several questions deal with patterns of communication.

23. When a rumor is passed from one person to another, it may in-
volve a great deal of interaction (discussion and clarification), or it may
involve very little interaction. Overall, how interactive was a typical dis-
cussion of the rumor? 1 NOT AT ALL INTERACTIVE, 2 A LITTLE INTER-
ACTIVE, 3 SOMEWHAT INTERACTIVE, 4 MOSTLY INTERACTIVE, 5
EXTREMELY INTERACTIVE.
24. Overall, how active (frequently transmitted) was the rumor?
1 INACTIVE, 2 A LITTLE ACTIVE, 3 SOMEWHAT ACTIVE, 4 MOSTLY
ACTIVE, 5 EXTREMELY ACTIVE.
25. A rumor may pass through an established communication net-
work (e.g., an office grapevine) or through a group of people who never
communicated before. Overall, how established were the communica-
tion channels of the group? 1 NOT AT ALL ESTABLISHED, 2 A LITTLE
ESTABLISHED, 3 SOMEWHAT ESTABLISHED, 4 MOSTLY ESTAB-
LISHED, 5 EXTREMELY ESTABLISHED.
26. The grapevine is an established informal communication net-
work. To what extent could the communication network through which
this rumor passed be considered a grapevine? 1 DEFINITELY NOT A
GRAPEVINE, 2 A LITTLE LIKE A GRAPEVINE, 3 SOMEWHAT LIKE A
GRAPEVINE, 4 MOSTLY LIKE A GRAPEVINE, 5 DEFINITELY A
GRAPEVINE.
Finally, a few details about yourself for statistical purposes.
27. You are (Circle one number): 1 MALE, 2 FEMALE.
28. Your present age: YEARS.
29. Your current title:
30. You are a (Circle one number): 1 CORPORATE COMMUNICA-
TIONS OR PUBLIC RELATIONS OFFICER, 2 COMMUNICATIONS OR
PUBLIC RELATIONS CONSULTANT, 3 OTHER (specify):
31. The primary industry or industries in which you work or consult:
32. Your total years of experience in communications: YEARS.
 Psychological Factors
in Rumor Spread

In the rancorous months preceding the 2004 U.S.

presidential election, one of us (DiFonzo) received this
false e-rumor about Senator John Kerry:
Subject: Kerry and John 16:3. The Lord has a way of
revealing those of us who really know him, and those
that don't! Think about it! Kerry gave a big speech last
week about how his faith is so "important" to him. In
this attempt to convince the American people that we
should consider him for president, he announced that
his favorite Bible verse is John 16:3. Of course the
speech writer meant John 3:16, but nobody in the
Kerry camp was familiar enough with scripture to
catch the error. And do you know what John 16:3
says? John 16:3 says, "They will do such things
because they have not known the Father or me." The
Spirit works in strange ways.
It turns out that the same false e-rumor about President
George W. Bush was being circulated at the same time.
—Mikkelson and Mikkelson, 2004

n this chapter we identify the psychological factors underly-

ing rumor spread. We focus on the individual-level psycho-
logical motivations involved—that is, what are people trying
to accomplish when transmitting a rumor? The psychologi-
cal literature on rumor has identified five variables related
to rumor transmission: uncertainty, importance or outcome-
relevant involvement, lack of control, anxiety, and belief
(Bordia & DiFonzo, 2002; Rosnow, 1991; Walker & Blaine,

69
70 R U M O R P S Y C H O L O G Y

1991). We review the literature pertaining to each variable. In addi-

tion, we address the following question: "Why do these variables affect
rumor transmission?" We argue that these variables represent funda-
mental goals in social cognition and behavior; people transmit rumors
to fulfill these goals. We apply the literature on motivations in social
behavior to understand the goals that motivate rumor spread. In addi-
tion, we present empirical evidence supportive of these motivational
underpinnings.
This motivational approach is beneficial for at least three reasons.
First, although past research has identified a collection of variables
related to rumor transmission, relatively less attention has been given
to theorizing why these variables predict transmission. Second, a
motivation-based approach helps us connect the rumor literature with
a broader social-psychological literature. For example, by considering
rumor transmission as a self-enhancing activity, we are able to draw
links with social identity literature and consider ways in which rumors
that are derogatory to the outgroup help boost esteem associated with
the ingroup. Third, the core-motivations approach notes a variety of
influences on rumor spread and helps us identify research gaps in the
current literature. For example, the relationship-enhancement func-
tion of rumor transmission has largely been overlooked.
What motivates any social interaction? In general, people interact
with the social world to fulfill one or more of the following three goals:
to act effectively, to build and maintain relationships, and to manage
favorable self-impressions (Cialdini & Trost, 1998; Wood, 1999, 2000).
In serving the goal of acting effectively, social interaction helps people
to acquire a socially validated sense of reality and enables effective
response to, and coping with, the environment. In serving the goal of
building and maintaining relationships, social interaction helps build
and maintain relationships that are vital for continued survival of hu-
mans as social animals. Finally, a more self-serving goal, and one that
may lead to several biases in information processing, is the goal of self-
enhancement. That is, in the context of social interactions, people seek
to affirm their sense of the self and use the social context in various
ways to boost their self-esteem.
In the context of rumor transmission, these goals are represented
in three motivations: fact finding, relationship enhancement, and self-
enhancement. In the following sections, we review the literature on
rumor spread under the rubric of these motivations. In each section
we begin by noting how the basic goal is served in social interaction
and then apply the motivation to the rumor context. Following this, we
discuss antecedents and consequences of these motivations for rumor
transmission. Finally, with the help of results from an empirical study,
we illustrate the role of these motivations in rumor transmission
intentions.
 Psychological Factors in Rumor Spread 71

Fact-Finding
Motivation

The goal of acting effectively leads to fact finding, often in a social

context. Effective response and coping with the physical and social
environment is necessary for our survival. This goal manifests itself in
a variety of human needs such as control (Bandura, 2001); competence
(White, 1959); and understanding (Fiske, 2003; Fiske, Lin, & Neuberg,
1999). To act effectively, we need valid and accurate knowledge of our
circumstances. This goal thus motivates the search for—and spread
of—accurate information that is essential for effective response to the
situation. When motivated by the goal of acting effectively, we seek
accurate information and we evaluate that information using available
strategies, such as comparing it with knowledge that we already have
or evaluating the credibility of its source. This information is often
available in our social context (Fiske et al., 1999). Sometimes we explic-
itly seek information from other people. Other times, we obtain infor-
mation in passive or covert ways, such as following the example of
others, conforming to group norms, or complying with persuasion
attempts, and generally are influenced to varying degrees by informa-
tion from the social milieu (Cialdini & Trost, 1998). Indeed, in the act
of establishing objective reality, validation of our experiences from
significant others figures largely (Hardin & Higgins, 1996).

FACT-FINDING MOTIVATION
AND RUMOR SPREAD
Of the variables that have been identified in the rumor literature as
precursors of rumor spread, the following reflect most the fact-finding
motivation: uncertainty, importance, lack of control, and anxiety (we
discuss belief under relationship-enhancement motivation). Uncertainty
is defined as a psychological state of doubt about what current events
mean or what future events are likely to occur (DiFonzo & Bordia,
1998). Uncertainty about issues of personal importance engenders feel-
ings of lack of control and anxiety. For example, not knowing the
precise nature and consequences of organizational restructuring and
its consequences for one's job (i.e., uncertainty about a topic of high
importance) leads to feelings of lack of control about how to prepare
for or cope with the change and may lead to considerable anxiety
among employees (Blake & Mouton, 1983; Hunsaker & Coombs, 1988;
Mirvis, 1985). People are motivated to reduce uncertainty and anxiety
and restore a sense of control over their circumstances (Ashford &
72 R U M O R P S Y C H O L O G Y

Black, 1996; Berger, 1987; Berger & Bradac, 1982); that is, a mix of
uncertainty, importance, lack of control, and anxiety generates a need
to know. In the absence of news from formal channels (e.g., manage-
ment, civil administration, or news media), people turn to informal
networks (e.g., office grapevine, friends, social groups) for information.
The informal interpretation arising out of this collective process be-
comes a rumor.
The role of uncertainty and ambiguity in rumor spread was noted in
some of the early theorizing on rumor spread. Belgion stated, "rumour
depends upon uncertainty" (1939, p. 12). Prasad (1935) claimed that
situations "of an uncommon and unfamiliar type" (p. 5) lead to rumors.
G. W. Allport and Postman (1947b) similarly noted that rumor spread
is directly proportional to ambiguity multiplied by the importance of
the topic. Caplow (1947) observed that rumors frequently increased
with uncertainty. Festinger and colleagues (1948) pointed out that
rumors pertain to issues that are shrouded in cognitive unclarity.
Schachter and Burdick (1955) demonstrated the effect of uncertainty
in a field study. They planted a rumor in a girls' preparatory school
and then exposed some students to a staged event aimed at creating
uncertainty. Rumors spread in this high-uncertainty group nearly twice
as much as among students who were not exposed to the staged event
(low-uncertainty group). In a similar way, uncertainty was positively
related to transmission among users of a suburban transit system that
was facing disruption from strike action (Esposito, 1986/1987).
As noted in chapter 1 (this volume), sociological work on rumors
has emphasized the role of collective sense making in situations of
uncertainty (Shibutani, 1966). In the absence of information from
formal channels explaining ambiguous events, group members may
engage in a collective problem-solving process; that is, group members
share and evaluate information that explains the ambiguous situation.
In a study of group problem solving that accompanies rumor spread,
we analyzed discussions of rumors on the Internet (Bordia & DiFonzo,
2004). A content analysis of over 280 rumor-related postings on
Internet discussion groups revealed that a large proportion of the inter-
action was devoted to fact finding. Group members sought and shared
information germane to the rumor, evaluated the information, and
made judgments about the plausibility of the rumor. In sum, the
rumor interaction served an uncertainty reduction and sense making
function.
The role of topical importance (also referred to as outcome-relevant
involvement; Rosnow, 1991) in rumor spread has also been demon-
strated empirically. Rosnow, Esposito, and Gibney (1988) studied
rumor transmission in the aftermath of a murder on a university cam-
pus. The proportion of people who reported transmitting rumors related
to the murders was twice as high on the university campus where the
 Psychological Factors in Rumor Spread 73

murder took place (high importance) as that of another university

campus in the same city (low importance). In a similar way, in Esposito's
(1986/1987) study of suburban transit system users, importance of the
rumor content was positively related to transmission. These findings
imply that outcome relevant involvement is necessary for people to
become involved in rumor spread. People are uncertain about many
issues, but they pursue uncertainty reduction only on those topics that
have personal relevance or threaten the goal of acting effectively.
Psychological control has also been implicated in rumor transmis-
sion in a couple of ways. As noted, situations marked by uncertainty
concerning important topics are also likely to elicit feelings of lack of
control (Ashford & Black, 1996; Bordia, Hobman, Jones, Gallois, &
Callan, 2004) and lead to anxiety. In an analysis of circumstances
leading to rumors heard by corporate public relations officers, we found
that the effect of uncertainty on anxiety was mediated by feelings of
lack of control (DiFonzo & Bordia, 2002a). With our colleagues (Bordia,
Hunt, Paulsen, Tourish, & DiFonzo, 2004), we similarly found that the
relation between uncertainty and psychological strain during a large-
scale organizational change was mediated, at least in part, through
feelings of lack of control.
Second, rumors have been conceived as attempts to acquire second-
ary control over situations marked by low primary control (Bordia &
DiFonzo, 2002; Walker, 1996; Walker & Elaine, 1991). Primary control
refers to action-focused coping responses: managing or changing one's
circumstances in the desired direction. Faced with an ambiguous re-
structuring, for example, I may increase my productivity to increase
my chances of job change survival. When primary control is not possi-
ble, people may resort to secondary control strategies; these involve
emotion-focused coping responses such as lowering expectations, pre-
dicting the worst to avoid disappointment, attributing events to chance,
and attempting to understand the meaning of events and accept them
(Rothbaum, Weisz, & Snyder, 1982; Walker, 1996; Walker & Elaine,
1991). Faced with a potential layoff, for example, I may eagerly partici-
pate in water cooler rumor discussions about why and when these
layoffs will occur. Walker (1996) content-analyzed 200 rumors col-
lected on a university campus to see if they contained themes related
to secondary control. All rumors contained secondary control themes,
with the most common being interpretive secondary control (explain-
ing the meaning of events; n = 93), followed by predictive secondary
control (predicting future events; n = 69). These findings provide some
evidence for the idea that rumors represent ways of regaining control
over uncertain and uncontrollable circumstances. However, this aspect
of rumor is underresearched and requires greater empirical attention.
Finally, both trait (dispositional) and state (situational) anxiety are
related to rumor spread (Anthony, 1973, 1992; Rosnow, 1991). Jaeger,
74 R U M O R P S Y C H O L O G Y

Anthony, and Rosnow (1980) measured trait anxiety of students in a

class with the Taylor Manifest Anxiety Scale, then planted a rumor
that some students had been caught smoking marijuana during exams.
High trait anxiety students spread this rumor more than did low trait
anxiety students. Walker and Beckerle (1987) manipulated state anxi-
ety. Participants were invited to help the instructor improve exam
questions by answering a mock exam. While the participants waited
to do this, confederates planted two rumors by saying, "I heard two
rumors about what is actually going on here" (p. 356). One rumor
(anxiety enhancing) claimed that the real purpose of the study was to
catch cheaters; the other rumor (anxiety alleviating) claimed that the
professor wanted to assess the effectiveness of classroom demonstra-
tions. Participants in the high state anxiety condition then watched
what was apparently a live event, but was in fact a videotape, in which
a confederate was being rigorously questioned; those in the low state
anxiety condition merely read some test questions. After a while, an-
other confederate asked the participant, "What is going on here today?"
The dependent variable was the number of prompts needed for the
participant to report a rumor. Participants in the high state anxiety
condition on average needed only about two prompts to repeat a rumor
whereas the participants in the low state anxiety condition needed up
to four prompts.1
It is interesting that Walker and Beckerle (1987) also found that
participants in the high state anxiety condition were marginally less
accurate in reporting the rumor than were participants in the low
anxiety condition. This finding raises the possibility that although un-
certainty about topics of high importance may induce the fact-finding
motivation, anxiety may hinder fact-finding accuracy. This possibility
agrees with research on the role of anxiety in information processing.
Anxiety colors one's view of the world by directing cognitive resources
to search for threatening stimuli (Calvo & Castillo, 1997) and overem-
phasize the threat from ambiguous signals (MacLeod & I. L. Cohen,
1993). An anxious individual may find the doom and gloom of dread
rumors congruent with his or her emotional state and may favor the
interpretation provided by these rumors. This idea may also explain
why, in general, dread rumors are more widespread than are wish
rumors (S. R. Kelley, 2004; R. H. Knapp, 1944). The anxiety associated
with uncertain events of high personal significance biases rumor con-
tent toward negative and threatening interpretations. These rumors

1
The findings should be interpreted with caution as the audiovisual stimuli in the
manipulation of anxiety may have also made the rumors more salient and increased
recall as compared with the control condition. Future research could replicate this study
with more comparable stimuli.
 Psychological Factors in Rumor Spread 75

also help justify the anxiety being experienced (Festinger, 1957). We

return to a fuller exploration of these ideas in chapter 7 (this volume)
when we discuss how rumors become more—and less—accurate.

Relationship-
Enhancement
Motivation

Social relationships are vital to the survival of humans and serve,

enable, or enhance important goals such as mate selection, child rear-
ing, food gathering, protection from predators, and survival during
resource scarcity (Baumeister & Leary, 1995; Kenrick et al., 2002). The
goal of building and maintaining relationships is reflected in a variety
of everyday behaviors in the domain of interpersonal communication,
such as seeking the company of others, compliance with norms and
persuasive appeals to please others (Cialdini & Trost, 1998), impression
management and self-presentation tactics such as ingratiation (Leary,
1995); avoidance of conversational topics that might create conflict
(Knobloch & Carpenter-Theune, 2004) or hurt the conversation part-
ner (Rawlins, 1983); use of humorous or awe-inspiring topics to hold
others' attention (Guerin, 2003); and even deception (DePaulo &
Kashy, 1998). For example, lesser and Rosen (1975) demonstrated
that people are reluctant to transmit bad news for fear it will generate
negative affect in the recipients and the recipient may evaluate them
negatively. This withholding of negative information has been labeled
the minimize unpleasant messages (MUM) effect.
The context of the relationship influences the manner in which
this goal is expressed. For example, in short-term relationships or in
early stages of relationship formation, people are keen on making a
good impression and pleasing others. In such circumstances, honesty
may be sacrificed for other relationship-oriented goals such as generat-
ing liking or positive affect in others, and so information that helps
attract or hold the attention of a desirable audience may be passed
on without much care for its authenticity. However, in longer-term
relationships with greater personal involvement, accurate information
sharing is likely to be emphasized (Stevens & Fiske, 1995). In addition,
the anthropological literature on storytelling notes how narrative
content is manipulated to further interpersonal (attracting attention
of the target); group (maintaining status differences); or intergroup
76 R U M O R P S Y C H O L O G Y

(excluding people by making stories meaningful to only a few in the

audience) goals (Sugiyama, 1996).

RELATIONSHIP-ENHANCEMENT MOTIVATION
AND RUMOR SPREAD
Existing literature on rumors has paid little explicit attention to the
relationship-enhancement goal2 (cf. Guerin, 2003). One exception is
research on the spread of positive versus negative rumors. Applying
the MUM effect (lesser & Rosen, 1975), Kamins, Folkes, and Perner
(1997) predicted that people will refrain from passing negative rumors
(compared with positive rumors), fearing that these rumors might gen-
erate negative affect in the recipient—a relationship-enhancing goal.
Their results supported this prediction: Participants were more willing
to transmit a positive rumor (that the ranking of their business school
will rise) than a negative rumor (that the ranking of their business
school will fall). However, transmitting negative rumors may also serve
the cause of maintaining relationships—especially close, long-term
relationships—because rumors predicting negative events can often
help people cope with these events. For instance, Weenig, Groenen-
boom, and Wilke (2001) found that negative information was more
likely to be transmitted to a friend when the information was considered
useful in averting harmful consequences for the friend. Close friends
are expected to share helpful information, even if it is negative. Thus,
the context of the relationship can influence the sharing of negative
or positive rumors.
The relationship-enhancement motivation has been invoked to
explain the finding that belief in a rumor is positively related with
transmission (Rosnow et al., 1988). A reputation as a credible and
trustworthy source of information is vital for acceptance in social net-
works (Caplow, 1947; Guerin, 2003; Stevens & Fiske, 1995). One way
to ensure such a reputation is to share information that is accurate and
believable. Belief in rumor was strongly related to transmission among
university faculty experiencing a labor dispute (Rosnow et al., 1986);
among students at a university campus where a student had been
murdered (Rosnow et al., 1988) or had suddenly died of meningitis
(Pezzo & Beckstead, 2006); and among students at a campus in the
vicinity of the Washington, B.C., sniper murders (Pezzo & Beckstead,
2006). Kimmel and Keefer (1991) noted that disbelief in the rumor

2
At a mundane level, of course, following the social rules of communication (Hig-
gins, 1981), individuals are unlikely to share a rumor with someone we know is not
even remotely interested in the topic of the rumor (e.g., people might avoid conversation
about work over an intimate dinner with a date).
 Psychological Factors in Rumor Spread 77

was a reason why rumors about AIDS were not transmitted. And
Rosnow (1991) reported a moderate mean effect size (r = .30) between
belief and transmission.
Possessing and sharing valued information is also a way to heighten
status and prestige in the view of others in one's social network (Brock,
1968; Fromkin, 1972; Lynn, 1991); one's higher status often leads
others to like, value, or respect one more. During times of uncertainty
and threat (e.g., war or natural disasters), information is even more
valuable. In their eagerness to further their social standing, people may
unwittingly pass on rumors. G. W. Allport and Postman (1947b) provide
an example of such dissemination. A few members of an Italian Ameri-
can community owned transistor radios during World War II. In their
desire to demonstrate their status of being "in the know," they unwit-
tingly spread propaganda being broadcast on the radio. Being consid-
ered in the know by spreading rumors is thus one way to increase the
liking, value, and respect that other people have for oneself.3

Self-Enhancement
Motivation

The self-enhancement goal refers to the need to feel good about oneself.
People seek to maintain a positive self-image and engage in cognitions
that bolster self-esteem (Kunda, 1999; Steele, 1988). Self-enhancement
biasing effects on thinking and judgment are well known (Kunda,
1999). For example, most people think they are better than average
on a variety of skills and abilities: When people are led to believe
that certain traits (e.g., extraversion) or skills (e.g., driving) are more
desirable, they rate themselves more highly on these traits and are
more likely to recall instances in which they behaved in this desirable
manner (Kunda, 1990). A second example of self-enhancing bias: One's
own characteristics, as compared with those of others, are considered
more likely to lead to desirable outcomes (e.g., leadership position,
happy marriage, etc.; Kunda, 1987). A third example: People are gener-
ally more resistant to information that is counter to an attitude they
hold when it threatens their self-image; however, when people are
feeling good about themselves (on unrelated issues), they are less

3
Being considered in the know may at the same time be self-enhancing; further
research is needed to tease apart these motivations (see chap. 10, this volume).
78 R U M O R P S Y C H O L O G Y

defensive in evaluating counterattitudinal information and more easily

persuaded (G. L. Cohen, Aronson, & Steele, 2000).
These self-enhancing biases extend to groups people belong to or
identify with. A part of our sense of self derives from the groups we
belong to, such as demographic (e.g., ethnicity, gender), professional
(e.g., academics, engineers), or social (e.g., hobby clubs, student associa-
tions, honor societies) groups. We derive meaning and a sense of self-
worth by identifying with groups we consider high in status and prestige
(Hogg & Abrams, 1988). This identification may bias our judgments so
as to favor the group we are part of (the ingroup), often at a cost to
the group we are not linked to (the outgroup). For example, we are
more critical of negative information about the ingroup than we are
of positive information about the ingroup (Dietz-Uhler, 1999). In a
similar way, we readily adopt positive—but not negative—information
about the ingroup as representative; in addition, we attribute outgroup
success to external causes and outgroup failure to stable, internal fea-
tures (Beal, Ruscher, & Schnake, 2001; Pettigrew, 1979). Negative
reactions to outgroups are particularly strong when ingroup interests
are threatened (Bobo & Kluegel, 1993) and outgroup derogation is a
way we boost our self-esteem (Fein & Spencer, 1997). For example,
negative stereotypes about Blacks (e.g., "Blacks have low intelligence")
were believed more by Whites who perceived threat from Blacks than
by Whites who perceived no such threat (Quist & Resendez, 2002).
Negative stereotypes of outgroups serve the need to maintain favorable
ingroup evaluations (Goodwin, Operario, & Fiske, 1998) or justify unfa-
vorable outcomes for the outgroup (Quist & Resendez, 2002).

SELF-ENHANCEMENT MOTIVATION AND

RUMOR SPREAD
The self-enhancing motivation can range from blatant and conscious
spread of rumors for personal gain to unconscious selection and spread
of self-serving rumors. In this section, we begin with an acknowledge-
ment of conscious and motivated spread of rumors for self-gain and
then discuss the role of self-enhancement in the cognitive processing
of rumors.
Although this aspect is underemphasized, the rumor literature has
shown that rumors may originate or spread as part of a motivated
and malevolent propaganda tactic (G. W. Airport & Postman, 1947b;
DiFonzo & Bordia, in press; Kapferer, 1987/1990; Rosnow, 2001; Sinha,
1955; P. A. Turner, 1993). When much is at stake (e.g., election to
political office or product sales in a highly competitive consumer mar-
ket), rumors afford tangible gain. In wartime, rumors are intentionally
spread to demoralize enemy troops (G. W. Airport & Postman, 1947b;
 Psychological Factors in Rumor Spread 79

Mihanovic, Jukic, & Milas, 1994). Sales agents steer consumers away
from rival products and toward their own by using rumors, and rumors
become the grist for the word-of-mouth advertising mill (Kapferer,
1987/1990; P. A. Turner, 1993). During elections, rumors (or "whisper-
ing campaigns"; G. W. Allport & Postman, 1947b, p. 184) sully the
reputation of the opposing candidates (Kapferer, 1987/1990; Sinha,
1952). In elections in the Indian state of Madhya Pradesh, the Congress
Party spread false allegations that the Indian prime minister, who be-
longs to the rival Bharatiya Janata Party, eats beef. Cows are sacred to
Hindus and beef eating is abhorrent. The allegations created enough
difficulty that the prime minister had to announce dramatically: "I
would rather die than eat beef" (Verma, 2003). And Fine (2005) has
suggested that malicious rumors are a way for deceivers to spread lies
without actually facing the malevolent nature of their act; they take
comfort in the uncertainty of rumor.
In spite of the variety of contexts in which conscious rumor spread
may occur, it has generally been ignored as a variable of interest (but
see Pratkanis & Aronson, 1991, 2001, as notable exceptions). One
reason could be because conscious and malicious rumor spread reveals
a repugnant characteristic of human nature and forms "the ugly under-
belly of interpersonal life" (Leary, 1995, p. 9). Also, malicious intent
in rumor birth cannot sustain rumor growth (Horowitz, 2001). For a
rumor to take hold, it must find fertile ground and catch the imagination
of several people; that is, it needs to serve one or more of the motivations
in rumor spread. A rumor may originate or even acquire an occasional
fillip from mischievous agents, but to widely circulate, the rumor needs
to fulfill fact-finding, relationship-enhancing, or self-enhancing moti-
vations. Nonetheless, we maintain that the use of rumors by conscious
manipulators of public sentiments needs greater attention (see DiFonzo
& Bordia, in press), particularly in the context of public education
regarding rumors and their effects. Knowledge of the use of rumors
by propagandists may bring about watchfulness among people and
prevent their falling prey to the machinations of rumor peddlers (i.e.,
people must learn to distrust those who sow distrust).
A second way in which self-enhancement motivates rumor spread
is as follows: Spreading rumors may boost one's self-esteem by boosting
one's social identity. As discussed earlier, groups prefer interpretations
that portray the ingroup in a favorable light and are derogatory of the
outgroup. It is not surprising then that rumors derogating the outgroup
are much more prevalent than are rumors negatively portraying the
ingroup. For example, in a study conducted by R. H. Knapp (1944)
during World War II, Reader's Digest magazine readers were invited to
submit rumors they had heard. Of the 1,089 rumors collected, over
60% of the rumors were derogatory of some social group and were
80 R U M O R P S Y C H O L O G Y

labeled wedge-driving rumors because they contributed to inter-

group discord.
Self-enhancement motivations are also operative when rumors are
spread to rationalize self-enhancing attitudes. When evaluating rumors,
people are more likely to favor those that support or justify existing
prejudices. Unjustified prejudice is a distasteful notion, but the preju-
dice acquires legitimacy through rumors that are consonant with the
prejudiced viewpoint (Van Dijk, 1987, p. 62). In other words, rumors
aid in the process of "justification construction" (Kunda, 1990, p. 483)
for the desired belief structures. G. W. Allport and Postman (1947b)
noted that "rumor rationalizes while it relieves" (p. 37), implying that
rumors derogatory of the outgroup vent the prejudice and also justify
it. For example, rumors among Whites that Blacks are overly aggressive
and violent justify discrimination. Another dramatic illustration of the
self-enhancement bias occurs when the same ambiguous event is inter-
preted in different but self-serving ways by different groups of people.
The race of the perpetrator of violent acts gets transposed, depending
on the race of the narrator of the rumor (Black perpetrator when the
rumor circulates in the White community, but White perpetrator when
the rumor circulates in the Black community; Rosnow, 2001). Whereas
American sources attributed the destruction of a hotel in Baghdad
during the Iraq war to a bomb planted by terrorists, the rumor among
local Iraqis was that the destruction was caused by a wayward American
missile (Shanker, 2004).

Contextual
Determinants of
Rumor-Spread
Motivations

In general, situations that threaten a goal will activate the associated

motivation. When people's ability to cope effectively with the environ-
ment is threatened, fact finding is activated: We search for accurate
information. For example, on hearing about merger talks between their
organization and a rival company, employees are likely to seek accurate
information about the consequences to the structure of their organiza-
tion as a result of the merger. In a similar way, when the self or ingroup
is threatened, self-enhancement motivation may be activated and we
are likely to favor information that makes us feel good about ourselves.
In such situations, we are less likely to be concerned about the accuracy
 Psychological Factors in Rumor Spread 81

of the rumor and more concerned about its self-enhancing value. At

other times, characteristics of the person or the situation may make a
particular goal (and the associated motivation) more salient. For exam-
ple, a young adult keen on developing romantic relationships is likely
to be driven by relationship-enhancement motivations and is likely
to share a rumor that helps attract or hold the attention of his or
her audience.
Sometimes, motivations work in tandem. In particular, the
relationship-enhancement and self-enhancement motivations are eas-
ily served by the same rumor. Spreading a rumor that sheds positive
light on the ingroup is likely to be self-enhancing (it will boost my self-
esteem) and will also help form a good impression in other ingroup
members (it will enhance our relationship). In a similar way, rumors
that are derogatory of the outgroup are self-enhancing and also aid
in the development of solidarity and cohesion (Kakar, 2005) among
ingroup members. At other times, however, the motivations may com-
pete. For example, when the recipient is an outgroup member, sharing
a positive ingroup rumor may be self-enhancing but may not serve
relationship-enhancement goals as well as sharing a positive outgroup
rumor would.
Consider the case of a college student who hears the rumor that
in the next year's ranking of colleges by U.S. News & World Report, his
or her college is going to be ranked lower than it currently is. The
rumor may create anxiety and arouse the need to ascertain its validity
(i.e., the fact-finding motivation) and consequently be shared with an
ingroup member. However, this rumor will reflect badly on the person
and is therefore unlikely to be transmitted to an outgroup member.
However, a positive rumor (that the ranking will increase next year)
is self-enhancing and is therefore more likely to be transmitted to the
outgroup member. What if the rumor happens to be about the outgroup
(i.e., rise or fall in ranking of a rival college)? A positive rumor (rise
in ranking) will likely be transmitted to a friend from the outgroup as
this rumor will serve relationship enhancement. However, a negative
rumor about the outgroup (fall in ranking) is more likely to be transmit-
ted to an ingroup member as it will be self-enhancing.
To test some of these ideas, we conducted a study in which we
manipulated rumor valence (negative or positive), target (rumor about
the ingroup vs. about the outgroup), and recipient (an acquaintance
from the ingroup vs. the outgroup). We adapted the procedure for this
study from Kamins et al. (1997) but extended it in one important way:
We clearly identified the recipient of the rumor as belonging to the
ingroup or outgroup. The rumor referred to the rise or fall in ranking
of a school in the next U.S. News & World Report ranking. The participants
were undergraduate students from Rochester Institute of Technology
82 R U M O R P S Y C H O L O G Y

(KIT). Thus, RIT formed the ingroup. The outgroup was another school
from the same city, the University of Rochester (UofR). The study had
a 2 (valence: rise or fall in ranking) x 2 (target: rumor about RIT or
UofR) x 2 (recipient: acquaintance from RIT or UofR) design. The
participants were given a hypothetical scenario in which a friend (from
RIT) tells them: "I don't know if this is true, but I heard that RIT's [or
UofR's] ranking in U.S. News & World Report will fall [or rise] by four
positions next year."
They then were asked to imagine that they happen to meet an
acquaintance, another student from RIT [or UofR]. With the following
two items, the participants were asked how likely they were to share
the rumor with this acquaintance (i.e., likelihood of transmission):
"How likely are you to tell the student the statement about the U.S.
News & World Report ranking?" and "Is this statement something that
you would mention to other—similar—acquaintances?" The two items
were strongly related (r = .90) and were combined as a measure of
likelihood of transmission. To explore the role of motivations underly-
ing the intention to transmit, we also measured fact-finding motivation
("rate the extent to which you were motivated to figure out whether
or not this statement was true or false" and "rate the extent to which
you were motivated by a desire to see if the acquaintance knew if it
was true or false"; r = .53), relationship-enhancement motivation ("rate
how you think the acquaintance will feel about you if you tell him or
her this statement" and "rate how you think the acquaintance's level
of respect for you will change if you tell him or her this"; r = .65), and
self-enhancement motivation ("you were motivated to say something
that would make yourself feel good, not bad" and "you were motivated
to create a pleasant mood—and not an unpleasant mood—in yourself";
r = .71). All ratings were on a 9-point scale.
Figures 3.1 to 3.3 present the results for fact-finding, relationship-
enhancing, and self-enhancing motivations. Overall, the patterns of
motivations were as expected. The fact-finding motivation was highest
in the case of a negative rumor about the ingroup when the recipient
was also from the ingroup (see Figure 3.1). The relationship-
enhancement motivation was highest for the condition in which the
rumor was positive and about the outgroup, and the recipient was
an outgroup member (see Figure 3.2). Finally, the self-enhancement
motivation was highest when the rumor was positive and about the
ingroup, and the recipient was from the outgroup (see Figure 3.3).
Motivational strength varied as a function of rumor valence, rumor
target, and rumor recipient.
The results for the likelihood of transmission are presented in
Figure 3.4. Several interesting patterns are visible. First, rumors about
the ingroup were more likely to be transmitted to ingroup recipients. In
 Psychological Factors in Rumor Spread 83

FIGURE 3.1

• Rumor Recipient Ingroup Member (RIT)

6- n Rumor Recipient Outgroup Member (Uof R)
5.36
4.90

Negative Rumor Positive Rumor Negative Rumor Positive Rumor

About the Ingroup (RIT) About the Outgroup (Uof R)

Fact-finding motivation in conditions of positive versus negative rumor about the

ingroup or the outgroup when the recipient is from the ingroup or the outgroup. RIT =
Rochester Institute of Technology; UofR = University of Rochester.

general, the ingroup audience was the preferred target in all conditions,
except when the rumor was positive in valence and about the outgroup.
Second, contrary to the MUM effect, when the rumor was about the
ingroup and the recipient was a member of the ingroup, both positive
and negative rumors were equally likely to be transmitted. In other
words, participants did not hesitate to transmit negative rumor to an
ingroup recipient. We expected the fact-finding motivation to underlie
this effect. To test this idea, we conducted a mediation analysis that
tested the effect of rumor recipient (ingroup vs. outgroup) on likelihood
of transmission, when the rumor was negative and about the ingroup.
We predicted that participants were more likely to transmit a negative
rumor about the ingroup to ingroup recipients (as compared with out-
group recipients) because they wanted to know if the rumor was true.
Our prediction was partly supported: The effect of rumor recipient on
likelihood of transmission was partially mediated by the fact-finding
motivation.4

4
The standardized regression weight of the relationship between rumor recipient
and the likelihood of transmission dropped from -.57 to -.41 after the mediator was
84 RUMOR P S Y C H O L O G Y

Rumor Recipient Ingroup Member (RIT)

Rumor Recipient Outgroup Member (UofR)
5.46

Negative Rumor Positive Rumor Negative Rumor Positive Rumor

About the Ingroup (RIT) About the Outgroup (UofR)

Relationship-enhancing motivation in conditions of positive versus negative rumor

about the ingroup or the outgroup when the recipient is from the ingroup or the
outgroup. RIT = Rochester Institute of Technology; UofR = University of Rochester.

Third, when the outgroup was the recipient of the rumor, the
MUM effect did operate; that is, positive rumors (compared with nega-
tive rumors) were more likely to be transmitted to outgroup recipients.
We expected the relationship-enhancement motivation to underlie this
effect: That is, positive rumors (compared with negative rumors) were
transmitted to outgroup members in the hope that the positive rumors
would generate liking for the narrator. Once again, we tested the medi-
ating effect of the relationship-enhancement motivation on the effect
of valence (positive vs. negative) on the likelihood of transmission to
an outgroup member. Results supported our prediction.5

controlled for; however, it remained significant. The Sobel test (R. M. Baron & Kenny,
1986) for the indirect effect of the independent variable through the mediator was
significant (Z = -.205; p = .04).
5
The effect of valence on likelihood of transmission (.31) became nonsignificant
(.17) when relationship-enhancement motivation was in the equation. Moreover, the
Sobel test for the indirect effect through the mediator was significant (Z = 2.69; p = .007).
 Psychological Factors in Rumor Spread 85

FIGURE 3,3

• Rumor Recipient Ingroup Member (RIT)

n Rumor Recipient Outgroup Member (UofR)

Negative Rumor Positive Rumor Negative Rumor Positive Rumor

About the Ingroup (RIT) About the Outgroup (UofR)

Self-enhancement motivation in conditions of positive versus negative rumor about

the ingroup or the outgroup when the recipient is from the ingroup or the outgroup.
RIT = Rochester Institute of Technology; UofR = University of Rochester.

Finally, we took a look at conditions in which motivations might

compete. Consider the situation in which the outgroup member is the
recipient of the rumor. Positive rumors about the ingroup might be
more likely to be communicated to this outgroup member for self-
enhancement reasons. However, relationship-enhancement motiva-
tion would be better served by the transmission of a positive rumor
about the outgroup. The patterns of means for the likelihood of trans-
mission, self-enhancement motivation, and relationship-enhancement
motivation for the two conditions (positive rumor about the outgroup
vs. the ingroup) when the recipient is in the outgroup condition are
shown in Figure 3.5. As the figure shows, the self- and relationship-
enhancement patterns go in opposite directions. However, likelihood
of transmission is in the same direction as relationship-enhancement,
which suggests that, in the context presented in the current study,
relationship-enhancement motivation was more influential. To test this
idea we conducted a mediation analysis; relationship-enhancement
motivation did indeed mediate the effects of rumor target (about
86 RUMOR PSYCHOLOGY

o • Rumor Recipient Ingroup Member (RIT)

'55 _
to 6 a Rumor Recipient Outgroup Member (UofR)
I
in 5.07 5.08
4.93
4.77

Negative Rumor Positive Rumor Negative Rumor Positive Rumor

About the Ingroup (RIT) About the Outgroup (UofR)

Likelihood of positive versus negative rumor transmission about the ingroup or the
outgroup when the recipient is from the ingroup or the outgroup. RIT = Rochester
Institute of Technology; UofR = University of Rochester.

ingroup vs. outgroup) on likelihood of transmission to an outgroup

member.6

Conclusion

Rumors serve several goals. At times they are part of a search for
valid information. Other times rumor interactions help in forming or
strengthening relationships. Still other times, rumors assuage a threat-
ened sense of self-worth or prop a prejudicial viewpoint by derogating
an outgroup. The strength and influence of a motivation depend on
several contextual features, including characteristics of the narrator,

6
The relationship between rumor target (ingroup vs. outgroup) and likelihood of
transmission became nonsignificant after the relationship-enhancement motivation was
controlled for. Also, the indirect effect was marginally significant (Sobel test Z = 1.89;
p = .051).
 Psychological Factors in Rumor Spread 87

• Self-Enhancement
Relationship Enhancement
6- Likelihood of Transmission
O)

'oc•s
5-

4-

About Ingroup About Outgroup

Positive Rumor
Self-enhancement motivation, relationship-enhancement motivation, and likelihood of
transmission to an outgroup member of positive rumor about the ingroup versus
the outgroup.

the recipient, the relationship between them, the content of the rumor,
and so on. In this chapter, we have described the origins of the motiva-
tions in rumor transmission and reviewed rumor literature pertaining
to each motivation. We also presented some empirical evidence regard-
ing the role of these motivations in transmission intention.
We began by considering conceptual and descriptive issues related
to rumor. In this chapter we provided a motivational framework for
understanding rumor transmission. Next we turn to processes underly-
ing belief in rumor.
 Factors Associated With
Belief in Rumor

In 2003, All Karim and Hussein All, both residents of

Fallujah, Iraq, were quite angry. Karim had heard that
"US President George W. Bush wants the US Army to
kick everyone out of their homes [and] force them to
live in tents" (Slackman, 2003, p. Fl); furthermore,
Fallujah would become a military camp. Ali, a coffee
shop owner, heard that the United States plans to
install a king to rule Iraq. Other Fallujians heard
equally inflammatory negative rumors about the
United States: American soldiers were spying on
Muslim women with night-vision goggles and handing
out pornography to children. To Americans, rumors
such as these seem strange, yet to Fallujah residents in
2003, they were believable.

umors of all shapes and sizes often enter by way of the

R auditory canal, exit through the oral cavity, and make an

important stop along the way: the cerebrum. These anatomi-
cal locations conveniently symbolize three broad steps
thought to be involved in the life of a rumor (DiFonzo,
Bordia, & Rosnow, 1994,-Rosnow, 1988,1991). In the gener-
ation stage, participants become motivated to receive rumor
in an attempt to relieve anxiety or dispel uncertainty. In
the transmission period, rumors are communicated through
social interchanges. Between these periods is the evaluation
phase, in which a rumor's likelihood of being true is assessed.

89
90 R U M O R P S Y C H O L O G Y

This assessment has implications for the life of any rumor. As noted
in the previous chapter, people are more inclined to pass along a rumor
they believe is true than one they believe is false (Rosnow, Yost, &
Esposito, 1986). It has been theorized that disinclination to share a
rumor perceived as false stems from a relationship-enhancement mo-
tive: penalties will be exacted from any false messenger. If the rumor
describes a pipe dream (R. H. Knapp, 1944) the messenger risks becom-
ing a source of disappointment. If the rumor excites fears, the messenger
becomes a stimulus for undue anxiety. Either way, the reputation of
one who raises false hopes or triggers false alarms is tarnished. There-
fore, relationship-enhancement reasons sometimes coincide with fact-
finding motivations, and people are desirous of accurately evaluating
the rumor's credibility or veracity.
Although inclined to discern the truth, people are sometimes noto-
riously bad at the task. False or fabulous rumors often enjoy widespread
acceptance. Consider these examples: Some Iraqis believe the United
States plans to install a king in Iraq. The false rumor that the Procter
& Gamble Corporation tithed to the Church of Satan generated 15,000
calls per month to the besieged corporation as well as a threat of product
boycotts (Austin & Brumfield, 1991; Blumenfeld, 1991; Cato, 1982;
Marty, 1982); many people believed this rumor. And more than one
third of a sample of African American church members believed the
rumor that "the AIDS virus was produced in a germ warfare laboratory
to be a form of genocide against Black people"; another third were
"unsure" about it ("Black Beliefs," 1995, p. Bl).

Probabilistic Mental
Models: Using Cues to
Assess Authenticity

It is easy to see how people might impute credence to a plausible rumor,

but widespread belief in such incredible ones as these begs explanation.
In broader terms, how do individuals engaged in social exchange infer
authenticity of rumors? Because a thorough literature search (as de-
scribed later) yielded neither an explicit theory pertaining to belief in
rumors nor a review of this topic, we examined the literature on belief
in comparable entities. A relevant theory was proposed by Gigerenzer,
Hoffrage, and Kleinbolting (1991; cf. Day, 1986). These authors pro-
posed probabilistic mental models (PMMs) to explain how people arrive
at judgments of confidence in statements of general knowledge. Confi-
 Factors Associated With Belief in Rumor 91

dence judgments are subjective estimates of the probability that a proposi-

tion is true, and are conceptually similar to strength of belief ratings
in a rumor.
PMM theory is a form of Egon Brunswik's lens model of judgment
(1952). Brunswik's lens model posits that people infer judgments on
the basis of cues. We judge a person's age, for example, on the basis
of cues such as presence or absence of facial wrinkles, hair, and age
marks. These cues form the lens through which we perceive distal
attributes (e.g., age). In PMM theory, cues are variables through which
we infer a statement's veracity. For example, recipients of a persuasive
message on the effectiveness of pesticides are more persuaded by a
chemist than a chemistry student (Rajecki, 1990); here the cue is source
expertise. In a similar way, persons may express high confidence in
the statement "Buffalo has over a million people" presumably because
Buffalo has a football team in the National Football League (NFL) and
populations in NFL cities usually exceed 1 million; here the cue is the
presence or absence of an NFL team (Gigerenzer et al., 1991).
Certain cues may work well in some situations and poorly in others
(Gigerenzer et al., 1991). If the front page of the newspaper states
"Dewey Wins," then one may (incorrectly) infer that Dewey, and not
Truman, has won the 1948 U.S. Presidential election; the cue here is
"I read it in the newspaper." Of course, not everything one reads in
the newspaper is true; journalists and pollsters sometimes err. The
extent to which they are in error will affect how often the use of the
"read it in the newspaper" cue will result in a flawed inference.
Within the general structure of PMM theory, the question of how
rumor veracity is assessed may be framed as follows: What cues are
used by recipients of rumor to assess its truthfulness? We address this
question by first reviewing and meta-analyzing literature considering
factors associated with belief in rumors. We set forth these factors in
propositions about what elements of a rumor lead to its acceptance,
and then we interpret these elements as cues that recipients use to
assess veracity. We then present results from a set of interviews con-
ducted with brokers in which these cues were explored.

Literature Search and

Meta-Analytic Methods

We began the search by generating literature from 20 major academic

electronic indexes. In light of rumor's conceptual ambiguity (see
chap. 1, this volume), the following search terms were used: rumor*,
92 R U M O R P S Y C H O L O G Y

rumour*, grapevine, bruited about, gossip*, hearsay, whispering campaign*,

mudslinging, scandal, and scuttlebutt (S. D. Knapp, 1993; asterisks indicate
any letter combination, for example, rumor* will search for rumors,
rumoring, and rumored). All articles relevant to the topic of rumor
were retained and the reference sections of each were culled to obtain
additional studies. This procedure produced over 170 manuscripts on
the topic of rumor, 37 of which bore directly on the subject of belief
in rumor. Where sufficient statistical information was reported, prod-
uct-moment correlations (rs) were calculated for effect sizes (see
B. Mullen, 1989, chap. 6; R. Rosenthal & Rosnow, 1991, chap. 22;
Shadish & Haddock, 1994) for factors thought to influence belief. Eigh-
teen separate effect size rs were computed from 10 independent investi-
gations, all of which measured belief.1

Consistency With
Attitudes

A substantial quantity of descriptive and correlational research points

toward belief in rumors that are consistent with the hearer's currently
held attitudes. These conclusions are intuitively appealing. If a rumor
supports or accords with what the actor already holds to be true, it is
plausible that the actor would assign greater credence to it. These
conclusions also dovetail with a long line of attitude research findings
showing the effects of attitude on judgment. For example, attitudes
affect judgments of information that is relevant to the attitude
(Pratkanis & Greenwald, 1989). In a similar way, partisan
identification—an attitude—has been shown to bias the evaluation of
arguments on a political issue (Lord, Lepper, & Ross, 1979), and judg-
ment of the veracity of the Clinton-Lewinsky scandal allegations was
strongly predicted by one's prior Clinton approval rating (Fischle,
2000). It is a short step to therefore state that attitudes also affect belief
judgments about rumors. Before we consider the evidence for this,
however, it is important to clarify how the term attitude is used here.
Although attitudes have been conceptualized as having three sepa-
rate components (affect, cognition, and behavior; see Rajecki, 1990,

'For example, Jaeger, Anthony, and Rosnow (1980) asked subjects to rate their
belief in a planted rumor on a 4-point scale ranging from 0 (no-confidence) to 3 (complete
confidence). The number of belief scale alternatives varied from two (e.g., believe vs. do-
not-believe; Goggins, 1979) to 11 (e.g., 0 = no confidence in truth of rumor to 11 = strong
confidence in truth of rumor, Esposito, 1986/1987; Kimmel &• Keefer, 1991).
 Factors Associated With Belief in Rumor 93

chap. 2), few attitude component distinctions were explicitly made in

the rumor literature. However, affective or cognitive attitudinal compo-
nents were implicitly indicated from the context. For example, the
"Eleanor Club" rumors we referred to in chapter 1 (this volume) are
consistent with anti-African-American and anti-Roosevelt sentiment
and the fear of inversion of status, which are affective in nature (G. W.
Airport & Postman, 1947b, p. 175). However, those who considered
wartime rationing programs to be unfair or unnecessary believed World
War II rumors of waste and special privilege more than did those
who considered them fair or necessary (F. H. Airport & Lepkin, 1945);
consideration is a cognitive activity. Perhaps this approach stemmed
directly from consistency theory; even the cognitive of Festinger's (1957)
cognitive dissonance seems to include affect as well (Rajecki, 1990). Fur-
thermore, although G. W. Allport and Postman (1947b) differentiated
between the "intellectual and emotional contexts" in which rumors are
transmitted, they maintained that these contexts are "indistinguishably
blended" (p. 100). In this discussion, then, attitude is conceptualized
primarily in terms of its cognitive and affective components.2
Descriptive evidence consisted mainly of case studies and collec-
tions of rumors. These rumors are presented in Table 4.1, along with
the attitudes with which the rumors are presumed to be consistent.
Perhaps the most exemplary set of speculations comes from the bench-
mark work on prejudicial rumors by Gordon W. Allport and Leo Post-
man (1947a, 1947b). These researchers asserted that racial stereotypes
and animosities lay behind the prevalence of racist rumors. In a similar
way, Robert H. Knapp speculated that widespread inconvenience and
political or prejudicial sentiment produced the disproportionate per-
centage (65.9%) of World War II wedge-driving rumors in the United
States (e.g., "the Catholics in America are trying to evade the draft,"
Knapp, 1944, p. 24).
Some attitudes are so deeply rooted in a group that they are more
rightly considered as culture. One of R. H. Knapp's (1944) characteris-
tics of a thriving rumor was that it accorded with the cultural traditions
of the group. Prasad (1935) noted that rumors gained currency when
they were consistent with local superstitions. In a later (1950) study,
he characterized rumor as a picture completion task, with attitude
(largely made up of cultural traditions) painting the picture. London
and London (1975) similarly speculated that widespread false variations
of a rumor that President Nixon had stolen a teacup from Chairman
Mao during his visit to China sprang from elements of Chinese national

2
In chapter 10 (this volume) we propose how future researchers may fruitfully
distinguish these components.
94 RUMOR PSYCHOLOGY

Descriptive Evidence for Belief in Attitude-Consistent Rumors

Reference Rumor(s) Attitude with which rumor is consistent
Abalakina-Paap Conspiracy rumors (e.g., AIDS is a Authoritarianism, anomie, and
& Stephan plot to wipe out minority groups, powerlessness.
(1999) the government is storing
information in computer files to
against its citizens, and
fluoridating drinking water will
hurt people).
G. W. Allport & Rumors of incidents of "the White supremacy. For example,
Postman (1947a, Negro's criminal and disloyal "What can we do with a black
1947b) tendencies... Negro stupidity, man who is disloyal, criminal,
gullibility, and laziness, [and] .. clownish, stupid, menacing, and
myriad tales of Negro sexual immoral—except to keep him in
aggression" (1947b, p. 177). his place just as we are now
doing?" (1947b, p. 177).
G. W. Allport & "Eleanor Club"3 rumors circulating Anti-African American and anti-
Postman (1947a, in southern states in 1943; the Roosevelt sentiment and fear of
1947b) theme of these rumors was that inversion of status. "White rumor
large numbers of African American spreaders find their feelings of
women were banding together to economic and social insecurity to
reverse the social order (e.g., "A some extent explained and
White woman was away for a relieved by these stories" (1947b,
while, and when she returned she p. 176).
found her colored maid sitting at
her dresser combing her hair with
her comb," 1947b, p. 175).
G. W. Allport & Rumors of incidents of Black men Self-vindication with regard to
Postman (1947a, having sexual relations with sexual morality. For example,
1947b) unwilling White women. "Even if we are not blameless, yet
his misdeeds (as recounted in
rumor) are overt and worse than
ours. We need not feel guilt at our
peccadilloes" (1947b, p. 179).
G. W. Allport & World War II wedge-driving Religious and racial prejudice.
Postman (1947b) rumors (e.g., "The Jews get the
cushiest jobs in the Army").
Ambrosini (1983) A psychiatric hospital patient "had Internal aggressive and sexual
Case 1 sexual relations with a man on the conflicts. The rumors allowed
ward and ... would suffocate patients and psychologists to
herself while in the hospital" rationalize their own sexual and
(p. 76). aggressive impulses.

(continued)
 Factors Associated With Belief in Rumor 95

T A B L E 4,1 (Continued)
Descriptive Evidence for Belief in Attitude-Consistent Rumors

Reference Rumor(s) Attitude with which rumor is consistent

Ambrosini (1983) Recurrent rumors that "the clinical Internal sense of separation and
Case 2 psychology department of a anxiety. The rumor meshed well
prestigious university was about to with new graduate students' sense
be disbanded" (p. 77). of disillusionment and depression
after their initial elation at being
accepted into the program (e.g.,
"This program isn't so great after
all").
Ambrosini (1983) A therapy group's co-leader Internal sense of separation and
Case 3 (a psychiatrist) was "imminently loss. The rumor meshed well with
departing from the group" (p. 78). group members' belief that the
group would eventually disband.
Festinger et al. A researcher who was stimulating Opposition to the project and anti-
(1948) community activities in a tenant's communist sentiment.
association was an avowed
communist.
Hicks (1990) Satanic rumor panics that are The "satanic model that has
believed and propagated by emerged at conferences of
police. therapists, police, and cult
survivors... obtained 'through
hours of networking' between the
various professionals" (p. 383).
Jung (1959) Recurrent rumors of flying saucers. The collective archetypes of the
self, the masculine-feminine
antithesis, the antithesis of what is
"above and below," the antithesis
of "unity and quaternity," and the
antithesis concerning the
"enigmatic higher world and the
ordinary human world" (p. 16).
Kelley (2004) Iraqi hostility rumors. Ethnic and political sentiment
(e.g., Anti-U.S./Coalition, Anti-
Israel, Anti-Sunni, Anti-Saddam,
Anti-Shiite, Anti-Turk attitudes).
R. H. Knapp Wedge-driving rumors. Religious and racial prejudice.
(1944)
Knopf (1975) Rumors circulating during 20th- Racial hostilities.
century race riots.

(continued)
96 RUMOR P S Y C H O L O G Y

: :
;t (ton tiniffif>'
Descriptive Evidence for Belief in Attitude-Consistent Rumors
Reference Rumor(s) Attitude with which rumor is consistent
London & Variations of a widespread rumor Elements of Chinese national
London (1975) that President Nixon visited China character: "superiority and face
because he needed Chairman saving." For example, "Nixon
Mao's praise to be re-elected, that visited China because he needed
he had stolen a valuable "nine- Mao's support" (China is superior);
dragon" teacup and denied it, and "we retrieved the cup"
that a magician had deftly or (superiority) "without
cleverly retrieved the cup from embarrassing him" (face saving).
Nixon.
Nkpa (1977) Rumors of mass poisoning Biafran religious belief that those
collected in Biafra in a period killed by poison would not be
encompassing the Nigerian civil reincarnated. Nkpa speculated that
war (1967-1970). this belief was responsible for the
low percentage of poison rumors
in comparison with other types of
rumors.
Prasad (1950) Rumors collected after a massive Myths and legends. For example,
Indian landslide that contained the "rain of blood is mentioned in
recurrent themes such as the the great epics of Valmiki,
disappearance of rivers, eclipses Ramayana, and Mahabharata"
and planets, the destruction of a (p. 140).
capital town, a rain of blood,
strange animals, and the end of
the world.
M. Rosenthal Racial rumor circulating in Detroit, Racial stereotypes. When repeated
(1971) Michigan, in 1967 through 1968, in the White community, the gang
falsely alleged that a child using a was said to be Black and the victim
shopping mall lavatory was White. When told in the Black
castrated by a gang of teenage community, the gang was said to
boys. be White and the victim Black.b

(continued)

character. Finally, Jung (1959) ventured beyond culture as the mold

in which some rumors are cast; he viewed rumor generation as a kind
of projective mechanism giving evidence of presuppositions that are
handed down from our ancestors through a collective unconscious.
Correlational evidence for attitudes that bear specifically on the
rumor in question is consistent with these ideas. Studies in which belief
in rumor was related to attitudes are presented in Table 4.2. One
investigation in this set was performed by Floyd H. Allport and Milton
Lepkin (1945) during World War II. A rationing program caused short-
ages of many commodities and inconvenienced many citizens to one
 Factors Associated With Belief in Rumor 97

TAB.ti 4.1"|f
Descriptive Evidence for Belief in Attitude-Consistent Rumors

Reference Rumor(s) Attitude with which rumor is consistent

Scheper-Hughes Brazilian shantytown children are Class conflict attitudes, such as

(1990) routinely abducted and mutilated that the poor are given
by American or Japanese agents to inadequate medical treatment and
obtain profitable body organs for are unfairly treated, and that
transplants. children of the poor are prey to an
unscrupulous "baby trade."
Victor (1989) Rumor panic about dangerous Dissatisfaction with changing
satanic cults alleging ritual societal values, as well as fear of
meetings, animal sacrifice, blood the effect of "dangerous new
drinking, and the planned influences" (p. 39) on children.
kidnapping of a blond-haired,
blue-eyed virgin.

Note. aNamed after Eleanor Roosevelt, "Eleanor Club" rumors fused anti-Roosevelt and anti-African American
themes. bBelief was not measured. We assume that the presence of a rumor indicates some belief as well. For ex-
ample, the absence (in the White community) of the rumor that the teenage castrators were White is taken to
mean that this rumor was not believed in that community.

degree or another. Of concern to government officials were certain

wedge-driving rumors that commodities were being wasted or were
freely accessible to those of higher status (e.g., "A certain prominent
government official has 3 cars and a large underground storage tank
filled with gasoline for his own use," p. 5). Because these allegations
were either false or unsubstantiated, Allport and Lepkin wondered
what caused people to believe in these rumors of waste and special
privilege. Among other variables, they measured strength of belief in
each of 12 rumors as well as attitudes that Allport and Lepkin posited
might cause belief.
The relations in Table 4.2 are arranged in order of ascending
strength within each study; correlations vary widely. The strongest
associations between belief and attitudes are for those attitudes that
are highly specific to the rumors. For example, belief in rumors of
waste and special privilege was more highly correlated with unfavorable
attitudes toward the rationing program than with adverse attitudes
toward the Roosevelt administration (F. H. Allport & Lepkin, 1945).
That is, the attitude that "the rationing program is unfair or unneces-
sary" is highly specific to such rumors as "A certain prominent govern-
ment official has 3 cars and a large underground storage tank filled
with gasoline for his own use." We note that the Pearson-r correlation
for this relationship was weak to moderate (.19). A similar attitude-
specificity relationship increase occurred in the DeClerque et al. (1986)
98 RUMOR PSYCHOLOGY

_Q)
-Q
(C
Q.. o
. Q. .
rr, m

-a 01
>• s
i
5J
l Is
i ll
iil! i
lI
Consisten Rumo

c
I T3 lii
C Q.
_c ro a)1
0) -
i01
tt °3
|
5 t
CD O
O o J-
H-

= dl
i_ 01
(D <U
ID
O
+s
_re
eo (M Q. CTi
— —
 Factors Associated With Belief in Rumor 99

o
(N en _
Lr J3
01 II
00 -,
g " 01 <
oS
II fM ra
E € O £ j;
c OJ OJ M-
I"- II o
T <-» fM •B 11 <u
.2 * CD ro
Q CD 4-1 C

l/i -r?; Cn to
fD C QJ

o
i O -Q
O _ +•• 'F
c
0 •o
0) S g»S e
o "S S * |
o
Q.
Q.
a
o
£ £ o3 l||||
Q. "5 *~ O
C 4- w-*~
Q.TO ^
O

P C»|^
(jl O i- tn <U

OJ
2 1- * & ? ^
•2 -c <1) 2 O
|
I *
.V
>
m u "- t»
T3 .E "5 •§ -5
-8 C « = a, c o
•g __. 2 -Q "
ID ^r 0) -n

5 ro
58 o 3 o = < 01 2
i/l <! . CD CU
•?i .
>, cp
"
'•C '>
+.; (D
'>O 6w O 01
w <-
<( -c 1/1 a. 01 +j

1C
00
kQ
00
S £ tf
OJ a> 41 O) r- u> -*-1
t/)
..
i/i 3 S-g
m

^
se

Cl 3 Dl D
c (U c If 5
o > ~m o "<5
E E > .^y,
in
>,
rtj
Q. S
'-4-»
(D 'S. «
<b 0)
(D
Ol
jj>> aj £
gth o

O
00
u O
00
S
m 3 S >•-OJ
O ^
'i/i
ai 2 a~
4-»
Ol L_
4-»
CT •^
•B fc ^
CU
dJ C 01 ro £ O
c 0 U C 0 U
"^ u 01 u 01 -^ &
si

"
ds

01 o. fo c j;
Ol
c "S
£
c 2 S fD O..B-5
o o
^
%<u ^

J5 s JS ^ o "i I" 3
.£ 3 ,_
3 <o c D OJ c ^ E M- t>n n
u £ _*
^~ 4-» n
t;.C
+^ ro
_* -Q D -n ra *-
ro i-
u t/t HI 'u t/) Ol £ £13
k_
O .2 ? k_ c
O .2 §
1-2 o g ^
D.~ O
s; I i
causes

causes

'+3
E a. E '4-1
D Q.
belief
Wote.

D >,
i_ k_ >, -D S{ i3
01 Ol 5
< LLJ < LU ts -° V,
o .Si,
100 R U M O R P S Y C H O L O G Y

study: Having the attitude that the pill is "very harmful" was similarly
associated (r = .17) with the rumor circulating among Egyptians that
oral contraceptive use causes weakness. The corresponding increase in
belief rate in a Binomial Effect Size Display (R. Rosenthal & Rosnow,
1991, pp. 280-283) between people possessing attitudes consistent with
the rumor and those not possessing such attitudes is 17%—an impor-
tant practical effect. 3
These results suggest an association between belief in rumor and
attitudes specifically consistent with that belief (Proposition 1). No
studies offered evidence that attitudes caused belief in particular ru-
mors, although almost all implied such a conclusion. This conclusion
is plausible, given that the attitudes described and measured seem more
stable than do the beliefs in rumor (implying temporal precedence). It
is also plausible, however, that rumors influence attitudes. This associa-
tion was hypothesized in only the DeClerque et al. (1986) study; these
researchers posited that an Egyptian rumor that the birth control pill
causes weakness (loss of physical strength) had led to negative percep-
tions of the pill.

Rumor Source
Credibility

A long line of research in persuasion has shown that source credibility

is linked to attitude formation and change (Hovland & Weiss, 1951;
Petty & Cacioppo, 1981). In a similar way, it is likely that rumors heard
from more credible, authoritative, or well-positioned sources (e.g., lay-
off rumors heard from a manager) would tend to be more strongly
believed than would rumors heard from less credible, authoritative, or
well-positioned sources (e.g., layoff rumors from a fellow worker). And
indeed, the evidence supported this idea. To begin with, researchers
assembling collections of rumors have consistently observed that attri-

3
An r-value of .17 is typically considered weak to moderate (J. Cohen, 1988), but
even weak effect sizes may indicate important effects (R. Rosenthal & Rosnow, 1991;
see also R. Rosenthal, 1991, pp. 132-136). For example, the effect of aspirin on reducing
heart attacks is r = .034 (Steering Committee of the Physicians Health Study Research
Group, as cited in R. Rosenthal & Rosnow, 1991). Although this correlation accounts
for only 0.11 % of the variance, the result is of great practical importance, as the increase
in the percentage of lives saved by taking one aspirin per day versus not taking aspirin
is 3.4% (i.e., over three lives per hundred treated). With regard to DeClerque's investiga-
tion, an effect size r of .17 accounts for "only" 2.89% of the variance, yet a 17% increase
in the belief rate is hardly inconsequential.
 Factors Associated With Belief in Rumor 101

bution to a credible source is part of the typical rumor formulation

(e.g., Bird, 1979; Blake, McFaul, & Porter, 1974; R. H. Knapp, 1944).
After compiling and cataloging 1,089 World War II rumors in Septem-
ber 1942, Knapp (1944) concluded that "successful" (flourishing) ru-
mors tended to be attributed to authoritative sources: "From whatever
humble beginning a rumor may spring, it is soon attributed to a high
authoritative source. This gives the rumor . . . the appearance of verac-
ity" (p. 30). In a similar way, folklorist Donald Airport Bird (1979)
amassed and analyzed the stylistic characteristics of an immense collec-
tion of rumors from printed sources, archives, and student question-
naires, and found that rumors are frequently ascribed to a high-status
community member or someone "in the know" (chap. 2). Blake et al.
(1974) posited and found evidence that, under normal conditions (i.e.,
low collective excitement), rumors gained plausibility by the addition
of an authoritative citation and a media source from which the rumor
was supposedly heard. When questioned about a rumor that a hatchet
murderer "would kill several coeds at a small college in Ohio" (p. 7),
94% of respondents attributed the source of the rumor to Jeane Dixon4
and 51 % stated this prediction was published in newspaper or radio
media (Blake et al., 1974).
The only direct investigation measuring rumor source credibility
and rumor belief fully accords with these conclusions. Porter (1984)
found that rumor communicator credibility was moderately to strongly
related to belief in negative rumors about birth control in the Dominican
Republic (r = .40; pp. 27, 29). Taken all together, these findings imply
an association between credibility of source and belief in rumor
(Proposition 2).

Hearing Repeatedly

Persuasive circumstantial evidence points toward the proposition that

repetition (the number of times a rumor is heard) is associated with
belief. A series of experiments investigated the confidence placed in
responses to questions that are repeated. The original study in this
series was by Hasher, Goldstein, and Toppino (1977). On three separate
occasions, they asked subjects to answer true-false items of general
knowledge and rate their confidence in their responses. Some of the

"We suppose that in the domain of prophetic statements, reference to self-

proclaimed seer Jeane Dixon may be considered an authoritative citation, although in
fact her record of accurate prophecies is abysmal (Donnelly, 1983).
102 R U M O R P S Y C H O L O G Y

items were repeated over the three occasions. The researchers found
that subjects placed increasing confidence in their responses to repeated
items only and concluded that "If people are told something often
enough, they'll believe it" (p. 112). Later research replicated this illusory
truth effect (Bacon, 1979; Begg, Anas, & Farinacci, 1992; Boehm, 1994).
If being presented with and responding to a true-false item increases
confidence in the response to that item, then perhaps hearing and
evaluating a rumor repeatedly may increase belief in the rumor.
Direct evidence is also supportive. Weinberg et al. (1980) found
that undergraduates tended to receive a rumor about a campus hit-
and-run accident two or three times before passing it on. F. H. Allport
and Lepkin (1945) found that those who had previously heard the
rumors presented in their survey believed them more than did those
who had not previously heard them.5 These findings mesh well with
R. H. Knapp's (1944) intuition regarding the effect of rumors on public
opinion: "Once rumors are current, they have a way of carrying the
public with them. Somehow, the more a rumor is told, the greater is
its plausibility" (p. 27). In other words, rumors, when repeated, seem
more believable. All together, the available evidence implies a third
proposition: There is an association between repetition (the number
of times a rumor is heard) and belief (Proposition 3).

Rumor Rebuttal

In light of rumor's long-standing ignominy as a saboteur of morale and

reputation (summary in Bird, 1979, chap. 1), it is not surprising that
a number of studies have explored how to effectively combat rumor.
Effect sizes for studies that have investigated how rebuttal affects belief
in a rumor are presented in Table 4.3. Earlier studies used a between-
groups design in which all subjects received rumor, but some also
then received a rebuttal of the rumor; belief was then measured and
compared between the rebuttal and no-rebuttal groups. Our own inves-
tigations used a within-groups design in which belief in the rumor
would be measured both prior to and after the rebuttal. Results from
eight rebuttal studies6 were meta-analytically combined (R. Rosenthal
&Rosnow, 1991,p. 505), yielding an ]V-weightedravg. = .33—amoderate

5
f(5216) = 21.5, Puaiied = 1.29E-98, r = .29; see Table 4.2, note b.
'The studies in Table 4.3 were used, excluding F. H. Allport and Lepkin's (1945)
results because they violate the independence-of-observation assumption (see Table 4.2,
Note b).
 Factors Associated With Belief in Rumor 103

effect. 7 Caution is in order, however, because these eight effect sizes

were quite heterogeneous,8 although in light of the variation in rebuttal
sources and study methodology, heterogeneity is to be expected.
It is important to note that we have been discussing the effect of
rebuttal overall and have not considered moderating variables; these
may be quite important in enhancing or negating the effect of rebuttal
on belief (indeed, the heterogeneity of our meta-analysis points toward
this). For example, Koller (1992) found mixed evidence for an interac-
tion between method of rebuttal (positive advertising, denial, or no-
comments) and prior knowledge of the rumor. Iyer and Debevec (1991)
found that conciliatory rebuttals were more effective than were in-
flammatory ones. They also found that credibility of the rebuttal source
interacted with the origin of the rumor (positive, neutral, or negative
stakeholder). In our own research we found that appropriate (Bordia,
DiFonzo, & Travers, 1998), honest (Bordia, DiFonzo, & Schulz, 2000),
and high-credibility (Bordia, DiFonzo, Haines, & Chaseling, 2005) re-
buttal sources enhance rebuttal effectiveness. We review in more detail
the moderators of denial effectiveness—on belief and other outcomes—
in chapter 9 (this volume). Overall, however, these results lead to a
fourth proposition, that there is an inverse association between belief
in rumor and the presence of a rebuttal (Proposition 4).

Other Factors

A small potpourri of other factors associated with belief appeared in

our review of the literature. Kimmel and Keefer (1991) reported a
jumbo-sized association (r = .70) between importance and belief; they
posited importance to be a mediating variable between belief and trans-
mission. F. H. Airport and Lepkin (1945) reported a nonsignificant
relation between education and belief, but Kapferer (1989) reported a
moderate negative (r = -.32) relationship. We turn now to recent
empirical evidence of factors associated with belief.

7
The corresponding decrease in belief rate in a Binomial Effect Size Display
(R. Rosenthal & Rosnow, 1991, pp. 280-283) caused by rebuttal is 33%. With R. Ro-
senthal's (1979) file drawer analysis, 341 unpublished studies averaging null results
would have to exist to bring the overall puaiw = .0003 (based on ravg. = 3.42) to a "just
significant" p - .05 level.
8 2
X (7) = 46.44, p = 7.19E-8 (see R. Rosenthal & Rosnow, 1991, pp. 500-501).
104 RUMOR P S Y C H O L O G Y

o
o
r\l
q
rvi 00
en
10
n
ST
r-.

CTI
o
r-
I I
LLJ
o 00
q q
ivi
n II
a Q. Q. Q.
J3 ^ *
en in oo ID
in oo
II II "

01

01
01
Q.
01 03
Q. <U

ai
_c
fO
0; ~O
n <U
_ t
t;
OJ

~ 8. 0) O
i=
+j -a
<D v, 01
-Q C °*
O 01 >-
1 §
E ^
"(D 1
- <D > g. II
+J tt
C "^
oj a; .
a,
cn.E I-S 1 -= a. is
01
ro J3
E £
2 § g S -^-g E £
o _
E £
o _
Ol > (D u—
CO 01 X ro o
O .a
1 § II a.^ 2
01
U1 0) LLJ C £• Q. < 'g

o
tt
•O 08
01
cc 01
C .E T3 c
o ra ^ '<TJ
Q. I

1
-Q
01
08 O in
N o
c o ^
O rsi ^T-
o"
N

O <N
in
o
o
or O m
.t *"*
*^ -9-Q. ai
o Q ?§ c

I CD
01
ai
ra S.
_
o ^
CO U
nj
U
 Factors Associated With Belief in Rumor 105

II

ID"

n
(o

o
in

E
i-
73

0^
D .
<" OJ
OJ >

£-0
•£ -a

IS

o. g
|S

irt f^

.£ i?
5^
ss
§s
Lj- ^^
Q g1

II
O JT
03 U
106 R U M O R P S Y C H O L O G Y

Field Interviews
With Brokers

To explore cues used to infer veracity in field settings, we conducted

field interviews with a sample of 10 stockbrokers in Philadelphia,
Princeton, and New York who were willing to be interviewed (DiFonzo,
1994, Study 2). Subjects were contacted via personal acquaintance and
snowball sampling (participants were asked if other brokers they knew
would provide additional interviews). All were employed in several
major investment firms and typically possessed 8 to 12 years of experi-
ence. A critical incident technique (Flanagan, 1954) was used to query
subjects about actual instances of rumors: Participants were asked,
"Think of a rumor that you have heard that is relevant to a security
that is important to you" and "What was the source of the rumor?"
To assess belief, we asked participants, "On a scale of 0 to 10 where
0 means 'no confidence' and 10 means 'complete confidence,' how
confident were you, at about the time you first heard it, that this rumor
was true?" To explore what cues were being used to arrive at that
confidence judgment, we asked participants, "How did you arrive at
that judgment?" To assess the repetition cue, we asked, "From how
many independent sources did you hear this rumor?" To assess the
source credibility cue, we asked the participants to identify the source
of the rumor. We assessed cue validity of the source of the rumor by
asking participants, "In general, what percentage of rumors that you
hear from [the source of this rumor] are true?" A total of 18 separate
rumors were discussed; these are presented in Table 4.4. Although
takeover rumors were most numerous, the set deals with a diverse
range of content areas (e.g., staff changes, crop-planting patterns, cover-
up attempts). Most of the rumors were recent; 64% of the rumors had
been heard within 8 months prior to the interview. Brokers were also
asked if the rumor had proven true; the set was roughly evenly divided
between the three possibilities of true (n = 8), false (n - 5), or uncertain
(« = 5).
Cues used fell into seven fairly distinct categories, three of which
matched our propositions. First, investors used consistency with hear-
er's attitude. For example, one trader inferred some degree of likelihood
to admittedly "radical" and "extreme" rumors that then-U.S. President
Clinton wanted to devalue the dollar (as compared with the yen) so
as to reduce the apparent United States-Japanese trade deficit. Such a
rumor meshed with the broker's attitude that the government likes to
take the "shabby, easy way out" of difficult problems. Second, investors
 00
ID
k.
T3
O)

W»
•o
BO

C
so>

VI
S

01
Field Interview Rumors

u
re

1
1

a.
JS

3 S
ee
GO

0
in

3
O

•8
5

c
3

0
7

TJ
Source of rur

e
I
I

,_
«-
l_

^
i-
_
J
Ol

*~

<u

0)
> E
O

o
*•

I/I
^. 0
President of Company X

u
1. Fit the pattern: Past instances

1 fci
"

i_ O

nj o
~

.E*

<2
o- m
just flew somewhere to

§.i
had occurred in which this

^ .c
3
fix a major problem. president traveled and fixed

13 -:r
Q; ID c
4-< J
=

IN
major problems.

i-
^

i_ O
U. A_,

^ -C
2. Source credibility*1: Source had

JJ
'+-• O ' +-
- 0 0 .

5 12 1/1 o e v, .21
c *• c c
Of ra

x? .y Ji -? .y J{
•»-<

sfci s
a* > u to > u
0 •" 0 o <" 0
nothing to gain by spreading

D 'r:
3
rumor, was trusted, and was
close to the situation.

s
-o

'-
IN
<V
sO
*~
(N

-a
0^

+j
"S

ID
Small-company product Source credibility': Source had Investor associ
p will be purchased by something to gain by spreading with company

:^
vj
_Q
large Company X. rumor. prior to its pul
sale of stock.

«-
^_

^c
V

_Q
Cl

<u
"c

.c
<u

m
4-*
<D

U
c
+-•

-o
(V
ID

i/i
£

n:
o
in

^
s
E

u
u
Company X will obtain Can't rememb

ra
Ol

l_

8g
new contracts. exact source.

i E

,_
*t

IN
*~"

"c
0)

<u
<u
^2

+J
tu
in

ID
ij
m

lj
E

Company X is "frying the

I

Analyst opinion that rumor was Can't rememb

u
books" (using illegal false.
accounting procedures).
f—

IN
IN
^

SS
(N
o
o
m

S
m

John Sculley is leaving Low confidence in source1 (views Dan Dorfman

Spectrum Technologies. Dorfman as "incompetent"). (CNN).

^
ID

ro
p»
in
^

Vince Foster had

ra

Not plausible that White House

l
ID _ro
^^ OJ
*•

"ID
§§
c 2
m 1/1

'ij
m 2:

. c

committed suicide in a staff involved in homicide.

o|

l
,X
a* I o

V rag

lo§a s«
at-"
01 mT ^
m IN +;

£ ^ u

location different than C ^ oj

•= C "a?
"o ra x,

what official accounts

noted (negative
implications for White
House staff).

IN
iZ
V

m
en

E
IN
o

>
£
o

1/1

Fit the pattern: Trading behavior Traders, then

-g- 5
•K =

S1*-
5o
c ^
(U V*
~ fM
of persons associated with the media.

1 §
firm was unusual and made the
Factors Associated With Belief in Rumor

rumor very plausible.

_C
Q
•g
3

8
107
 U)
•a
!5

•o

00
c
s

u.

2
ield Interview Rumors

re
1
1

1
o
„

1"
•s

ac
k.
'is

5
3
3

in
o
5
8E

oe S
M

5
41
m

M
01

3
.21

1
i
i

SX

«-
E
<U
ro
ro

,_
^

S
a;
o

4-1
^

^
o

^
ro
u
Japanese banks are . Fit the pattern: "It explained

a.
1

'
m m

c
•
I/I

ro
•^ro <Lvi

t
ro

?
unusual things 1 saw" (e.g., wP

:>.
Q.
bankrupt and are selling

Q.
f

~
2 ? 'o
C 2 o?

off assets. Ministry of is the yen so high?).

rsi
Finance is helping them . Compatible with attitude that

<"
J3
through this crisis. Clinton administration would 1
Clinton wants to push likely to take a politically easy
yen a lot higher to solution (government likes to
reduce apparent trade take the "shabby, easy way
deficit. out").
108 R U M O R P S Y C H O L O G Y

"O
c

oi

4-1
o
5

c
^

o
-

(N
3

^
There will be massive . Meshes with Clinton policy.

ro

1
1
•• S

i- (N
. There are "other reasons" wh;

••»»
planting of corn over
OirVj- m'
oo .- +,

beans. it is plausible.

.c
k_
JO
"o
rsi

,_
^
in

o
01
O

in
<
*

Checkers drive-in fast- it the pattern: Stock is in indust

£,

u. +j
food restaurant chain lat is consolidating.
will be bought out by
Pepsi-Cola.
f

_Q

T-
OJ
£
rsi

O
^

£o
in
in

fM
<

in
Baltimore Bank Corp. . Fit a trend: Super-regional
will be taken over. commercial banks are buying
small banks.
. Saw increased insider trading.
. Saw price increasing.

IN m *r
c
. Possible merger was featured
a financial report.
1

£
f-

10
ID

in
4

<u
in

in

ro
+j

CO
O

.«

to
0

in
ro

1
1
'01 ff
C
3T oi

c
O ro

§-3

Si <°

O
ro >
< E

c <D
O)
"ro >>
C 0)

§-1
c 'u
M
c
-fcj "TO JS

LJ_ U (/)
P "ro
o E
s *>%
»
s^
0
(N

Ji

,-
U)
2
7

Germantown Savings ow confidence in source1 but fit

ro
u

(U -^
'sQ. •-
"at J2
c c

2
^^

Bank will be taken over. rend: Super-regional commercia

_l fi -D
ianks are buying small banks.
 „

U3
A

(N
rvi

(N
tm

^
Baltimore Bank Corp. 1. Low confidence in source.'

2
•p
"o> E

c
01 -*
£=5
Cl OJ

-c 2
c £
will be taken over.

^_^

•> 01
2. Possible merger was featu

s~ S
£,-"
a financial report.
3. Fit a trend: Super-regional

C
commercial banks are buy

Ol
small banks.

r^
I

^
0)
(N

"c
in

0)

J-»
J2

o
Q
(M

2!
ij
in
^

E
John Sculley will become "Seemed possible to people
CEO of Spectrum familiar with Apple and
Technologies. Spectrum."

^
_r-

l<
•5;

IN
•-

•fc

•"
<U
i5
"""
Q

•V
Q
T

«!

tJ
£
5

^o
m
m
Knew something about comj

•§>

^
C 5

l/l
O .—

O)
that "made rumor seem likel

TJ
—
fw

0 >*

£ ii!^
Qfc c
C Q)
>
>
m -2 e
i-S
^.
•§- fc

H—

M- —i S: ^
W ra ?
Sg.s
! «• -

Jls.f?
D n) ra
>• 01 !r
O O u-
^ C ^

"§j^g>.E
.

0 & .-D

J
^-
I

,-
cd
c3

"v
J3

•o

2c
c
c
o
^
Federal government will

?
Didn't fit the pattern (i.e., Cl

C
S. £
t/i —

C
£

not permit AT&T merger economic mandates).

S
(o ra

«.2 fc
C
in "O ID

"" n C
0 > J2

with McCaw Cellular

ff

< &i
"i _>,
i- OJ To

because of anti-trust
laws.

,-
oS
00

\s
OJ

_Q
in

~2
IN

C
Jr.
£
in

01

01
2
o
Company X's earnings 1. Frequency: testimony of ni

O)

vt

,*£
'

% E
would be 6 cents per brokers.
share when only 1 cent 2. Fit the pattern: Recent

'c
a!
per share was expected. acquisitions of two compa
generated revenues.
3. Credible sources.1

o
in
^
o

Purposely unaware of
rumors and doesn't
trade on rumors because
they are unreliable.
rti

QC
i^
U)

>~
(~vl
Qj

0)
<u
-I-l

£
"P
iu

C
i/l

in
nor /D is composed of broke
_, ;_ ~°

oS'"1
s

u
-,-i

O> C
^

c ._

if\ •—
i_ —

X
f^
_Q
-Q .£

'£ .£
^
-S £

o
M_ "O

^
8 _ -g

E u
"o
O

l/l ^
QJ CJ
^ c

OJ
ro S

-i^
^ c
f",

^
-c ^

<U On
§ -

^ T3
fD

i^
2^
'o a
^ <u

fc 0
i/i in
S£

.? «
^ ro
13 U

JD U
0 ^
Jr OJ
o u

ui

= 5

01
0 T|

°?

confidence) scale, indicates

l-i
S|
i/i • —
u 11
>,

5 £1
§1

2 >,
*1

i S

I c

Ol

So?
3 a £

2 a SB
gth of belief (see text).

". "B. tn
Factors Associated With Belief in Rumor
109
110 R U M O R P S Y C H O L O G Y

used credibility of rumor source as a veracity cue. One investor felt the
rumor source was credible because "he [the source] was close to the
president [of the company]." Most sources were word-of-mouth tips
from market participants less highly positioned. It is not surprising that
published sources, such as the Dick Davis Digest, the Johnson-Smick Report,
and the "Heard on the Street" column, were generally assigned greater
credibility than were nonpublished sources. Source credibility was used
in six cases; a Pearson-r calculated between cue validity and strength
of belief for this small a number of cases can be viewed only as sugges-
tive, but it was very large (r= .88).'Third, there was anecdotal evidence
that frequency of hearing related to belief in the rumor. One investor
was initially doubtful about false rumors implicating White House
involvement in former staff member Vince Foster's suicide; on hearing
the rumor "lots and lots" of times, he adopted a "let's-wait-and-see
attitude." In a similar way, a Pearson-r calculated between number of
sources and strength of belief was moderate but not significantly differ-
ent from zero (r - .24). 10
Fourth, one investor used stakeholder status as a veracity cue. Tips
from tippers with "something to gain by spreading the rumor" were
viewed skeptically; rumors from disinterested sources were viewed
more favorably. Fifth, rumors that "fit the pattern" tended to be per-
ceived as veridical. The pattern referred to here could be a larger trend
exemplified in the rumor. Larger trends exemplified in specific rumors
included the following: super-regional commercial banks are buying
small banks, fast-food industry is consolidating, Clinton economic pol-
icy favors expansion of communications industry, acquisitions of
smaller companies increase revenues, and problems get fixed when
president of Company X flies to problem site. The pattern could also
refer to the configuration of unusual and unexplained events that the
rumor attempts to explain. Instances of such unexplained configura-
tions included unusual trading behavior of persons associated with a
large industrial firm and the unusually high valuation of the yen. Sixth,
investors used consistency with emerging data—in this case, congruent
price changes—to infer veracity. One broker stated that a rumor that
Company X would obtain lucrative contracts "had to be true [because]
the price kept doubling." Other investors typically heard a rumor and
then monitored price changes to verify the rumor. Seventh, brokers
used expert consensus to infer veracity. Rumors "spurred research,"
caused an investor "to talk with the people familiar with Apple and

>,.,,„ = .01, n = 6.
"1p,.lailed = .17, «= 18.
 Factors Associated With Belief in Rumor 111

Spectrum," and spawned "phone calls to my colleagues to see what

they thought." Analysts' opinions were also highly valued.
Finally, it is important to note the influence of cues that changed
over time. Although all cues could, theoretically, change over time,
repetition or congruent price changes seemed prone to change the most
quickly. A rumor could be repeated throughout the course of a trading
period; prices, obviously, can change. In some instances judgments
changed under the influence of these cues.

Summary

Taken as a whole, the literature and our interviews with brokers suggest
that rumors are believed to the extent that they (a) agree with recipi-
ents' attitudes (especially rumor-specific attitudes), (b) come from a
credible source, (c) are heard or read several times, and (d) are not
accompanied by a rebuttal. Cues follow naturally from these proposi-
tions: How well does the rumor accord with the hearer's attitude? How
credible is the source perceived to be? How often has the hearer heard
the rumor? Has the hearer not been exposed to the rumor's rebuttal?
These cues should lead to greater acceptance of the rumor. In addition,
other cues have been suggested, including the extent to which (e) the
denial source has nothing to gain from the rebuttal, (f) the rumor fits
a pattern already in place, (g) the rumor is consistent with emerging
data, and (h) the rumor agrees with expert consensus. Overall, the
results shed light on the process of rumor evaluation and are consistent
with a probabilistic mental model framework. Would-be believers "hear
it on the street" but also listen for cues to guide them in their assessment.
In this chapter we have explored factors associated with belief using
a lens model of judgment. In the next chapter, we use social-cognitive
models to explore how groups make sense using rumors.
Rumor as Sense Making

In mid-1991 a rumor claimed that Prodigy, an Internet

service provider, might be tapping the hard drives of its
subscribers' computers. In early May of that year,
someone posted the following to an online discussion
group: "I've heard that Prodigy subscribers can have
their hard drives tapped when they're logged onto the
service. .. . Has anyone else heard of this?" Over the
next week, the group members discussed the rumor's
plausibility, shared personal experiences and
information (including media coverage on CNN and
The Wall Street Journal; M. W. Miller, 1991), expressed
belief and disbelief, and analyzed the technical details
that might allow Prodigy access to information on the
hard drives. The group discussion vividly demonstrated
the sense making that accompanies the emergence and
spread of rumors.

umors help people to make sense of the world. Rumor does

R
this at at least two levels: individually and collectively. Indi-
vidual sense making involves how people make sense of the
world within their own minds, and has much to do
with individual cognition. Collective sense making has to
do with how people make sense of the world by interacting
with others, and has much to do with group processes. In
this chapter we explore rumor sense making at both levels.
To link rumor and individual sense making, we explore
connections between rumor and explanation theory, causal
attribution, illusory association, and prediction. To link

113
114 R U M O R P S Y C H O L O GY

rumor and collective sense making, we consider the content, functions,

and flow of rumor discussions.

Rumor and Individual

Sense Making

Individuals use rumors to make sense. In this section we explore this

idea by discussing how rumor affects explanation processes and, in
particular, causal attribution; we also explore the significance of these
effects for certain classes of judgments and behaviors. Rumors draw
attention to events, set forth initial explanations, activate knowledge
structures with which to interpret events, and motivate the examina-
tion of an explanation. Stable-cause attribution is especially involved
in these processes. As a consequence, rumors often lead to associations
that are in fact illusory, and rumors also engender systematic biases
in prediction.

RUMORS AND THE EXPLANATION PROCESS

Anderson, Krull, and Weiner (1996; see also Krull & Anderson, 1997)
proposed a model of explanation integrating a number of cognitive
processes that are undoubtedly affected by rumors. In this model, an
event is an occurrence or characteristic for which a person seeks under-
standing or predictive meaning. In the beginning stages, the event must
be noticed and interpreted, and an initial explanation is set forth; these
stages involve relatively automatic processes. A person must first notice
the event and its potential causes; the noticeability of these causes
influences the likelihood that they will be adopted as explanations.
Factors influencing noticeability include novelty, visual dominance,
relevance to one's goals, and temporary or chronic activation of particu-
lar categories (Hilton & Slugoski, 1986). The event is then interpreted.
The event itself and prior expectations (e.g., stereotypes, primed or
chronically accessible categories, contextual cues) are factors affecting
interpretation. These activities are successful if they produce an initial
explanation. The person may then, if so motivated and if not con-
strained by time or cognitive resources, engage in more effortful pro-
cesses that iteratively test and ultimately select a final explanation.
Each stage within Anderson et al.'s (1996) explanation theory
is guided by the activation of knowledge structures, which are mental
representations of an object, event, or construct. Examples include
 Rumor as Sense Making 115

category representations, scripts, procedures, episodic memories, and

understandings of specific persons. Knowledge structures are conceptu-
ally similar to "explanation kernels" (Abelson & Lalljee, 1988), "cogni-
tive structures" (Sedikides & Skowronski, 1991), "causal mental mod-
els" (Jungermann & Thiiring, 1993), and "causal schemas" (H. H.
Kelley, 1973). Knowledge structures must first be activated before they
are used in accomplishing explanation tasks such as noticing and inter-
preting an event and evaluating and selecting a proposed explanation.
Activation of a knowledge structure depends "upon salience, availabil-
ity, or accessibility" or similarity (Anderson & Slusher, 1986, p. 272).
Changing the structure, then, will change the subsequent outcome of
the explanation task and result in the law of cognitive structure activation:
Ambiguous stimuli will be encoded as consistent with the most salient
knowledge structure and thereby affect relevant judgments and behav-
iors (Sedikides & Skowronski, 1991).
Rumors influence this explanation process at several junctures.
First, rumors make people notice an event. False rumors of Continental
Bank's impending bankruptcy because of large loans to underdevel-
oped nations (Koenig, 1985) informed people of these considerable
debts. Second, rumors offer an initial explanation for noticed events.
The false rumor that the soft drink Tropical Fantasy contained a
substance that would make Black men sterile provided an explanation
as to why the soda was sold primarily in minority neighborhoods
(Freedman, 1991). Rumors often perform both of these functions
synonymously. False rumors that Procter & Gamble Corporation do-
nates to the Church of Satan may have drawn attention to celestial
components in the company's logo at the time and encouraged their
interpretation as Satanic (Marty, 1982). Third, rumors activate knowl-
edge structures that direct the search for further information. For
example, rumors of an impending loss in company profits may compel
a review of next-day stock-price changes (DiFonzo & Bordia, 1997,
2002b). Fourth, rumors often convey anxiety-provoking information
that may sustain motivation to prolong the examination of explana-
tions. The rumor that Prodigy (an Internet service provider) taps users'
hard drives for marketing purposes was proposed, evaluated, and
debated on an electronic bulletin board discussion; it was apparent
that anxiety propelled the discussion over several days (Bordia &
Rosnow, 1998). Thus, rumors may materially affect the explanation
process at several stages.
At each stage, rumors inform or activate knowledge structures
that determine the course of explanation throughout the process. One
knowledge structure that seems central in this regard as it pertains to
rumor is causal attribution.
116 R U M O R P S YC H O L O GY

RUMOR AND STABLE CAUSE ATTRIBUTIONS

Rumors help people make sense of the world because they are often
embedded with ready-made attributions of stable causes for events.
Some causal attributions are derived from ready-made explanations
available from knowledge structures about why Person A slapped Per-
son B) or nonperson (Pop Rocks candy exploded in Mikey's stomach)
events occur. Put another way, causal attributions are explanations
about why events occur.1 In this discussion then, the term explanation
is used synonymously for causal attribution.
The question of whether and how often people engage in effortful
causal analyses to arrive at explanations has received attention (Fiske
& Taylor, 1991; E. R. Smith, 1994). The explanation process proceeds
differently depending on whether or not one has knowledge of a perti-
nent and specific domain (Hilton & Slugoski, 1986). This prior knowl-
edge may automatically yield explanations, dictate the search for fur-
ther information (Trope & Thompson, 1997), or suggest particular
explanations for evaluation (Anderson et al., 1996; Hilton & Slugoski,
1986). If no knowledge is present, then simple rules of causality are
used (e.g., those of H. H. Kelley, 1973, analysis of variance model;
cf. Einhorn & Hogarth, 1986). An alternative that has received little
attention, however, is that people may simply adopt a ready-made
explanation that becomes available to them in a social milieu. The
explanation may come to them, as it were, prepackaged. As Fiske and
Taylor (1991) put it, "While one may argue that social contact is a
vehicle for filling out missing information in an attributional search . . .
it is also evident that sometimes the motive is simply to acquire a ready-
made causal explanation from someone who is better informed" (p. 63).
We suggest here that rumor may be the quintessential ready-made
explanation in the attributional process.
Many rumors are obviously causal attributions, that is, they are
explanations of why a person or nonperson event occurred. For
example, "I heard that Manny was being laid off because he was
working too quickly on the assembly line and making the rest of us
look bad." Now consider an example of a nonperson event explanation:
"Goodyear stock plummeted today; I heard it was because Goodyear
profits are down." These rumors offer explanations for phenomena.
Some rumors do not explain, of course; they simply offer information:

1
Although attribution theorists have traditionally been concerned about how people
interpret their own or others' actions (B. Harris & Harvey, 1981; Heider, 1958; Jones &
Davis, 1965), the focus of attribution has broadened to include causation in general
(E. R. Smith, 1994; cf. Antaki & Fielding, 1981).
 Rumor as Sense Making 117

"I heard that Manager X is resigning, the plant is being shut down,
Prodigy taps hard drives, Michael Jordan is returning to basketball,
and Tropical Fantasy causes sterility in Black men." However, even
these information statements tend to possess an explanatory flavor
in that they are causal attributions in the process of being constructed
or they explain one's own feelings (Festinger, 1957). Here is an
example of the former: The rumor "I heard that Manager X is resigning"
is quickly replied to with "Really? That's interesting! I wonder why.
Perhaps he wants to start his own business." The rumor quickly
evolves into a causal explanation: "Manager X is resigning because
he wants to start his own business." Here is an example of a rumor
explaining one's own feelings: "I feel anxious and suspicious [these
feelings are unstated]; it must be because Prodigy [a large corporation]
is tapping into my hard drive!" In a similar way, as we noted in
chapter 4 (this volume), racist rumors can explain one's own feelings
of being threatened: "I feel threatened; it must be because "an [out-
group] man castrated an [ingroup] boy in a shopping mall lavatory
and left him in a pool of blood" (M. Rosenthal, 1971). These observa-
tions are not new (although they are more fully elaborated here);
G. W. Allport and Postman (1947b) stated that rumors contain causal
attributions: "In ordinary rumor we find a marked tendency for the
agent to attribute causes to events, motives to characters, a raison d'etre
to the episode in question" [italics in original] (p. 121).
An important aspect of causal attribution is how stable versus unsta-
ble the cause of an event is judged to be (Anderson et al., 1996; Weiner,
1985). A stable cause, such as one's disposition, is relatively permanent.
An unstable cause, such as chance, is relatively temporary. A fruitful
line of research into depression, for example, has discovered attribu-
tional differences between individuals with depression and those with-
out depression (Seligman, Abramson, Semmel, & von Baeyer, 1979).
Stable-cause attributions in the face of a failure event ("I got a D on
the midterm because I'm stupid") are part of a depressive and pessimistic
explanatory style; however, unstable-cause attributions ("I got a D
because I didn't study") form part of an optimistic explanatory style
(Struthers, Menec, Schonwetter, & Perry, 1996).
We speculate that most of the causal attributions conveyed by
rumors are stable in character. That rumors contain stable-cause expla-
nations can first be inferred from Heider's (1958) observation that
people prefer stable causal explanations because they need to see the
world as understandable and predictable. As discussed in chapter 1
(this volume), rumors fulfill this need for understanding and predict-
ability. We reason that rumors wouldn't be very successful at enhancing
understanding and predictability if they didn't possess stable-cause
118 R U M O R P S YC H O L O GY

attributions. Second, a perusal of rumors collected from a variety of

investigations suggests that causal explanations embedded in rumors
are almost always stable (e.g., dispositional, unchanging), rather than
unstable (e.g., random, chance-like), in character. Racist rumors among
Whites often included the notion that African American men possessed
excessive (stable) sexual and aggressive desires (G. W. Airport & Post-
man, 1947b). Rumors about government—such as that government
officials had unfair access to rationed commodities during World
War II—proposed that this was so because of (stable) negative disposi-
tional qualities such as greed (F. H. Allport & Lepkin, 1945). Rumors
of downsizing because of (stable) declining profits (DiFonzo & Bordia,
1998), rumors of CEO turnover because of their (stable) desire for
autonomy (DiFonzo, Bordia, & Rosnow, 1994), and rumors of Internet
provider's illegal tapping of user hard drives because of (stable) corpo-
rate marketing incentives and (stable) corporate avariciousness (Bordia,
DiFonzo, & Chang, 1999; Bordia & Rosnow, 1998) are representative
examples from our own research. However, we are hard-pressed to
think of an unstable-cause attribution set in a rumor. Even rumors of
random acts of gang violence contain the stable-cause dispositional
attribution that gang members are evil (Vigoda, 1993).
Experimental evidence also supports this idea (DiFonzo & Bordia,
1997, 2002b): "Investors" playing the computerized stock-market
games described in chapter 2 (this volume) were exposed to rumors
that were not predictive of daily stock-price changes, yet they strongly
attributed price changes to these rumors. For example, the rumor that
"Goodyear profits are up" was judged to be the cause of stock-price
increases on the day the rumor appeared and the next day also; the
cause was stable enough to last at least 2 days. The interview study of
brokers described in chapter 4 (this volume) also qualitatively sup-
ported this idea. Brokers unanimously felt that the effect of real stock-
market rumors lasted for 2 to 3 days (DiFonzo, 1994, Study 2). Put
another way, real brokers judged rumors to be the cause of stock-price
changes, not just that day, but the day after and the day after that as
well. Thus, some—perhaps most—rumors are embedded with stable-
cause attributions.
Stable-cause attributions have been implicated in a number of sys-
tematic judgment biases. We explore two of these here: illusory associa-
tion and antiregressive prediction.

RUMOR AND ILLUSORY ASSOCIATION

Stable-cause attributions have produced illusory associations: erroneous
judgments that two characteristics are associated (Jennings, Amabile,
 Rumor as Sense Making 119

& Ross, 1982). Undergraduates and clinicians, after viewing random

pairings of Draw-A-Person test drawings and a psychological diagnosis,
thought that an emphasis on the eyes (in the drawing) covaried with
paranoia; this result occurred even though the diagnosis of paranoia
was not associated with large or salient eyes in drawing (Chapman &
Chapman, 1969). Why then did participants associate the two? A stable-
cause attribution—drawing large eyes is caused by a paranoid sensitivity
to being watched—led subjects to associate these variables. Stable-
cause attribution leads people to see relationships even when they
don't exist.
In a similar way, rumors lead people to see relationships even if
they are nonexistent; that is, stable-cause attributions embedded in
many rumors lead to illusory associations. For example, rumor has
long been associated with the formation and maintenance of racial
stereotypes (G. W. Allport & Postman, 1947b; R. H. Knapp, 1944;
Knopf, 1975; P. A. Turner, 1993). Rumors embedded with stable-cause
race-characteristic attributions encourage these illusory associations.
Children hearing multiple variations of such rumors as "I heard that
Johnny Black stole a car" (because Blacks are thieves) or "I heard
that Officer White beat him unnecessarily with his billy club" (because
Whites are brutal) will be led to associate African Americans with
theft and European Americans with aggression. Experimental evidence
accords with these ideas. The investors in the computerized stock-
market simulation game described earlier were exposed to price
changes that were uncorrelated with rumors yet they thought that
these price changes were associated with rumors (DiFonzo & Bordia,
2002b). Investors perceived relationships that were nonexistent.

RUMOR AND ANTIREGRESSIVE PREDICTION

Stable-cause attributions have also produced systematic errors in pre-
diction of categories (e.g., Is this person an engineer or a lawyer?) and
sequential events (e.g., How likely is it that football team x will win
their next game?). In these prediction situations, people tend to rely
on causal information to the neglect of often more predictively valid
base-rate information (Ajzen, 1977). Base-rate information is typically
the proportion of specified outcomes in a possible population of events.
For example, when presented with a fictitious name ostensibly sampled
randomly from a population in which 70% of the people are lawyers
and 30% are engineers, control subjects tend to use the engineer base
rate—30%—when predicting how likely it is that the person is an
engineer. However, when given additional causal information in the
form of a character sketch of the person (e.g., "He has a need for order
120 R U M O R P S Y C H O L O G Y

and clarity, and for neat and tidy systems"; Kahneman & Tversky, 1973,
p. 238), subjects abandon these base rates, rely on the character sketch,
and predict that the person is an engineer. They reason that a stable
dispositional cause—the need for order—manifests itself in the desire
to become an engineer and hence they estimate a higher probability
of being an engineer.
Causal attribution effects have also been found for predictions about
events in a sequence (Kahneman & Tversky, 1982; Matthews & Sand-
ers, 1984). Given a stable cause for recent events in a sequence (e.g.,
stock prices are decreasing because of poor management), people tend
to predict that the next event in the sequence will reflect the recent
trend (continued decrease) rather than the often more predictively
valid base rate of past outcomes. Predictions that the recent trend will
continue are necessarily not regressive toward the central tendency of
previous events and are therefore called nonregressive or even antiregres-
sive predictions. For example, subjects presented with both the long-
term average proportion of football team game wins—say, 50%—and
the recent sequence of team outcomes—win-loss-win-win-win-win-
win—would tend to predict that the team would win at their next
game, presumably because their recent wins are because of a stable
cause (e.g., team talent, or the team is "hot"; Gilovich, Vallone, &
Tversky, 1985). However, the base rate—50% in this example—would
be more predictive.
In a similar way, stable-cause attributions in rumors lead to nonre-
gressive predictions. Aggregate-level studies of the effect of rumor on
stock trading have shown that rumors are responsible for nonregressive
deviations from randomness in stock prices (Lazar, 1973; Pound &
Zeckhauser, 1990; Rose, 1951). Simply stated, rumors affect stock
prices' changes in a nonregressive fashion. Individual-level experiments
also agree with this. Hearing the rumor "Goodyear profits are up" led
investors to predict that tomorrow's Goodyear stock price will rise
and to abandon more predictively valid base-rate information about
tomorrow's price change (DiFonzo & Bordia, 1997, 2002b). In the
stock simulation research described earlier, investors seized on causal
information embedded in rumors and ignored more predictively valid
base-rate information when trading stock. They did so despite rating
the rumors as noncredible, untrustworthy, and unbelievable. In other
words, rumors did not have to be believed or trusted to powerfully affect
trading; they simply had to make sense. As a consequence, investors
exposed to rumors engaged in less profitable trading strategies than
did those who received no rumors. If stock-market price changes are
indeed unpredictable (Fama, Fisher, Jensen, & Roll, 1969; Malkiel,
1985), then hearing rumors may be bad for one's portfolio!
 Rumor as Sense Making 121

Rumors and Collective

Sense Making

Groups also use rumors to make collective sense. In this section we

explore this idea by examining the content, functions, and flow of
statements made during rumor discussions on the Internet; this species
of rumor has much to tell us about rumor discussions in general (Fisher,
1998). We show that these rumor discussions are rich conversations
in which a variety of statements are uttered, most of which are related
to or concerned with the task of sense making. To fulfill the functions
necessary to make sense, people adopt transient roles during rumor
discussions, and the rumor episode typically passes through a multistage
group sense-making process ordered around the assessment of rumor
veracity and meaning.
In this section, we draw heavily from our quantitative content
analysis of 14 rumor discussions on the Internet (Bordia & DiFonzo,
2004; see also Bordia & Rosnow, 1998). In this study, we gathered
rumor interaction episodes (RIEs) by searching archived discussions on
computer networks and by observing ongoing computer discussion
groups. Each rumor discussion had to contain five or more postings
spread over at least 2 days and reflect some earnestness on the part of
participants. The rumors touched on a variety of domains, including
health ("Ibuprofen increases susceptibility to flesh-eating bacteria,"
"Smallpox-infested blankets were distributed in Indian reservations"),
information technology ("A virus titled 'Good Times' is being spread
by an electronic mail message," "Certain features will be included on
Windows 95," "Prodigy taps hard drives of its subscribers' computers"),
conspiracy ("Republican party was involved in the explosion of the
space shuttle"), and sports ("Michael Jordan is returning to professional
basketball"; Bordia & DiFonzo, 2004, p. 38).

CONTENT OF RUMOR DISCUSSION

STATEMENTS
What types of statements constitute rumor discussions and what is
their relative prevalence in a typical rumor episode? To address these
questions, we, along with our colleagues, developed a coding system—
the Rumor Interaction Analysis System (RIAS; Bordia & DiFonzo, 2004;
Bordia & Rosnow, 1998)—and used it to content analyze each RIE in
the sample. We began by parsing all text into thought units using
122 R U M O R P S Y C H O L O G Y

guidelines provided by Wheelan, Verdi, and McKeage (1994). Often

these statements were in the form of simple sentences, such as "I am
not sure if this is true or not," "It happened in 1968," and "I hope this
happens." The reliability of this unitizing procedure was assessed by
an independent judge and was very high (93.07% agreement). The
development of the RIAS was guided by the conceptualization of rumor
as a collective problem-solving process as described in chapter 1 (this
volume). That is, in conditions of ambiguity, rumors are working
hypotheses of the group that attempt to make sense, manage threat,
and restore a sense of predictive or interpretive control; they are impro-
vised news. This more sociological perspective emphasizes the collective
sense-making function of rumor.
We identified 14 types of statements: prudent, apprehensive, au-
thenticating, interrogatory, providing-information, belief, disbelief,
sense making, directive, sarcastic, wish, personal-involvement, digres-
sive, and uncodable. The RIAS, along with the overall percentage of
each statement type, is presented in Table 5.1. Table 5.1 shows that
these rumor discussions overall were composed of statements that were
directly or indirectly useful to the group as they attempted to make
sense of their situation. The most frequent type of statement was in
fact sense making (29.4%), by which discussion participants directly
attempted to solve the problem of whether or not the rumor was true.
In other words, rumor discussants devoted a plurality of statements to
interpretation, analysis, and inference. In addition, most of the remain-
ing statements were indirectly related to sense making. Participants
provided information, asked questions, authenticated credentials and
information, shared relevant personal experience, expressed belief and
disbelief, and attempted to persuade others. All of these activities sup-
port the collective goal of sense making. Thus, the content of these
diverse Internet rumor discussions was a rich social sense-making inter-
play involving an exchange of hypotheses, opinions, views, proposals,
arguments, and emotions.

CONTENT AND COMMUNICATIVE POSTURES

OF CONTRIBUTIONS
In any discussion, people take turns communicating; one person contri-
butes a set of statements, then another provides an additional set of
statements. Discussions can be thought of as serial contributions of
statement sets; understanding the content and function of such state-
ment sets affords a better understanding of the nature of the discussion.
Rumor discussions are no different. For example, a rumor discussant
may contribute a statement that provides information (Pi) then, still
holding the microphone, propose a sense-making explanation (Sm).
 tj<B
&
V)

<0
*t
•o
<U

C
0)
»-
_c
VI
VI

J2
tj

o
o
ui

se.

VI
"o
0

<
in
Rumor li

S
S

5
iystem Statement Types and Overall Percei

'5.

Ol

1
I
Statemen
"8*

01 „

X
2«
> <s
O *
(abbrevia
f•
Definition

.1
a
rr\

<~>
4_i
in

0^
-c-
Prudent

;~
itionary statements used to qualify what

im> >*-
Ei/i
ra In c

<= £
^ s.

"o
.:£
O 0)
3 4->

(D JI
2 0
O ID
u E
£ 01
ows as being "hearsay."

>*

i/l

i_
.c
'l/»
<r

0)
<u
^>
^<
0)

<SZ
00
(N

+•*

01

<
«J
•o

a.
tements expressing rumor-related fear, dre

a.
"S.
a

flj

(D £
H^

-C
a « £ +-> -p
t! ^
:= T3

tJ
~ >> .5>E
in ±: ro T3 O

Ol o
^01+-
$

f §
iety, or apprehension, or a sense of feelinc

r£
.*!

r^
>>

1 « -Q r" >»
>, Q.

> 0)
<u E ft2 Q.

a
5 nj££
reatened."

,c
*r

4->

'•(-»
00

4-J
•?

m
(U

O
Authent

u
"3
a*

<
tements expressing the speaker's attempt 1

01
TO

a
k_
*—

in as
5-8
.E S
-a <D
<v E
£ (D

nj u,
2 S
1 credibility to what he or she was saying.

±: a
1
Is:
||
a

a
tt

k_
+^
2(D

<L>
r^-

C
03
<ji

2
!c
(N

_0)
ip

03

(D
-C
T3

QJ

4-*
4-1

T3
TJ

Q
O

4-1
I/I
i/i

•^«
01

C^
sstions seeking information (not including

>,
CT

01
fQ
^, ^^

i^
:astic remarks or persuasion attempts).

a
a

(0
_c
r»

^T

+-"
id

(D
c
00

QJ
IO

cu

+j
^
Providini

^
tements providing information relevant to

ro
Q.
a
CTl

'cZ
t

c
C31 '.-".
informal lor being discussed.

CO
M-
CO
S S X

"S
i_

as

.*
r\i
(N

(/)
4-i
o!

*»*
a^
3
tements indicating belief in the rumor.
s

h|_

X
vt

(0
00

4->
T3

ID
t
O
o

a^
fi

Disbeliel
s
01
u

tements indicating disbelief in the rumor.

10
5 !5
—

o>
(D
c
•*

i^

Sj
•=
(N

E
in
oi

a *

O)
tements reflecting attempts to solve the
• a

a s
$r
c -jf

o
Q. 5
O) ">

j
Q. >
ro ^_i
<" 0
.E «

blem of whether or not the rumor is true; ^£

Is
lj
a|

ro -=
tatements that analyze, dispute, disagree '

"> w
J: .t: «
Jraw inferences from what someone else h

•^ 5 <5

"
1; statements that justify one's own views,
ons, and beliefs; and statements describing

+3
™

x &...*& a 1-S
ision rules and heuristics.

<J
'•C
c
"S

8
Rumor as Sense Making
123
124 RUMOR PSYCHOLOGY

IX)
00

8 »
rvi (N

o
c
O
'•c
t;
3 Ol
Ol
c
O
T3 C £'
2 oo O 01
& Ol
o- c
I or, O c o dj
01 X fr o CD 01
o * S ,
O) O
ll 01

mor.
orm
C O) ns
ce o VI

* 01
O J2 o. 01
0.-1- O J3 03
SEE (D O) to TJ «
s^
permiss
ial Cog
Statements

.fil
23 $2
<u
Q.
.2IS S§
0 s
io
a. c fe
•5 "5
'"
4-
T3
Qj

i-S
Ol
0>
o c
3 £ °
i
S.
01
D.

01 §3
0) N
O Ol O I!
u Ol u 'O
•o c m o
ctly relevan

Ol
Oj +-• QJ f>
ra Ol
in
S.
Q.
O
ll
x E
S
JJo
£ ro
__ U
O3 'C
"O 4*
^ (D . Ol 3
. *l
uggesting

Dig
«£
ng

+rf C c
St I*
0>
01 5
'>»
> o-
3
-Q T=
'k- H- 1»
I C Olb! o -M J3
Statements not

S c •SS 0 E
c
P
Ol (0
l/t 0C i_8 £ £ J "§>
omments.

"*S
tatement

§ c o « I:
I 01 O
l 01 OJ
?3
rumor.

C fli E S
>>
I/I 12 S-c
JSS* w ^
S w E m
U) S «
i/i O K.S i/> -Q ^1
O) ro
ction

!™ S

1 o>
>
K II
Ol
^
state
Psyc

Ol
o
 Rumor as Sense Making 125

To better understand the nature of rumor discussion, we asked, "What

is the content of statement set contributions during a rumor discus-
sion?" That is, what typical combinations of statement sets are presented
during turns taken by rumor discussants? We were interested in the
make-up of participants' turns because they reveal the function of each
contribution at that point in the rumor discussion; we have dubbed
these functions communicative postures (Bordia & DiFonzo, 2004).
Communicative postures are conceptually similar to the "commu-
nication roles" or "communication styles" posited by Shibutani (1966)
such as the "messenger, the person who brings a pertinent item of infor-
mation to the group" and the "interpreter, the person who tries to
place the news in context, evaluating it in the light of past events and
speculating on implications for the future" [italics in original] (1966,
p. 15; see also R. H. Turner & Killian, 1972). Each role performs some
function in the context of the discussion (e.g., bringing information,
interpreting data). We coined the term posture, however, to better reflect
Shibutani's idea that these roles are transient in nature: A speaker's
posture seems less permanent than his or her role or style. For example,
in an Internet discussion I may post a message that brings information to
the discussion; it therefore performs an information-delivering function
(and I have exhibited an information-delivering posture)—but later I
may post a skeptical message that performs a disbelieving function
(and it displays a disbelieving posture). Postures can change during the
course of a discussion. We wished to empirically explore these ideas
and generate a more definite conceptualization of the transient
contributions—the postures—made to rumor discussions.
In our Internet rumor study, rumor discussion contributions con-
sisted of 281 message postings from the 14 Internet RIEs (Bordia &
DiFonzo, 2004). To identify common patterns of statement sets in each
posting, we performed a hierarchical cluster analysis on these postings.
Cluster analysis groups cases on the basis of their similarity along vari-
ous dimensions; in this investigation, we grouped postings according
to their similarity on RIAS statements (these groupings are called dus-
ters) . In other words, we used cluster analysis to discover categories of
participant postings to the discussion that had similar RIAS statement
profiles. Thus, the unit of analysis here is the postings submitted to the
rumor discussion—a higher level than the RIAS statements in the
previous section, but a lower level than the individuals who composed
the postings.
Eleven clusters were identified; each had a statement profile con-
sisting of the average number of RIAS statement types (Pr, Ap, Au,
etc.) contained in postings belonging to the cluster. We used these
profiles to interpret the communicative posture of that cluster. For
example, Cluster 1 exhibited a high average number of prudent (Pr)
and providing-information (Pi) statements but low means for all other
126 R U M O R P S Y C H O L O G Y

types of statements; we interpreted this cluster as an explanation-

delivering posture whose function was to present the rumor. Cluster 2
displayed a very high mean number of sense-making (Sm) statements
relative to all other statement types; we interpreted this cluster as an
explanation-evaluating posture whose function was to interpret the
rumor. Statement profiles for all 11 clusters are presented in Table 5.2
along with our interpretation of their associated communicative
postures.
We were guided here in part by Anderson et al.'s (1996) explana-
tion theory. As discussed at the beginning of this chapter, an event is
noticed, an initial explanation is generated, and—depending on the
motivation to engage in effortful processing—the explanation is itera-
tively tested. We adapted this cognitive model to a collective frame-
work, and the fit was quite natural. Groups seeking to explain events
must perform similar sense-making activities: They must notice the
event, generate initial explanations, test explanations, decide whether
or not to continue searching for alternate explanations, and direct the
gathering of information. The 11 identified clusters naturally displayed
these types of activities. To wit, in addition to explanation presenting
and explanation evaluating, two additional postures functioned to ana-
lyze the rumor and voice agreement (explanation verifying) and dis-
agreement (explanation falsifying); a fifth posture simply voiced ac-
ceptance of the rumor (explanation accepting). Two postures per-
formed information-sharing (information reporting) and questioning
(information-seeking) functions. One posture contributed information
and suggested courses of action (directing). Two postures functioned
to sustain motivation by voicing hopes (motivating by considering
gains) and fears (motivating by considering losses). Finally, one posture
displayed noninvolvement in sense making (casual participation). Thus,
during the turns taken in Internet rumor episodes, participants typically
contribute to the discussion in 1 of 11 ways, and most of these ways
can be understood in relation to collective sense making.

DYNAMICS OF RUMOR DISCUSSION

STATEMENTS
How do rumor discussions flow over their lifetimes? In more specific
terms, how do postures and statement types change over the course
of the discussion? These questions concern the dynamics of RIEs.
Once again using the 14 RIEs (Bordia & DiFonzo, 2004), we ex-
plored these questions in a couple of ways. First, we divided each RIE
into "quarters" by dividing the number of postings chronologically
by 4; the type and number of postures during each quarter were then
analyzed. Clear posture trends—in line with collective sense making—
were observed: Explanation-delivering and directing postures were
 Rumor as Sense Making 127

T3
in (N 0
S1
O Qj
d d d d d
.C
c
10 ,_ 0 ,_ o 8
a. d d d
d d

o o 0 o o
^
3 d d d d d
Ol
c

I
c
VI
rM
d
(N
d
o
d d
o
d

Ia o
^
m
£ d d d d d

UD **! ^. <n
Of
+* E •^ ^ d
VI ^J ^O
S ^

•5 ,_ o o o LO o
1a d d d fvi d

3
C Q
_. ^ IS
c OQ »—
II d m tY) d
in 1
_O « ._ <£> <t p 0
'in »P ^ ^ (V^ d d T— d
in g
3 a
"""
a (N r\i O m o
Q I
g d d d d d
o *
E 5
in
3 3 ^ p o p o
cc d <N d
+J
c
a o (N o o o
1 d d d d
o
in p o
01 £ N d d d

1
("Here is an explanation

Accepts the explanation

unicative Postures of P

("Here's an explanation
that 1 heard; 1 am not

Analyzes and disagrees

cu *- c
that 1 believe for the
Posture description

with the explanation

rumor—to the group

fc- Q) O
Analyzes and agrees

following reasons.")

("Here's why 1 don't

(and example)

sure if it is true.")
explanation — the

v
with explanation

4-* — C
C C (D

i-B s-
("1 believe the
explanation.")

explanation.")
Presents brief

C (T3 <U
believe the

S-=S
^ QJ O
< £ (D
E
E
o O)
c
u c C c c c
o
-a >! fO' C_ O
'i3
D)
C '+-" Ol
O
'f 01
•Is
ro 1*
lo
(0
C
JS
p
HI
D
<D C
c -^,
_ra ^
C •£
«j Q.
Q. ro
11 S-li X > x <D X "m
Q. S
x u
c ^ LU ID LU > LU 42 LU m

u
^
0 is
u a
^_; oi ni Lfi
 128

+rf
_c
tc
V

+^

(C

c
•o
s>

4J
c

o
ISI

I/I

I/I
'5

O
Ul

o
i/i
O

E
E
2
3

E
unicative Postures of P

u
u

u
D)

1
c
C

a
c

1/1
*

S
•sk

*;
3
tn
-M
c

flj
c
0

x
&
"3
Posture description

!
1

._

PO

3
Q
1 I/I
(and example)

k
O
Q,

I/I
Q

§1
S1

SI
Ol

l| oa
a ^-

(6
<N
<—
T—

c
o
io

IN
(N
fN

d
in

< d
d
d
(N
Shares information and

N
d
f*i
10
1C

d
o

d
d
d
d

ro c

£ £
E t
k o
personal experiences

| ££ Q.
•^ 01
("Here's what 1 know
related to the
explanation.")

r~
T-
1C

•*
00

o
"i
,_

d
d
d
^
Seeks information
IN

0
d
d
0
rri

d
d
q
d
d
d

0. (U
£ Si
o '2
("Here's what is needed
RUMOR P S Y C H O L O G Y

01
to be known to generat

0.
or evaluate an
explanation.")

00
V—

c
TJ

'•Qjd
ID
,—

00

(N

IN
<N

5
r\i

in

d
«i
«~

d
d
d
^

Provides information ani

d
d
o

Ol
q
in

d
d
d

suggests a course of
action ("Here's what 1
know and someone
should do XYZ to gathei
more information.")

4->

^
T-^

oi
.0
o

o
in
IN

d
LO

d
d
Justifies explanation,
o

q
0

d
•N

d
d
O

q
d

d
q
o
0

d
d

V*
0.01

u
hoping for certain

C.

!§'!
1'L
consequences or things

O)
("Here's a hoped-for
consequence of this
explanation, which is
why we should continue
to evaluate it.")
 Rumor as Sense Making 129

CO «^ •£? rs
••i x i 'n
at <u X
~ -o £
<" -n
1- o ils"
W ^~v O
d d QJ r- fN

5 i o
S | E!
o «-
d d > y s°
£-5
"ro CD -;
u c 2
0 ^- §
D ^
rj "?
C
d d E "TO ro
C TO
O.E^
u
-Dm0
(N
d
v-

d
c
QJ 3
5! - n
vi Q-
2Q. CU
a; -Q^
OJ g -
00 f>
d N -a "0 c
c c o
nj fT3 ^j
t/> "D C Q
<ij aj en •-
0 o -0 .N o $
o u o •;=
d d .a = _ E
Q- 2 ro ir
<u •- -u g.
c S> 0 °-
o o O r5
d d '^ S ™ ^
S = g^
.E <J § Q.
to CO
d *~ c ^
0 ™^ <
4-;^
c m s
5 c = c
^ — 01 .Q
(N o g S ^ IS
d d to. — - n
OJ >? <U o
"t^ -C i/i
^ •*-<-(-• tn

^r £ S <
^- (N - E ° ro
d d 3 * c-^.
^ S°o
-5i "> tj -o
f •S "o 2 'u
N d l?l«
u ro — c
c p
3 01 —
ro ro
u
0 o Si f 8'fc
oj o> on £
d O 5 -§1 c <
S± o>^

E ° S o ~ S S.£-is
.E o "o g
were derived from 276 pos!

IB posture. From "Problem !

67, pp. 44-46. Copyright 2
why we should contin
this explanation, whic
a feared consequence
its consequences ("He
Posture reflects conce
and fear of the rumor

to statement type abbreviat

indulges a little in
problem solving.
Easily distracted,
to evaluate it.")

"i >* -^ \-
>>
_Q en
c
^oj"^ -^
<" '- -F
i_ ^ to
+o
01 £> c
3 o '-
C .= '^ TJ >2 E §
ro
S|
> T3 w
_ .9-
ro 'u
---20
•- u-i o
V» wi <f ^ '-P 0V OJ J=t 01
i/> T- 15
Isl
d
ro ra
U Q.
^
QJ ^
* (U i ^
Mi
"?^ •*-

•§ on
< <" ti
On ^Q,
130 R U M O R P S Y C H O L O G Y

more common in the first quarter of rumor interaction, explanation

evaluation peaked during the third quarter, and casual participation
peaked at the end. In other words, the discussion first centered on
presenting the rumor and directing efforts to make sense of the rumor,
then proceeded to sifting and analyzing the rumor; sense making then
subsided as people lost interest. A second quarterly analysis—this time
of the incidence of common RIAS statement types—accorded with this
interpretation. Interrogatory statements peaked during the first quarter,
disbelief in the second, sense making during the third, and casual
participation during the fourth. Providing-information statements were
frequent in all but the third quarter. Thus, collective Internet rumor
sense making seems to proceed in a multistage process of bringing the
rumor to the attention of the group (Stage 1), sharing information
(Stage 2), evaluating explanations (Stage 3), and then resolving prob-
lems (Stage 4).

GENERALIZABILITY TO FACE-TO-FACE
RUMOR INTERACTION EPISODES
It is important to remember that our conclusions regarding collective
sense making stem mainly from our analysis of Internet rumor epi-
sodes. These computer-mediated types of RIEs may differ in substantial
ways from face-to-face episodes. First, we selected only RIEs in which
participants displayed some earnest interest in the discussion and thus
importance was probably high; other computer-mediated or face-to-
face episodes may not be so involving. Second, the nature of computer-
mediated networks means that each discussion contribution was poten-
tially read by every person in the group; in other words the group
was closely—rather than diffusely—connected. These characteristics
seem to encourage what H. Taylor Buckner (1965) dubbed "multiple
interaction": the process by which a rumor is actively recirculated
within the same group of people (see chap. 7, this volume). Less
important rumors transmitted through more diffuse social networks
may not exhibit the content, postures, and dynamics we observed in
computer-mediated episodes (see chap. 6, this volume, for a similar
discussion of the differences between serially transmitted and collabo-
rative rumor activity).

Conclusions

In this chapter we explored the sense-making function of rumor from

both social-cognitive and collective levels of analysis. Rumor affects
 Rumor as Sense Making 131

the individual's attempt to explain events at a variety of junctures

such as noticing the event, providing an initial explanation, motivating
further exploration, and activating cognitive structures with which to
interpret the event. We considered how rumors are often embedded
with stable-cause attributions and how the activation of these types of
cognitive structures may lead to illusory associations and nonregressive
prediction. Rumors also influence the collective sense-making process.
We explored the content of statements that are made during rumor
episodes, as well as the content, functions, and flow of the contributions
made to rumor discussions.
We've considered here the sense-making processes—both individ-
ual and interpersonal—involved in rumor activity. How efficacious are
people at this important task? We consider this question next.
 Rumor Accuracy:
Patterns of Content Change,
Conceptualization, and
Overall Accuracy
Negative rumors abounded among the 75 employees in
a division of a large corporation near Rochester, New
York, in the late 1990s (see chaps. 2 and 8, this
volume). The division was a tight-knit community of
professionals who had heard that drastic layoffs were in
the offing. Indeed, fully half of the division was
eventually let go. One week prior to the official
announcement of who would be laid off, workers
circulated a rumor listing the names of targeted
persons. Remarkably, the rumor was 100% accurate.

uring the American Civil War, telegraph cables strung across

D
trees in the manner of an agricultural grapevine carried
intelligence messages (K. Davis, 1972). Today, the "intelli-
gence" efforts of organizational members often rely heavily
on rumors passed through the communications grapevine
(Burlew, Pederson, & Bradley, 1994; K. Davis, 1972;
Harcourt, Richerson, & Wattier, 1991; Newstrom, Monczka,
& Reif, 1974; Smeltzer & Zener, 1992; Walton, 1961;
Zaremba, 1989). Indeed, substantial portions of a national
sample of middle managers in a recent survey "rated the
grapevine as superior to formal communication" and as the
best source of information for certain topics such as

133
134 R U M O R P S Y C H O L O G Y

promotion opportunities, company future plans, departmental future

plans, and salaries and raises (Harcourt et al., 1991, p. 357; see also
Modic, 1989).
How effective are these informal intelligence efforts? That is, how
well do groups fare in ferreting out the facts? The question is of both
academic and practical interest. As was seen in chapter 2 (this volume),
some organizational rumors have disrupted productivity, sullied repu-
tations, and eroded trust (K. Davis, 1975; DiFonzo, Bordia, & Rosnow,
1994; Zaremba, 1988). When rumors are false or distorted, such effects
are sadly unwarranted. In addition, from the "rumor consumer's" point
of view, normative questions of how much credence ought to be placed
in any bit of information carried by a channel of communication hinge
on the overall validity of that channel. If rumors are generally accurate,
then it would seem that they are worthy of belief. However, if rumors
are usually inaccurate, they ought not to be trusted.
This chapter and the next address several questions related to accu-
racy. First, we review patterns of change in rumor content. Next, we
explore the concept of accuracy and its measurement. Third, we assess
the overall accuracy of rumors, with a special focus on organizational
rumors. In the next chapter, we will examine the mechanisms by which
rumors become more (or less) accurate. Finally, we will address how
such mechanisms are likely to operate in organizational contexts. At
several points we introduce our own recent empirical research bearing
on accuracy.

Patterns of Rumor
Content Change

It is instructive to review theory and research on overall change patterns

in rumor content. In general, four types of changes have been identified:
leveling, sharpening, adding, and assimilation. As we show later, the main
controversy in this corpus of literature is whether leveling or adding
occurs in everyday rumor activity. The debate stems partly from differ-
ences in serial transmission (ST) versus collaborative (COL) approaches
to studying and conceptualizing rumor and whether both approaches
generalize to real-life rumor situations. With this in mind, we review
patterns of rumor content change here with special attention to ST
versus COL conceptual and methodological issues.
 Rumor Accuracy 135

FOUR PATTERNS OF RUMOR

CONTENT CHANGE
How does rumor content change over time? For example, does the
corpus of details grow larger or smaller? Does the content of the rumor
become more or less exaggerated? Does the rumor begin to conform
to widely held ideas? Researchers have addressed these questions and
identified four patterns of rumor content change: leveling, adding,
sharpening, and assimilation.

Leveling
Leveling refers to the loss of detail and the reduction in length at each
successive transmission so that the rumor is more easily grasped, espe-
cially during early transmissions. Rumors are leveled to "short concise
statement^]" (G. W. Airport & Postman, 1947b, p. 81) in the sense
that complex buildings are leveled to simple and small heaps. Kirkpat-
rick (1932) called the process "condensation." For example, of 20 de-
tailed statements in an original stimulus description, 15 may be leveled
and only 5 remain.

Adding
Adding is our term for addition to rumor content in the form of new
material or additional details. Adding has been referred to as "snowball-
ing" (Rosnow, 1991), invention and elaboration (G. W. Allport & Post-
man, 1947b), "compounding" (Peterson & Gist, 1951), "embroidering"
(G. W. Allport & Postman, 1947b), and "fabrication" (Sinha, 1952).
For example, after observing a photograph of a mob of people without
weapons, clubs were supplied to that mob in successive serial retellings
of the details of that photograph (G. W. Allport & Postman, 1947b,
pp. 116-121). Peterson and Gist (1951) also found that new themes
were added—not leveled—in a set of rumors about a murder. Rosnow
(1991) described the snowballing of the false "Paul McCartney is dead"
rumor. Note that adding is to leveling as addition is to subtraction.

Sharpening
Sharpening refers to the accenting and highlighting of certain details in
the rumor message. This accenting may occur as a result of leveling;
certain details are brought into sharpened focus necessarily by the
clearing away of other details. For example, the odd term remonstrated
was retained—presumably because it was an odd term—throughout
136 R U M O R P S Y C H O L O G Y

serial retellings of a certain tale that became successively shorter (G. W.

Allport & Postman, 1947b); the term was thus emphasized. Sharpening
may also occur as a result of changes such as exaggeration (e.g., as when
one Negro became four Negroes; G. W. Allport & Postman, 1947b); the
idea "a Negro was present" was therefore highlighted (see also Firth,
1956; R. H. Turner & Killian, 1972).

Assimilation
Assimilation refers to the shaping of rumor content—through leveling,
adding, and sharpening—so as to be in greater accord with personal
cognitive schemas. Assimilation may be a relatively cool (i.e., cognitive)
process, such as when details are leveled, added, or sharpened so as
to make the rumor more thematically coherent and plausible (e.g., an
ambulance became a Red Cross station to fit a battle theme; see also
Kirkpatrick, 1932), to complete incomplete data (e.g., Gene Antry be-
came Gene Autry), to simplify a complex stimulus (e.g., a set of subway
posters became lots of advertising], and to fit our expectations and linguis-
tic conventions (e.g., kilometers on a road sign was changed to miles
by Harvard students; G. W. Allport & Postman, 1947b, pp. 99-104).
Assimilation may also be a relatively hot (i.e., defensive or motivated)
process, such as when rumor change occurs because of personal inter-
ests (e.g., details about clothing or occupation predominated among
subjects interested in clothing or occupation, respectively), self-interest
(e.g., a group of police officers focused favorably on the police officer
in a story), and prejudice (e.g., hostile behavior was imputed to minority
figures; pp. 105-115).
Whereas leveling, adding, and sharpening are more elemental pat-
terns of content change, assimilation is akin to overall pattern fitting.
Assimilation is thus a more holistic and higher level pattern of change.
It is the guiding of these subpatterns of change so as to be in accord
with personal schemas.

GENERALIZABILITY OF PATTERNS
Consistent evidence points to sharpening in the service of assimilation
in real-life rumor situations (Buckner, 1965; Peterson & Gist, 1951;
Rosnow, 1991; Shibutani, 1966; R. H. Turner, 1964, 1994; R. H. Turner
& Killian, 1972). However, disagreement persists regarding the extent
of leveling versus adding in real-life rumor situations (Rosnow, 1991;
Shibutani, 1966; Turner & Killian, 1972). Leveling, rather than adding,
has tended to occur in ST laboratory situations and in planted-rumor
field study situations characterized by ST-like passing of information.
Our position is that leveling occurs mostly in some real-life rumor
 Rumor Accuracy 137

episodes marked by low ambiguity and in which a group is primarily

engaged in ST-like information diffusion (i.e., simply passing along
the information). However, adding, rather than leveling, has occurred
mostly in field observation studies with high ambiguity and high-
importance contexts such as catastrophes and murders. It occurs in
real-life ambiguous situations of importance in which the group is
quite interactive and collaborative. We explore these findings and our
position more fully first by examining ST and COL approaches to the
study of rumor, and then by considering evidence and arguments for
the generalizability of both leveling and adding.

Serial Transmission and Collaborative Approaches

If we conceive of rumors as being serially transmitted, then distortions
occur at each transmission node, at each telling of the rumor down
the line (Shibutani, 1966). This approach tends to highlight the changes
that occur during transmission as a result of cognitive (e.g., narrowing
of attention, limits to memory, perceptual biases) and motivational
(e.g., fact-finding, relationship-enhancement, and self-enhancement)
factors at work in the individuals involved at each node (these factors
are discussed in chap. 7, this volume). The ST framework has been
primarily, although not exclusively, used in the laboratory setting.
Laboratory studies in this vein have reported a three-part embedding
process; embedding refers to leveling and sharpening guided by assimila-
tion (G. W. Allport & Postman, 1947b'; Bartlett, 1932; Higham, 1951;
Kirkpatrick, 1932; McAdam, 1962). In particular, leveling—not add-
ing—occurred in these studies. In fundamental terms, such studies are
mostly about passing information along, and they generalize to real-
life rumor situations that are mostly characterized by some type of ST.
If we conceive of rumors as the tentative and changing hypotheses
produced as persons in a group collaborate more actively in an ambigu-
ous situation, then the focus tends to shift to group, network, and
situational features (also discussed in chap. 7, this volume). This frame-
work assumes that rumor is a collaborative, not a serial, activity; that
is, each person makes a collaborative rather than additive contribution
(Shibutani, 1966). For example, persons A and B discuss a rumor about
impending downsizing and who will be laid off; A and B compare notes
on the basis of information they each bring to this discussion and they
quickly modify the rumor. Out of this collaboration, a predominant
interpretation of an ambiguous situation arises. The COL framework

1
G. W. Allport and Postman (1947b) also discussed invention and elaboration, but
felt that these were so infrequent as to presumably be not part of the embedding process.
138 R U M O R P S Y C H O L O G Y

has used primarily field observational settings. These studies have gar-
nered support for sharpening, assimilation, and especially adding—but
not leveling (Peterson & Gist, 1951; Shibutani, 1966; R. H. Turner,
1964; R. H. Turner & Killian, 1972). Studies investigating this type of
rumor activity are fundamentally about interactively collaborating in
an ambiguous and important situation and generalize to real-life rumor
situations that are characterized by collaborative sense making.
We note that both types of real-life rumor episodes—those charac-
terized by ST-like transmission and those best described as COL-type
activity—involve sense making. However, ST-like transmission tends
to produce leveling whereas COL-type activity tends to produce adding.
Let's consider evidence and arguments for the generalizability of each.

Generalizability of Leveling
G. W. Allport and Postman (1947b, pp. 134-138) argued that the
embedding process (which includes leveling) is representative of what
happens with everyday rumor. To support this statement, they offer
the example of a rumor arising from an incident that occurred in Maine
in 1945 shortly before the surrender of the Japanese: A Chinese teacher
on vacation asked for directions to a scenic outlook. Within short order,
an assimilated rumor was being actively transmitted in the community:
"a Japanese spy had ascended the hill to take pictures of the region" [italics
in original] (p. 134). They state that this story had been leveled and
sharpened according to a dominant interpretive schema (i.e., the Japa-
nese spy motif). G. W. Allport and Postman (1947b) also supported the
generalizability of the embedding process by pointing to the similarity of
their ST results with that of Gestalt memory studies on geometric shapes
and projective test studies (memories and projections are similarly lev-
eled and sharpened in the service of assimilation).
But the argument that leveling was observed in ST, Gestalt memory,
and projective test studies indicates only that the results of each type
of study were similar, not whether ST study results can be generalized
to all (or some) everyday rumor episodes. In addition, ST research has
been criticized as not being mundane realistic (i.e., the lab experience
does not match real-life experience in key ways; DiFonzo, Hantula, &
Bordia, 1998; Rosnow, 1980) in such a way as to predispose leveling
(Bordia, 1996; Shibutani, 1966; R. H. Turner & Killian, 1972). At least
three main arguments in this vein have been raised; we present each
here and critically examine them.
First, Buckner (1965) posited that distortion in ST occurs mostly
because of memory limitations: "In Allport and Postman's experiments,
the words are leveled out because of the difficulty of remembering
twenty or so new and discrete items of information" (p. 59). G. W.
 Rumor Accuracy 139

Allport and Postman (1947b) themselves noticed that ST participants

are instructed to be accurate and would therefore be prone to not relay
statements of which they are the least bit unsure (p. 76). Laboratory
ST subjects similarly have no opportunity for questioning, challenging,
and probing and may thus be prone to encoding failure. However,
subsequent laboratory ST experiments that allowed discussion at each
transmission node—ST plus interaction—exhibited less distortion (see
chap. 7, this volume), but leveling still occurred (Leavitt & Mueller,
1951; McAdam, 1962); that is, statements became shorter and more
easily grasped. Such "ST plus interaction" situations are more mundane
realistic and thus more generalizable than are pure laboratory ST
procedures.
Second, ST participants are much less emotionally involved (and
therefore not motivated to speculate) than are real-life rumor partici-
pants (Shibutani, 1966; R. H. Turner, 1964). Although ST laboratory
experiments that did use more involving rumors found greater trans-
mission accuracy (see chap. 7, this volume), leveling still occurred
(Higham, 1951).
Third, R. H. Turner (1964) stressed that ST cannot accommodate
the invention stage of everyday rumor. Invention occurs when people
create hypotheses to explain an ambiguous situation; it is then that
rumor often snowballs. Laboratory ST studies and field studies in
which formed rumors have been planted therefore preclude invention
by not supplying the requisite undefined situation. Ambiguity in a
situation of importance does seem to lead to invention. Published
stock-market takeover rumors, for example, seem to "derive from
market professionals' interpretations of unusual firm-specific trading
activity" (Pound & Zeckhauser, 1990, p. 306). "Seldom does the
rumor . . . precede the unusual price and volume activity. Usually
rumors follow this activity as market observers seek to explain its
cause" (Pound & Zeckhauser, 1990, p. 306). Thus, according to Turner's
crisis model, (a) a crisis occurs and people seek information;
(b) information is lacking or the sources of formal information are
not trusted, and therefore people feel frustrated and seek information
from informal channels; and (c) if no information available, people
engage in affirmative rumoring, that is, they speculate on the basis of
whatever evidence and framework of understanding they possess;
invention occurs at this stage.
There is evidence supporting Turner's objection. Most field studies
reporting leveling tend to be planted-rumor studies that do not seem
to involve ambiguous and important situations. One such study is self-
labeled "a community diffusion experiment" (De Fleur, 1962, p. 51).
Seventeen percent of 249 households were told the slogan "Gold Shield
Coffee—Good as Gold." Then 30,000 leaflets were airdropped on the
140 R U M O R P S Y C H O L O G Y

community. These leaflets stated that every house would be

interviewed and free coffee would be given to those who knew the
slogan. Leveling (as well as sharpening and assimilation) occurred after
two tellings of the information. A second example follows: W. L. Davis
and O'Connor (1977) planted information in an academic department
that Davis's wife was pregnant; the information contained a number
of details (e.g., the doctor's name, the child's possible name) that were
leveled. Here is a third, more recent, example: A planted rumor con-
densed around the main fact that "next year's T-shirt may cost the
students money" (Sedivec, 1987, p. 37). The point is that situations
involving coffee slogans, office birth announcements, and student
T-shirt costs seem to be far less ambiguous and involving than, say,
situations that threaten safety, health, job, or well-being.
However, this objection (that invention is precluded) presupposes
that all real-life rumor episodes are highly and interactively collabora-
tive; they do not address the possibility that some real-life rumor epi-
sodes are ST-like in their transmission pattern. Some leveling clearly
has occurred in field observational situations as a result of ST-like
transmission. Scanlon (1977) traced a single rumor in a disaster situa-
tion through a serial chain. Although he found clear evidence of adding
more detail—as we would predict in a COL situation—he also found
some evidence of leveling (some details were lost in the transmission).
In addition, most rumors in Caplow's (1947) field study tended toward
"simplification" (p. 301); Caplow noted that most rumors contained
three statements, many contained one or two, and few contained more
than three. This simplification is a form of leveling. Furthermore, Nkpa
(1977) reported a pipe-dream wartime rumor that underwent leveling
as it spread: A detailed story about how Northern Nigerian General
Gowon had been killed by Nigerian rulers dissatisfied with his perfor-
mance became "I heard that Gowon has died in Kaduna" (p. 32). Again,
some leveling occurs in real-life rumor transmission; this may happen
simply because some portion of rumor transmission is serially
transmitted.

Generalizability of Adding
Most field studies about high-ambiguity and high-interest rumor epi-
sodes show adding rather than leveling (Peterson, unpublished, cited
in De Fleur, 1962; Peterson & Gist, 1951; Schachter & Burdick, 1955).
In Peterson and Gist's (1951) field study of rumors surrounding a very
high interest baby-sitter murder, the central theme was not distorted,
but was compounded with additional speculations. That is, in this high-
ambiguity rumor episode, details and variations proliferated rather than
leveled out. Rumors following public disclosure of the scandalous loan
 Rumor Accuracy 141

practices of an important bank were "completed and enriched" (Roux-

Dufort & Pauchant, 1993, p. 238). Again, adding occurred in this high-
ambiguity episode.
Although adamant that snowballing is a "misconception," even
G. W. Allport and Postman (1947b, p. 153) observed some invention
(i.e., adding) in their laboratory ST demonstrations and offered one
possible situation in which rumors might snowball: After highly
emotionally straining events, people may tend to perseverate, "mull
it over, talk about it endlessly, [and] explore in fantasy all possible
consequences" (p. 154). Furthermore, Schachter and Burdick's (1955)
field experiment strongly supports this idea. This study manipulated
importance while creating a highly ambiguous situation: The percent-
age of girls reporting new rumors in the high-importance conditions
was much higher (70%) than in the low-importance conditions (15%).
In addition, the diversity of rumors was also much greater in the
high-importance conditions (average of 12 different rumors in each
condition vs. average of 1.5 different rumors in the low-importance
conditions). The high-importance condition groups were composed
of friends of the girls who had been suddenly removed from a classroom
setting; these friends were undoubtedly earnestly and interactively
collaborating about the event and thus prone to invention. The low-
importance groups were composed of girls who were not acquainted
with the accomplices; these girls were presumably passing along
(serially transmitting) interesting bits of information. Again, adding
seems to occur in real-life situations that are interactively collaborative
in character.

CONCLUSIONS
In sum, what can we say about patterns of content change? First,
certain real-life field situations, particularly those possessing high-
ambiguity and high-interest contexts, predispose adding. These rumor
episodes are about making sense of an important and ambiguous sit-
uation. Other real-life situations, particularly those characterized by
low-ambiguity and low-interest contexts in which the rumor is simply
being transmitted, seem to engender leveling. These rumor episodes
are about serial diffusion of a message. Of course, even some high-
importance-high-ambiguity situations may involve ST ("There is a fire
in the building! Leave now!" will probably be passed along quickly
without discussion, interaction, or collaboration), but we speculate
that ST-like situations are less frequent in real life than are COL-type
situations. Second, in addition to adding or leveling, sharpening
also occurs. Finally, all three types of changes occur in service of
assimilation.
142 R U M O R P S Y C H O L O G Y

Thus far we have considered broad categories of rumor content

change. We next examine patterns of changes in rumor accuracy. We
begin by more clearly conceptualizing the term rumor accuracy.

Accuracy
Conceptualization and
Measurement

What exactly do we mean when we say that a rumor is accurate? In

this section, we sharpen the conceptualization of accuracy and discuss
how accuracy is measured.

RUMOR VERITY AND RUMOR PRECISION

Two senses of the construct rumor accuracy bear exploration. One refers
to the degree to which a rumor corresponds with fact, reality, and
truth. When we say a rumor is accurate in this sense, we mean that
it corresponds to the facts. We term this sense of accuracy rumor verity.
The opposite of accuracy here is falsehood. Both true and false rumors
may be generated, sifted, and evaluated when groups attempt to define
ambiguous situations. A veritable rumor circulated in the aftermath of
the Hiroshima bombing: The devastation was caused by energy released
when small particles were split (D. L. Miller, 1985). This true rumor
circulated along with many other false ones as people attempted to
comprehend what had happened. R. H. Turner and Killian (1972)
described rumors generated as a crowd milled after discovering the
body of a man in a car parked near a hotel. The original rumor generated
and accepted by the group (that the man had been murdered by the
owner of the vehicle) turned out to be false; the veritable one (that
he died of alcohol poisoning) was proposed but rejected. Sinha (1952)
categorized rumors circulating after a landslide as truth, exaggerations,
or "outright falsehoods"; accuracy in the sense of verity is the dimension
that incorporates the first and last of these characterizations. Accuracy
here is akin to the concept of validity in psychometrics: Does the rumor
faithfully represent some real state of affairs?
The second sense of accuracy refers to the degree to which the
rumor corresponds with an original perception or message. When we
say a rumor is accurate in this sense, we mean that it corresponds
closely to some original version. We call this sense of accuracy rumor
precision. The opposite of accuracy in this sense is distortion, which
 Rumor Accuracy 143

refers to the degradation of or reduction in quality of some original

message during transmission. G. W. Airport and Postman (1947b) re-
ported rumor distortion as rumors were serially transmitted in their
laboratory studies; the terminal reports invariably differed from the
original stimuli. All 96 girls in a girls' school reported hearing an undis-
torted version of a planted rumor in Schachter and Burdick's (1955)
field experiment; note that the rumor was false to begin with, although
accurate in the sense of being precisely transmitted. In Sinha's (1952)
three-part classification scheme, accuracy in this sense is the dimension
that incorporates the term exaggeration (an exaggeration contains distor-
tions in the direction of greater quantity or quality). Accuracy here is
akin to the concept of reliability in psychometrics; how consistent is
the rumor with its original version?
It is useful to momentarily consider six of the possible combinations
of these senses of accuracy, which are presented in Table 6.1. Some
rumors are true and were precisely transmitted; these stars seem unaf-
fected by potential forces of distortion and change. Other rumors are
false and were precisely transmitted; these counterfeits began false and
don't change (like counterfeit coins). Some rumors are true but were
not precisely transmitted; they are either converts that changed for the
better during their lifetime orgrainies that became slightly distorted (as a
grainy photograph). Finally, certain rumors are false and were distorted
during transmission; some are fallen stars that started well but ended
badly and others are hopefuls that started false but show signs of slight
improvement. This categorization sketches the careers by which true
and false rumors obtain and frames some of the questions pertinent
to rumor accuracy. First, both veritable and false rumors can be gener-
ated, but how is verity measured, how often does it occur, and what
processes are involved in such generation? Second, rumors can change,

Verity-Precision Rumor Categories

Rumor precision
Rumor
verity Precise Distorted toward truth Distorted toward falsehood
True Stars: True rumors that Converts: True rumors Grainies: True rumors
were precisely that were distorted that were distorted
transmitted (toward veracity) in (toward falsehood) in
transmission transmission
False Counterfeits: False Hopefuls: False rumors Fallen Stars: False rumors
rumors that were that were distorted that were distorted
precisely transmitted (toward veracity) in (toward falsehood) in
transmission transmission
144 R U M O R P S Y C H O L O G Y

and they can change toward or away from verity. How is such change
measured, how often does it occur, and what processes are involved
in such change? We begin with measurement.

MEASUREMENT OF VERITY AND PRECISION

Rumor verity accuracy has tended to be measured in field studies
approaching rumor from a collaborative (COL) perspective. These in-
vestigations typically involved collecting rumors or variations of a
rumor within a particular field setting and assessing what percentage
of these rumors was true. For example, verity accuracy in organizational
grapevine research was measured as the percentage of communication
bits in a rumor or set of rumors that were true (K. Davis, 1972; Marling,
1969; Rudolph, 1973; Walton, 1961; Weinberg & Eich, 1978). The
percentage of takeover rumors published in The Wall Street Journal
that proved true (Pound & Zeckhauser, 1990) assessed verity. More
qualitative measures of verity include Sinha's (1952) three-part catego-
rization and Caplow's (1947) global recollection of the overall truth of
military rumors.
Rumor precision accuracy has tended to be measured in research
using variants of the ST paradigm in a lab or field-experiment setting.
Lab studies typically involve observing an original stimulus (e.g., a
drawing, photograph, or videotape) and transmitting a description of
that stimulus through a chain of participants without discussion. In
field settings the original stimulus is a planted rumor in an actual
organization; transmission here, of course, involves discussion. Rumor
precision accuracy is the percentage of the final report that corresponds
to the original stimulus. G. W. Allport and Postman's (1947b) classroom
demonstration studies exemplify this approach in the laboratory (see
also Higham, 1951; Lyons & Kashima, 2001; Werner, 1976). Other
laboratory studies allowed discussion during transmission (Leavitt &
Mueller, 1951; McAdam, 1962).
Field studies using planted rumors have also tended to measure
rumor precision rather than verity: All 96 girls in Schachter and
Burdick's (1955) study reported an undistorted version of a planted
rumor. Accuracy was the extent to which the planted rumor had re-
sisted distortion rather than the quality of the groups' efforts to make
sense of the situation. This point is underscored by the fact that many
new rumors arose, some quite bizarre. Accuracy was not measured as
the percentage of rumors that corresponded to the actual facts (the
sudden removal of girls from the classroom was part of a psychological
experiment) or even the planted rumor (the girls had been removed on
suspicion of stealing tests). Precision accuracy has also been measured as
a continuous and a dichotomous variable. Sedivec (1987) first measured
 Rumor Accuracy 145

accuracy as the percentage of items that participants recalled out of an

original set of seven parts of a planted rumor in a student organization;
missing, distorted, and added lines each counted as inaccurate state-
ments. Using the same data, he also measured the accuracy of recalled
statements as either containing or not containing the one main fact.
Although researchers—especially those in the laboratory ST
tradition—may have measured rumor precision, rumor verity is what
all rumor researchers have been ultimately interested in. ST researchers
measuring precision accuracy tended to assume that rumor precision
was synonymous with rumor verity in real-life situations because they
used their results to explain why rumors were often false. For example,
G. W. Airport and Postman (1947b) concluded that "so great are the
distortions . . . that it is never under any circumstances safe to accept
rumor as a valid guide for belief or conduct" (p. 148). Like our rumor
researcher forebears, we are likewise interested in rumor verity. We
are concerned with rumor precision to the extent that it sheds light
on some of the processes involved in rumor verity. Therefore, unless
otherwise specified, we use the term accuracy to refer to rumor verity.

How Accurate Are

Rumors Overall?

The term rumor connotes inaccuracy. Subjects in our studies typically

stated that they would risk far less money on rumors than on news
and viewed them as neither credible nor trustworthy; these ratings
applied to both unpublished and published rumors (DiFonzo & Bordia,
1997; see also G. H. Smith, 1947). Rumor clearly has a bad name, but
is this reputation deserved? We address this question by first presenting
results of empirical investigations of rumor accuracy in the literature
and from our own recent unpublished studies.

RUMOR ACCURACY LITERATURE

A handful of studies have gauged rumor accuracy in field settings; these
are summarized in Table 6.2. These investigations typically involved
collecting rumors or variations of a rumor and assessing what percent-
age of these rumors (or rumor components) was true. Table 6.2 summa-
rizes the rumor samples collected and arranges them in descending
order of accuracy percentage (the percentage of each sample that was
true). We conclude from these summaries that rumor accuracy varied
substantially, but certain field settings seemed to produce accurate
146 R U M O R P S Y C H O L O G Y

Summary of Accuracy Studies

Accuracy
Reference Rumor sample and setting N subjects %• N rumors
Caplow Grapevine rumors in military. nr Almost nr
(1947, p. 301) 100%
Marling Grapevine rumors in a vertical 451 98.42% 15
(1969, p. 123) segment of management and non-
management employees in a mid-
sized electronics manufacturing firm.
Rudolph Grapevine rumors in a public utility. 124 96% nr
(1971, p. 187;
1973)
Davis Grapevine information in industry nr 80-99%b nr
(1972, p. 263) for noncontroversial company
information.
Walton Information attributed to the <101 82%c 12
(1961, p. 48) grapevine at the Naval Ordinance
Test Station, China Lake, CA.

Pound & Financial takeover rumors published na 43% 42

Zeckhauser in The Wall Street Journal "Heard
(1990, p. 293) on the Street" column.
Weinberg & Eich Rumors collected on a hotline nr 16.2% nr
(1978, p. 30) during a university strike by
graduate students.
Prasad Rumors collected after a nr 9%d 23
(1935, pp. 1-4) catastrophic flood.
Sinha (1952) Rumors collected after a nr Very nr
catastrophic landslide. low

Note, nr = not reported, na = not applicable. ^Refers to the overall percentage of communication details that
could be assessed as true or false, which were true in a rumor or set of rumors. bRefers to Davis's summarization
of his own research (i.e., several studies). 'Refers to the percentage of correct responses of those attributed to
grapevine information on a 12-question quiz administered to employees. Choices included a "don't know" op-
tion, however, which garnered between 35% and 77% per question (M = 52%), thus limiting the 82% accuracy
figure to those responses for which the employees felt "reasonably" certain of their answers (Walton, 1961,
pp. 48-49). dPrasad (1935) presented a "representative set" of 30 rumors, 23 of which were verifiable (i.e., dealt
with empirical as opposed to metaphysical events).

rumors. We consider these factors later, but note here that rumors
within established organizational settings, especially those character-
ized as grapevine rumors, tended to be very accurate. This rinding
accords with Hellweg's (1987) conclusion in her literature review of
organizational grapevine research: Grapevine information (including
rumor) tends to be accurate, although incomplete.
 Rumor Accuracy 147

Our own, more recent empirical studies are consistent with this
conclusion. We report here three studies we conducted to investigate
questions related to rumor accuracy in organizations. The first was a
set of field interviews with organizational communications personnel;
the others were questionnaire studies with samples of employed stu-
dents. All three sets of results are consistent with the conclusion that
rumor accuracy varies widely, but rumors within organizations tend
to be accurate.

1996 FIELD INTERVIEWS

The first study—herein referred to as the 1996 Field Interviews—used
an in-depth critical-incident methodology (Flanagan, 1954) in organi-
zational settings. This method asks participants to recall a specific
event—the critical incident—that exemplifies the phenomenon under
investigation and then answer questions about that incident. In 1996
we conducted field interviews with organizational communications
personnel from several corporations in the metropolitan area of a U.S.
city as part of a multistage study investigating the harmful effects of
organizational rumors and how managers dealt with them. These per-
sonnel were typically directors of communication or vice presidents of
public relations. Each interviewee was asked to recall a specific incident
of a harmful or potentially harmful rumor that he or she had responded
to. Of the 18 largest publicly traded corporations contacted (as listed
in a local trade publication), 6 consented to an in-depth interview (the
remainder refused for a variety of reasons including a reluctance to
discuss the organization's rumors and managerial time constraints).
Most of the organizations represented in this sample were multinational
in scope as well as recognizable household names. Each interview was
audiotaped, transcribed, and analyzed with respect to various hypothe-
ses related to accuracy. Accuracy was determined by asking participants
to estimate the percentage of the rumor (or rumors) that was (or
were) true.
Table 6.3 presents the rumors, their publics (the groups within
which they circulated), and their accuracy percentages in descending
order. All of the rumors but one (brittle reactor vessel) circulated within
an established organizational setting. All but the last two had survived
over a period of time (see table notes). An inspection of the findings
shows that rumor accuracy varied widely, but like the rumor studies
within organizational settings reported in the literature, these organiza-
tional rumors tended to be accurate. It is important to note that organi-
zational communications personnel were not asked to provide true
rumors; however, most did. Furthermore, if we consider only those
rumors that survived over time, overall accuracy was even greater.
148 R U M O R P S Y C H O L O G Y

pill" I'M' 6 vi" •" v ' •;~:->:TVv :'?f :::' ^^••^^•>~--- r --: "* • ' " : ^' "••: ' : • •• y ; y r •: ;';;>::q
ire: '" ' ":{•.••-':-•- ••:••• ,~ . . ". ,- " - - '=. ... ' • • • , , • • ' ..-,,<• .' . - ,• - -! r.' ' :-' \ . - , - , - .• •' :.'.. ", ' ..<'•• - s °"-~

1996 Field Interviews Accuracy Data

Accuracy
Rumor Rumor public percentage
Two manufacturing centers would consolidate Employees at each plant 100
operations.
Two manufacturing plants would consolidate Managers at each plant 100
operations.
Subsidiary company would be sold. Subsidiary employees 100
A large operations center would be closed. Employees at targeted 80
center
University president was secretly working for University faculty and staff 60
government agency.
Company would be taken over.a Company employees 0
Utility had a brittle reactor vessel in danger of News media 0
cracking.6

Atofe. aThis rumor was quickly squelched during group interaction. bSquelched when news media contacted utility
to verify.

STUDENT RUMOR SURVEY 1

A second investigation collected workplace rumors from employed
students in 1996; this investigation is herein called Student Survey 1.
Fifty-six students in two upper level psychology courses were adminis-
tered a questionnaire (see Appendix 6.1). Rumor was first denned as
"an unverified bit of information about something of importance to
a group." Participants were asked to "Think of an instance when a
rumor occurred in a workplace setting. You may have simply heard/
read the rumor or you may have also passed it along to coworkers."
To calculate an accuracy percentage, we directed students to "Choose
a rumor that has since been proven true or false beyond a reasonable
shade of doubt; we are not interested in rumors that are still uncertain."
In an open-ended question, students were asked to "State the rumor."
Rumor accuracy was assessed by asking: "How accurate (true) did the
rumor prove to be?" Rumor accuracy trend was assessed by asking:
"From the time when the rumor started until when the rumor was
proven true or false, to what extent did the rumor tend to become
more accurate?"
Fifty-four usable questionnaires were obtained; 12 of these were
discarded because they described incidents of gossip rather than rumor
 Rumor Accuracy 149

(e.g., "Two employees were having an affair"). 2 The final sample of 42

workplace rumors is presented in Table 6.4. Accuracy ratings were
bimodal—as expected given the request for true or false rumors—
and negatively skewed, indicating that, as a whole, workplace rumors
tended to be very accurate. Most were 100% or nearly 100% true.3
This finding is again consistent with the literature review and results
from the 1996 Field Interviews. Accuracy trend ratings were also nega-
tively skewed, indicating that workplace rumors tended to become
more accurate over their lifetimes.4 Overall, then, workplace rumor
participants tend to be fairly good at ferreting out the facts of a situation.
Dummy coding represents each range of scores in a distribution with
a categorical or ordinal value (e.g., see note to Figure 6.1). By dummy
coding the accuracy and accuracy trend variables for this sample, we
assessed the frequency of the verity-precision combinations presented
in Table 6.1. These frequencies are presented in Figure 6.1 and identified
in Table 6.4. By far, most of the rumors in this sample resembled
converts; they proved all or mostly true and also became more accurate
during their lifetime. Some were stars in that they proved true but had
changed very little. And some proved all or mostly true but had been
distorted; these were grainies. Of those that proved false, most resem-
bled fallen stars in that they tended to become more degraded. A
handful were counterfeits because they proved false but had changed
very little. Very few were hopefuls (false rumors that had become more
accurate). In summary, the clear central tendency of this workplace
rumor sample was toward accuracy.

STUDENT RUMOR SURVEY 2

A third investigation collected workplace rumors from employed U.S.
college students from 1997 to 1998; this investigation is herein called
Student Survey 2. One hundred eighty-five students in lower- and
upper-level psychology courses were administered a questionnaire sim-
ilar to Student Survey 1. Rumor was first defined as "an unverified bit
of information about something of importance to a group. It is like
news in every way except that it is unverified." In an attempt to weed
out gossip, we instructed the participants that "Gossip, however, is
usually about something personal or private, and is usually meant

2
The final sample (M age = 22.56 years, SD = 3.32) was composed of 14 females
and 27 males (1 did not report sex).
3
Mean accuracy = 7.29 on a scale of 1 to 9, SD = 2.64, N = 42.
4
Mean accuracy trend = 6.21 on a scale of 1 to 9, SD = 2.34, N = 42.
150 R U M O R P S Y C H O L O G Y

Workplace Rumors From Student Survey 1

Type Rumor summary

CF The stolen petty cash was an inside job and the thief was a certain employee.
CF Kris is going to leave the company.
CV The candidates for a position of chairman at a college or department [will be so-
and-so].
CV Person A [of our group] gets paid the highest salary.
CV We would be getting new computers at work.
CV Bill Parcells is leaving the New England Patriots after the season to coach the New
York Jets.
CV A certain department would be laid off.
CV A coworker was fired.
CV A coworker was injured the night before.
CV Several coworkers were going to be fired, but no one knew who.
CV The supervisor was going to be fired and replaced.
CV [University] administrators have $287 million in undesignated funds but refuse to
spend it on student community because they are tightwads.
CV A coming lay-off of many workers in my company.
CV A fellow co-op student was fired from an automobile factory for taking a nap
inside a car in one of the test facilities.
CV Someone was leaving.
CV We will get to choose our work schedules for next quarter instead of having them
decided for us.
CV Someone [a coworker] was going to get fired.
CV A few of my sorority members were getting prank phone calls.
CV A rumor that I would be named battalion commander.
CV A really good cook was returning to work with us.
CV Coworker got fired for shoplifting from the bulk department.
CV Mary is leaving the company.
CV Certain persons were breaking into our restaurant and getting to our liquor.
CV The upstairs balcony of my fraternity house will be [turned] into a lounge.
CV Michelle is being let go.
CV There are going to be significant cutbacks (40%).
continued
 Rumor Accuracy 151

Workplace Rumors From Student Survey 1

Type Rumor summary

FS A coworker never shows up for meetings.

FS Our company was going to buy a division of another company.
FS A certain coworker slashed my coworker's tires.
FS A coworker is getting fired.
FS Students who used a business course as an engineering elective will not graduate
on time because the course won't be counted.
GN All workers will have to take 2 weeks off without pay so that the company will be
able to report better profits.
GN Boss cheats on taxes.
GN A female coworker was being stalked by a male coworker and harassment charges
were filed leading to termination.
GN The photography department was going to be cut by 50%.
GN Several persons would be "on their way out."
HP There was a rumor that I was quitting.
ST Coworker was demoted from a management position for leaving a cashbox sitting
out.
ST A coworker was taking scrap metal (that was still usable) and recycling it for cash.
ST Some people in higher management will quit their jobs.
ST A fellow worker sold drugs.
ST Rick, who just had an accident, is coming back to work.

Note. N = 42; CF = counterfeits; CV = converts; FS = fallen stars; GN = grainies; HP = hopefuls; ST = stars.

primarily to convey social standards or to entertain." Participants were

asked to "Think of an instance when a rumor (not gossip) occurred in
a workplace setting. You may have simply heard/read the rumor or
you may have also passed it along to coworkers." To sample equally
from rumors that had proven true and those that had proven false, we
asked students to recall a rumor that had proven true and then one
that had proven false (the order of the request was alternated).
The 185 questionnaires could have yielded a possible set of 370
rumors (two per questionnaire). Of these 370, 94 were gossip, 2 were
missing accuracy and accuracy trend ratings, and 30 did not describe
the rumor. The final sample consisted of 146 subjects—48 of whom
contributed one rumor and 98 contributed two rumors—for a total N
152 R U M O R P S Y C H O L O G Y

25- Accuracy Trend Converts

• More Distorted
H Little Change
D More Accurate
20-

o
0>

tr
o>
10-

Fallen Stars Grainies Stars

All or Mostly False All or Mostly True

Accuracy

Verity-precision combination frequencies (Student Survey 1). N = 42; Accuracy

dummy = proved all or mostly false if accuracy < 4; = proved all or mostly true if
accuracy > 6 (no values = 4, 5, or 6); accuracy trend dummy = more distorted if
accuracy trend < 4; = more accurate if accuracy trend > 6; otherwise = little change.

of 244 rumors.5 One hundred thirty-seven rumors were true; 107 were
false. To assess more closely the types of change occurring in true
and false rumors, we computed the frequency of the verity-precision
combinations for this sample; these are presented in Figure 6.2. The
results are very similar to those of Student Survey 1. We first observe
that the overwhelming majority of rumors were all or mostly true or
false; there appeared to be little middle ground for rumors that had
since been proven true or false. Of the all or mostly true rumors recalled,
most by far resembled converts. Of those that proved all or mostly

5
The final subject sample (M age = 24.96 years, SD = 7.64, 15 did not report age)
was composed of 118 females and 113 males (13 did not report sex).
 Rumor Accuracy 153

Accuracy Trend Converts

100-
• More Distorted
H Little Change
D More Accurate

80-

60

Counter-
Fallen fejts
40- Stars

20-
Stars

All or Mostly False Some False and True All or Mostly True
Accuracy

Verity-precision combination frequencies (Student Survey 2). A/ = 244. Accuracy

dummy = proved all or mostly false if accuracy = 1, 2, or 3; = some false and true if
accuracy = 4 or 5; = proved all or mostly true if accuracy = 6, 7, or 8. Accuracy trend
dummy = more distorted if accuracy trend = 1, 2, or 3; = little change if accuracy
trend = 4 or 5; = more accurate if accuracy trend = 6, 7, or 8. Both scales ranged
from 1 to 8.

false, most resembled either fallen stars or counterfeits. In summary,

rumors tended to change toward or away from verity. True rumors
especially tended to mutate toward accuracy; false rumors tended either
to become more false or to remain false.

LIMITATIONS AND CONCLUSIONS

Two limitations of these three studies should be considered. First, these
results may be caused by some form of recall bias wherein true rumors
154 R U M O R P S Y C H O L O G Y

are recalled more easily than are false ones. When true rumors become
fact, perhaps the ramifications of these facts (e.g., job layoffs, boss left,
company was sold) tend to serve as cues to remind people of the original
rumor. However, false rumors may be equally memorable if found to
be bizarre in content. Second, we note again that rumors that remained
unverified were procedurally excluded from the sample. Therefore, it
is necessary to qualify the applicability of these results to organizational
rumors that have been proven true or false.
Taken all together, however, the literature and our empirical results
suggest two conclusions. First, organizational rumors tend to be accu-
rate. The reputation of workplace rumor as inaccurate apparently is
itself inaccurate! The reason for this disparity is puzzling. If the over-
whelming majority of rumors that are recalled were true, why would
the overall impression of rumor tend to be not credible? We have
noticed this pattern repeatedly: When asked about rumor overall, peo-
ple classify it as false or low-quality information. When asked to recall
specific rumors, people tend to report true or high-quality information.
To explain this disparity, we offer two speculations. First, social desir-
ability bias may be operating; participants may assume that relying on
rumors is a less than acceptable behavior. It may conflict with their
self-image as intelligent persons. Second, the same cognitive processes
that result in stereotyping of persons may result in stereotyping of
rumors. For example, rumors may be vulnerable to illusory correlation:
Because rumors are probably a minority of the information that is
processed (e.g., news and information), false specimens of that minority
may become relatively salient and then falsely correlated (cf. Chapman
& Chapman, 1969). In an attempt to establish meaningful categories,
people tend to therefore associate rumor with false.
Second, the literature and results suggest that for those rumors
that prove to be true or false, the true tend to get "truer" and the false
either stay the same or become more false. Rumor verity trends seem
to bifurcate. We dub this intriguing possibility the Matthew accuracy
effect. The Matthew effect in science refers to the finding that well-
known scientists are accorded a disproportionately large share of credit
and access to new scientific ideas and information whereas lesser
known scientists are allotted fewer accolades and less access. Robert
K. Merton (1968) coined this term after the same general principle
expressed in a biblical text from the Book of Matthew: "For everyone
who has will be given more, and he will have an abundance. Whoever
does not have, even what he has will be taken from him" (Matt. 25:29,
New International Version). The idea is expressed in such adages as
"nothing succeeds like success" and "the rich get richer and the poor
get poorer." Variants of this idea have been noted in many domains;
in network science, nodes with many connections tend to obtain more,
 Rumor Accuracy 155

whereas those with few connections tend to lose what they have (New-
man, 2003).
In the next chapter we will explore the mechanisms by which
accurate-inaccurate rumor content is generated and changed.
156 R U M O R P S Y C H O L O G Y

Appendix 6.1
Student Survey 1: Request for Participation in a
Short Survey on Rumor Accuracy

Rumors are one of the ways that people obtain information in organiza-
tions, and rumors may impact us in a variety of ways. Of course, we
all know that rumors may or may not be true, but we must sometimes
make decisions on the basis of a rumor. That's fine if the rumor turns
out to be true, but could cause problems if it's a false rumor. For us to
be better "consumers" of rumors, it would be helpful to know how
likely they are to be true. Your participation in this study will help us
answer this question and thereby help us create guidelines for when
(if ever) it is appropriate to trust a rumor.
Kindly respond to the attached brief survey on an instance in which
a rumor occurred in the workplace. The survey takes only 10 minutes
to complete and your cooperation will be greatly appreciated. Your
participation is completely VOLUNTARY, and your returning this sur-
vey will be taken to mean that you have consented to participate in
this research study. You may, of course, stop at any time without
consequence or prejudice. Your responses will be ANONYMOUS (please
do not put your name or student number on the survey) and the data
will be kept confidential (only aggregates will be reported). At the end
of the quarter, I will report the results to our class.
THANK YOU VERY MUCH.
A rumor is an unverified bit of information about something of
importance to a group. Think of an instance when a rumor occurred
in a workplace setting. You may have simply heard or read the rumor or
 Rumor Accuracy 157

you may have also passed it along to coworkers. CHOOSE A RUMOR

THAT HAS SINCE BEEN PROVEN TRUE OR FALSE BEYOND A
REASONABLE SHADE OF DOUBT; we are not interested in rumors
that are still uncertain.
1. State the RUMOR here:
2. Please describe the SITUATION OUT OF WHICH THE RUMOR
AROSE:
3. How ACCURATE (true) did the rumor prove to be? (circle one
number): 1 100% False, 9 100% True.
4. From the time the rumor started until when the rumor was
proven true or false, to what extent did the rumor tend to become
MORE ACCURATE? 1 Became More Distorted, 9 Became More
Accurate.
Now remember the situation JUST PRIOR TO WHEN THE RU-
MOR WAS PROVEN TRUE OR FALSE. Answer all of the remaining
questions for the point in time JUST PRIOR TO WHEN THE RUMOR
WAS PROVEN TRUE OR FALSE.
5. Overall, how ANXIOUS (worried/concerned) did the CONTENT
OF THE RUMOR ITSELF make people feel who heard and/or passed
the rumor? 1 Not at all Anxious, 9 Highly Anxious.
6. Overall, how ANXIOUS did the SITUATION OUT OF WHICH
THE RUMOR AROSE make people feel who heard and/or passed the
rumor? 1 Not at all Anxious, 9 Highly Anxious.
7. Overall, how ANXIOUS (worried/concerned) were people who
heard and/or passed the rumor? 1 Not at all Anxious, 9 Highly Anxious.
8. Overall, how UNCERTAIN (filled with doubts and uncertainties
about the SITUATION OUT OF WHICH THE RUMOR AROSE) were
people who heard and/or passed the rumor? 1 Certain, 9 Uncertain.
9. Overall, to what extent did people KNOW WHAT TO DO in
order to minimize negative consequences associated with the rumor
or maximize positive consequences associated with the rumor? (Do
not answer if no consequences whatsoever or neutral consequences
only were associated with the rumor.) 1 Did Not at all Know What to
Do, 9 Knew Exactly What to Do.
10. Overall, how CONFIDENT were people who heard and/or
passed the rumor that this rumor was true? 1 Not at all Confident,
9 Highly Confident.
11. If the rumor had proven true, how POSITIVE (favorable) would
the probable consequences have been for most of the people who heard
and/or passed the rumor? 1 Very Negative, 9 Very Positive.
12. Overall, how IMPORTANT (significant, consequential) was the
RUMOR to people who heard and/or passed the rumor? 1 Not at all
Important, 9 Very Important.
158 R U M O R P S Y C H O L O G Y

13. Overall, how IMPORTANT (significant, consequential) was the

SITUATION OUT OF WHICH THE RUMOR AROSE to people who heard
and/or passed the rumor? 1 Not at all Important, 9 Very Important.
14. When a rumor is passed from one person to another, it may
involve a great deal of interaction (discussion and clarification), or it
may involve very little interaction. Overall, how INTERACTIVE was a
typical discussion of the rumor? 1 Not at all Interactive, 9 Very Inter-
active.
15. A rumor may recirculate through a group (most people in the
group hear the rumor from more than one person), or it may not
recirculate (most people hear it once and never pass it on). To what
extent did the rumor RECIRCULATE through the group? 1 Did not
Recirculate, 9 Actively Recirculated.
16. A rumor may pass through an established communication net-
work (e.g., an office grapevine) or through a group of people who
never communicated before. Overall, how ESTABLISHED were the
communication channels of the group of people who heard and/or
passed the rumor? 1 Novel, 9 Established.
17. The grapevine is an established informal communication net-
work. To what extent was this rumor passed through a grapevine? 0%
Passed Through Grapevine, 100% Passed Through Grapevine.
Finally, a few details about yourself:
18. You are: Male / Female (please circle one).
19. Your age is: (in years).
Mechanisms Facilitating Rumor
Accuracy and Inaccuracy

In 2005 the most devastating storm in U.S. history—

Hurricane Katrina—hit New Orleans. According to
rumors and news reports, the Big Easy became a
hotbed of anarchy: Children and adults were raped in
the Convention Center, violent gangs shot at rescuers
and police helicopters, cars were hijacked at gunpoint,
and hundreds of dead bodies floated in the shark-
infested waters that submerged the city. These stories
turned out to be almost entirely false. Although some
looting occurred along with isolated shooting incidents,
the tales of widespread chaos were gross exaggerations
and fabrications; they were largely inaccurate.
—Dwyer and Drew, 2005; Gillin, 2005

n the previous chapter we saw evidence suggesting that

organizational rumors tend toward accuracy, and cata-
strophic rumors—as those about Katrina illustrate—tend
toward inaccuracy. Why might this be? In more specific
terms, what are the mechanisms by which accurate (or
inaccurate) rumor content is generated or changed? In this
chapter we summarize evidence from studies investigating
rumor accuracy and present our own recent empirical
investigations when pertinent. As in the previous chapter,
accuracy signifies verity, and distortion pertains to lack of
precision during rumor transmission. Because distortion
may obviously affect accuracy, we examine literature per-
taining to rumor distortion also. For clarity, we parse

159
160 R U M O R P S Y C H O L O G Y

accuracy mechanisms into five sets: cognitive, motivational, situational,

network, and group mechanisms; we summarize these mechanisms
in Table 7.1. In reality, these processes may be inseparably blended,
such as when cognitive schemas are informed by cultural concepts.
Special attention is given to the neglected work of H. Taylor Buckner
(1965) whose theory of rumor accuracy focused on network and
motivational mechanisms. We also describe our own recent research
that investigated Buckner's ideas.

Cognitive Mechanisms

Cognitive mechanisms affecting accuracy refer to processes associated

with information processing. These include the narrowing of attention,
memory limits, and perceptual biases.

ATTENTIONAL NARROWING
Systematic rumor distortion—resulting in inaccuracy—results from
the narrowing of attention on the part of a person listening to rumors
told serially and with no discussion. Research on the teller-listener-
extremity effect in impression formation is relevant here: Subjects
serially transmitted—without discussion—a message about an actor's
culpable behavior (a football player's drunk-driving incident). The
teller heard central statements (e.g., the football player was driving
drunk) and statements about mitigating information (e.g., he didn't
know that someone had spiked the punch). Those with first-hand
information (tellers) rated the actor less extremely than did those
who recounted the actor's story (listeners). Why would listeners tend
to rate the actor more harshly? Tellers' accounts "tend to be more
disjointed, temporally disorganized, and incomplete" (R. S. Baron,
David, Brunsman, & Inman, 1997, p. 827). Baron et al. argued that
disjointed teller accounts are harder for the listener to process and
thus require increased attentional resources, which leads to ignoring
mitigating information, which in turn leads to more extreme judg-
ments. Baron et al. gathered support for this sequence by demonstrating
that the effect becomes more extreme in subjects that hear white
background noise while listening to the message (white background
noise diminishes attentional capacity).
Attentional narrowing effects seem especially pronounced in initial
tellings of the rumor—again in the context of serial transmission (ST)
with no discussion. Consistent with G. W. Airport and Postman's
 Facilitating Rumor Accuracy and Inaccuracy 161

T A B L E 7. 1
Mechanisms Involved in Rumor Accuracy
Mechanism Summary Example

Cognitive: In the context of ST with no Tellers' accounts of a football player's

attentional discussion, attentional drunk driving incident tended "to be
narrowing narrowing leads to distortion more disjointed, temporally
in favor of salient information. disorganized, and incomplete" and
taxed listeners' attentional resources;
listeners therefore focused on central
features of these accounts (Baron et al.,
1997, p. 827).
Cognitive: memory In the context of ST with no Leveling of details in a series of ST
limits discussion, memory limits lead demonstrations was not random; those
to distortion that favors easily details that were more difficult to
encoded or salient encode or retrieve were dropped (G. W.
information. Allport & Postman, 1947b).
Cognitive: Perceptual biases lead to False rumor content that a community
perceptual biases selective perception and worker (in charge of community social
interpretation so as to cohere activities) was communist seems to have
with existing cognitive been made plausible by a rapid increase
structures such as stereotypes in news media reporting on the threat
and schemas. of communism (Festinger et al., 1948).
Motivational: Accuracy motivation tends to Ego-involved subjects—subjects for
accuracy produce more accurate rumors. whom the topic was important—leveled
details less than neutral subjects did in
a ST laboratory study (Higham, 1951).
Motivational: The desire to enhance or Change in rumors surrounding a man
relationship- maintain relationship with found dead in the backseat of an
enhancement others promotes rumor automobile; the actual cause of death
content that creates positive was resisted by those who first found
affect and positive impressions, the body and were thus "in the know";
or is stereotype-consistent. the original rumor was, however,
inaccurate (R. H. Turner & Killian, 1972).
Motivational: Self-enhancement leads to MBA students were more willing to
self-enhancement rumor content that reinforces transmit rumors that reflected well on
existing beliefs, attitudes, their own school and negatively on a
wishes, biases, suspicions, and rival school (Kamins et al., 1997).
desires or derogates outgroups
to boost self-esteem.
Situational feature: High collective excitement Because of high anxiety, suggestibility
high collective leads to "suggestibility" and after catastrophic landslides was
excitement diminished critical ability. heightened, people set aside their
critical set, rumors were not scrutinized,
and people did not desire to verify
them (Sinha, 1952).
continued
162|RUMOR P S Y C H O L O G Y

Mechanisms Involved in Rumor Accuracy

Mechanism Summary Example
Situational feature: With an accuracy motivation, Military rumors in Caplow's (1947) study
capacity to check the capacity to check leads to were accurate partly because military
increased accuracy. personnel could ask superiors about the
veracity of rumors (superiors could then
confirm that the rumor was false).
Situational feature: For groups capable of ferreting False rumors in the 1996 Field
time out the facts, time increases Interviews (see chap. 6., this volume)
accuracy; otherwise time leads were quickly discarded.
to inaccuracy.
Group: conformity Conformity pressures lead to Once consensus among prison inmates
rumors consistent with the was achieved about who were
consensus. "snitches," data were reorganized and
conformity was demanded (Ackerstrom,
1988).
Group: culture Rumors tend to be consistent Rumors that earthquakes resulted from
with cultural axioms. planet alignments and Himalayan deity
indignation recurred over time (Prasad,
1950).
Group: epistemic Higher group standards of Groups characterized as "skeptical" by
norms evidence lead to accuracy. public relations personnel tended to
produce more accurate rumors than did
those characterized as "gullible"
(DiFonzo & Bordia, 2002a).
Network: interaction Interaction increases precision Messages transmitted under conditions
accuracy. in which participants could ask
questions retained precision accuracy
whereas conditions prohibiting
interaction exhibited distortion (Leavitt
& Mueller, 1951).
Network: ST and cluster configurations Group skepticism moderated the
transmission tend to promote distortion; relationship between Ml and accuracy
configuration multiple interaction promotes in a sample of workplace rumors in
accuracy for critical set group Student Rumor Survey 2 (this chapter).
orientations and inaccuracy for
uncritical set orientations.
Network: channel Channel age is associated with Caplow (1947) attributed high accuracy
age accuracy. rates in part to the ability of
increasingly established channels of
communication to label the quality of
information sources.

Note. ST = serial transmission.

 Facilitating Rumor Accuracy and Inaccuracy 163

(1947b) ST leveling curves, R. S. Baron et al. (1997, Study 1) replicated

the teller-listener-extremity effect for first- to second-generation trans-
missions, but not for second- to third-generation transmissions. The
authors proposed that lower number of events (because of omissions)
in the second to third generation permitted attention to mitigating
information.
R. S. Baron et al. (1997) also speculated that attentional narrowing
effects may be exacerbated by anxiety. Consistent with this idea, Ellis
and Zanna (1990) showed that arousal narrows attention to salient
stimuli and thus increases the impact of salient information on causal
attribution. Because anxiety is closely linked with arousal, anxiety may
therefore cause people to focus on salient parts of a situation (e.g.,
behavioral action rather than situational factors) in the rumor construc-
tion process. G. W. Allport and Postman (1947b) found similar effects:
ST without discussion in front of an audience rather than without an
audience resulted in greater leveling of rumor content. Although All-
port and Postman accounted for this finding as the result of greater
motivation to be accurate (participants passed on only that information
about which they were absolutely confident), anxiety and attentional
narrowing are also quite plausible.
Thus, attentional narrowing—in the context of ST with no
discussion—seems to result in inaccuracies characterized by emphasis
on salient or central information. Often this information may be about
behavioral actions rather than situational constraints; such effects are
likely to be exacerbated by anxiety or arousal.

MEMORY LIMITS
Memory limits led ST—with no discussion—participants to level the
number of details recalled from an initial set of 20 to a final set of 5
(G. W. Allport & Postman, 1947b). Leveling of these details was not
random; those details that were more difficult to encode or retrieve
were dropped. For example, proper names and titles were almost always
deleted (see also Bartlett, 1932). However, "odd, perseverative word-
ing" (e.g., "There is a boy stealing and a man remonstrating with him";
Allport & Postman, 1947b, p. 89) tended to hang on. Items pertaining
to movement (e.g., "There is a window with three flowerpots, one
falling out," p. 95) and size (e.g., "big warehouse," p. 96) also tended
to remain presumably because they are more vivid and thus more
easily encoded. Allport and Postman also noticed that ST subjects, who
were instructed to be as accurate as possible, tended to level to an easily
remembered phrase. Thus, the limits and biases of memory itself—in
the context of ST with no discussion—result in inaccuracies character-
ized by salient or easily remembered information.
164 R U M O R P S Y C H O L O G Y

PERCEPTUAL BIASES
Activating listeners' cognitive structures, such as stereotypes and
schemas, may result in selective perception and interpretation of suc-
ceeding stimuli so as to fit the activated structure (Sedikides & Ander-
son, 1992); distortion of rumors in the direction of established struc-
tures (also known as assimilation) may result (G. W. Allport & Postman,
1947b). Trope and Liberman (1996) have noted this confirmation bias
in social hypothesis testing; once a hypothesis is generated, there is a
tendency to restrict attention and information processing. The hypothe-
sis lends a frame to the evidence and people tend to look for evidence
consistent with their hypothesis. For example, false rumor content that
a community worker (in charge of community social activities) was
communist seems to have been made plausible by a rapid increase in
news media reporting on the threat of communism (Festinger et al.,
1948). The communism rumor was elaborated over a 2-week period
and other information was reinterpreted to agree with this rumor (it
should be noted that no rebuttal efforts were made during this period).
A similar phenomenon in collective behavior has been labeled symbol-
ization (R. H. Turner & Killian, 1972). Symbolization is one way that
crowds selectively define a situation by focusing tensions and actions
in a simplified way on one object, as with a scapegoat. For example,
the statement "White professors A and B are behind the [civil rights]
sit-ins" tends to select and highlight evidence that points toward how
these professors are the ones truly responsible for the sit-in. In like
fashion, "evidence" supporting (false) rumors of Paul McCartney's
death was found on a Beatles album cover: Paul wore no shoes, and
in Britain, the deceased are typically interred without shoes
(Rosnow, 1991).
The effect of perceptual biases in race rumor formation and perpetu-
ation is well known (Bird, 1979; R. H. Knapp, 1944; P. A. Turner,
1993). The same stereotypes documented by G. W. Allport and Postman
in 1947 are influential today. They posited that stereotypes that are
characteristic of some rumors are a means of easily condensing—or,
to use a more modern cognitive term, chunking—a number of ideas.
One of us (DiFonzo) periodically asks students to anonymously list
rumors they have heard about other races; such rumors invariably
conform to stereotypes. Race rumors often seem to mutate to reflect
the racial stereotypes of the rumor public (Maines, 1999). Rumors
circulating in the African American community—such as that a boy
had been mutilated in a shopping mall lavatory—specified the perpetra-
tors as White; the same rumors circulating in the White community
specified them as Black (M. Rosenthal, 1971). Stereotypes not only
influence how evidence is interpreted but also lead to a premature
 Facilitating Rumor Accuracy and Inaccuracy 165

cessation of evidence gathering (Trope & Liberman, 1996). Of course,

reliance on stereotypes when interpreting individuals or instances is
likely to result in inaccuracy. Anxiety seems to strengthen reliance on
cognitive structures. In the context of interpersonal communication,
Gudykunst (1995) proposed that high anxiety leads to reliance on
stereotypes when one makes judgments regarding a stranger. Anxiety
was associated with agreement of the rumor with group biases in our
survey of rumor episodes recalled by top public relations personnel
(DiFonzo & Bordia, 2002a). Anxiety thus seems to increase dependence
on activated cognitive structures, especially stereotypes.
These perceptual biases sometimes lead to false confirmation when
testing hypotheses, but there is evidence that people are cognizant of the
diagnosticity of the evidence (Trope & Liberman, 1996). For example,
Caplow (1947) observed that statements delimiting the probability of
the rumor's truth (e.g., "this is probably not true . . .") were attached
to the rumor. We note that perceptual biases have occurred in both
ST laboratory (G. W. Allport & Postman, 1947b; Lyons & Kashima,
2001) and field situations in which not much information is known
(Festinger et al., 1948) or formal sources of information are not trusted
(P. A. Turner, 1993). The individual tendency to originate or distort a
rumor toward activated schemas may be augmented or inhibited by
situational and group processes (discussed later).

Motivational
Mechanisms

Motivational mechanisms affecting accuracy refer to the goal of the

interpersonal interaction that involves rumor. That is, what is the
rumor-spreader trying to accomplish? Following the outline developed
in chapter 3 (this volume), these processes include accuracy, relation-
ship enhancement, and self-enhancement.

ACCURACY
People are often motivated to achieve an accurate picture of a situation;
recall the fact-finding motivation introduced in chapter 3 (this volume).
One implicit rule of conversation is that information transmitted be
trustworthy (Grice, 1975; Higgins, 1981). When anxiety is not high,
for example, and if participants are attempting to define a situation
"realistically," then accuracy is a "primary consideration" (Shibutani,
1966, pp. 72-76). In these situations, the group checks reliability of
166 R U M O R P S Y C H O L O G Y

information and tests hypotheses. When the group is given the proper
resources to ferret out the facts, such efforts have been successful:
"When accuracy is important to those participating in the experiment,
there tends to be little distortion of any kind" [italics in original] (Shibu-
tani, 1966, p. 92).
Accuracy motivation—as well as consequent hypothesis testing and
salience of diagnostic information—increases with the importance of
the topic. People are motivated toward accuracy when the costs of false
confirmation are greater. People are also more aware of how diagnostic
the information is when the costs are higher; they are careful to form
conclusions based on valid evidence. This careful attention to accuracy
even occurs in ST studies: Higham (1951) found that ego-involved
subjects (subjects for whom the topic was important) leveled details
less than did neutral subjects.
Accuracy motivation also increases when people are held person-
ally responsible for what they say. Grapevine participants care about
their reputation among sustained relationships; they may not care
about it among strangers (Shibutani, 1966). The implication is that
among grapevines consisting of sustained relationships, people are
more motivated to be accurate. G. W. Allport and Postman (1947b)
noted that ST in front of an audience rather than without an audience
resulted in greater leveling of rumor content and they attributed this
result to accuracy motivation: Subjects passed along only those items
about which they were certain. Johan Arndt (1967) similarly noted
that message distortions in word-of-mouth communications hinge on
the ability to evaluate the message and the "rewards associated with
precise transmission" (p. 65). Arndt speculated that in product rumors,
"the knowledge that the receiver of word of mouth can buy the product
and thus check the veracity of the message would appear to discourage
extreme exaggerations. After all, the communicator has his reputation
as a reliable source at stake" (p. 66).

RELATIONSHIP ENHANCEMENT
As noted in chapter 3 (this volume), people are often motivated to
build and maintain relationships. Often this motivation manifests itself
in an inclination to say something that is likely to make the hearer
feel good, and a disinclination to share a rumor that will diminish the
hearer's mood. This is the minimize-unpleasant-messages effect (Tesser
& Rosen, 1975) already discussed. Business school students resisted
passing negative rumors (compared with positive rumors) because such
rumors might generate negative affect in the recipient (Kamins, Folkes,
& Perner, 1997). Selective transmission of rumors in this fashion fosters
inaccurate content by promoting the survival of only socially acceptable
 Facilitating Rumor Accuracy and Inaccuracy 167

rumors that enhance relationships with one's ingroup (here

relationship-enhancement motivation works in tandem with self-
enhancement motivation, which is described later). However, as we
discussed in the Rochester Institute of Technology (RIT)-University of
Rochester (UofR) study presented in chapter 3 (this volume), when
people seek to maintain close or long-term relationships, relationship
enhancement may also lead to accuracy motivation; people try to be
accurate so as to maintain their reputation. In that case, relationship-
enhancement and accuracy goals coincide.
Relationship enhancement similarly manifests itself in a desire to
foster a favorable impression of oneself in others. Scientists often be-
come invested in their theories (Kuhn, 1996); so it is with rumors
and their transmitters. The desire to manage and enhance a favorable
impression acquired by being "in the know" may affect content; those in
the know may resist rumor revision that would discredit their proposed
version of the situation. R. H. Turner and Killian (1972) observed such
resistance in rumors surrounding a man found dead in the backseat
of an automobile; the actual cause of death—alcohol poisoning—was
resisted by those who first found the body and were thus in the know.
Spreaders of the baby-sitter murder rumors (Peterson & Gist, 1951)
also modified rumors so as to enhance their own prestige. In a similar
way, exaggerations (e.g., "The whole family had perished") and drama-
tizations (e.g., "a house had come rolling down" [italics in original])
occurred in rumors surrounding a massive landslide (Sinha, 1952).
Why? Sinha proposed that in the lab, the transmitter is accuracy-
motivated, but in the field, one transmits partly for effect; that is,
the story is shared partly for diversion. Rumor content, especially in
situations that resist definition, may thus resist change toward accuracy
to enhance or maintain the impression one has formed in others
(Arndt, 1967).
Another aspect of the relationship-enhancement motivation is
more cognitive in nature: Despite implicit communication rules to
transmit accurate and truthful messages, accurate transmissions may
conflict with the goal of sharing a coherent message, that is, one that
is understandable, plausible, and acceptable to the hearer (Ruscher,
2001). Stereotype-inconsistent material, for example, may therefore
be dropped in the interests of "a tidy story" (p. 68). Peterson and Gist
(1951) proposed such a motivational interpretation of how snowballing
occurred in a rumor concerning murder of a baby-sitter: Transmitters
selectively emphasized or de-emphasized aspects of the rumor, drawing
from their beliefs about what was probable so as to make the rumor
seem more plausible. Shibutani (1966) has similarly proposed that
even accuracy-motivated rumor content tends to be consistent with
"presuppositions of the public" (p. 86). Relationship-enhancement—
168 R U M O R P S Y C H O L O G Y

by coherence—motivation may be most apparent in situations in which

the group is unable to ascertain the facts; rumor publics in the baby-
sitter murder and landslide rumors did not have reliable evidence
available.

SELF-ENHANCEMENT
Self-enhancement, also discussed in chapter 3 (this volume), is the
desire to defend against threats to one's sense of self and maintain
one's self-esteem. One way that self-enhancement motivation operates
is to ensure the circulation of rumors that are complimentary to one's
existing beliefs and attitudes. G. W. Allport and Postman (1947b)
noted that rumor "firms pre-existing attitudes rather than forming
new ones" (p. 182). F. H. Allport and Lepkin (1945) found that World
War II rumors of waste and special privilege were more likely to be
believed—and presumably continue to be transmitted—by people who
opposed the Roosevelt administration (see chap. 4, this volume). In
other words, rumor content may change to advance the process of
rationalizing and justifying existing beliefs. This is a relatively cool
(cognitive) process.
Another warmer (emotional) way that self-enhancement motiva-
tion operates is to promote the circulation of rumors that are comple-
mentary to existing unfulfilled wishes, biases, suspicions, and desires,
especially in situations in which no consensus develops (Shibutani,
1966). Rumor distortions have been viewed by a vein of researchers
as projections of repressed impulses introduced at each transmission
node. This research is best typified by Jung's (1910/1916) analysis of
a girls' school rumor about a student-teacher affair; the rumor repre-
sented a wish-fantasy. Through the mechanism of projection, rumors
vent underlying emotional tension (i.e., they relieve, justify, or explain
anxiety; Wilkie, 1986). For example, Lowenberg (1943) noted that
psychotic patients display a fear of poisoning and that in times of
national crisis, rumors of mass poisoning are common in "normals."
He concluded that—as with the psychotics—rumors of mass poisoning
are a projection of shock and fear. Why poisoning? Lowenberg pointed
to psychoanalytic theories that shock and fear became associated with
the oral zone during childhood weaning. In today's terms, we might
say that the child formulates an illusory correlation between any illness
or discomfort he or she experienced during this period and the act of
ingesting new types of food. Adult anxiety in a crisis then rouses this
early association and even normal adults become primed to believe
rumors of mass poisoning. In more broad terms, another psychoanalyst,
Ambrosini (1983), posited that rumors project intrapsychic anxieties
on extrapsychic objects. Likewise, rumors have been explained as a
 Facilitating Rumor Accuracy and Inaccuracy 169

justification for anxiety through the mechanism of cognitive dissonance

(Festinger, 1957; Prasad, 1950; R. H. Turner & Killian, 1972). Disso-
nance is the psychological tension arising from contradictory attitudes
or actions; for example "I feel anxious but there is nothing to feel
anxious about." At a more conscious level, rumor distortion depends
partly on what a person desires or expects to be true (Turner & Killian,
1972). For example, friends tended to transmit favorable rather than
unfavorable rumors regarding their friend who had been abruptly with-
drawn from class; nonfriends showed the reverse pattern (Schachter
& Burdick, 1955).
Self-enhancement motivation also operates by fostering rumor
content that derogates outgroups and thereby enhances the prestige
of the ingroup. A rumor sketching a negative characterization of them
makes us feel better about we—and, by extension, me. The Kamins et
al. (1997) study discussed earlier in this chapter and in chapter 3 (this
volume) supports this idea: MBA students were more willing to transmit
rumors that reflected negatively on a rival school (fall in ranking) than
they were willing to transmit rumors that the rival school's rankings
rose. The RIT-UofR study presented in chapter 3 (this volume) margin-
ally replicated this finding: With rumor recipient as ingroup and target
of rumor as outgroup, negative rumors were marginally more likely
to be transmitted than were positive rumors' (see Figure 3.4). In a
similar way, rumors of violent acts and atrocities typically portray per-
petrators as outgroup members. For example, P. A. Turner (1993)
documented rumors circulating among both African and Caucasian
groups that the other group was cannibalistic. Self-enhancement moti-
vation explains why people typically encounter a large number of
wedge-driving rumors—rumors that are hateful or hostile toward out-
groups—but rarely hear rumors that are negative or critical of the
ingroup. On the whole then, the evidence indicates that people prefer
to transmit negative rumors about the outgroup.

Situational Features

Situational features refer to the conditions and circumstances of the

rumor episode that bear upon accuracy. These features include collec-
tive excitement, ability to check rumor veracity, and time.

'((45) = 1.56, p Mailed = .06.

170 R U M O R P S Y C H O L O G Y

HIGH COLLECTIVE EXCITEMENT

We have discussed how individual trait and state anxiety may exacer-
bate cognitive processes that usually inhibit accuracy. Collective
excitement—situations in which many or all individuals in a situation
are anxious—may intensify such effects by increasing suggestibility (dis-
tortion of perception) and diminishing critical ability. Crowd milling,
for example, may catalyze restless individuals into an excited mob that
immediately acts on an inaccurate rumor (R. H. Turner & Killian, 1972).
Sinha (1952) observed that suggestibility after catastrophic landslides
was heightened because of high anxiety: People set aside their critical
set, did not scrutinize rumors, and did not desire to verify. Shibutani
(1966) proposed two general sorts of rumor deliberation patterns that
hinge on anxiety: deliberative ("If unsatisfied demand for news is moder-
ate, collective excitement is mild, and rumor construction occurs
through critical deliberation"; p. 70) and extemporaneous (in situations of
intense collective excitement, rumor construction becomes behavioral
contagion). Mausner and Gezon (1967) provided an example of extem-
poraneous rumor construction: A grade school temporarily closed be-
cause of unfounded fears of an outbreak of vaginal gonorrhea among
girls (only 3 out of 173 girls had it, but many exhibited some symptoms);
this rumor exemplified contagion in that it was marked by high collec-
tive excitement and suggestibility.
High collective excitement may also enhance the development of
less stringent norms of verification. In critical situations when formal
lines of communication are closed, informal networks that are tempo-
rary and unstable form. New norms for evaluating information and
deciding behavior may emerge: "This is what everybody is saying!"
(R. H. Turner & Killian, 1972, p. 32). Less stringent norm development
is enhanced in close groups rather than in situations that promote
communication, as in a crowd in which people may become sensitized
to others in the group. However, people may already be sensitized
because of existing group ties. In such "organized" groups, emergence
of group norms in a fear situation (smoke billowing into a room) was
quicker and more intense than in "unorganized" groups; organized
groups showed more fear and reacted more quickly (Turner & Killian,
1972, pp. 38-41).

CAPACITY TO CHECK
Given an accuracy motivation, people strive to check the validity of
information; rumor accuracy increases when they can do so (Shibutani,
1966). Military rumors in Caplow's (1947) study were accurate partly
 Facilitating Rumor Accuracy and Inaccuracy 171

because military personnel could ask superiors about the veracity of

rumors (superiors could then confirm that the rumor was false). In
deliberative situations, people strive to check the veracity of informa-
tion. For example, people called a rumor control center during a gradu-
ate student worker strike to check on the veracity of rumors; the center
became a new source of external information and many inaccurate
rumors were presumably squelched in this way (Weinberg & Eich,
1978). Even during catastrophes, people check the quality of the source
of the rumor; for example, Port Jervis, New Jersey, residents had heard
rumors that the dam would burst but left only when these rumors
were spread by the fire department—an authoritative source (R. H.
Turner, 1964).
Veracity checking is not possible or may be greatly encumbered in
a number of situations. ST without discussion obviously exemplifies
one of these situations. False eyewitness perceptions that are unduly
trusted constitute another. Rosenberg (1967) described a news account
that was inaccurate from the outset and remained inaccurate because
newspapers that picked up the story failed to check its accuracy. News-
papers similarly failed or were unable to check rumors of mass anarchy
following Hurricane Katrina (Dwyer & Drew, 2005). Situations with
novel channels of communication constitute yet another situation in
which ability to check veracity is difficult or impossible. Exaggerations
and fabrications arose after a devastating earthquake had destroyed
normal channels of communication; these reports could not be checked
for a long time (Prasad, 1935). In a similar manner, military units
undergoing radical personnel changes saw a flurry of initial inaccurate
rumors until more normal lines of communication could be reestab-
lished (Caplow, 1947).
The perceived urgent need to act before information can be checked
constitutes a fourth situation in which checking is constrained (Prasad,
1935; Shibutani, 1966). People in situations in which delays may have
adverse consequences may perpetuate inaccurate rumors because the
consequences of taking time to check rumor veracity may be extremely
negative should the rumor prove true. One must act quickly, for exam-
ple, to alert one's friends that a computer virus, identified only by the
presence of a teddy bear icon, has infected one's computer and spread
to all addresses on one's address list ("JDBGMGR.EXE", 2002; see also
Weenig, Groenenboom, & Wilke, 2001). If the rumor is true, damage
to a friend's computer may be averted. (Concerned associates of ours
actually spread this false rumor out of a perception that they had to
act quickly.) The headlights hoax described in chapter 2 (this volume)
spread quickly for the same reason: Almost all of the graduate students
and faculty we interviewed immediately spread the rumor to friends
172 R U M O R P S Y C H O L O G Y

and loved ones in part because they felt they had to act quickly to
prevent violent death. Keep in mind that in these last two examples,
the hoaxes spread among very skeptical people: academic psychologists!
A fifth situation in which checking is constrained is one in which no
firm information is available (Buckner, 1965). The baby-sitter murder
rumors persisted because police shrouded their investigation in secrecy
(Peterson & Gist, 1951). Even when firm information is available, the
information source must be trusted for rumors to become accurate
(R. H. Turner & Killian, 1972). Rumors surrounding the death of John
F. Kennedy, for example, almost certainly persist among those who
distrust government authorities and the Warren Report. However, the
false rumor that Procter & Gamble Corporation contributes to
the Church of Satan has been periodically and successfully squelched
by the publication of "truth kits" containing statements from trusted
religious leaders, such as Billy Graham, stating the rumor to be a false-
hood (Green, 1984; Koenig, 1985).
One's ability to check may also be constrained by rumor partici-
pants' proximity to valid sources of information. Accuracy of 12 planted
rumors was related inversely to distance from original source; employ-
ees hearing it firsthand held more accurate rumors than did those
hearing it second- or thirdhand (Walton, 1961). G. W. Allport and
Postman (1947) similarly observed that some secondhand reports are
accurate because someone in the rumor chain had access to firsthand
knowledge, was motivated toward accuracy, or had opportunity to
verify.
Closely related to a group's ability to check the validity of informa-
tion sources is the group's ability to internally check consistency of
information by comparing rumors. Rumors may be compared with one
another to successfully deduce accuracy under the assumption that
more authentic portions of the rumor will appear more often. For
example, a dozen Korean War veterans recently recalled how they
killed approximately 300 South Korean civilians in 1950 at the village
of No Gun Ri (Choe, Hanley, & Mendoza, 1999). The accounts differed
only in details; by retaining common elements of these accounts, news
reporters were able to create an accurate reconstruction of this event.
Such comparison is possible if a rumor is actively recirculating within
a group (Buckner, 1965; DiFonzo & Bordia, 2002a).

TIME
The findings related to situational feature of time are mixed; with
time, true rumors sometimes surface. For example, rumor survival was
related to the accuracy of Caplow's (1947) military rumors. Yet Hershey
found no relation between persistence and accuracy (as cited in Hell-
 Facilitating Rumor Accuracy and Inaccuracy 173

weg, 1987, p. 217). Buckner (1965) proposed that for groups that
possess the ability and motivation to achieve accuracy, time tells all.
Inaccurate rumors in such groups may initially proliferate during an
invention stage (R. H. Turner, 1964) but, as we saw in the 1996 Field
Study, false rumors are quickly discarded. However, for those groups
that possess neither the inclination nor the ability to achieve accuracy,
time results in further inaccuracy.

Group Mechanisms

Group mechanisms refer to aspects of the social entity through which

the rumor spreads; especially those processes associated with group
identity, norms, and influence. Those processes most pertinent to rumor
accuracy include conformity processes, culture, and group epistemic
norms.

CONFORMITY
Once consensus is formed, conformity is demanded (Festinger et al.,
1948; Firth, 1956; R. H. Turner & Killian, 1972). In an attempt to
ascertain which prison inmates snitch, hypotheses are tested and
sources checked as in the rumor construction process; once a hypothesis
is accepted, however, data are reorganized and conformity is demanded
(Ackerstrom, 1988). In a similar way, in crowd formation, if some
group members assent to one definition of a situation over others, it
becomes difficult to advance other proposals (e.g., if someone in the
crowd yells, "Police brutality!"; Turner & Killian, 1972). To the extent
that such formulations are incorrect, inaccuracy is perpetuated.

CULTURE
Shibutani (1966) noted that, even though people retain critical ability in
both deliberative and extemporaneous rumoring, the emergent rumors
tend to be plausible to the rumor public—thus they tend to agree with
cultural axioms. Rumor content is thus partly shaped by group biases.
Information is invented, distorted, or ignored to fit the main theme of
the rumor. Prasad (1950) categorized rumors surrounding earthquakes
from 1934 and from 1,000 years prior, and from countries other than
India. It is surprising that he found common content and themes.
He posited that such commonality could not come from individual
projection of "complexes" (p. 129), emotional stress, or archetypes.
174 R U M O R P S Y C H O L O G Y

Rather, he argued they arise from a common "attitude" (p. 129)—

drawn heavily from cultural heritage—arising from the earthquake
situation. Rumor construction is like a picture-completion task, with
attitude driving the picture. In earthquake situations in which the
causes of earthquakes were unknown, speculations based on "the tradi-
tional and cultural heritage peculiar to the group" (p. 7) were adopted.
These common cultural attitudes explain rumor content; for example,
the earthquake resulted from planet alignments and Himalayan deity
indignation. Cultural effects are not limited to uneducated persons.
Even scientists and intellectuals entertained these rumors.
Rumor content is thus, at least in part, anchored to group beliefs
because social representations (collective understandings of general
topics) are anchored to group beliefs (Lorenzi-Cioldi & Clemence,
2001). Organizational culture informs individual schemas and thus
affects the individual's sense-making process (S. G. Harris, 1994), and
therefore the content of rumor construction. Shibutani (1966) posited
that skeptics are not immune to cultural anchoring: In deliberative
rumor construction activities, the content may change so as to convince
skeptics. For example, an authority may be cited, and the rumor may
accord with "interests, sensitivities, and beliefs" (p. 85) of the group.
Cultural anchoring of rumors is well documented. Knopf (1975)
posited that rumor content is the crystallization of culturally sanctioned
racial hostility. Kapferer (1989) stated that rumor content is the
expression of collectively shared but repressed fears and beliefs. G. W.
Airport and Postman (1947b) noted that "all rumors are liable to ...
cultural assimilation" (p. 157). They propose that as rumors circulate,
they lose their individual features and come more to resemble broadly
cultural ideas and familiar linguistic phrases (i.e., they become conven-
tionalized). They also note that rumors are elaborated to conform to
cultural frameworks. For example, in Bartlett's (1932) ST work, Hindu
subjects were prone to adorn a tale with a moral in accord with
Hindu fable.

GROUP EPISTEMIC NORMS

Closely related to culture, group norms of what evidence is counted as
acceptable have been theorized to be associated with accuracy. Buckner
(1965) posited that groups that possessed "meager standards of evi-
dence" (p. 57) were less likely to achieve accurate rumor hypotheses.
However, those groups characterized by skepticism tend to arrive at
more accurate conclusions. We found correlational evidence supportive
of Buckner's thesis (DiFonzo &• Bordia, 2002a); groups characterized
by PR personnel as skeptical tended to produce more accurate rumors
than did those characterized as gullible.
 Facilitating Rumor Accuracy and Inaccuracy 175

Network Mechanisms

Communication networks are organizational structures of relations be-

tween entities: usually individuals, but also organizations and groups
(Monge & Contractor, 2000). Attributes of the relations between these
entities, rather than attributes of the entities themselves, are thus the
focus of network analysis. Emergent communication networks are in-
formal and naturally occurring (e.g., the grapevine), as opposed to
formal networks, which usually correspond to organizational structure
charts. Information diffusion—including rumor transmission—occurs
on, within, or through these communication networks. In this section
we consider how three aspects of information diffusion on communica-
tion networks—the extent to which discussion occurs in transmission,
the pattern of information transmission, and the age of the information
channel—each affect accuracy.

INTERACTION
Interaction refers to discussion that occurs between sender and receiver
during transmission (Buckner, 1965). Interaction may involve redun-
dant communication (repeating the message), clarification, compari-
son, and interpretation. It has also been labeled "free feedback" (Leavitt
& Mueller, 1951) and "reciprocity" (D. L. Miller, 1985); we sometimes
referred to it earlier as discussion. Interaction is generally associated
with more precise transmission (Buckner, 1965, McAdam, 1962; R. H.
Turner & Killian, 1972). Messages—consisting of descriptions of geo-
metric patterns—transmitted under conditions of free feedback re-
tained accuracy whereas serially transmitted messages exhibited distor-
tion (Leavitt & Mueller, 1951). However, ST—without interaction—
invariably leads to distortion (G. W. Allport & Postman, 1947b; D. L.
Miller, 1985; Peterson & Gist, 1951). When people are allowed to
verbally interact, even ST distortion is reduced (McAdam, 1962).

TRANSMISSION CONFIGURATION
Rumor transmission configurations can differ substantially. An ST net-
work consists of members transmitting information from one to another
along a single chain. Cluster (C) transmission patterns refer to transmis-
sion in which information is told to a cluster of people, some of whom
don't pass it along, and some of whom pass it along to other clusters
(K. Davis, 1972). Those who pass it along are called liaisons. Multiple
interaction (MI) refers to transmission in which "many people hear the
176 R U M O R P S Y C H O L O G Y

rumor from more than one source" (Buckner, 1965, p. 62). MI awaits
a precise conceptualization but it clearly includes interaction and recir-
culation of the rumor. We have operationally denned MI as conceptu-
ally similar to rumor "activity" (DiFonzo & Bordia, 2002a). When rumor
activity is high, many people hear versions of the rumor from multiple
sources (MI), they interact (discuss) with one another, and the versions
often recirculate (e.g., X tells Y, Y tells Z, then Z tells X). Most grapevine
studies within organizations have found a C rather than ST or MI
pattern of transmission (K. Davis, 1972; Hellweg, 1987). (In the follow-
ing discussion, remember that C stands for cluster pattern of transmis-
sion, ST for serial transmission pattern, and MI for multiple interac-
tion pattern.)
Transmission configuration has at least two implications for accu-
racy. First, ST and C patterns may grant liaisons more influence over
rumor accuracy than do MI patterns. Keith L. Davis found that approxi-
mately 20% of the average network were liaisons (1972, p. 264). Some
persons in the network—usually liaisons—contribute more to rumor
construction and therefore affect content more than do others (R. H.
Turner & Killian, 1972). Thus, in ST and C transmission, content toward
or away from accuracy may especially depend on characteristics of the
liaison subset. MI may diminish the influence of this subset.
Second, ST and C patterns of transmission are likely to show some
distortion because of cognitive mechanisms (discussed previously) at
each node of transmission. As noted, such distortion is mitigated
through interaction, especially if the topic is outcome relevant to partici-
pants. MI patterns of transmission have the capacity to correct or accen-
tuate such distortions. Buckner (1965) proposed that group orientation
plays a moderating role in the relation between MI and rumor accuracy.
Group orientation encompasses several situational, motivational, and
network factors that result in either a critical or uncritical set. Critical
set orientation refers to an unspecified combination of factors such as
ability to check, high standards of evidence, and established channels;
uncritical set orientation refers to the lack of these attributes. For groups
with a critical set, MI should be associated with accuracy; more interac-
tion and recirculation in these fortunate groups should result in more
valid hypotheses. This idea is consistent with Nisbett and Ross's (1980,
p. 267) suggestion that group interaction has the capacity to greatly
mitigate common inferential biases. For groups with an uncritical set,
MI should be negatively associated with accuracy; more interaction
and recirculation in these groups will result in inaccurate and bias-
laden hypotheses.
Reanalysis of data from a recent study of PR officers (DiFonzo &
Bordia, 2002a; see Appendix 2.1) is partially consistent with the idea
that MI and accuracy are associated in groups with a critical set orienta-
 Facilitating Rumor Accuracy and Inaccuracy 177

1996 Field Interview Group Orientation Attribute Measures

1. Knowledgeability. Most of the group was knowledgeable about the subject matter of the
rumor (e.g., many witnessed the alleged event or nonevent; many were informed by the
plant manager as to whether or not the company is laying workers off).
2. Situation familiarity. Most of the group is familiar with the situation in which a rumor such
as this typically arises (e.g., journalists know that a few days before a national election,
rumors that a candidate leads an immoral life often spring up and therefore journalists tend
to be quite skeptical of these rumors).
3. Channel age. Most of the group heard the rumor within a stable interaction system and
thus know the reliability of the teller from past experience (e.g., if Bill Smith tells a rumor
and Bill Smith has been accurate or inaccurate in the past).
4. Defense motivation. Believing or disbelieving the rumor fills an emotional need for most
people in the group; that is, they wantto believe or disbelieve (e.g., rumors that the company
will be handing out large bonuses this year).
5. Situation urgency. Most of the group was in an urgent situation and was unable to take
the time necessary to investigate a rumor (e.g., "The dam has broken!").
6. Channel novelty. Most of the group was in a situation in which stable rumor channels were
disrupted and hence the reliability of the rumor transmitter could not be evaluated (e.g.,
crisis situations in which many people were milling about, all interested and involved in the
situation and seeking information).
7. Information nonavailability. Most of the group was in a situation in which absolutely nothing
was known (e.g., such as when the police would not comment on whether or not an
incident occurred).
8. Gullibility. Most of the group had very meager standards of evidence (e.g., they tended to
accept what people said without questioning it).

Note. All attributes are adapted from Buckner (1965). Attributes 1 to 3 are conducive to critical set; attributes 4
to 8, to uncritical set.

tion. These professionals recalled harmful or potentially harmful rumor

incidents and rated them on a variety of indices including MI and
accuracy. MI in this sample was associated with rumor accuracy. In
Buckner's framework, such a relationship ought to apply only to
critical-set-oriented groups. To test this idea, we reanalyzed the data
to assess critical set orientation. Buckner proposed that group skepticism
and established communication channels are elements of a critical set
orientation (see Exhibit 7.1: Group skepticism would be indicated by
low group "gullibility" and how established channels of communication
are by channel age and novelty). The group, as a whole, was neither
skeptical nor gullible, but did possess established channels of communi-
cation.2 Results of an experiment performed by Komarnicki and Walker

2
We conducted one-sample t tests against the neutral (midpoint) scale value for
group skepticism and how established group channels were: skepticism t(60) = 1.10, p =
.28, established channels £(59) = 3.31, p = .002.
178 R U M O R P S Y C H O L O G Y

(1980) are also consistent with these ideas. As in G. W. Allport and

Postman's (1947b) studies, participants were asked to pass along infor-
mation they heard to the next person in line. Those communicating
in line structures yielded the usual ST-type distortions. However, par-
ticipants communicating in more complex structures in which they
could change locations in line and immigrate to other lines yielded
precision accuracy. More complex network structures mitigated
ST-type distortions.
Buckner (1965) further proposed two group-level variables affect-
ing MI: group structure and outcome-relevant involvement. First, MI
is affected by group structure, which is how close or diffuse the group
is. In diffuse groups, contact tends to be through chains and therefore
multiple interactions are not likely. An individual in a diffuse network
might hear a rumor only once. In close groups, members enjoy close
and continuous contact; therefore MI is more likely. Close groups may
be temporary (e.g., crowds) or long-standing (e.g., small towns, college
fraternities, cliques, mental wards). In close groups the individual is
likely to hear the same rumor many times. Empirical evidence supports
this statement. People closely involved in friendship networks were
more likely to have heard a rumor that a community worker was a
communist (Festinger et al., 1948). Military rumors tended to diffuse
within established groups rather than between them (Caplow, 1947).
Second, Buckner proposed that MI is related to group involvement
with the topic; groups that are highly interested in a rumor are more
likely to exhibit MI.
Buckner's two-factor theory of MI may be summarized as follows.
First, close groups with high involvement in the rumor should produce
much interaction and recirculation—hence the high levels of MI.
Caplow's (1947) study of military rumors, Schachter and Burdick's
(1955) field experiment at a girls' school, Peterson and Gist's (1951)
study of babysitter murder rumors, and R. H. Turner and Killian's
(1972) studies of crowd behavior exemplify this situation. Second,
diffuse groups with high rumor involvement should produce serial
chains of communication and a few patches of recirculation resulting
in moderate levels of MI. We speculate that some Internet rumors
exemplify this category. Third, close groups with low rumor involve-
ment will result in a few small serial chains and patches of recirculation,
again producing only moderate MI. Finally, diffuse groups with low
rumor involvement would produce very short serial chains and low
ML Sedivic's (1987) planted rumor did not disseminate widely in part
because of low interest (involvement) and physical barriers to dissemi-
nation (departments were located far from one another and were thus
diffuse). Additional support for this two-factor idea comes from Back
et al. (1950) who conducted a participant observation study in which
 Facilitating Rumor Accuracy and Inaccuracy 179

seven rumors were planted in an organization over 4 months. Two

rumors about lost data (morale committee questionnaires) spread
quickly to morale committee members, to whom it was important;
the remaining rumors were relevant to the organization but varied in
their spread.

CHANNEL AGE
Shibutani (1966) noted two types of communication channels: institu-
tional (formal) versus auxiliary (normal everyday informal contacts).
In deliberative (low anxiety) rumor construction situations, people use
auxiliary channels. In extemporaneous (high anxiety) rumor construc-
tion situations, people use auxiliary channels and any other sources of
information they can obtain. Channels are likely to differ with respect
to age. Institutional and auxiliary channels are likely to be established
or stable; other channels are likely to be novel or unstable.
Established, rather than novel, channels may lead to accuracy be-
cause information sources could be easily tagged for validity (Buckner,
1965). For example, Caplow (1947) attributed high accuracy rates in
part to the ability of increasingly solid (established) channels of commu-
nication to label the quality of information sources (e.g., "This came
from Joe so don't trust it; he never gives us good information"). Caplow
noted that unreliable informants were excluded from the network,
which thus increased rumor accuracy.

Recent Empirical
Evidence

In this section we specify hypotheses based on the previous discussion—

especially Buckner's (1965) framework—and report research testing
these. In these investigations, we focused on factors related to group
orientation. We hypothesized that rumors in groups possessing a critical
set orientation ought to be more accurate; rumors in groups possessing
an uncritical set ought to be less accurate. Because Buckner did not
specify how critical set variables might interact, no hypothesis was
made regarding the possible combinations of critical set variables (e.g.,
groups able to check but not possessing established channels). We
also hypothesized that group orientation variables would moderate the
Mi-accuracy relationship.
180 R U M O R P S Y C H O L O G Y

1996 FIELD INTERVIEWS

We used data from the 1996 Field Interviews (described in chap. 6,
this volume) to explore the first set of hypotheses and collected quanti-
tative data regarding rumor accuracy and orientation. These data were
subjected to qualitative analyses only (because of a small ri) and served
as a springboard for further discussion during the interviews. We deter-
mined accuracy by asking participants to estimate the percentage of
the rumor (or rumor variants) that was true (accuracy rates were
presented in Table 6.3). Group orientation attributes were explored
with a series of questions that described each orientation attribute and
asked respondents how characteristic the attribute was in the rumor
incident they were describing. All attributes were adapted from Buckner
(1965) and are presented in Exhibit 7.1. All items were followed by
the open-ended query "What indicated this to you?" in an attempt to
evaluate responses in terms of observable events.
Qualitative results were consistent with the hypothesis that a criti-
cal set orientation is associated with rumor accuracy. One case exempli-
fied a less-than-critical-set orientation: Members in this organization,
whose president was rumored to be working secretly for a government
intelligence agency, were rated as not knowledgeable about the subject
matter of the rumor, unfamiliar with this type of rumor situation, and
not part of an established interaction network. In addition, most people
in the group felt an emotional need to either believe or disbelieve the
rumor, and were in a situation in which absolutely nothing was known
(and thus checking could not occur). Rumors persisted for extended
periods and remained substantially distorted compared with the re-
maining episodes (40% of the rumors were false).
The remaining cases exemplified the efficacy of groups with critical
set orientations in discerning veracity. For example, false rumors among
concerned workers that a parent company intended to sell a subsidiary
eventually became more accurate as they circulated and recirculated
among the subsidiary's workers. In this case, organizational members
had been presented with this type of rumor situation in the past (this
rumor concerned the latest in a series of company reorganizations),
and, like the journalists in Buckner's (1965) example, they were ac-
quainted with situations in which this type of rumor might arise.
Another case also illustrated the power of the group to use the
rumor mill to collectively procure the facts. Prior to the announcement
of a major reorganization that resulted in the closing of an entire
operations center and the relocation of hundreds of employees, workers
were obsessed with gathering information and rumors percolated inces-
santly throughout the organization. The rumors also became increas-
ingly accurate, making the eventual formal announcement redundant.
Leaks presumably spouted in the face of the irresistible force of employ-
 Facilitating Rumor Accuracy and Inaccuracy 181

ees' "need to know" to plan accordingly. One manager visiting the

operations center remembered "getting a lot of questions asked of me."
Even though she was unable to give out any information, she marveled
at the productive use of rumors in collective reasoning:
One person would see something, for example, that they're not
going to fill this vacant position out here but we've got more
people here—this doesn't make sense—why are they doing
this?—and put this together with another piece of it—kind of
talking out loud to one another—and they'd start saying "Geez,
that makes sense!"
Another manager described it as the fitting together of "little pieces of
a puzzle." In Buckner's terminology, this group was composed of stable
communication channels in which each member's reliability was
known; MI probably enhanced this group's ability to produce accu-
rate rumors.

STUDENT RUMOR SURVEY 2

Data from Student Rumor Survey 2, which is described in chapter 6
(this volume), was also used to explore these hypotheses. Group orien-
tation variables were group skepticism and how established communi-
cation channels were,3 We performed hierarchical moderated regres-
sion analysis to test the moderating role of skepticism and (separately)
established channels on both accuracy and the Mi-accuracy relation-
ship. Results—displayed in Table 7.2—indicated that skepticism pre-
dicted accuracy in both the main effects and moderated models. More
skeptical groups tended to produce more accurate rumors overall. Yet
in addition, as hypothesized, skepticism moderated the Mi-accuracy
relationship, as illustrated in Figure 7.1 by the slopes of regression
lines linking MI to accuracy under high, average, and low levels of
skepticism.4 In high-skepticism groups, accuracy increases with MI;
more interaction and recirculation brought about greater accuracy. In
low-skepticism groups, accuracy did not improve with increased MI.
Established channels also predicted accuracy in both main and moder-
ated effects models. Contrary to hypothesis, however, established chan-
nels did not moderate the relationship between MI and accuracy.

J
We reverse-scored group skepticism by asking participants an 8-point version of
question 22 in Appendix 2.1. How established communication channels were was as-
sessed with question 16 in Appendix 6.1. MI was obtained by averaging responses to
items 14 (interaction) and 15 (recirculation) in Appendix 6.1.
4
These slopes were computed with Table 7.2 moderated model regression coeffi-
cients under different values of the moderator variable skepticism (Aiken & West, 1991).
The values used to reflect high and low skepticism were one SD above and one SD below
sample means, respectively.
182 R U M O R P S Y C H O L O G Y

Results of Hierarchical Moderated Regression Analysis Predicting Accuracy

Variable Main effects model Moderated model

Moderating variable: skepticism

Ml .21M.12) .17 (.12)

Skepticism .30** (.10) .29** (.10)
Ml x Skepticism .12* (.06)
AR2 .02*
Model R2 .04** .06***
Adjusted R2 .04 .05

Moderating variable: established channels

Ml .12 (.12) .12 (.12)

Established channels .21* (.10) .21* (.10)
Ml x Established channels .05 (.06)
A/?2 .004
Model R2 .03* .03*
Adjusted fi2 .02 .02

Note. Ml = multiple interaction (see text). Ap < .10. * p<.05. **p < .01. ***p < .001.

Overall, these exploratory results suggest mixed support for the

moderating role of group orientation variables such as skepticism. Addi-
tional research in this area should more clearly conceptualize and oper-
ationalize group orientation variables as well as the MI variable.

Implications for
Organizational Rumors

We speculate that some of the previously mentioned mechanisms af-

fecting accuracy are regularly involved in organizational contexts. First,
for those rumor situations in which information is simply passed along
(e.g., "I heard that the boss's wife had a baby today"), attentional
narrowing and memory limitations may distort transmission. However,
such distortion is likely to be counteracted by simple interaction because
of redundancies and feedback. In sense-making situations (e.g., "I heard
that our division is being downsized; what did you hear?"), organiza-
tional members are likely to be motivated toward accuracy, knowledge-
able about issues surrounding the rumor topic (e.g., profitability of the
division), and able to check (e.g., by using leaks from well-positioned
sources). These features predispose accuracy. To the extent that com-
 Facilitating Rumor Accuracy and Inaccuracy 183

8- Skepticism
"" Low
• •ii Average
7-
— •High

6-

Iu 5 -

3-

2-

1-
[ I
-1SD OSD +1 SD
Multiple Interaction

Computed slopes of regression line (predictor: multiple interaction; outcome: accuracy)

at sample low, average, and high skepticism.

munication networks are established rather than novel, and when

group norms favor skepticism, accuracy is also favored. Furthermore,
high outcome-relevant rumors (e.g., about layoffs) are likely to be
subjected to multiple interaction, especially in close networks; MI in
these situations will enhance accuracy.
Thus far in this volume we have updated, explored, and expanded
a number of enduring themes in rumor research: rumor concepts,
effects, transmission, belief, sense making, and accuracy. We have ad-
dressed these themes from research done in a variety of contexts. In
the next two chapters, we focus on two recurring themes that occur
especially within organizational settings. The first of these is trust; the
second, managing rumors.
Trust and Organizational
Rumor Transmission

For 9 hours on a hot Sunday afternoon in Sydney,

Australia, police in riot gear battled 200 Aboriginal
youths (Chulov, Warne-Smith, & Colman, 2004).
Bricks, bottles, and firecrackers rained down on the
police, 40 of whom were injured during the melee. The
spark that ignited this tinderbox was a rumor that
police cars had chased 17-year-old Thomas Hickey on
his bicycle, causing him to fall and fatally impale
himself on a steel-spiked fence. Police insisted they had
been cruising the area looking for a bag-snatcher,
found the impaled youth, and attempted to resuscitate
him. Although a hot summer day and alcohol had
contributed to this rumor-sparked riot, distrust played a
foundational role.

friend whose company was facing the prospect of severe

A
downsizing mentioned to one of us (DiFonzo) that rumors
were rampant. "What has management said about it?" I
asked. He replied, "I wouldn't trust anything they say." We
have often noticed that rumor seems to thrive where there
is a dearth of trust. Yet this variable is rarely mentioned in
rumor transmission literature (see chap. 3, this volume). We
wondered, How is trust related to rumor transmission? In

An earlier version of this chapter can be found in DiFonzo, Bordia,

and Winterkorn, 2003. We thank Rob Winterkorn for his invaluable assis-
tance in collecting data analyzed in this chapter.

185
186 R U M O R P S Y C H O L O G Y

this chapter we approach this question by continuing to report the

results of a longitudinal investigation of the division of a company
undergoing a drastic downsizing. In chapter 2 (this volume) we intro-
duced this study and reported the correlates of hearing rumors; in
this chapter we address a different question: "How does trust in the
organization affect rumor transmission?"
This question has not hitherto been quantitatively explored. We
think it is important for organizational rumor transmission research,
of course, but it is also important for another topic of research in applied
psychology: namely, the role of trust in organizational phenomena.
Interest in this latter topic has increased in recent years (Dirks & Ferrin,
2002; McEvily, Perrone, & Zaheer, 2003; Robinson, 1996; Rousseau
& Tijoriwala, 1999). In particular, Dirks and Ferrin (2001) recently
proposed two ways—a main effects model and a moderating effects
model—by which trust affects organizational phenomena. We use these
models to frame our understanding of how trust affects transmission.
We begin with a discussion of how we think trust directly affects trans-
mission, then proceed to how trust moderates the by now well-known
relationships between uncertainty, anxiety, and rumor. We then pres-
ent our study investigating these effects. Our findings suggest that
trust—or more specifically distrust—is not peripheral to rumor trans-
mission; indeed, it seems to play a central role.

Direct Effects of Trust

on Rumor
Transmission

Trust is the willingness to be vulnerable because a person thinks some-

one has his or her best interests at heart (Rousseau, Sitkin, Burt, &
Camerer, 1998). Researchers have proposed slight variations of this
definition; trust has often been operationalized as both a specific and
generalized expectation of the benign nature of others' actions (Creed
& Miles, 1996; Kramer, 1999). Distrust, naturally, is the obverse of
this: It is the expectation of the malignant nature of others' actions.
Distrust has been linked to rumor activity. After qualitative analysis
of numerous rumor episodes, Shibutani (1966) concluded that when
formal information is not trusted, people compensate with informal
speculation, or rumor. For example, rumors in the former Soviet Union
were spawned by distrust of formal news sources (Bauer & Gleicher,
1953). Despite these findings, we know of no follow-up studies quanti-
 Trust and Organizational Rumor Transmission 187

tatively investigating trust's role in rumor transmission, especially

within the organization.
Dirks and Ferrin (2001) proposed a main effects model in which
trust acts directly on organizational variables and a moderator model
in which trust moderates relationships between organizational vari-
ables. In the main effects model, trust acts simply to produce positive
attitudes and cooperative behavior. Individuals' expectations of the
benign nature of management lead to more positive appraisals of ambig-
uous events and to behaviors congruent with these appraisals. In a
similar way, high trust in management should reduce rumor activity:
"Management told us work assignments are slow lately because of a
new central office accounting system. I trust them; they wouldn't lie
to us." However, low trust undermines formal communications and
increases the need for collective (informal) sense making. "Manage-
ment would lie to us—they are heartless and mean-spirited. I don't
believe their explanation; they intend to use this slowdown as a pretext
for layoffs." Therefore, low trust should lead to more rumor activity.
These ideas mesh well with Jean-Noel Kapferer's (1987/1990) emphasis
on rumor as unofficial or unsanctioned information, and with that part
of our definition (see chap. 1, this volume) focusing on rumor as a
means of managing risk. In sum, people rely more heavily on and
participate more frequently in the informal sense-making apparatus
(the rumor mill) if they don't trust or if they feel threatened by the
formal sense-making network (the boss or management).
Rousseau and Tijoriwala (1999) found evidence consistent with
this straightforward role of trust in management during organizational
change. Management explained a complex restructuring to registered
nurses in a U.S. hospital as being necessary to improve quality of patient
services. Trust predicted acceptance of these explanations. Nurses with
low levels of trust tended to discount these official accounts for change
or accounts focused on economic factors and to accept alternate expla-
nations based on self-serving managerial motives. These researchers
concluded, "High trust is likely to be associated with acceptance of
information provided by management and a reduction of information-
gathering efforts" (p. 524). Such alternate explanations are tantamount
to rumors; the information-gathering efforts that surround them are
tantamount to sense-making rumor activity. In short, distrust led to
rumor activity.
In a similar way, recent organizational rumor research has inciden-
tally noted that trust inhibited rumor activity (DiFonzo & Bordia, 1998).
In interviews with corporate management, we observed an association
between attitudes of distrust and rumor activity, whereas attitudes of
trust seemed to reduce rumor activity. For example, corporate employ-
ees concerned about a potential major reorganization exhibited high
188 R U M O R P S Y C H O L O G Y

levels of rumor activity and a lack of trust in management. Communica-

tions officials in one instance were considered almost malevolent by
employees; this company experienced rampant rumor activity. How-
ever, another communications officer claimed that trust had been con-
scientiously built up over the years; this company experienced relatively
short bouts of rumor and seemed to place faith in formal communiques.
We therefore propose that trust is negatively associated with the fre-
quency of rumor transmission.

Moderating Effects
of Trust

In the moderator model (Dirks & Ferrin, 2001), trust enhances or

inhibits relationships between other variables. Dirks and Ferrin offered
this example of the moderating effect of trust: High trust permits the
relationship between conscientiousness (a dispositional trait) and orga-
nizational citizenship behavior (OCB). Conscientious personality
should covary with OCB. Low trust, however, would inhibit the
strength of such a relationship; even people who are dispositionally
inclined to OCB will not exhibit such behavior in situations of low
trust because to do so would violate social exchange norms. Under
low trust then, conscientious personality would be unrelated to OCB.
However, under conditions of high trust, conscientious behavior should
be a good predictor of OCB; that is, the two variables would be related.
In this example, trust plays the role of a moderator variable, a catalyst,
an amplifier, a necessary but not sufficient condition.
It is important at the outset to note that a variable may have both
direct and moderating effects. If we continue with the Dirks and Ferrin
(2001) example, we see that trust may directly affect OCB (trust in
management leads to being a good organizational citizen) and it may
moderate the relationship between conscientiousness and OCB. We
return to this point later in the discussion.
Dirks and Ferrin (2001) proposed that one mechanism by which
trust moderates relationships is by affecting "one's interpretation of
another's past action or events relating to the past action: Under high
levels of trust, one is more likely to respond favorably to a partner's
action than under low levels of trust" (p. 459). Trust thus helps people
to interpret ambiguous actions by other people as friendly rather than
hostile in intent. However, low trust engenders a hostile, rather than
friendly, interpretation of events; even unambiguous and non-anxiety-
producing events may become infused with hostile intent.
 Trust and Organizational Rumor Transmission 189

In the context of rumor, trust may moderate the relationship be-

tween rumor transmission and its antecedents. Uncertainty and anxiety
have been linked with rumor transmission (see chap. 3, this volume).
Trust should moderate these relations in a similar way; this is depicted
in Figure 8.1. First, when trust is high, uncertainty and anxiety predict
rumor transmission. When people trust management, they engage in
rumor discussions only when they feel anxious or uncertain. Yet when
trust is low, uncertainty and anxiety may not predict rumor transmis-
sion because distrust of management—"management is evil (or unjust
or incompetent)"—leads to rumor transmission even when uncertainty
and anxiety are minimal. When trust is low, management's actions are
perceived to be hostile, even a small amount of anxiety and uncertainty
becomes magnified and leads to rumors. Distrust of management may
also lead to wedge rumors (rumors that are uttered mostly out of self-
enhancement reasons). Again, uncertainty and anxiety result from

• Low Trust • High Trust

1.0

0.9
0.8-
0.7-
0.6-
0.5-
0.4-
•O
O 0.3
o
0.2
0.1-1
0

Low High
Uncertainty or Anxiety

Hypothesized moderating role of trust on uncertainty-transmission and anxiety-

transmission relationships.
190 R U M O R P S Y C H O L O G Y

interpretations of other's actions, and these interpretations hinge on

trust.
Recent organizational rumor research findings are consistent with
the idea of trust as a moderator; trust may have also played a moderating
role in organizational rumor episodes we analyzed (DiFonzo & Bordia,
1998). In this research, distrust of the corporation may have engendered
perceptions that were conducive to the generation of rumors regardless
of levels of uncertainty or anxiety. That is, even small levels of uncer-
tainty and anxiety may have found fertile ground in attitudes of distrust;
thus a little uncertainty or a little anxiety—among attitudes of distrust
in the company—may have been increased through negative interpre-
tations and thus resulted in rampant rumor activity. High levels of
uncertainty and anxiety, of course, would also lead to much rumor
activity. In low-trust conditions then, the relationships between uncer-
tainty and rumor, and between anxiety and rumor, would be weak,
because people would be rumor active all the time.
Now consider a rumor episode we studied that seemed to exhibit
high-trust conditions: Layoff rumors quickly abated at a large consumer
products manufacturer after management limited uncertainty by out-
lining the general extent of layoffs, committed to a time line when
further information would be forthcoming, and explained why it could
give no further details (DiFonzo & Bordia, 1998). In interviews with
company employees, we noted that employees trusted management's
announcements because of previous trust-generating experiences such
as frequent supervisor-to-employee communication and quarterly
company-wide meetings. Some rumor activity was present, but quickly
diminished. Trust in the company may have reduced the effects of a
little uncertainty and anxiety. In other words, a little uncertainty or a
little anxiety—among attitudes of trust in the company—did not result
in rumor activity. However, high levels of uncertainty and anxiety
would of course lead to much rumor activity. Put another way, under
high trust, the relationships between uncertainty and transmission,
and between anxiety and transmission, appeared to surface.
In sum, we propose both a direct and moderating role of trust in
rumor transmission. In regard to the direct effect, trust should act to
inhibit rumor transmission. In regard to the moderating effect, low
levels of trust may lead to much rumor transmission regardless of
uncertainty and anxiety; under high levels of trust, however, uncer-
tainty and anxiety should predict transmission. We tested these ideas
in two of the four waves of the longitudinal sample of employees from
a division of a company undergoing radical downsizing. We gave an
overview this study in chapter 2 (this volume); here we describe it in
greater detail.
 Trust and Organizational Rumor Transmission 191

Sample, Procedure, and

Instrument

Study participants consisted of 75 employees in a division of a subsidiary

company of a large corporation.1 Two employees distributed and col-
lected each of the four waves (Tl, T2, T3, T4) of the questionnaire.
Each wave was administered approximately 1 month apart. Identifying
information was removed from each survey. Work slowdowns and
restructuring (merging of some departments in this and other divisions)
were occurring prior to the first and second wave of surveys. Layoff
announcements occurred between the second and third wave. Approxi-
mately 50% of employees were laid off between the third and
fourth wave.
Uncertainty, anxiety, and trust were measured with a 7-point scale
(see Exhibit 2.2). The number of different rumors heard during the
past month was recorded. Of these, the number of different rumors
passed was also recorded. By dividing the number of rumors passed
along by the number of rumors heard, we assessed the proportion of
heard rumors transmitted; this variable is the likelihood of transmission
(LOT). LOT operationalizes rumor transmission as the propensity to
pass along a rumor that one hears. LOT builds on previous measures
of rumor transmission that listed rumors that participants had heard
as well as whether or not they were passed along (Rosnow, Yost, &
Esposito, 1986; see also K. Davis, 1972; Esposito, 1986/1987; Rosnow,
Esposito, & Gibney, 1988; Schachter & Burdick, 1955). To maximize
the reliability and validity of LOT, we had participants record a summary
of the rumors that they had heard. As with previous measures of
transmission (e.g., DiFonzo & Bordia, 2000), LOT was correlated with
anxiety and uncertainty (see Table 8.1), which evidences the reliability
and validity of this measure. Furthermore, the components of LOT—
number of rumors heard and passed—both peaked at T2, which agreed
with informal impressions conveyed by organizational informants.

1
All 75 employees in this division received the questionnaire during wave 1 (Tl);
61 (81%) returned it completed. Seventy-two employees received the questionnaire
during T2; 48 (67%) returned it. Accurate response rates for T3 (« = 40) and T4 (« =
29) could not be calculated; however, estimates on the basis of layoff numbers at T3
and T4 indicated that at least 50% of surviving employees responded. Sample age group
F(3,163) = .22, p = .88, tenure f(3,163) = .17, p = .92, and gender proportions X 2 (3) =
2.76, p = .43, did not differ by wave.
192 R U M O R P S Y C H O L O G Y

Zero-Order Correlations and Alpha Coefficients

i 2 3 4 5 6 7 8
1.T1 LOT
2. T1 uncertainty .36** .77
3. T1 anxiety .39** .69** .87
4. T1 trust -.47** -.41** -.32* .86
5. T2 LOT .60** .22 .23 -.49**
6. T2 uncertainty .32* .52** .46** -.33* .28 .84
7. T2 anxiety .40** .61** .68** -.22 .27 .65** .96
8. T2 trust -.55** _ 44** -.30* .83** -.61** -.41** -.37** .87

Note. N = 60 for T1-T1 correlations; N = 46, 47, or 48 for all other correlations. LOT = likelihood of transmission
(proportion of heard rumors transmitted). Uncertainty and anxiety were transformed prior to correlation calcula-
tions. Alpha coefficients are in the diagonal. No alpha coefficients for T1 or T2 LOT could be computed because
these were single-item measures. *p < .05. **p < .01.

Operationalizing rumor transmission in this way is appropriate

given a number of different rumors in circulation over a period of
time—a condition typical of organizational rumor episodes. In addition,
LOT is independent of the number of rumors that a particular individual
hears. The advantage of this independence becomes apparent when one
considers that the alternative operationalization—number of rumors
passed—depends in large part on the number of rumors that one hears;
LOT accounts for this confound by being a within-subjects variable. By
extension, LOT also accounts for factors known to affect the number
of rumors heard, such as whether or not one is a liaison (K. Davis,
1972) and whether or not one is part of a close network (Buckner,
1965). Therefore, LOT affords the advantage that the results that obtain
in this investigation cannot be caused by factors associated with the
number of rumors heard.

Results

Raw variable means by wave are portrayed in Figure 8.2. Tl was a

time of mounting tension. Work slowdowns and some departmental
mergers had occurred; uncertainty and anxiety were at high levels.
Employees were clearly hearing layoff rumors with specific and consis-
tent numbers (25-60 employees), dates (first quarter), and departmen-
tal targets, as well as speculation about whether the division would be
sold or outsourced. There seemed to be some confusion as to work
 Trust and Organizational Rumor Transmission 193

FIGURE 8.2

8- ^" Rumors Heard

• ••i Rumors Passed
• • LOT
7-
— Uncertainty
— Anxiety
6- • - Trust

5-

4-

3- „"••••"
2-

1-

0-

Wave

Mean number of rumors heard, number of rumors passed, likelihood of transmission

uncertainty, anxiety, and trust by wave.

tasks also: Participants indicated in open-ended responses that work

assignments were "not detailed," "indefinite," and "nonspecific." The
picture presented is that of employees perceiving a dearth of work,
speculating as to the meaning of this for the organization and for their
jobs, and a reticent management. Comments evidenced a sense that
employees knew the division was in trouble but "hoped against hope"
that management would lead them with "a clear, well-thought-out,
plan." (We note at this point that one's level of trust in management
would greatly influence how these conditions were interpreted.)
T2 represents a climax; suspense peaked. T2 occurred prior to layoff
announcements. Rumor activity, uncertainty, and anxiety peaked, and
trust faltered. Employees' comments indicated the same queries as in
Tl, but also indicated increasing distrust—"The company made twice
194 R U M O R P S Y C H O L O G Y

as much money as they expected," the parent company is not commit-

ted to the division, management should "show us the commitment to
its programs," management should have "answered this question asked
directly to them; instead they deferred," and management is "keeping
everyone in the dark until the very end"—and increased anxiety—"Is
my job secure or not?" Many employees were dismayed that they had
not been told "the truth sooner."
T3 and T4 (after the layoffs) represent anger and moving-on. Uncer-
tainty and anxiety fell, and rumor activity plunged (although the LOT
remained stable). Some residual rumors of future layoffs and sale of
the division occurred. Although on average trust rose, comments indi-
cated that many employees were bitter: Management is not "honest,"
"Management tells its employees only what they want to in order to
keep them productive," "Employees are really being jerked around,"
"Who cares?" and "No sources can be trusted."
In the remainder of this section we describe results of tests for
effects on LOT. We specifically addressed the following questions: What
are the main effects of uncertainty, anxiety, and trust on LOT? What
are the moderating effects of trust? We performed eight hierarchical
moderated linear regression analyses.2 We used only Tl and T2 data
in these regressions because of lower sample sizes at T3 and T4. Four
regressions tested the main effects of uncertainty and the main and
moderating effects of trust on an uncertainty-LOT relationship; results
are presented in Table 8.2. Another four regressions tested the main
effects of anxiety and the main and moderating effects of trust on an
anxiety-LOT relationship; results are presented in Table 8.3. Of each
of the four sets of analyses, two were cross-sectional and two were
longitudinal.

MAIN EFFECTS OF UNCERTAINTY

AND ANXIETY
Main effects for uncertainty (Table 8.2) and anxiety (Table 8.3)
obtained. When uncertainty was the sole predictor of LOT in each

2
Missing data patterns were first analyzed longitudinally and no systematic patterns
of attrition were discovered. For example, participating in both Tl and T2 (versus only
Tl) questionnaire administrations was not correlated with Tl uncertainty, anxiety, LOT,
or trust. All possible wave combinations (Tl vs. T2, Tl vs. T3, T2 vs. T3, etc.) were used.
After means and SDs were computed for each variable by wave, the dataset was screened
for correlation and regression analyses with Tabachnick and Fidell (1996/2001). Outliers
(those values whose ?-scores were beyond ±3 SD from the mean) were changed to the
next most extreme score (1, uncertainty; 3, anxiety; 6, no. rumors heard, and 5, no.
rumors passed values). Multicolinearity was investigated with Mahalanobis distances;
one data point at Tl was deleted. Zero-order correlations were then calculated (Table
8.1 presents the Tl and T2 intercorrelations); all correlations did not go beyond +.85,
 Trust and Organizational Rumor Transmission 195

cross-sectional analysis, it predicted LOT; hence there was a main effect

for uncertainty. This result replicates previous research showing a rela-
tionship between uncertainty and transmission (Rosnow, 1991). The
effect disappeared, however, after trust was added, indicating that dis-
trust accounted for a large part of the LOT variance resulting from
uncertainty. Put simply, in a matchup between distrust and uncertainty,
the distrust variable wins hands down.
Anxiety displayed a similar pattern of results; when anxiety was
the sole predictor of LOT, in each cross-sectional analysis, it predicted
LOT. Hence there was a main effect for anxiety, and this result also
replicates previous research showing the anxiety-LOT relationship
(Rosnow, 1991). The effect disappeared or was greatly diminished,
however, after trust was added, indicating that distrust accounted for
a large part of the LOT variance resulting from anxiety. In a matchup
between trust and anxiety, the distrust variable again wins
convincingly.
These results point toward a more central, perhaps a more proxi-
mal, role for trust than heretofore theorized. With trust, we can account
for a much higher proportion of LOT variance than without it. These
results also extend previous rumor transmission research to a new
operationalization of transmission: LOT.

MAIN EFFECTS OF TRUST

Results suggest strong support for the main negative effects of trust on
both current and longitudinal rumor transmission, as predicted. As
shown in Tables 8.2 and 8.3, in the cross-sectional analysis for Tl, trust
was a significant negative predictor of LOT in the main effects models
that included either uncertainty or anxiety. It remained a significant
negative predictor in the moderated effects models. In other words, trust
dampened same-time LOT over and above the effects of uncertainty,
anxiety, or trust-moderation. This pattern also obtained in the T2 cross-
sectional analysis. Note that this same pattern also obtained for both
longitudinal analyses.3 Trust dampened future LOT over and above
the effects of same-time or future-time uncertainty, anxiety, or trust-
moderation. In sum, trust dampened both same-time and future LOT

indicating no bivariate collinearity. Bivariate scatterplots appeared to be elliptical, indicat-

ing linearity and homoscedasticity.
3
Tl trust was a significant negative predictor of T2 LOT in the main effects models
that included either Tl anxiety or uncertainty, and it remained a significant negative
predictor in the moderated effects models. Tl trust was even a significant negative
predictor of T2 LOT in the main effects models that included either T2 anxiety or T2
uncertainty, and it remained a significant negative predictor in the moderated effects
models.
196 RUMOR PSYCHOLOGY

Hierarchical Moderated Regression Analysis Predicting LOT at T1 and T2 Using Uncertainty

T1 Cross-sectional analysis: moderating variable: T1 trust

Outcome variable: T1 LOT (N = 60)
Variable Main effects models Moderated model
T1 uncertainty .09*** (.032) .05 (.033) .05 (.033)
T1 trust -.13*** (.043) -.13*** (.043)
T1 uncertainty x T1 trust .02 (.022)
.12*** .009
Model R2 .13*** .25**** .26****
Adjusted R2 .11 .22 .22

T2 Cross-sectional analysis: moderating variable: T2 trust

Outcome variable: T2 LOT (N = 47)
Variable Main effects models Moderated model
T2 uncertainty .06A (.032) .02 (.028) .02 (.028)
T2 trust -.18**** (.039) -.16**** (.043)
T2 uncertainty x T2 trust .03 (.021)
Atf .30**** .03
Model R2 .08A 38**** 41****
Adjusted R2 .06 .35 .37

Longitudinal analysis A: moderating variable: T1 trust

Outcome variable: T2 LOT (N = 46)
Variable Main effects models Moderated model
A
T2 uncertainty .06 (.033) .04 (.031) .03 (.030)
T1 trust -.14*** (.042) -.12** (.043)
T2 uncertainty x T1 trust .04A (.025)
ig*** .04A
2 A
Model R .07 26**** .31****
Adjusted R2 .05 .23 .26

Longitudinal analysis B: moderating variable: T1 trust

Outcome variable: T2 LOT (N = 46)
Variable Main effects models Moderated model

T1 uncertainty 05 (.035) .002 (.035) .003 (.036)

T1 trust -.15*** (.046) -.15*** (.047)
T1 uncertainty x T1 trust .003 (.020)
19*** .000
Model R2 .05 .24*** 24**
Adjusted R2 .03 .20 .19

Note. Data reported are beta coefficients with standard errors in parentheses. LOT = likelihood of transmission
(proportion of heard rumors transmitted). T1 = Time 1. T2 = Time 2. Ap < .10. **p < .01. ***p < .005. ****p <
.001.
 Trust and Organizational Rumor Transmission 197

Hierarchical Moderated Regression Analysis Predicting LOT at T1 and T2 Using Anxiety

T1 Cross-sectional analysis: moderating variable: T1 trust

Outcome variable: 77 LOT (N = 60)
Variable Main effects models Moderated model
T1 anxiety .13*** .09* (.038) .09* (.037)
T1 trust -.13*** (.041) -.11** (.040)
T1 anxiety x T1 trust .07* (.029)
A/?2 .13*** .06*
Model R2 .15*** .28**** 34****
Adjusted R2 .13 .26 .31

72 Cross-sectional analysis: moderating variable: 72 trust

Outcome variable: 72 JLO7 (N = 47)
Variable Main effects models Moderated model
T2 anxiety .09A (.048) .04 (.041) .05 (.041)
T2 trust -.18**** (.038) -.16**** (.039)
T2 anxiety x T2 trust .06 (.040)
A/?2 .31**** .034
Model R2 .07A 39**** 42****
Adjusted R2 .05 .36 .38

Longitudinal analysis A: moderating variable: 77 trust

Outcome variable: 72 LOT (N = 46)
Variable Main effects models Moderated model
A
T2 anxiety .09 (.049) .07 (.044) .09* (.044)
T1 trust -.14**** (.040) -.14**** (.040)
T2 anxiety x T1 trust .08A (.042)
A/?2 .21**** .05A
Model R2 .07A 28**** 33****
Adjusted R2 .04 .25 .28

Longitudinal analysis B: moderating variable: 77 trust

Outcome variable: 72 LOT (N = 46)
Variable Main effects models Moderated model
T1 anxiety .06 (.042) .03 (.039) .06 (.038)
T1 trust -.14**** (.042) -.11** (.042)
T1 anxiety x T1 trust .08* (.037)
A/?2 20**** .08*
Model R2 .05 .25*** .33****
Adjusted R2 .03 .22 .29

Note. Data reported are beta coefficients with standard errors in parentheses. LOT = likelihood of transmission
(proportion of heard rumors transmitted).
A
p< .10. *p < .05. **p < .01. ***p< .005. ****p < .001.
198 R U M O R P S Y C H O L O G Y

even when uncertainty and anxiety were accounted for; these main
effects were strong and consistent.

MODERATING EFFECTS OF TRUST

Results in Table 8.3 give strong support for the moderating effects of
trust on both current and longitudinal anxiety-LOT relationships, as
predicted.4 In the cross-sectional analysis for Tl, the addition of the
interaction term between anxiety and trust accounted for an additional
6.1% of the variance in LOT. The beta coefficient for this interaction
term was both positive and significant, indicating that high levels of
trust lead to a strong anxiety-LOT relationship, whereas low levels of
trust lead to a weak anxiety-LOT relationship. Figure 8.3 illustrates this
relationship by showing the slopes of regression lines linking anxiety to
LOT under high, average, and low levels of trust.5 As trust throughout
the sample was low, it should be noted that these values are not
absolute; indeed, high trust fell at approximately the scale midpoint.
Examination of Figure 8.3 illustrates that at low levels of trust, LOT
was uniformly high; low-trust participants passed along most of the
rumors they heard regardless of levels of anxiety. At high levels of
trust, however, anxiety covaried with LOT. This same relationship was
evidenced in T2 cross-sectional analysis, but was not significant.6
In the longitudinal analysis evaluating the moderating effect of
Tl trust on the Tl anxiety-T2 LOT relationship, the addition of the
interaction term accounted for an additional 8.3% of the variance, and
was also significant; this interaction is illustrated in Figure 8.4. The same
pattern as in the cross-sectional plot is evident here. The remaining
longitudinal analysis evaluated the moderating effect of Tl trust on the
T2 anxiety-T2 LOT relationship; the same pattern was again evident
but only marginally significant at conventional alpha levels.7 Even at
conventional significance levels, however, we observed that Tl trust

4
Hypothesized interaction effects for Tl and T2 were also tested in the hierarchical
linear regressions. In this procedure, the predictor and moderator variables are centered
before the interaction term is computed (Aiken & West, 1991). Centering refers to subtract-
ing the mean of the variable from each value. Interaction effects—indicating
moderation—are then tested in regressions after all main (centered variable) effects have
been entered.
5
These slopes were computed with Table 8.3 moderated model regression coeffi-
cients under different values of the moderator variable trust (Aiken & West, 1991). The
values used to reflect high and low trust were one SD above and one SD below sample
means, respectively.
>p = .12.
7
p = .07.
 Trust and Organizational Rumor Transmission 199

FIGURI 8,3

1.0- T1 Trust
^— Low Trust (-1 SD)
••••i Medium Trust (0 SD)
0.9- - • High Trust (+1 SD)

0.8-
ansmiss
P
~J
i

•8
1
P
O)
i

0.5-

0.4-

0.3-

1 T
-1SD OSD +1 SD
T1 Anxiety
Computed slopes of regression line (predictor: T1 anxiety; outcome: T1 likelihood of
transmission [LOT]) at sample low, average, and high T1 trust in the company.

moderated the relationships between Tl anxiety and both Tl LOT and

T2 LOT.
Results (presented in Table 8.2) assessing the moderating effect of
trust on both current and longitudinal uncertainty-LOT relationships
evidenced similar patterns, but were weaker. Cross-sectional interac-
tion terms were not significant; longitudinal interaction terms were
also not significant at conventional levels.8

s
p = .11 and p = .14.
200 RUMOR PSYCHOLOGY

1.0- T1 Trust
^•i Low Trust (-1 SD)
••in Medium Trust (0 SD)
0.9- • • High Trust (+1 SD)

c 0.8-
O
'55
_<o

I0'7
f
•5
•o 0.6-
o
o
0>
JC
0.5-

0.4-

0.3-

1 1
-1 SD OSD +1 SD
T1 Anxiety

Computed slopes of regression line (predictor: T1 anxiety; outcome: T2 likelihood of

transmission [LOT]) at sample low, average, and high T1 trust in the company.

In sum, trust moderated anxiety-LOT relationships: During Tl it

moderated the relationships between anxiety and Tl LOT; during T2
it marginally moderated the relationship between anxiety and T2 LOT.
Tl trust even longitudinally moderated the relationship between Tl
anxiety and T2 LOT, and marginally moderated the relationship be-
tween T2 anxiety and T2 LOT. Similar but nonsignificant patterns
obtained for uncertainty.
 Trust and Organizational Rumor Transmission 201

Trust as a Key Variable

in Rumor Transmission

Results suggest that trust is perhaps a more important—and more

foundational—variable in rumor transmission than is uncertainty or
anxiety. Strong support was found for the main and moderating roles
of trust in both cross-sectional and longitudinal anxiety-LOT relation-
ships. Trust negatively predicted same-time and future-time rumor
transmission. Trust during one time period also moderated the relation-
ships between anxiety from the same time period and both current
and future rumor transmission. Although in the predicted direction,
results were nonsignificant for the moderating role of trust in
uncertainty-LOT relationships.
Results not only extend the corpus of literature—rapidly increasing
in size—on the various roles that trust may play in organizations
(Kramer, 1999); they also are consistent with the framework proposed
by Dirks and Ferrin (2001) to explain when trust would manifest itself
as a main or moderating effect. These researchers proposed that trust
would manifest itself as a main effect in situations of "weak" strength—
that is, in situations that do not "provide clear or powerful cues that
lead individuals to interpret the events in a similar way" (p. 462). In
situations in which situational strength was in the "mid-range," they
proposed that trust would recede to a moderating role. Situational
strength in some situations may be in between weak and mid-range
and may thus manifest itself as both a main and a moderating variable.
Our results offer support for this conceptualization in two ways. First,
uncertainty-transmission effects (although practically important) have
been found to be weak to moderate (r= .19; Rosnow, 1991); we should
therefore expect trust to manifest itself as a main effect only, which
indeed occurred in every analysis we performed. Second, recalling that
anxiety-transmission relationships are strong (r = .48; Rosnow, 1991),
we should therefore expect trust to manifest itself as a moderator or
as both a main effect and moderator variable. The latter indeed occur-
red: Trust always manifested itself as a main effect and sometimes as
a moderating variable.
These results provide further insight into the high levels of rumor
activity typical of situations in which formal sources of information are
not trusted. Distrust may provide fertile soil for rumors that cannot
easily be squelched by dispelling uncertainty and reducing anxiety.
Such an instance occurred in the case of the external organizational
202 R U M O R P S Y C H O L O G Y

rumor plaguing the soft drink Tropical Fantasy in 1991. Remember

that false rumors circulated among the African American community
that the soda was owned by the Ku Klux Klan and contained substances
that would sterilize Black men. The rumor caused sales to drop by
70%. Commenting on the popularity of similar rumors in the African
American community, Lorraine Hale, an African American psycholo-
gist, stated,
Having come from a slavery background, where we were so
brutalized for so long, the sense of fear we have as a people is
very real. There's a mass paranoia that the objective here is to
kill us out, as easily and quickly as possible. We don't articulate
it, but we act on it. This leads to watchfulness and caution and
suspicion, enough to question the contents in a soft drink, (as
quoted in Lerbinger, 1997, p. 159; see also P. A. Turner, 1993)
In our terminology, distrust in a slightly uncertain situation led to high
levels of rumor transmission. The rumor may have resisted media and
government agency attempts to dispel uncertainty until then-mayor
of New York City David Dinkins, himself an African American, publicly
drank the soft drink (Freedman, 1991). In our terminology, trust in
the formal source of information was restored for this situation. The
point is, again, that trust was key.
Results also support the view of trust as a stable variable. Evidence
points toward stability of trust attitudes over time. Previous longitudinal
research by Robinson (1996) found a significant moderate correlation
(r = .34) between trust in employer measured at initial employment
and then after 30 months. We note that our 1-month Tl trust to T2
trust intercorrelation was the highest of all study variable pairs (r =
.83; see Table 8.1).
Trust may also be a self-perpetuating attitude aided by the social
mechanism of rumor. Robinson's (1996) research showed that trust at
time of hire exerted both main and moderation effects 18 and 30
months later. In particular, initial trust predicted and moderated later
interpretations of breach of psychological contract. Consistent with this
research, our results also evidenced longitudinal effects of trust for
impending layoffs that were often seen as unfair. Indeed, low-trust
individuals engaged in more rumor transmission. Higher rates of trans-
mission of negative rumors among low-trust individuals may help en-
trench interpretations that ensure subsequent deteriorations in trust
(cf. similar effects of gossip in Burt & Knez, 1996). As our study was
mainly concerned with rumors about negative outcomes, future re-
search should test this idea among organizations experiencing rumors
about positive outcomes. Such positive rumors presumably would help
to enhance trust over time. We also note that future research should
strive to identify other modes of measurement besides self-report.
 Trust and Organizational Rumor Transmission 203

Practical implications of these results include the importance of

addressing attitudes of trust in formal sources of information (see also
DiFonzo & Bordia, 2000). Dispelling uncertainty and reducing anxiety
may not dampen rumor activity without attention to levels of inter-
group trust. Rumors of plant closings and downsizings, for example,
may not be diminished by announcements to the contrary, unless the
source of the rebuttal is judged to be honest and trustworthy (Bordia,
DiFonzo, & Schulz, 2000). These implications seem particularly relevant
to situations involving organizational change, the focus of this study.
Explanations by management may effectively become impotent when
given to employees who do not trust management. Interpretations—
based on employee perceptions of the company as caring, honest, and
trustworthy—may determine the extent to which official explanations
are accepted and acted on.

Conclusion

In summary, the results of our study suggest that trust is a key variable
in rumor transmission and is likely to play a central role in organiza-
tional rumor activity in at least two ways. First, distrust in the organiza-
tion is likely to fuel rumor activity. For example, if an employee per-
ceives the company to be uncaring and dishonest, he or she is unlikely
to rely on their explanations to account for recent changes in personnel
that affect the quality of his or her job. Second, trust is likely to alter
the relationships between uncertainty, anxiety, and rumor. When trust
in the company is low, employees may be especially prone to engage
in rumor discussions regardless of their levels of uncertainty or anxiety;
when trust is high, such rumor discussions are necessary only under
conditions of high uncertainty or anxiety. For example, if I, as an
employee, perceive the company as uncaring and dishonest, even small
amounts of uncertainty and anxiety are enough to make me concerned.
I am then likely to participate in rumor discussions because I think
that my coworkers in the rumor mill—but not the management—have
my best interests at heart. Even rumors appearing during times of
quiescence and stability would receive lots of my attention because
they might protect me from dreaded consequences that the company
did not care about. However, when I trust the company, there is no need
to pay much attention to rumors because the company explanation can
be relied on; I need turn to rumors only when the company is unable
to quell my uncertainty or anxiety. Future research should seek to
replicate these patterns in other arenas—both field and experimental—
204 R U M O R P S Y C H O L O G Y

and should seek to further clarify the nature of the relationship between
trust and transmission. Future transmission research—in whatever
venue—should routinely measure trust in formal communication
sources.
In the next chapter, we continue in an organizational vein and
discuss empirical evidence pertaining to the management of rumor.
 Rumor Management

In the late 1990s, we were hired by the Procter &

Gamble Corporation (P&G)—a major manufacturer of
household products—as expert witnesses in a court
case involving perhaps the most famous external
corporate rumor of the 20th century (Fine, 1992;
Green, 1984; Marty, 1982; Mikkelson, 2003; Pinsdorf,
1987; we introduced this rumor in chap. 1, this
volume). Having first targeted the McDonald's
Corporation in the late 1970s, the famous—and false—
Satanism rumor has bedeviled P&G since 1981
(Koenig, 1985, p. 42). The rumor, which is entirely
false, states that the president of P&G appeared on The
Phil Donahue Show to say that he was "coming out of
the closet" with regard to his Satanism. The rumor, in
all its forms and facets, is categorically false, yet at
times its spread was wide enough to result in
approximately 15,000 phone calls per month to P&G.
P&G's response to this rumor has been multipronged.
Enlisting the help of high-ranking religious leaders,
they compiled a "truth kit" to debunk it, which they
sent to individuals who contacted them and to
churches in areas where the rumor was breaking out.
At times they marshaled a nationwide press campaign
denouncing the rumor. Finally, they sued—often
successfully—competitors who had spread the rumor.
These efforts have dampened rumor activity.

205
206 R U M O R P S Y C H O L O G Y

e have shown in this volume how rumors can be a problem. The false
rumor that McDonald's uses worm meat in its burgers led to a drop in
sales of up to 30% in some areas (Tybout, Calder, & Sternthal, 1981).
As noted earlier, the false rumor that Procter & Gamble is associated
with Satanism led to 15,000 calls a month on the consumer help-line
(Austin & Brumfield, 1991). Rumors of layoffs were associated with
increased change-related stress among employees of a metropolitan
hospital undergoing restructuring (Bordia, Jones, Gallois, Callan, &
DiFonzo, in press). Rumors that a police chase led to the accidental
death of an Aboriginal boy led to rioting and attacks on police in Sydney,
Australia (Chulov, Warne-Smith, & Colman, 2004). In situations like
these, effective management and control of rumors is vital to the man-
agement of the crisis situation. How can rumors be effectively managed?
In this chapter we review the literature on rumor management, includ-
ing our own empirical work, and present guidelines for preventing and
neutralizing rumors. We begin with a general review of the literature
on rumor-quelling strategies. Next, we explore in more detail the role
of rumor rebuttals (or denials) in reducing belief in rumors. Finally,
we provide some general guidelines on rumor management.

Rumor-Quelling
Strategies

In understanding the role of rumor-quelling strategies in curbing

rumors, it is useful to consider each strategy in the context of the life
cycle of rumors. Recall that the life cycle of rumors can be divided
into three stages: Generation, evaluation (or belief), and transmission
(DiFonzo, Bordia, & Rosnow, 1994). Rumor is generated in times of
uncertainty and anxiety regarding topics of high importance. Next, the
rumor is evaluated for plausibility or belief; this process occurs at the
individual and group levels. A rumor that meets the standards of plausi-
bility gets widely circulated in the transmission stage. The rumor-
quelling strategies can be aimed at each stage. Some strategies can
reduce generation (and thereby prevent rumors or reduce the inclina-
tion to participate in a rumor discussion); others reduce belief in the
rumor; and still others target transmission.
Table 9.1 provides a summary of rumor-quelling strategies drawn
from a review of psychological and business literature. From each paper
listed, we gleaned recommendations aimed at curbing rumors. The
specific recommendations were then condensed into general categories
of quelling strategies. The articles summarized in Table 9.1 include
 ec

c
"o
^

01
"3

o
g
3
VI

5,
itrategy Literature

O>
fi

Ji
«
(0
'5
Referei

DC

^
S
E

a
1

<*
O
t
C
A;
.c

*-
+J

dji
"!5

O
s

Q.
D

Q.
mor clinic": publication of rumor and st

O)

Ill f—
DC *>
4—

3 i5
CD

.Q
QJ

01
a:
ence.

•f-
1
00
og
c
Q

H-
r^

CJ
5i

D
2
E
2
1 CD

^
"^
CD
QJ

<U
5
<U

g
^ebut with facts from neutral stakehold

1
:

_Q £

3 3

CD

>
CD
O

oc oe
J2
in
C
ile lawsuits for libel; use private detect!

•o TJ
TO

n
'•4-*

c SS
c

1_

^ rvi rn
o
Remove trademark or change trademarl

o.
TO

*
<H
0)
(U

CD
CD
Df negative cues),

^~
+->
CD
0)

-a -a
CD

<D
.£3 .Q
flj

DC o:
3 D

*
'5

g
S
^ebut to consumers in affected regions

s -g ?
^

*
r

Q-

,_

IB
Q
O!
in

&
5J

CM"

,*d
ent them by
C

Cu
CD

.
>
Droviding information regarding import
g?

.
CD
Droviding emotional and economic secu
TO >»

TO 'C
DCOCQ:
DfD-Q

~S"S13

.
C C d J

3 3 3
CD
-PH-
fostering a cooperative environment.
tu.9^11

CC
4-
ite them by
CD
/anguishing the rumor with truth,
CD

lot repeating the rumor,

(U "^ '^

CD
C C
g* O O

•ebutting early,
C *wi '*/>

tQ j£^ E

CD

•ebutting by appropriate spokesperson,

0)

•ebutting face-to-face,
TO
C

CD

istening for gaps in information,

^T1

T—
oeocceccocococ
c T O T O f l j o t c c
^nooj.^

CD

"S"S"S"S"S^"S
3 3 3 3 3 3 3
D ^ ^ ^ ^ I D T O

istening for feelings.

>.
4_ H- t ^

rf(
S

LU
DO
00

o
*"

8
o
m

0
g
m
01

Q.
rnal rumors:
QJ
^

•1

(eep employees informed.

•S*

C
o
c
Q>

8
x

u
•W

Heed rumor (assess and address underiyi

D)
iu .fc

D TO i;
o
•£

<U
C

\ct promptly; don't repeat,

TJ T3 T> -a
3 3 3 3

^s **J
i* tu o ci
0:0:0:0:

inlighten personnel.
'—
-Q
c c 5 <u
flJ

(U

LLJ
rnal rumors:
'—

-a
3

oj
QJ

o:

tebut.
oi
-Q
!•& *
CD
.C
73

C
3

o
u
Rumor Management
207
208 R U M O R P S Y C H O L O G Y

c
o
'l/i
4-
E
\A
Ol c
_Q ra
"£ •a ••^ c
'a C
ra
•o o ,,_
4«
o « ra !/! .92
"5
a! *-" E
'x ~ c T3
3 c o> ra c
ra ra
c ^ c C C C C C o -,, c
_ o to & O .p^ O O ^ c o •4-"

4-» '^ C '^

^ I/I .—. t/1
"l/l *W1
t/1 t/t C 'wi C 'M 'M
.— vt . — l/l (/I
C JJv
.— !/! c
^
•jg E | E'El/l 'E 4_
V) QJ ^
01 O J ^ O l ^ ^ O l O l . — »
H— 4—
01 <D
4—
g;
4-
Oi
E
£
£
•— c Ol
u
u ra c ra ra ra "5; "oi c ra c " ra "^ c c c ~ol ~QJ "oi oi 2 c
C ^.i 3 £z
3
+j 4-1 ^ 3-t->3£j.M_QDra ra o o .a £ D
fa a> d Ol Ol Ol
If 0101010101010101 Ol Oi 01 Ol <y Q) (U
^ 3 D D 3 D D D 3 3 D D D D D D 3 D 3 D 2
^ 3
9-T3 "D T3 T3 T3 T3 "O T3T3T3T3T3TJT3T3 T3 T3 T3 •a
•— at ai ai flj Q) QJ Ol Ol Ol Oi (U (U
O DC DC DC DC DC DC DC DCOCDCOCDCDCDCDC DC DC DC DC o£ o£ DC

«/i j_,
^2 *^
O ^~ t~

i^ ra £ _
0 .y D §
E . § ^ -°
D 01 ^ •— ra
5 E
.E 53
Ol

D
u
t 2 01 8 n *
•D
01 d E£ o -^ d ^ ^
Campaign against rumors; ridicule rumor i
Don't use public address system to rebut r

Develop faith in the source of managemei

Present facts about the topic rather than <

, Distract people's attention from rumor are

reflect?" Then relieve tension by correctin

RQS investigated or propo:

Give "prohibitory orders" banning assemb

. Ask the question "What anxiety or attitud

Keep normal communication channels op«
Integrate old leaders into new activities.

. Educate supervisors in rumor dynamics.

Leader rebuts rumor in a public forum.

Arrest or detain opposition leader.

Prevent idleness and monotony.

Increase counterpropaganda.
Give detailed information.

Remove rumor instigator.

1 Strategy Literature

. Police investigations

ra
Q.
T3
£
O
. No rebuttal

a>
caused it.

. Rebuttal
. Rebuttal

ra
JD "ra
£
'E JD
w» Ol £
«- rvj on «tf «- IN m O ^rMm^invorvoooi «- (M DC <- fM a.

"3 o
Ol
f Ol
00
o
3
s^™
ra
E
oi
Ol
*
oc r D u" c
^ Ol o
TZ vo Oi o o 00 Ol
*J? in ^ 00
o "flj
08 O Ol o
^
"c~ ^**
Ol Ol

Ol
o s OJ
01
E
CT1 .£
QJ
Ol
Q
< g- ai
k~ ai
1_
4-» ra
-C oa aj o 01

s s •s.
0) t/i "OL
"•P
c O 2 OJ ra
ee Ll_ u I ^ ^
•2? DC ^
 Rumor Management 209

a!
'x
c
ID
TJ
C
c
C C Ol
>» >» >^ >> c c
£• 4J
c 4J 4J O O £• E £• +j ^-| +j -M +J o o
t~
•ID1 — — 's 's"^ e~ Ol
l/l
_c C C C C _c "vi
in
"vi
(/)
ID >. ID ID "^ (0 fO fO TO ID
t
Ol
•t- -M "H 4-* t
Ol •— 01 0) "*
c
*"
t *-
01 Ol
M-
Qj
^- M-
GJ QJ
•£
Ol
1 111
01 01 Ol Ol
t E
J2
4- E
Oi "!
4-
Ol
u o 1^ — > ~ u u u u u ^ '~ L^
c
3
Ol
c c c c 5 2
3
O)
ID 3
01 01
3 43 43
Ol Ol Ol
C Ol
3 .O
Ol 01
(U "O <U
A <* A
m CD m
c
3
c c c c
3 3 3 3
C
3
Ol
43
01
^ +3
Ol Ol
1
3 3 3 3 3 3 3 3 3 3 V" 3
Ol
3
Ol Ol Ol Ol
3 3 33
2)
3
TJ TJ TJ TJ TJ TJ TJ TJ TJ TJ '3i TJ TJ TJ TJ T3 TJ TJ TJ TJ TJ
Ol
CC
Ol Ol Ol Ol Ol Ol
ac. tz OL tz cc o:
Ol 01 Ol O Ol
0:0.0:
"8
CE
Ol Ol Ol 01 Ol
QC
Ol
DC
Ol Ol
or o:
Ol
o:

"a. 1/1
!E tS -^
k_
12 ^j § Ol .c </>
Q. •!-•
X ID
"aj
£ c
S OI+3
C C ID IIJ « o
^_ ••;; J^ *-*
u
Ol
. . fll
t; •£
'43 N
~
C
O ID S
V C -R
2
U
k_
o
"c c ra 03 E
<u B c 1
^, c
c M- -9 "0 3
~o -75 o -c ^2 E k_
J C • ^
J^
3 o 5
Q. 0 "° 01 O "w" Ol "S O Ol
•£5 •— — c .C c
JQ

si i lie S Ol
* to ID O
Q- 9-
_c
01 TJ
, Prevent idleness, monotony, and pers

e- . C Oj
. Give complete and accurate informat

4J -i« ^3
. Reduce uncertainties about the disea

. Make information accessible (e.g., ed

Install a toll-free phone line for inqui

. Campaign deliberately against rumor

investigate them, and issue a point-b;

Remove doors (improve access to mai

Widely disseminate this rebuttal throi

se key communicators to spread accun

Hold regular monthly town meetings

. Open business rumor control centers

u
Collect rumors, assemble independen
. Create and sustain faith in leaders.
. Tell news quickly and completely.

•S-

I
o
If)
, Have a skeptical outlook.

3
Ol
Positive advertising

C
jmor control center
Intervene quickly.

ID
a general denial.
Confirm rumors.

_C
No response

1
U.
O
Rebuttal

*— E
•M *—

-Q
OJ
^
Iii
>•; Ol
»- T— IN m ^ LT» ID i- IN «- IN m 3 «- IN m <g- CE «- IN m tZ u S1
ro
1/1
en
oi 00 c
Ol "5
01
c 5 Ol f"
7
.[ 01 Ol o
Ol
Ol LO
Ol ro ilj E
d 00 (N Ol
Ol r^ lj" 08
CO
l^
a Ol Ol h^ Gl Ol M
oo ID *~ Ol c c 1
c* ? E to
~ai oi ID *^
Ol
_c
3
I—
Ol
A
a
C£L
C ^=
<D
X 1
E X
1
Ol
o "o U
o I
Q: Q. of
210 R U M O R P S Y C H O L O G Y

strategies aimed at all three stages in the rumor life cycle. The most
common recommendation for reducing generation of rumors is to re-
duce uncertainty. This goal can be achieved by providing accurate
and timely information and having open channels of communication.
However, only a few authors suggest reducing anxiety, which perhaps
reflects the difficulty in achieving this goal; K. Davis (1975) and Hershey
(1956) recommended reducing anxiety by providing emotional and
economic security to employees, and R. H. Knapp (1944) suggested
inculcating trust in leaders.
The most frequent recommendation aimed at reducing belief in
the rumor was the use of rebuttal, including suggestions to strengthen
rebuttals by garnering the support of neutral spokespersons. For exam-
ple, R. H. Turner (1994) found that Chinese earthquake rumor abate-
ment strategies include assembling a local panel of experts to evaluate
the rumors and issue a point-by-point rebuttal rather than a general
denial; these methods have been reported to be nearly universally
successful. Belief could also be reduced by increasing skepticism (Koe-
nig, 1985), fostering a cooperative, trusting environment (K. Davis,
1975), and even using counterpropaganda (Goswamy & Kumar, 1990).
Finally, several rumor-quelling strategies were aimed at reducing trans-
mission. These included strategies that dissuade people from spreading
the rumor, including punitive steps such as police investigations (Kap-
ferer, 1989) and lawsuits (Austin &Brumfield, 1991). Of course, reduc-
ing generation and belief should also indirectly reduce transmission
of rumors.
We wondered how commonly these strategies were used and how
effective they were in quelling rumors. These questions were part of
our investigation of 74 highly experienced public relations (PR) profes-
sionals introduced in chapter 2 (this volume; DiFonzo & Bordia, 2000).
Recall that the respondents had on average over 26 years of experience
in PR or corporate communication roles and came from several sectors,
including automotive, aerospace, banking and finance, health care,
retail, and transportation. In this part of the study, we presented the
PR professionals with a list of 17 strategies to prevent or neutralize
(i.e., counteract) rumors. These strategies (see Figure 9.1 and Appendix
2.1) represented an exhaustive list gleaned from existing literature on
rumor management and from the results of our own interviews with
managers and PR officers. For each strategy, participants were asked
if they had used the strategy to prevent or neutralize rumors; each
participant then rated the effectiveness of the strategy. Results showed
that these strategies were commonly used and 15 of the 17 strategies
had been used by over a third of the participants. Rumor prevention
strategies, such as stating the values and procedures that will guide
organizational changes, were the most commonly used. Strategies that
 Rumor Management 211

E B ro 42 in
O ro j= u •-
£ u +-• a; >
^ISE^I
-o

2.50
.o
i/i o

"QJ

1.75 2.00 2.25

Mean Effectiveness
°.i 2 oS
&S - u ^
5 £ ^ >, Q.

tTOr\i
oi 4i ^

-o
1.50

t%
Iffll
03 '
?r "
1.25

LO \p LO O CD
.

I
TO U LO QC
OJ O) •" LO ,
ti 3= "> Ji! -^
OJ OJ OJ p -Q
00

-o oe
C Q-
a-~S.
- o o
OJ O CD
§ g * 0 -^ -^ 5 -I
>

I* 1 1? | 5 m 3 45 C

? 5 1 = S 9 s
I I S
£
g ? 8
P• I S .Ca a. re
5 3

£ 8 5 - II

c = ~
111 o r,
i- ro - Ol
L^ O
c
l

! ia I I s S!*
II! I £ ^
ffl ^
•B £
11 ^2
c _-
O t
~*~^ OJ
•+3
JS
^ *- rrt
2 ra
& c "
O) C
01 g

Ol T3
CU 1= ro .—
S S.
'01JZ -^
<u _ F it
S? § o
LO

t I > JZ T3 Q.
Z> fN 2 c C O)
QC r^ _ QJ 03 Qi
212 R U M O R P S Y C H O L O G Y

involved counter-offensive measures, such as searching for or seeking

to punish people who planted the rumor and spread counterrumors,
were the least likely to be used.
To group the strategies into common underlying dimensions, we
conducted a principal components analysis on the effectiveness ratings.
The analysis revealed two underlying dimensions: structuring uncer-
tainty and enhancing formal communications. Structuring uncertainty
means to limit it in some way. Uncertainty is a major cause of rumors
and structuring it is not surprisingly a way to prevent or control rumors
(Davis, 1972; Hirschhorn, 1983). These strategies included stating val-
ues and procedures that will be used to guide the upcoming changes
and providing time lines by which official announcements will be made.
These strategies restrict the range of uncertainty ("I don't know what
will happen, but I will find out next week" or "I don't know if I will
be affected by the restructuring, but I know that they will take work
unit performance into account before deciding to break them up").
Strategies that structured uncertainty were rated overall as more effec-
tive than were strategies that enhanced formal communications.
The enhancing formal communication strategies involved denying the
rumor by internal or external sources, or confirming the rumor. Con-
firming the rumor was rated as highly effective, albeit sparingly used;
this strategy may be applicable only in rare instances, but is very effec-
tive in neutralizing rumors. Once confirmed, the rumor generally loses
its unique informational value and is likely to subside. We now turn
to the strategy of rebutting rumor and the effectiveness of rebuttals in
rumor quelling.

Rumor Rebuttals

One of the most common strategies used in response to a harmful

rumor is to deny the rumored allegation (Kimmel, 2004b). This denial
may take many forms, such as a press release, full-page newspaper
advertisements, statements by CEOs on company Web sites, testimoni-
als and endorsements by neutral third parties, and so on. Although
commonly used, rebuttal is a controversial strategy. Business writer
Owen Edwards (1989) asserted, "About the only way not to counter
a rumor is to deny it, since any denial tends to give rumor added clout.
The more vehement the denial, the more credible the story becomes"
(p. 228). However, Professor of Sociology Frederick Koenig strongly
recommended it: "If a company is the target of a rumor, it should
 Rumor Management 213

deny it immediately as forcefully and publicly as possible, showing the

evidence that proves it is unfounded" (quoted in Goleman, 1991).
As we showed in chapter 4, rumor rebuttals reduce belief in rumor.
Academic literature suggests, however, that some variables may moder-
ate this effect. Iyer and Debevec (1991) exposed participants to allega-
tions of harmful effects of environmental tobacco smoke (ETS). The
allegations were attributed to one of three sources: a neutral source (a
TV station), a negative stakeholder (R. J. Reynolds), and a positive
stakeholder (American Cancer Association). Next, the allegation was
rebutted by a neutral source (The Boston Globe newspaper) or by a low
credibility source (R. J. Reynolds) in an inflammatory tone or in a
conciliatory tone. Iyer and Debevec also had a condition involving no
rebuttal. The dependent measure was attitude toward ETS. The results
showed that the conciliatory denial was effective (more positive atti-
tudes toward ETS) when the allegations arose from a positive stake-
holder. A neutral source rebuttal was similarly more effective when
the source of the allegation was neutral. Yet when the source of the
allegation was a negative stakeholder (R. J. Reynolds), not rebutting
the allegation led to the most favorable attitude. This result is not
surprising, as a rebuttal from R. J. Reynolds after an allegation from
the same source (i.e., R. J. Reynolds) would be a strange occurrence.
Thus, on balance, Iyer and Debevec (1991) found support for the effec-
tiveness of rebuttals, but the results suggest that denial source character-
istics (neutral vs. low credibility source) may render a denial more or
less effective.
In our own research, we have approached rumor rebuttals as per-
suasive messages; their aim is to persuade the audience to disbelieve the
rumor. The social-psychological literature on persuasion and attitude
change has addressed characteristics of the source and content of the
message that determine the persuasiveness of the message (Petty &
Wegener, 1998). For example, a message delivered by a source of high
credibility is more persuasive than is a message delivered by a low-
credibility source. In a similar way, a message with high-quality argu-
ments is more likely to persuade the audience than is one with low-
quality arguments. With the aid of this framework, we now discuss
two sets of moderating variables that we investigated: denial source
characteristics and denial message characteristics.

DENIAL SOURCE CHARACTERISTICS

Rebuttal effectiveness is enhanced when the source of the rebuttal
matches the scope of the rumor. For example, in an organizational
context, if the rumor is about a departmental policy or process, the
department head is the appropriate source to rebut the rumor. A line
214 R U M O R P S Y C H O L O G Y

employee would be too low in the official structure, not likely to be

in a decision-making role, and therefore not be a persuasive source.
However, the CEO of the organization would be too far removed from
departmental matters and therefore not expected to comment on a
department-level rumor. In fact, the involvement of the CEO may even
raise suspicion among staff that matters may be grave. As one PR
officer observed,
If the building is on fire, the chairman should go and take over
because it's very serious. If a manager on the sixth floor has a
fire, you don't send the chairman, because what ends up
happening is that it's just a little fire, and if he runs in there, he
will get this fire smell all over his clothes, and it takes a long
time for the smell to go away. (DiFonzo et al., 1994, pp. 59-60)
We tested these ideas in the context of a rumor about student entry
requirements in a psychology department in a university in Australia
(Bordia, DiFonzo, & Travers, 1998). Places in the undergraduate pro-
gram were in high demand and students were often anxious about
getting admitted. Rumors about entry requirements were widespread.
The study was conducted in a laboratory setting. Participants were
exposed to a rumor that the grade point average required for entry
into second-year undergraduate courses would be going up next year.
The participants were first-year students and therefore the rumor was
realistic and important. The rumor was denied by one of four sources: a
fellow student, a lecturer in the department, the head of the department
(HOD), or the CEO of the university (known as the vice chancellor or
VC). We expected that the HOD would be the most effective in reducing
belief and anxiety associated with the rumor because the HOD would
be the one to make the decisions about entry requirements and there-
fore the most knowledgeable about departmental matters. Therefore,
we also hypothesized that the HOD would be rated as the most appro-
priate source to deny this rumor. The student and lecturer sources are
not in decision-making roles and therefore lack the credibility required
to refute the rumor. The VC would be too high in the university admin-
istration hierarchy and therefore not likely to be relevant as a source
on a department-level matter. Finally, we hypothesized that the denial
would be successful in reducing belief and anxiety associated with the
rumor; belief and anxiety were measured before and after presentation
of the denial statement. As expected, overall, denial did reduce belief
and anxiety associated with the rumor. Furthermore, the HOD condi-
tion had the greatest reduction in belief and anxiety, as shown in
Figure 9.2. In addition, Figure 9.2 shows that the HOD was rated as
the most appropriate source.
Although our prediction about source appropriateness was sup-
ported, we wondered what underlies a judgment about appropriate-
 Rumor Management 215

l Student 0 Lecturer & Head of Department H Vice Chancellor

9 8.25
8 -
7 -

f6
4-1
(0
OC 5 H
(0
4 -

3 -
2.75 2 54
'
2 1.52 1.5
1 H
0.23 ^
0
Belief Reduction Anxiety Reduction Source Appropriateness

Mean belief reduction, anxiety reduction, and source appropriateness ratings for rumor
denials issued by sources that varied in appropriateness. Rumor that the grade point
average required for entry into second-year undergraduate courses will be going up
next year. Data from Bordia et al., 1998.

ness. And, as we saw in chapter 8 (this volume), rumors often circulate

in tense and conflict-ridden settings, such as organizational restructur-
ing. In such situations, employees can be suspicious of management
intentions and not believe rumor rebuttals. Thus, perceptions of source
honesty should be particularly important for effective rebuttal. Indeed,
research on source characteristics in persuasion notes that perceptions
of source honesty play a very important role in assessments of message
accuracy (Priester & Petty, 1995). We conducted another study to test
these ideas (Bordia, DiFonzo, & Schulz, 2000). In this study we pre-
dicted that sources high on perceived honesty and appropriateness
would be effective in reducing belief and anxiety associated with the
rumor. The study was conducted on an Australian university campus
at which a rumor was spreading among the undergraduate students
that the undergraduate library was being closed down. We rebutted
this rumor by one of the following three sources in a between-groups
design: a library staff member on the loans desk, the librarian, or the VC.
The results on denial effectiveness replicated Bordia et al. (1998):
Both belief and anxiety were significantly lower after the denial.
216 RUM O R P S Y C H 0 L O GY

Moreover, as predicted, the denials were the most effective in reducing

belief in the rumor when the sources were perceived to be high on
honesty and appropriateness (see Figure 9.3; note that honesty was
not a manipulated variable; it was simply measured). In other words,
honesty and appropriateness had an additive effect in reducing belief
in the rumor. With regard to anxiety reduction, only source honesty
had a main effect; that is, anxiety was reduced most when the source
was perceived to be honest. These results dovetail nicely with our
findings regarding the moderating role of trust on rumor transmission
reported in chapter 8 (this volume). When the information source is
trusted (i.e., honest), a rebuttal is effective in dispelling uncertainty
and anxiety; when the source is not trusted, the rebuttal is less effective.
The results also provided insights into what may underlie assess-
ments of source appropriateness. Appropriateness was positively corre-
lated with knowledgeability ("How knowledgeable do you think this
source is regarding the rumor?") and status in the management struc-
ture. Because this rumor alleged the closure of a university facility and
was therefore an organization-level topic, the VC was considered the
most appropriate source.

a High Honesty, High Appropriateness a High Honesty, Low Appropriateness

a Low Honesty, High Appropriateness = Low Honesty, Low Appropriateness

2-i
1.71
1.47
1.5-
O)
_c
tp
1-
C 0.61
(0
0.5- ^XX
?^xx °-25
^XX :. -0-13 0.02
o- I
Belief Reduction Anxiety Reduction
-0.5 -
Mean belief reduction and anxiety reduction for rumor denials issued by sources high
or low in appropriateness and honesty. Rumor that undergraduate library will close
down. Data from Bordia et al, 2000.
 Rumor Management 217

DENIAL MESSAGE CHARACTERISTICS

Another set of studies provided insights into message characteristics
required for belief and anxiety reduction. First, denials that assist recipi-
ents in attaining a sense of control over potential threats can relieve
anxiety (Bordia, DiFonzo, Haines, & Chaseling, 2005). In these studies
we tested the effectiveness of rebuttals in reducing belief and anxiety
associated with the Good Times virus rumor. This rumor, which spread
over the Internet in late 1990s, alleged that a virus was being circulated
by e-mail and merely reading a text message would infect the computer.
The rumor caused concern among e-mail users and warnings of the
alleged virus were forwarded to friends and coworkers, creating in-
creased Internet traffic and concerned inquiries to technical support
staff. Accurate information debunking the rumor was also available on
the Internet.
In our studies, we presented the hoax e-mail to participants and
measured belief and anxiety associated with this rumor. Next, we pre-
sented another e-mail that rebutted the rumor and again measured
belief and anxiety. In the first study we found that the denial was
successful in reducing belief and anxiety associated with the rumor.
Moreover, this denial was more effective than a control condition in
which participants were exposed to a message that was unrelated to
the virus rumor. In the second study, we compared the denial with a
control message that was on the topic of computer viruses but did not
explicitly deny the rumor. Instead, the message provided information
on how to safeguard against computer viruses. It is interesting to note
that although the denial was significantly better than this control mes-
sage in reducing belief in the rumor, the difference in anxiety reduction
was small and not statistically significant. The control message, by
providing information and coping strategies, alleviated some anxiety.
In the third study, we incorporated these anxiety-alleviating elements
into the denial ("Being informed about viruses is the best defense for
future protection of your computer files. A computer virus is a snippet
of computer code that must infect a host program to spread. The best
way to detect these pathogenic computer codes is to install an antivirus
program checker"). This denial, which aided participants in regaining
a sense of control, was effective in reducing belief and anxiety; these
results are shown in Figure 9.4.
Second, another line of research has suggested that the inclusion
of rebuttal context may moderate denial effectiveness. These studies
examined the effects of a rebuttal in the presence or absence of a prior
accusation. Roller (1993) found that the rebuttal source is evaluated
more negatively when there is no prior accusation. In the absence of a
218 RUMOR PSYCHOLOGY

Denial Control

1.45

1.19

o>
35
n
tr

-0.17
Belief Reduction Anxiety Reduction
Mean belief reduction and anxiety reduction for denial of the Good Times rumor. Data
from Bordia, DiFonzo, Haines, and Chaseling, 2005, Study 3.

prior accusation, the audience finds the rebuttal puzzling and suspicious
and evaluates the rebutter negatively. In an early study demonstrating
this effect (Yandell, 1979), three experimental conditions were created.
First, an actor was accused of damaging a typewriter, and the actor
subsequently denied having done so. The second condition involved a
denial in the absence of an accusation. In condition three, the actor
confessed to damaging the typewriter. The actor was more likely to be
seen as being guilty in the second (denial-only) condition compared
with the first (accusation + denial). In fact, ratings of guilt in the second
condition were as high as when the actor confessed to damaging the
typewriter. Yandell concluded in attributional terms that the accusation
provided a situational explanation for the denial. However, a denial in
the absence of an accusation was attributed to guilty conscience and
led to an impression of guilt. A similar effect was noted by Wegner,
Wenzlaff, Kerker, and Beattie (1981) in their study on the innuendo
effect of newspaper headlines. They found that whereas the denial only
("Bob Talbert not linked with Mafia"; M = 3.73) led to less negative
impressions than did the assertion ("Bob Talbert linked with Mafia";
M - 4.25), the negative impressions were not as low as in the control
 Rumor Management 219

condition ("Bob Talbert celebrates birthday"; M = 3.00). In fact, the

denial condition was not significantly different from the assertion or
the control condition.
How might we understand these results? Holtgraves and Grayer
(1994) applied attribution theory and Grice's (1975) maxims of conver-
sation to the study of denials in a courtroom setting. Grice's maxim of
quantity states that a speaker should be as informative as the situation
demands. Holtgraves and Grayer postulated that a denial, in the absence
of an accusation, leads to the violation of the quantity maxim. When
faced with an overinformative denial, the audience is motivated to
explain this violation, which results in an attribution of guilt and low
ratings of the person on characteristics such as trustworthiness. Their
findings supported the predictions. Thus, rebuttal context affected re-
buttal effectiveness.
This line of research has primarily been conducted in the context
of allegations aimed at individuals. We applied this research to the
context of rumors of product contamination and rebuttals by corpora-
tions (Bordia, DiFonzo, Irmer, Gallois, & Bourne, 2005). We investi-
gated whether the inclusion of rebuttal context moderated the effec-
tiveness of denial. First, we wanted to compare a message that states
the rumor and then rebuts it (rumor + denial) with a message that
rebuts without repeating the rumor (denial-only). Second, we also
wanted to compare these rebuttals with a condition that follows the
advice of some PR practitioners and responds only with "our policy is
not to respond to rumors" (i.e., a no-comments approach). Third, we
wanted to see how these three strategies (denial-only, rumor + denial,
or no comments) fared in comparison to a rumor-only condition.
The study was conducted in a laboratory setting and participants
were presented with a fictional newspaper story that reported a state-
ment by a consumer products manufacturer, PBR. The story noted that
PER was plagued by a rumor that one of the ingredients in its products
was harmful to consumers. Four conditions were created. The text in
the four conditions is presented in Exhibit 9.1. In the rumor-only
condition, the newspaper story reported the rumor about PBR. In the
rumor + denial condition, the rumor was reported followed by a state-
ment by PBR that repeated the rumor and then rebutted it. In the
denial-only condition, no mention was made of the rumor. Instead
PBR statement adopted a rebuttal stance and insisted that its products
were safe for consumers. Finally, in the no-comments condition, the
rumor was presented and the PBR statement said that PBR does not
comment on rumors. We measured the following responses to the
news story: the amount of uncertainty regarding the reasons for the
press statement, suspicion of PBR, attribution of the statement to an
220 RUMOR PSYCHOLOGY

g
10
< k u _
Ul ifl un3 T 3 S HQ- 2x-t-''D 'O_ 3,,,

]
iga

o
u

.0 £ E§
01

VI

O
•+:
'o
3
£
s
.2 !T O
 Rumor Management 221

internal (something about PER) versus an external (something about

the environment) cause, perceptions of a cover-up by PER, an overall
evalutation of PER, and finally intentions to purchase PER products.
The results of the study provide unequivocal support for the denial
strategy and no support for a no-comments strategy (see Figure 9.5;
all differences in conditions discussed here are based on statistically
significant differences). The no-comments condition was very similar
to the rumor-only condition on most of the variables. In fact, it was
even worse than the rumor-only condition on perceptions of cover-
up. That is, on reading that the company was declining to comment
on the rumor, participants felt that the company had something to hide.
Next, we compared the denial-only condition with the rumor +
denial condition. As expected, level of uncertainty for the reasons for
the press statement was the highest for the denial-only condition.
Moreover, the press statement was attributed to more internal causes
in the denial-only condition than in the rumor + denial condition.
Also, perceptions of the cover-up were higher, and evaluation of the
company lower, in the denial-only condition compared with the rumor

B No Comment • Rumor Only a Denial Only n Rumor+Denial

Uncertainty Suspicion Causal Cover-Up Evaluation Purchase

Attributions of PBR Intentions
Mean ratings of uncertainty, suspicion, external attributions, cover-up intentions, com-
pany evaluation, and purchase intention. Rumor that PBR products are dangerous. All
constructs measured on a 1 to 7 Likert-type scale. Data from Bordia, DiFonzo Irmer
et al., 2005.
222 R U M O R P S Y C H O L O G Y

+ denial condition. Finally, there was no difference in the two denial

conditions for purchase intentions.
Overall, these results suggest the following: (a) a no-comments
strategy at best has the same effect as saying nothing, and at worst
leads to greater perceptions of cover-up; (b) both denial-only and rumor
+ denial are significantly better than saying nothing (i.e., rumor-only
or no comments) for evaluation of the company and purchase inten-
tions; and (c) repeating the rumor in the denial provides a context for
the denial statement and thereby reduces the uncertainty regarding
the reasons for the denial. It also helps deflect the reasons for the denial
to external causes and may lead to an overall positive assessment of
the company.
Although these results provide some support for rebuttal strategies,
we do not mean to imply that rumor management is easy or that
rebuttals are a panacea for rumors. Indeed, rumors are notoriously
difficult to control or manage. As discussed in chapter 3 (this volume),
rumors serve a variety of needs and motivations and will therefore be
resistant to discontinuing information. Indeed, social cognition litera-
ture on belief perseverance has found that impressions, once formed,
are highly resistant to evidence to the contrary. In the next section,
we use belief perseverance and the processes underlying it to assist us
in understanding why some rumors are often difficult to refute.

Why Some Rumors

Resist Rebuttal

A series of studies has demonstrated that initial impressions, even when

thoroughly discredited, tend to persist (Anderson, 1983, 1985; Ander-
son, Lepper, & Ross, 1980; Ross, Lepper, & Hubbard, 1975). In these
studies, subjects formed either a positive or negative directional impres-
sion based on manipulated information. For example, subjects formed
impressions about their ability to discern real from fake suicide notes
(they were either successful or unsuccessful in their ability to discern
real suicide notes). In other experiments, subjects formed impressions
about the relationship between a paper-and-pencil measure of risk
behavior and success as a firefighter (they were led to believe that
either a positive or a negative association exists between risky behavior
and success). The manipulated information in these experiments was
in the form of feedback from the experimenter (you are "correct" or
"incorrect") or biased case histories about the risky behavior measure
and successful firefighting. A manipulation check was then performed
to measure the direction and strength of the impression.
 Rumor Management 223

Subjects were then debriefed, that is, informed in some way that
the data on which they had based their initial impression was totally
predetermined. In the earlier examples, debriefing consisted of the
experimenter telling subjects that the feedback given was predeter-
mined and unrelated to their actual responses, or that the firefighter
case histories were fictitious. Subjects subsequently completed depen-
dent measures of the strength and direction of the impression. Results
uniformly showed that although weakened, initial impressions per-
sisted (i.e., persevered) after debriefing. These results are similar to our
findings on rumor denials. For example, with regard to the belief that
"I am good [bad] at detecting true versus false suicide notes," people's
strength of belief (a) is highest when they are given feedback consistent
with the belief and (b) is reduced, but not eliminated, when the evi-
dence is discredited. We similarly found that belief in a rumor (a) is
highest when the rumor is given and (b) is reduced, but not eliminated,
when the rumor is rebutted (in general). In other words, mud sticks—
but not completely. Beliefs and rumors, when discredited, still tend to
persist, but in a weakened state.
This belief perseverance phenomenon has been explained by three
mechanisms: confirmation bias, causal inference making, and denial
transparency. Confirmation bias is the tendency to conform incoming,
contradictory data so that it does not challenge existing biases (Nisbett
& Ross, 1980). Confirmation bias in this case discounts evidence contra-
dictory to the first impression (subjects discount discrediting evidence
so as to maintain their belief about their ability to discern suicide notes).
Confirmation bias also acts here to selectively interpret concurrent
information (reactions to false feedback) so as to bolster impressions
(Ross et al., 1975). Once formed, impressions become relatively autono-
mous, that is, independent of the evidence on which they were
founded. "The attributor doesn't 'renegotiate' his interpretations of the
relevance or validity of impression-relevant information" (Ross et al.,
1975, p. 890). Thus, at least on paper, it is easy to see how impressions
become immune to logical challenges. In addition, as we noted in
chapter 3 (this volume), rumors serve various psychological needs, and
people are likely to engage in motivated reasoning to hold on to or
legitimize cherished beliefs.
Causal inference making is the predilection to attribute unwarranted
cause-effect relationships to merely contiguous events (Nisbett & Ross,
1980). Explaining an event increases its subjective likelihood (Ross,
Lepper, Strack, & Steinmetz, 1977). In a similar way, as we showed in
chapter 5 (this volume), rumors influence behavior because they pro-
vide ready-made causal explanations and lead people to deduce cause-
effect relationships even in the absence of such an effect.
Finally, Wegner, Coulton, and Wenzlaff (1985) posited denial trans-
parency as a more parsimonious account of persistence phenomena.
224 R U M O R P S Y C H O L O G Y

Denial transparency is a term that expresses the ineffectiveness of denial

in negating propositions (denials are transparent in that people see
through them; i.e., they are ineffective). The theory contends that
persons classify all propositions as true initially, no matter how briefly.
Propositions deemed false are then appended with a not true tag which
is generally more difficult to remember than the affirmative proposi-
tion. For example, in the absence of other information, "Bob Talbert
is not connected with the Mafia" will tend to be recalled as "Bob Talbert
is connected with the Mafia." A denial does not erase a proposition;
rather, it accompanies it.
For a denial to be effective, it should be reliably recalled with the
target and the accusation. Given that recall of negations take more
cognitive effort, Wegner et al. (1985) reasoned that if the denial ("I
am not a cheat") is proposed in the affirmative ("I am honest"), it has
higher likelihood of being recalled. In other words, victims of rumors
should strive to be associated with positive impressions to replace the
negative impressions created by the rumor. A similar idea was proposed
by Tybout et al. (1981) who found that attitudes toward McDonald's
restaurants in response to the rumor that the burgers contained worm
meat were improved when worm meat was relabeled as a French
delicacy or when participants were asked to think of certain details of
the McDonald's restaurant they visited (e.g., whether it had indoor
seating or not). In other words, rumor control should be a multipronged
approach that may include an outright rebuttal and other reputation-
enhancing strategies. Indeed, most effective rebuttals adopt this ap-
proach. Consider the example of Stroh Beer (Koenig, 1985). In 1983,
the Stroh Brewery Company was plagued by rumors that the company
had donated money to Jesse Jackson's presidential campaign. The com-
pany took out advertisements in the Chicago Tribune and in addition to
stating that the rumors were "completely, totally false," the brewery
company provided information on its donations to the Statue of Liberty
renovation effort.

Recommendations for
Managing Rumors

With these research insights in mind, we can answer our original ques-
tion: How can rumors be effectively managed? Rumors can be pre-
vented by reducing uncertainty and anxiety in the workplace. Managers
need to anticipate events that may lead to uncertainty, anxiety, and a
 Rumor Management 225

loss of a sense of control (and result in rumors) and strive to address

these with a systematic communication strategy (Clampitt, DeKoch, &
Cashman, 2000). Although managers may not be able to address all
sources of uncertainty (all decisions may not have been made), strate-
gies aimed at structuring uncertainty are likely to prevent rumors (Di-
Fonzo & Bordia, 1998). As shown in chapter 8 (this volume), trust plays
a vital role in how formal and informal communication is perceived. In
an environment of distrust, even small amounts of uncertainty are
threatening and may increase credulity. Yet in a climate of trust, dread
rumors are unlikely to find a foothold. Therefore, trust-building mea-
sures (e.g., open and participative communication practices) are likely
to reduce credulity and spread of negative rumors.
To curb rumors, one must act quickly. The longer a rumor circu-
lates, the harder it is to control because of the following two features
of rumor dynamics: First, as a rumor circulates, it tends to evolve into
a more believable version and therefore becomes harder to contain.
Second, the more times people hear the rumor, the more likely they
are to believe it (the repetition effect noted in chap. 4, this volume).
Therefore, people responsible for communication management need
to be plugged into the informal networks and be aware of rumors as
quickly as possible. In light of the fact that management can often be
disconnected from line employees, by the time management hears of
the rumor, it may have already circulated for a while. One manager
reported using a "sonar man," a trusted subordinate who informed the
manager of rumors circulating among employees.
When a rumor outbreak is being faced, the first decision to be
made is whether to ignore a rumor or do something about it. As
Figure 9.1 shows, over 75% of respondents to the survey of PR profes-
sionals indicated that they had ignored rumors. However, this strategy
is relatively ineffective (Smeltzer, 1991). A rumor should be ignored
only when it is considered harmless and expected to die on its own
(DiFonzo et al., 1994). In particular, if asked for a response, managers
should always comment on the rumor. As described earlier, a no-
comments response conveys that one has something to hide and in-
creases the uncertainty surrounding the issue. For example, Frith
(2001) described how a no-comments policy, enacted by the AMP
Corporation, fanned rumors of merger. If the rumor is true, it (or at
least the elements of it that are true) must be confirmed. This strategy
was rated as highly effective in our survey of PR professionals. Rumors,
although exaggerated, often contain a grain of truth. Confirming the
part that is true can reduce the uncertainty. In one case, a bank acciden-
tally mailed a few credit cards to the wrong address. Rumors began
spreading that large numbers of credit cards were mailed to the wrong
226 R U M O R P S Y C H O L O G Y

address. The bank quickly informed the customers, confirming that

some cards had been mailed incorrectly but also noting that the rumor
exaggerated the problem (DiFonzo et al., 1994).
In many instances, a denial is required to reduce belief in the
rumor, inoculate future recipients of the rumor, control the damage
caused by the rumor, or just set the record straight (indeed, the absence
of a denial may be taken as evidence of guilt). It is important to base
the denial in facts. For one, it is the ethical thing to do. Acceptance of
responsibility and ethical action improve the evaluation of the com-
pany. Also, organizations are rarely able to keep secrets for long. Leaks
will occur, the facts will emerge, and management will lose credibility.
In one company, a costly human error had occurred. The PR office
was not informed and thus could not prepare a response for the press.
A journalist discovered the incident and published a story, without
proper response from the company. The PR officer described the inci-
dent as follows:
Monday morning, a reporter called me and said that she
understood that we had a real problem. I checked a little bit
with the management and though they knew there was a
problem, they did not mention it to anyone. The next day the
reporter called me again. She really seemed to have some facts
here. On the fourth day she told me that it was quite serious,
she knew the numbers, the error would cost hundreds of
millions, so I went back to the management and said, "What's
going on?" They said, "Yes, we do have this problem." She ran
the story that afternoon. (DiFonzo et al., 1994, p. 59)

The effectiveness of persuasive messages depends on the source

credibility and message quality; so too with rumor denials. Sources
that match the scope and seriousness of the rumor should be used. The
source of the denial should also be perceived as honest and trustworthy.
Indeed, neutral outside sources are likely to be very effective. For
example, Alpac Bottling Company's efforts to counter allegations that
syringes were found in soft drink cans got a boost when an independent
bottling inspector announced, "I've inspected this type of plant machin-
ery for years, and I know it just couldn't happen. It's a hoax" (Fearn-
Banks, 2002, p. 233). The messages should be clearly worded, be easy
to understand, and contain evidence (if available) discrediting the ru-
mor. For example, denials from the designer Tommy Hilfiger that he
made racist comments on the Oprah Winfrey Show included statements
from Oprah Winfrey that not only did he not appear on her show, but
she has never even met Tommy Hilfiger ("Tommy Rumor," 1999).
Finally, if it is likely that the denial will be received by people who
have not heard the rumor, the denial should provide information on
the reasons for and context that has led to the denial (i.e., acknowledge
 Rumor Management 227

the existence of rumor). For example, the Stroh Brewery Company,

when denying rumors of donations to a presidential campaign, clearly
stated the rumor and denied it: "There are rumors in the Illinois and
Indiana areas that the Stroh Brewery Company is making contributions
to presidential candidates. These rumors are not true" (Koenig, 1985,
p. 61). The half-page advertisement in the Chicago Tribune went on to
provide information on the company's contribution to nonpolitical
programs, the slanderous nature of the rumor, and the reward of
$25,000 for information on the origins of this rumor.
In this chapter, we have reviewed research pertaining to rumor
management and have explored the implications of this research for
those attempting to prevent or manage rumors. In the next and final
chapter, we summarize the concepts outlined thus far by addressing
the questions raised at the outset and proposing a more comprehensive
research agenda.
 Summary, Model, and
Research Agenda

Perhaps the most bizarre rumor to surface in the

aftermath of Hurricane Katrina, which devastated New
Orleans in 2005: Dolphins, specially trained by the U.S.
Navy to guard docked submarines by shooting poison
dart guns at terrorist divers, had been swept out to sea
during the hurricane. Because of this, according to the
rumor, all divers and Gulf Coast bathers were at risk
(Mikkelson, 2005). The rumor—judged "probably false"
by the hoaxbuster Web site http://www.snopes.com—
has enjoyed widespread circulation.

n this volume, we have addressed a series of questions—

some old, some new—that surround rumor. We related
these to current social and organizational psychological
theory and we grounded much of our conclusions in empiri-
cal research. Where does this leave us? In this chapter we
first summarize our conclusions on each question. We then
propose an integrated model of rumor phenomena. Last, we
set forth a detailed research agenda.

Summary

WHAT IS RUMOR?
In chapter 1 (this volume), we defined rumor as unveri-
fied and instrumentally relevant information statements in

229
230 R U M O R P S Y C H O L O G Y

circulation that arise in contexts of ambiguity, danger, or potential

threat, and that function to help people make sense and manage risk.
This definition specifically addresses the content, contexts, and func-
tions of rumor, and contrasts these with those of gossip and urban
legends. Rumor content is unverified and instrumentally relevant infor-
mation, whereas gossip is evaluative social talk and urban legends are
interesting stories. Rumors arise in contexts of ambiguity, danger, or
threat—either physical or psychological—and function to make sense
of ambiguity or manage risk. In contrast, gossip arises in the context
of social network formation and maintenance, and allows groups to
become more cohesive and to define their membership, norms, and
power structure. Urban legends arise in more socially cohesive contexts
and serve to entertain and to convey group mores, norms, and cultural
truths. We presented the Information Dimension Scale (IDS) to differ-
entiate six dimensions of information. Four of the dimensions pertained
to content: extent to which the information has evidentiary basis, is
important, is about individuals, and is slanderous; two pertained to
function: extent to which the information is entertaining and useful.
We then measured the information dimensions of exemplars of rumor,
news, gossip, and urban legends. Our exemplars exhibited different
information dimension patterns in line with our conceptualization of
each genre of hearsay. For example, our rumors were rated as low on
evidentiary basis and high on importance and usefulness. The rating
given to news mirrored these elements except for evidentiary basis,
which was rated as high. Our gossip, however, was considered by
participants as low in importance and usefulness, and high in slander-
ous content about individuals and entertainment value. And our urban
legends were rated low on evidentiary basis, importance, and useful-
ness, but high on entertainment.
The use of the information dimensions to differentiate species of
hearsay is a helpful advance in the study of rumor. It implies that, as
with biological species and types of mental illness, classification depends
on the presence of a number of characteristics and that nebulous forms
exist. It is sometimes difficult to characterize information as either
rumor or gossip; the information may indeed exhibit patterns character-
istic of both. "The boss is having an affair with the CEO" may be like
a rumor in some ways—it is rated low on evidentiary basis (there is
no proof) and high on importance (because it affects work processes)—
but like gossip in others—it is about individuals, slanderous, and enter-
taining. The IDS anticipates these nebulous forms and offers a way of
characterizing information with continuous rather than discrete yard-
sticks. Instead of classifying information as either rumor or not rumor,
we say that it exhibits an information dimension pattern that is strongly,
moderately, or weakly characteristic of rumor.
 Summary, Model, and Research Agenda | 231

WHAT ARE THE CATEGORIES OF RUMOR?

In chapter 2 (this volume), we posed basic descriptive questions. The
rumor pie has been sliced several different ways, but most often accord-
ing to the motivational tension characterizing the rumor: dread, wish
fulfillment, or wedge-driving (R. H. Knapp, 1944). In organizational
contexts, we offered our own categorization schemes on the basis of
rumor's central functions of sense making and threat management.
Our schemes classified rumor according to rumor public (internal or
external to the organization), object of collective concern (e.g., job
security, stock value), and content (e.g., turnover, costly errors). We
also classified internal rumors about organizational change on the basis
of four aspects of change that employees want to make sense of or
manage the effects of: changes to job and working conditions, the
nature of the organizational change, poor change management, and
consequences of change for organizational performance. Categorization
schemes such as these are important because they reveal what people
are concerned about, they reveal underlying attitudes and beliefs, and
they help to prepare practitioners concerned with rumor prevention
and management.

HOW FREQUENT ARE RUMORS?

Although research on rumor frequency is scant, the best available
evidence indicates that organizational rumors are rather frequent:
harmful or potentially harmful rumors reached the ears of top-level
corporate public relations (PR) officers almost once per week on average
(DiFonzo & Bordia, 2000). The majority of these corporate rumors
seemed to be internal and concerned with personnel changes and job
security; external rumors were most often concerned with organiza-
tional reputation and stock prices. A recurrent theme throughout this
research was that organizational rumors tend to occur in the context
of change such as downsizing, mergers, new technology, and restruc-
turing. Our analysis of internal change rumors at a large public hospital
(Bordia, Jones, Gallois, Callan, & DiFonzo, in press) found that the
largest portion—almost half—of internal rumors about change was
concerned with changes to job and working conditions. In addition,
the overwhelming majority of internal change rumors were driven
by fear of dreaded consequences (i.e., they were dread rumors). We
interpret these frequency patterns as further evidence of the sense-
making and threat management functions of rumor: Employees, con-
cerned about the ramifications of change, participate actively in the
rumor mill to prepare actively or emotionally for potential negative
events.
232 R U M O R P S Y C H O L O G Y

WHAT EFFECTS DO RUMORS HAVE?

Rumors clearly play a causal or contributory role in a variety of attitudi-
nal and behavioral outcomes. A long line of rumor research has noted
how rumors have contributed to panic, affected economic activity,
exacerbated racial tensions, incited riots, and affected health behaviors,
even when these rumors were not strongly believed. We reported how
a false rumor about the deadly consequences of blinking headlights at
nighttime motorists whose lights were off resulted in the nearly univer-
sal diminution of this prosocial practice of traffic culture, despite the
fact that our sample was a highly educated, skeptical, and good-natured
group of psychology faculty and graduate students. These findings ac-
cord with theory that people are especially sensitive to loss (Kahneman
& Tversky, 1979) and attempt to prevent loss even when such negative
events are extremely unlikely. In a similar way, rumors supposedly
heard from "brother-in-law Harry" were rated as noncredible, but sys-
tematically affected individual stock-market trading behavior as much
as did the same information presented as front-page Wall Street Journal
news. These rumors drew "investors" away from a buy-low-sell-high
trading strategy and therefore resulted in reduced profits compared
with control participants (DiFonzo & Bordia, 1997). The notion that
rumors have effects was also supported by results of our study of 74
highly experienced corporate PR officers; this sample rated rumor
effects overall as moderately severe (DiFonzo & Bordia, 2000). Three
general categories of effects were derived: external ramifications, inter-
nal attitudes, and internal behaviors. Of these, the effects of rumor on
internal attitudes were rated as the most severe. This finding is sup-
ported by Bordia et al.'s (in press) hospital change study results suggest-
ing that employees hearing negative rumors were more stressed than
were those not hearing rumors.
We also reported the results of a more comprehensive investigation
of the effects of rumor in a longitudinal study of an organization that
underwent a radical downsizing. The number of negative rumors heard
was associated with uncertainty and anxiety. The number of negative
rumors heard was inversely associated with key employee attitudes:
positive ratings of formal communication quality, trust in the company,
job satisfaction, and organizational commitment. The number of nega-
tive rumors heard was also associated with intention to leave and—
intermittently—lower productivity. Furthermore, the cumulative
number of rumors heard was also associated with these outcomes and
was generally a better predictor of these outcomes than was the number
of recent rumors heard. This pattern of evidence suggests that hearing
negative rumors is associated with a variety of negative outcomes. In
addition, the evidence suggests that hearing negative rumors month
 Summary, Model, and Research Agenda 233

after month materially contributes to a number of negative organiza-

tional attitudes, intentions, and behaviors.

WHY DO PEOPLE SPREAD RUMORS?

In chapter 3 (this volume) we reframed rumor transmission antecedents
within a motivational framework. Rumor transmission is motivated by
three broad psychological needs. The first need is to know and under-
stand one's circumstances so as to be able to act or cope effectively.
This need motivates fact-finding strategies. Uncertainty about a topic
of high outcome-relevance leads to a feeling of lack of control and
anxiety and motivates the search for information (i.e., fact-finding).
Rumors provide explanation and prediction and fulfill this need. Sec-
ond, the need to develop and sustain social relationships motivates
relationship-enhancement strategies. In the context of rumors, people
take into account the interaction context and may use rumors to bond
with the conversational partner and achieve other relationship -
enhancement goals. For example, people are more likely to transmit
rumors they believe to be true so as to enhance their status in the
social network as a reliable source of information (a relationship-
enhancement goal). Third, the need to further material and psychologi-
cal needs of the self motivates self-enhancement strategies. Motivated
by self-gain, people transmit rumors as propaganda to undermine oppo-
nents, present the outgroup in a negative light, and justify existing
(often prejudicial) beliefs and attitudes.
We also presented the results of an empirical study that explored
the role of motivations in people's intention to transmit rumors. We
manipulated the valence (negative or positive rumor), target (rumor
about the ingroup or outgroup), and recipient of the rumor (an acquain-
tance from the ingroup or the outgroup) and asked participants about
their intention to transmit the rumor and underlying motivations. We
found that the fact-finding motivation was highest when the context
was that of a negative rumor about the ingroup to be transmitted to
another ingroup member. Mediation analysis revealed that the fact-
finding motivation underlaid the intention to transmit a negative rumor
about the ingroup to another ingroup member (as opposed to an out-
group member). That is, people may transmit a negative rumor about
the ingroup to another ingroup member to ascertain the veracity of
this rumor. However, the relationship-enhancement motivation was
highest when the rumor was positive in valence, about the outgroup,
and being transmitted to an outgroup member. Mediation analysis
revealed that positive rumors about the outgroup were more likely to
be transmitted to outgroup members (compared with negative rumors
about the outgroup) and the relationship-enhancement motivation
234 R U M O R P S Y C H O L O G Y

underlaid this effect. It would appear that by being the bearers of good
news, people hoped to generate liking for themselves in a conversation
partner. Finally, the self-enhancement motivation was highest when
the rumor was positive and about the ingroup and the recipient was
an outgroup member. However, transmission intention in this context
was low, perhaps because this condition threatened relationship-
enhancement goals. Thus, in our experimental context, the self-
enhancement motivation took a backseat to relationship-enhancement
motivation. The motivation-based approach highlights the variety of
influences on rumor transmission, integrates previous research into a
tripartite model of motivational antecedents to rumor transmission,
and suggests that future research should incorporate the fact-finding,
relationship-enhancing, and self-enhancing motivations in arriving at
a more comprehensive understanding of rumor spread.

WHY DO PEOPLE BELIEVE RUMORS?

In chapter 4 (this volume), we addressed this question using Egon
Brunswik's lens model of judgment in which distal attributes are in-
ferred from proximal cues. That is, what cues do people use to infer
the veracity of a rumor? We reviewed and (when possible) meta-
analytically combined studies bearing on this question; we identified
four such cues. People are more likely to believe a rumor if the rumor
accords with the hearer's attitude (especially to the extent that the
rumor accords with a specific attitude), comes from a credible source,
is heard repeatedly, and is not rebutted. Rumors of government corrup-
tion and waste, for example, are more likely to be believed if the hearer
specifically opposes the current government administration (e.g., the
Roosevelt administration and not merely the government in general;
F. H. Allport & Lepkin, 1945), hears it from a source deemed credible,
hears it repeatedly, and does not hear a rebuttal of the rumor. We
reported a study of stock-market brokers and traders in which we found
that these and other cues were used to infer belief in stock-market
rumors. Other cues in this domain included stakeholder status of the
source (do they have anything to gain by spreading the rumor?), consis-
tency with a larger pattern, consistency with emerging data, and agree-
ment with expert consensus.

HOW DO RUMORS HELP PEOPLE MAKE

SENSE OF UNCERTAINTY?
In chapter 5 (this volume), we explored the psychological mechanisms
involved in the process of sense making at the individual and inter-
personal levels. At the individual level, we explicated rumor sense
 Summary, Model, and Research Agenda 235

making using frameworks within social cognition: explanation theory,

causal attribution, illusory correlation, and antiregressive prediction.
Rumors influence the individual explanation process by drawing the
attention of an individual to an event or a particular aspect of an
event, by offering an initial explanation for an event, by activating
knowledge structures that influence the search for further information,
and by motivating the sense maker—usually through anxiety-
provoking rumor content—to continue searching for a satisfactory
explanation. Cognitive knowledge structures are important at each
stage of the explanation process, and knowledge structures about
causes seem to be especially important in understanding how rumors
affect individual sense making. We proposed that rumors help people
make sense of the world by means of the ready-made and stable-
cause explanations that are often embedded in them. By ready-made
we mean that most explanations come from the social milieu—
often embedded in rumors—and are probably not derived from an
individual's own causal analysis. For example, people may observe
that the office is being renovated and hear through the rumor mill
a ready-made explanation for these renovations: The company will
be restructuring. By stable cause we mean that the nature of the cause
is more often lasting than temporary. For example, the restructuring
is occurring because management has a stable dispositional trait:
greed. We speculated—on the basis of a perusal of published rumors
and experimental evidence—that most rumors contain such stable-
cause attributions.
The significance of the idea that rumors are embedded with stable-
cause attributions is that such attributions have been implicated in
several well-known systematic judgment biases. We explored a couple
of these with regard to rumor. First, rumors often result in erroneous
judgments of correlation. They lead people to see relationships that in
reality do not exist. Stock-market "investors" presented with rumors
that were unassociated with stock price changes were strongly con-
vinced that these rumors covaried with these price changes (DiFonzo &
Bordia, 1997). Second, rumors often lead people to make antiregressive
predictions (the recent trend will continue) and abandon more predic-
tive base-rate information. Stock-market rumors have long been impli-
cated in such antiregressive trading behaviors (e.g., Rose, 1951). For
example, hearing a rumor that Goodyear profits are up led investors
to predict that the stock price would increase rather than regress toward
the mean of past prices. Our own experimental research found that
stable-cause attributions are at the heart of antiregressive trading be-
havior brought on by rumors (DiFonzo & Bordia, 2002b). It is interest-
ing that rumors in our studies did not have to be believed to have an
effect; they merely had to make sense.
236 R U M O R P S Y C H O L O G Y

We also explored the dynamics of collective rumor sense making

by examining the content, functions, and flow of statements made
during Internet rumor discussions. We first addressed the question,
"What types of statements constitute rumor discussions and what is
their relative prevalence in a typical rumor episode?" We reviewed the
Rumor Interaction Analysis System to content analyze rumor discus-
sions on the Internet. We found that the largest portion of discussion
content was about sense making: Participants attempted to ascertain
whether or not the rumor was true. Participants also provided informa-
tion, asked questions, authenticated credentials and information,
shared relevant personal experience, expressed belief and disbelief, and
attempted to persuade others; these activities support the collective
goal of sense making. We next addressed the question, "What is the
typical makeup of sets of statements contributed by rumor discussion
participants?" That is, when people contribute to a rumor discussion,
what is the relative composition of the types of statements they make
at any one point in the discussion? We found 11 distinctive patterns
of statement sets that discussants made at any one time; we dubbed
these communicative postures and interpreted these within a collective
sense-making framework. For example, the explanation-delivering
posture was composed of a high number of prudent statements (e.g.,
"I don't know if this is true, but I heard that. . .") and providing
information statements (statements that bring information to the dis-
cussion). Other postures included those that evaluated, verified, falsi-
fied, or accepted the explanation; shared or sought information; sug-
gested courses of action; and sustained motivation to continue sense
making. Thus, the posture that one takes at any one point in a rumor
discussion fulfills an important collective sense-making function.
We finally wondered, "How do postures and statements typically
change over the life of a rumor discussion?" Again, sense making is the
key to understanding the progression of postures over the life of a rumor
discussion. We observed that interaction episodes typically proceeded
as follows: Explanation-delivering and directing postures were more
common in the first quarter of rumor interaction, explanation evalua-
tion peaked during the third quarter, and casual participation (a posture
not associated with sense making) peaked at the end. A similar analysis
of statement types over rumor lifetimes yielded a similar pattern. Thus,
collective Internet rumor sense making seems to proceed in a multistage
process of bringing the rumor to the attention of the group (Stage 1),
sharing information (Stage 2), evaluating explanations (Stage 3), and
then resolving the problem (Stage 4). The idea that rumor discussions
are composed of various roles that may change over time is not new
(Shibutani, 1966; R. H. Turner & Killian, 1972); however, our discus-
sion advances, operationalizes, and tests these ideas.
 Summary, Model, and Research Agenda 237

Individuals and groups make sense using rumor, but this sense is
at times quite bizarre. The inaccuracies in sense making beg the ques-
tion, How accurate are rumors and how do they become more—or
less—accurate? In chapters 6 and 7 (this volume) we addressed several
questions related to rumor content change on the dimension of
accuracy.

HOW DOES RUMOR CONTENT CHANGE?

Four patterns of content change have been identified: Content is leveled
(the number of details is reduced), content is added (the rumor becomes
more detailed or elaborate), content is sharpened (certain details are
accentuated), and content is assimilated (content is shaped through
leveling, sharpening, or adding to be in greater accord with cognitive
schemas). Scholars generally agree that everyday rumors are sharpened
and assimilated, but there has been some difference of opinion about
whether rumors are generally leveled or added to. That is, do real-life
rumors tend to become smaller or larger? We argued that it depends
on transmission patterns stemming from situational ambiguity and
rumor importance. Leveling—but not adding—has tended to occur in
laboratory studies of rumor (in which participants are told to simply
pass information along) and in planted-rumor field study situations (in
which a rumor is planted and tends to simply be passed along rather
than discussed intensely). Leveling occurs in some real-life rumor epi-
sodes marked by low ambiguity and when a group is primarily engaged
in serially transmitting the information. In contrast, adding—but not
leveling—has occurred mostly in field observation studies set in high-
ambiguity and high-importance contexts such as catastrophes and mur-
ders. In other words, adding occurs in real-life ambiguous situations
of importance in which a group is attempting to interactively discuss
and collaborate on the rumor. Our argument highlights the sociological
versus psychological divide over rumor. Since the time of G. W. Allport
and Postman (1947b), psychologists have tended to view rumor activity
as serially transmitted information from one participant to another,
and the focus is on the individual's effect on rumor content at each
transmission node. Sociologists have tended to see rumor as transmitted
in more circular patterns for the collective purpose of explanation
construction, and the focus is on how the group as a whole collabora-
tively affects content. Our view is that rumor primarily functions to
help people make sense, but the dynamics of this sense making vary.
Some rumor episodes (we speculate the minority) are characterized
more by simple, perhaps even serial, transmission, whereas other rumor
episodes are better characterized as interactive discussions. The former
will tend to exhibit leveling; the latter, adding.
238 R U M O R P S Y C H O L O G Y

WHAT IS RUMOR ACCURACY?

In chapter 6 (this volume) we also presented a much needed conceptu-
alization of the term accuracy. We proposed two types of rumor
accuracy: Verity refers to correspondence with reality; precision refers
to correspondence with some previous version of the rumor. Rumor
accuracy varies on how true it is, and also on how precisely it has
been transmitted. Rumor verity is analogous to instrument validity
in test construction; rumor precision is analogous to instrument reliabil-
ity. We proposed a categorization scheme based on rumor verity and
distortion toward or away from verity. There are three types of true
rumors: Stars are those that were precisely transmitted, converts
became truer as they were transmitted, and grainies became slightly
distorted as they were transmitted (although they are still mostly
true). There are also three types of false rumors: counterfeits are those
that were precisely transmitted, fallen stars started well but ended
badly, and hopefuls started false but show signs of slight improvement
(although they are still mostly false).

HOW IS RUMOR ACCURACY MEASURED?

Field observational studies of accuracy have tended to measure rumor
verity. These investigations typically collected rumors or elements of
a rumor in circulation in a field setting, ascertained which rumors or
rumor elements were true, and calculated the percentage that were
veritable. Lab and field experiments have tended to measure rumor
precision. Lab experiments typically measured the distortions that
occurred in participants who serially transmitted—either with or
without discussion—a description of a stimulus such as a drawing;
these methodologies resembled the telephone game or whisper down
the lane. Field experiments usually planted a rumor within a setting,
collected the rumor in circulation at some later time, and compared
the terminal version with the original. It is important to note that
rumor precision was taken as a proxy for rumor verity. That is, in
the same way that test makers are interested in measurement reliability
because it affects validity, rumor researchers who measured precision
were ultimately interested in—and generalized their result to—rumor
verity. Although there are two different types of accuracy, we don't
think it will be helpful to promote the widespread use of these two
terms; rather, it is enough to be aware of the distinction when one
considers and designs rumor accuracy research. Therefore, when we
and other rumor researchers refer to rumor accuracy, rumor verity
is intended.
 Summary, Model, and Research Agenda 239

HOW ACCURATE ARE RUMORS OVERALL?

Is rumor's reputation as false information generally true? The answer
seems to be that rumor accuracy varies widely, but some settings—
notably organizational grapevines—generally produce accurate (i.e.,
true) rumors. We examined the handful of rumor accuracy field studies;
accuracy varied substantially, but those rumors circulating in organiza-
tional grapevines tended to be highly accurate. We presented results
of our interviews about rumor accuracy with managers and communi-
cations personnel in several organizations; their estimates of the accu-
racy of specific rumors were quite high, especially when the rumor
had been in circulation for a time. We reported two additional studies
in which we collected true and false workplace rumors from employed
students in the United States and Australia, assessed their accuracy,
and measured their accuracy trend (i.e., the extent to which they
become more accurate or less accurate over time). When given the
option of recalling a true or false rumor, most overwhelmingly recalled
true rumors; indeed, most recalled converts (true rumors that became
more accurate over time). When students were asked to recall both true
and false rumors, true rumors were again overwhelmingly composed of
converts; false rumors were either counterfeits or fallen stars. These
studies suggest that organizational rumors that have since been proven
true or false tend to be accurate: The reputation of workplace rumor
as inaccurate is itself inaccurate! These studies also suggest what we
have dubbed the Matthew accuracy effect: Of those organizational
rumors that prove true or false, the true tend to get "truer" and the
false either stay the same or become more false.

WHAT CAUSES ACCURACY

AND INACCURACY?
In more specific terms, what are the mechanisms by which accurate
(or inaccurate) rumor content is generated or changed? In chapter 7
(this volume), we reviewed literature pertinent to rumor content
change with regard to accuracy and proposed five sets of mechanisms
by which rumors become more accurate and less accurate; we proposed
cognitive, motivational, situational, group, and network processes.
Cognitive distortions result from limits to attention and memory, and
from schema activation: The narrowing of listener attention in the
context of serial transmission with no discussion results in inaccuracies
characterized by emphasis on salient or central information. Memory
limits—again in the context of serial transmission without discussion—
result in inaccuracies characterized by easily remembered information.
240 R U M O R P S Y C H O L O G Y

Perceptual biases accrue from the interpretation of stimuli to be more

in accordance with activated cognitive structures such as schemas
and stereotypes.
Motivational mechanisms also affect rumor content: Accuracy mo-
tivation on the part of rumor participants—given the needed
resources—is likely to result in more accurate rumors. Relationship-
enhancement motivation results in the selective transmission of rumors
that are likely to find a ready reception with the hearer; these rumors
may or may not be more accurate. Self-enhancement motivation results
in the transmission of rumors that boost the self-esteem of the teller,
put his or her ingroup in a favorable light, or denigrate his or her
outgroup; these rumors are often less accurate. Situational features
affecting accuracy include high collective excitement, capacity to check
rumor veracity, and time: High collective excitement leads to greater
inaccuracy because people become more suggestible and less critical
and adopt lower norms for the acceptance of information. The capacity
to check rumor validity not surprisingly increases rumor accuracy.
Constraints on this capacity occur in numerous ways: serial transmis-
sion without discussion, erroneous information from credible sources,
novel channels of communication, the urgent need to act quickly to
avoid threat, the lack of firm information on a matter, distance from
valid sources of information, and the inability to compare rumors with
other people. Time tends to result in either greater accuracy or
greater inaccuracy.
Group processes include conformity, culture, and epistemic norms.
Conformity processes—bringing individual behaviors and attitudes in
line with group consensus—may lead to greater or lesser accuracy,
depending on the veracity of group consensus. Rumors similarly tend
to accord with the beliefs, values, and practices contained in group
culture. Part of this culture is the epistemic norms—the group standards
for evidence; to the extent that group standards are high, rumor accu-
racy is more likely. Network mechanisms include interaction, transmis-
sion configuration, and channel age (i.e., length of time established
channels of communication have existed). Interaction—discussion that
occurs during rumor transmission—tends to produce more accurate
rumors. Three transmission configurations were identified: serial trans-
mission with or without interaction, cluster transmission (one person
tells persons in a group and some of these persons pass it on to other
clusters), and multiple interaction (MI; rumor is actively discussed and
recirculated). Serial and cluster transmission patterns are generally
less active than are MI patterns. We theorized that serial and cluster
transmission would lead to distortion at each transmission node because
of the cognitive processes discussed earlier. MI patterns of transmission
would tend to either mitigate or accentuate these and other errors.
 Summary, Model, and Research Agenda 241

Groups with what H. Taylor Buckner dubbed critical set orientation would
yield more accurate rumors with greater MI; groups with an uncritical
set orientation would yield less accurate rumors (Buckner, 1965). Group
orientation here refers to the combination of variables affecting a group's
capacity and inclination to ferret out the facts; these variables include
such factors as ability to check, high standards of evidence, and the
presence of established channels of communication. Finally, communi-
cation channel age was theorized to be related to accuracy; rumors
passed along established channels tend to be more accurate because
the tellers' credibility can be tagged. Novel channel credibility is natu-
rally less able to be ascertained.
In chapter 7 (this volume), we also presented new research findings
designed to explore and test some of the motivational, group, and
network mechanisms discussed here. Interviews with communication
personnel were consistent with the idea that a critical set orientation
was associated with greater rumor accuracy. When so motivated and
when the group is able to ferret out the facts of a situation, they do
so. Results from a survey of employed students suggested that channel
age and the group epistemic norm of skepticism each predicted rumor
accuracy. Furthermore, MI moderated the relationship between skepti-
cism and accuracy; that is, MI produced more accurate rumors in skepti-
cal groups and less accurate rumors in gullible groups. MI did not
moderate the relation between channel age and accuracy, however.
These tentative results are mostly consistent with key tenets of Buck-
ner's (1965) theory of rumor accuracy, although much work remains.

HOW DOES TRUST AFFECT RUMOR

TRANSMISSION?
We have often observed that where distrust grows, rumors flow; in
chapter 8 (this volume) we systematically investigated this idea. We
presented a longitudinal study of an organization undergoing a radical
downsizing. On the basis of a framework proposed by trust researchers
(Dirks & Ferrin, 2001) we hypothesized that distrust—expectations of
the malignant nature of management's intent—would engender more
negative appraisals of ambiguous events—negative rumors—and be-
haviors congruent with these appraisals. And indeed we found what
we expected: Distrust strongly predicted the likelihood of transmitting
negative rumors heard and was a better predictor than both uncertainty
and anxiety. When formal sources of information are not trusted, peo-
ple resort to sense-making rumor discussions. Furthermore, when man-
agement is not trusted, their actions are appraised as potential threats,
which increase the need to regain a sense of control and understanding;
employees attempt to do this through participation in the rumor mill.
242 R U M O R P S Y C H O L O G Y

Results also suggested that trust moderated the relationship be-

tween anxiety and rumor transmission. When trust is high, rumor
activity depends on anxiety; but when trust is low, rumor activity
is uniformly high. These same patterns were found for uncertainty,
although they were weaker. In other words, if a person distrusts man-
agement, he or she will tend to transmit rumors heard regardless of
his or her level of uncertainty or anxiety. Management's actions are
seen as antagonistic; even a small amount of anxiety and uncertainty
becomes magnified. Also, spreading a negative rumor that denigrates
management may be self-enhancing. These findings point toward trust
as perhaps a more important variable in rumor transmission than un-
certainty or anxiety. At the very least, it should be routinely measured
in future investigations of rumor transmission.

HOW CAN HARMFUL RUMORS BE MANAGED?

In chapter 9 (this volume), we reviewed the literature on rumor-
quelling strategies and summarized these in terms of whether they
intervened in the rumor generation (e.g., reduce uncertainty), evalua-
tion (e.g., reduce belief via rumor rebuttals), or transmission (e.g.,
threaten legal action) stage. The majority of the writing on prescriptions
for rumor management recommends strategies aimed at reducing gen-
eration of and belief in rumors. We also reviewed results of a survey
of highly experienced public relations (PR) professionals (with over 26
years of experience on average) on the prevalence and effectiveness
of rumor-quelling strategies. The results showed that rumor prevention
strategies (e.g., minimize uncertainty by stating values and procedures
that will guide organizational change) were the most popular, whereas
punitive strategies (e.g., identify and punish rumor-mongers) were the
least popular. Also, a priori strategies that structured uncertainty were
rated as more effective in managing rumors than was after-the-fact
rumor control via formal communication methods (e.g., rebutting the
rumor by internal or external sources). However, not all rumors can
be prevented, and rumor rebuttal remains an important rumor manage-
ment tool (Koenig, 1985).

HOW EFFECTIVE ARE REBUTTALS IN

REDUCING BELIEF IN A RUMOR?
We considered rumor rebuttals as persuasive messages; their aim is to
persuade people to disbelieve the rumor. We reviewed empirical re-
search suggesting that, as with other persuasive messages, rumor rebut-
tals are more effective when delivered by sources that match the scope
 Summary, Model, and Research Agenda 243

of the rumor, when sources are perceived to be honest, when the

rebuttal message discredits the rumor and provides useful (and anxiety-
alleviating) information on the topic of the rumor, and when the rebut-
tal message provides a context for the rebuttal (i.e., is framed in the
context of a prior accusation or rumor). Our findings also suggest that
a rebuttal is better than a no-comments strategy in combating a rumor.
However, rebuttals are not a panacea and some rumors refuse to go
away. This resistance of some rumors to rebuttals can be explained
by the belief perseverance phenomenon: Although belief in rumors
weakens when they are rebutted, some cherished beliefs endure and
the negative impressions created by a rumor persist. Rebuttals that
convincingly deny the rumor and associate the target with positive
characteristics seem more likely to restore positive attitudes toward the
rumor target.

Integrative Model

Figure 10.1 offers a pictorial representation of some of the main ele-

ments discussed in this volume. Let's describe each component of the
model. Recall first our definition of rumor: unverified and instrumen-
tally relevant information statements in circulation that arise in con-
texts of ambiguity, danger, or potential threat, and that function to help
people make sense and manage risk. The context of rumor activity—
symbolized by the background labeled Environmental Characteristics—is
ambiguity and threat. Such situations may pose physical or psychologi-
cal threat. In organizational settings, change is often the reason behind
ambiguity and threat; restructurings, downsizing, new technology, staff
changes, and other events like these raise questions among and carry
potentially deleterious effects for employees.
In the midst of change, ambiguity, or threat, groups and individuals
engage in sense making, which is represented by the large oval. Rumor
is the making of sense—represented in our model as a piece of the
puzzle that the group is attempting to assemble. The sense making
happens at both the group and individual levels—hence the labels
Group Sense Making and Individual Sense Making. At a group level,
rumors are information statements in circulation among and through
the rumor public; two arrows circling back on one another represent
this. In addition, characteristics of the group play an important role
in group sense making. These characteristics include group culture—
including beliefs, values, and group epistemic norms—and network
244 R U M O R PSYCHOLOGY

Environmental Characteristics
-Situational Ambiguity
-Danger or Threat
-Change

Cognitions
Psychological -Importance
States -Belief
-Trust
-Sense of Control

Characteristics of the Group

-Culture
-Network

Integrative model of rumor.

characteristics (see DiFonzo & Bordia, in press, for a discussion of net-

work characteristics and rumor transmission).
At the individual level, sense-making processes are affected by
motivations, psychological states, and cognitions. Motivations for ru-
mor transmission include fact finding, relationship enhancement, and
self-enhancement. Uncertainty and anxiety are aversive psychological
states of mind. Cognitions include the level of trust the individual has
in official information sources or in the targeted group, beliefs about
the veracity and importance of the rumor, and the loss of a sense of
control. Each element affects the others. For example, uncertainty
often leads to fact-finding motivation. Distrust of the target group may
engender self-enhancement motivations and lead to greater anxiety.
Relationship-enhancement motivation may lead to heightened belief
in a rumor sanctioned by the group.
 Summary, Model, and Research Agenda 245

Future Research
Agenda

What's next for rumor research? In this section we set forth an agenda
that highlights emerging trends, addresses gaps in knowledge, and
outlines specific investigations.

CONCEPTUALIZATION AND CLASSIFICATION

OF RUMOR, GOSSIP, AND URBAN LEGEND
In chapter 1 (this volume) we introduced the IDS to distinguish forms
of rumor from gossip and urban legend and tested these ideas in a
laboratory setting; we found that our small set of information exemplars
fit predicted IDS patterns. What IDS patterns obtain for information
communicated in nonlaboratory settings? This question may be ad-
dressed by collecting rumor, gossip, urban legends—and the many
intermediate and nebulous forms that exist—from archived sources
(e.g., snopes.com, urbanlegends.com, online discussions of rumor), by
recording conversations in public places, or by using diary methods,
and then rating them with the IDS. Special attention should be paid
to the pattern of functional and contextual differences we have hypoth-
esized. Such investigations would shed much needed light on the
roles—unique, complimentary, or common—that each form of com-
munication plays during ordinary social discourse.
Research in this vein would clarify the ongoing ambiguity associ-
ated with these forms. This research should also expand, refine, and
validate the IDS. We have hypothesized that the primary motivations
for participating in rumor discussions are to understand, to belong, and
to self-enhance, whereas the primary foci of gossip and urban legends
are to belong and to self-enhance. We have similarly noted that some
rumors survive and some don't; what IDS differences exist between
"fit" and "unfit" rumors? We think that the "fittest" rumors should be
those that most help groups understand, belong, and self-enhance.
Finally, we have raised the possibility that nonfalsifiable rumors are
more likely to survive; any expansion of the IDS should include a
falsifiability dimension.

RUMOR FORMS, FREQUENCIES, EFFECTS,

AND EFFECT MECHANISMS
In chapter 2 (this volume) we overviewed research on the types of
rumors that occur, their prevalence, and their effects. We tended to
246 R U M O R P S Y C H O L O G Y

focus on organizational settings. What rumor forms, frequencies, and

effects exist in other domains, such as religious communities, conflicted
ethnic or racial enclaves, populations engaged in behaviors adversely
affecting health, local communities considering environmental policy,
elections for public office, terrorist cells, protest crowds, populations
affected by natural disasters, Internet blogging networks, schoolchild-
ren friendship groups, and consumers considering product purchases?
Cataloging and counting rumors and rumor effects in such contexts
would illuminate the contours of rumor sense making in these arenas.
Perhaps of greatest practical concern are rumors that adversely affect
health and health detection behaviors, escalate ethnic conflict, interfere
with natural disaster responses, and are purposely planted as
propaganda.
Of special interest are the cumulative effects of hearing rumors.
The longitudinal study we presented indicated that hearing negative
rumors month after month affected—among other variables—job satis-
faction, trust, productivity, and withdrawal behaviors. This study
should be replicated and the analysis should be extended to positive
rumors: Positive rumors—heard repeatedly—should affect organiza-
tional attitudes and behaviors in an opposite manner. In this vein of
research, special attention should be paid to the mediating mechanisms
by which rumors affect outcomes. To wit, are rumor effects mediated
through social learning, equity calculations, or schema activation? Fi-
nally, our understanding of rumor propaganda effects on the evaluation
of both rumor target and rumor transmitter is currently limited (rumor
propaganda refers to rumors intentionally planted to influence public
opinion or behavior; DiFonzo & Bordia, in press). Rumors are often
spread in the context of rival, competing, or conflicted groups; how
does hearing negative rumors about Group A affect the evaluation of
Group B? Even when such rumors are not believed, do they have their
intended effect? Such knowledge would be timely and practically useful
in light of the heightened polarization and conflict of our post-9/11
world.

MOTIVATIONS IN RUMOR SPREAD

In chapter 3 (this volume) we presented a tripartite framework of
motivation in rumor spread: fact finding, relationship enhancement,
and self-enhancement. Most work has been done, albeit indirectly, on
the fact-finding motivation of rumor transmission. Several interesting
questions remain, however. First, it has been repeatedly asserted that
people transmit rumors to reduce anxiety or to indirectly cope with
an unfavorable situation by understanding and interpreting it (this is
a fact-finding motivation). Yet we are not aware of any research that
 Summary, Model, and Research Agenda 247

tests this anxiety-reduction function, that is, how effective rumors are
at reducing anxiety.
In contrast to fact finding, we know little about the relationship-
enhancement goal in rumor transmission. For example, what are the
effects of rumor transmission on others' evaluation of the transmitter?
A person may hesitate to share a negative outgroup rumor with a
member of the outgroup, for example, because he or she thinks it
impolite to do so. A second example: Sharing a negative ingroup rumor
with members of one's ingroup, especially in the context of a highly
polarized ingroup-outgroup conflict, may cost a person a great deal of
ingroup social capital. Investigating questions of this sort would lead
to a fuller understanding of the relationship-enhancement motivation
and would explore the social costs of sharing a rumor.
Much work also remains in exploring the self-enhancing motiva-
tion in rumor spread. First, the notion of self-enhancement is that
people, desirous of boosting their self-image, build themselves up by
putting others down in the form of negative outgroup tales. This idea
needs further systematical testing, especially with the aim of teasing
out relationship-enhancing motivations. Sharing a negative rumor
with another school chum about a rival school is at once self- and
relationship-enhancing; indeed, it may be that people share these
sorts of rumors mostly for the social benefits they accrue and not to
boost their self-esteem. These goals may perhaps be differentiated by
experimental manipulations that temporarily raise or lower par-
ticipants' self-esteem and expose them to self-enhancing and non-
self-enhancing rumors; participants with lowered self-esteem should
spread self-enhancing rumors more readily. Second, theorizing on
self-esteem suggests that it is multidimensional (Baumeister, Campbell,
Krueger, & Vohs, 2003). People with unstable or narcissistic forms
of self-esteem—typically based on group membership—may be more
prone to protect and enhance their self-image by spreading negative
outgroup rumors, whereas those whose self-esteem is based on a
realistic appraisal of accomplishments and relationships may not. For
the threatened narcissist, negative outgroup rumors may serve as a
form of aggressive self-esteem enhancement; this idea is ripe for testing.
Third, as discussed in chapter 3 (this volume), the self-enhancement
motive has often been capitalized on by those who have consciously
planted rumors to be spread to drive a wedge between groups. Yet
no attention has been paid to the motivations of those who orchestrate
and spread such rumor propaganda. What is the extent and the
effectiveness of rumor propaganda efforts, and how may their effects
be prevented or neutralized? We have outlined these intriguing—
and practically useful—questions elsewhere (DiFonzo & Bordia, in
press).
248 R U M O R P S Y C H O L O G Y

Several interesting hypotheses surround how these three motiva-

tions operate overall. First, in what circumstances is each motivation
generally operative? We speculate that relationship enhancement is
the conversational norm from which one might deviate only when the
possibility of tangible loss becomes salient (resulting in fact finding) or
when one's sense of self is threatened (resulting in self-enhancement).
Second, we posited that anxiety, uncertainty, and involvement lead to
fact-finding motivation (a direct effect) which in turn leads to rumor
transmission. In other words, fact-finding motivation mediates the
effects of anxiety, uncertainty, and involvement on rumor transmis-
sion. This hypothesis has yet to be empirically tested. Third, we posited
that relationship-enhancement motivation moderates the relationship
between belief and rumor transmission. When people are motivated
to enhance long-term relationships, they tend to pass along only those
rumors that they believe; in the absence of this motivation, belief may
not be closely related to transmission. This hypothesis also has yet to
be empirically investigated.
In the broader view, much work remains to be accomplished in
conceptualizing and investigating how each motivation is related to
each antecedent. We speculate that differing types of negative events
lead to differing motivations: Anxiety over possible loss of well-being
leads to fact finding, anxiety about possible relationship deterioration
leads to relationship enhancement, and anxiety over potential damage
to self-image leads to self-enhancement motivation. Anxiety thus pro-
vides the goal of the motivation. We further speculate that uncertainty,
involvement, and belief simply intensify motivation; for example, fear
of losing one's job from downsizing motivates fact-finding motiva-
tion, and this motivation is especially strong when the management
issues a no-comments statement (uncertainty is high), it is one's
department—as opposed to another—that is likely to be downsized
(outcome-relevant involvement is high), and downsizing seems plausi-
ble because management has downsized before in similar economic
climates (belief is high).
Of course, some of the most interesting questions involve the
effects—especially the cumulative effects over time—of each motiva-
tion on rumor selection, belief, and content change. To wit, we specu-
late that outgroup-negative rumors are most fit for survival over time
in the context of a group that is primarily motivated by self- and
relationship-enhancement, rather than fact-finding, motives. This hy-
pothesis could be tested within a broader research paradigm that varied
the motivation of experimental participants and then exposed them to
various ingroup-outgroup positive-negative rumors and measured the
likelihood of transmission to an ingroup-outgroup recipient. The re-
search methodology would be especially fruitful if it also included the
 Summary, Model, and Research Agenda 249

opportunity for participants to transmit rumors more than once and

to several neighbors; for example, give Participants A, B, C, and D—
all of the same ingroup—the opportunity to share either an outgroup-
positive rumor or an outgroup-negative rumor with each other indi-
vidually over several rounds of interaction. In conditions in which
participants were first motivated to self- or relationship-enhance, these
motivations should be accentuated by normative social influences (e.g.,
attitudinal conformity) and quickly result in the selection of (and higher
levels of belief in) outgroup-negative—as compared with outgroup-
positive—rumors. In conditions in which participants are motivated to
fact find, this pattern of results should be weaker.
We leave this section on a methodological note. In the research
we reported in chapter 3 (this volume), we saw that relative motivation
was affected by varying rumor valence (positive-negative), rumor tar-
get (ingroup-outgroup), and rumor recipient (ingroup-outgroup). For
example, the relationship-enhancement motivation was highest for the
condition in which the rumor was positive and about the outgroup and
the recipient was an outgroup member. This paradigm holds promise—
these manipulations may be used, for example, to test the effect of
motivation on rumor transmission and content—but we think these
procedures need refinement and replication with other scenarios.

CUES TO CONFIDENCE IN RUMOR

Belief is an understudied and interesting aspect of rumor phenomena;
several questions need to be addressed. First, a descriptive question:
How much confidence do people place in commonly circulated rumors?
In unpublished pilot work we asked college students to rate their belief
in a set of external organizational rumors selected from snopes.com.
We found very low levels of belief overall, but our sample of rumors—
and participants—was woefully narrow. Of particular significance in
future work would be confidence ratings of rumors related to specific
attitudes of interest, such as attitudes about competing groups. For
example, the false rumor that the AIDS virus was created in a western
laboratory and tested on 100,000 Africans is believed by one third of
a sample of African American churchgoers (another third are unsure
about it; "Black Beliefs," 1995); the high level of belief in this rumor
tells us much about the state of race relations in America today. Work
is also needed to describe in more detail the shape of the relationship
between belief and other outcomes; that is, we wonder how much belief
is necessary for rumors to be transmitted or to affect other attitudes and
behaviors. The nature of the relationship between belief and transmis-
sion may be linear or it may be catastrophic (a little belief affords large
effects). In a similar way, even low-confidence negative rumors may
250 R U M O R P S Y C H O L O G Y

affect behaviors, as when people decline to purchase a product rumored

to contain carcinogens.
Second, why do people believe the rumors they do? In chapter 4
(this volume) we applied Brunswik's lens model of social judgment to
frame this question and concluded that certain cues—agreement with
attitude, source credibility, repetition, and absence of rebuttal—were
operative. We first note the dearth of experimental evidence on the first
three cues (agreement with attitude, source credibility, and repetition);
much work could (easily) be done to experimentally validate the use
of these cues in rumor confidence judgments and to derive the relative
weight each cue is accorded. Second, in line with the previous chapter
on motivations, the agreement with attitude cue might be fruitfully
divided into cooler (the rumor agrees with a currently held attitude
that is not closely tied to one's self-identity) versus warmer (the rumor
agrees with an attitude that is closely tied to one's self-identity) cue
subsets. The key marker between a warm versus cool cue is thus
whether or not it bolsters a defensive reaction against a perceived
threat; warm cue rumors are therefore self-enhancing. For example,
strongly identified Democrats in 2004 may have been more likely to
believe the false rumor that George W. Bush hypocritically misquoted
the Bible because the rumor was self-enhancing, that is, it was a defam-
atory story attacking a perceived threat (the same could also be said
of Republicans who felt threatened by Senator John Kerry). These cues
should be experimentally tested; we speculate that self-enhancing cues
are stronger than cues that indicate mere agreement with one's attitude.
In broader terms, rumors that self-enhance are probably more likely
to be believed than are rumors that don't, especially by people with
low, unstable, or narcissistic self-esteem. In lens model terms, it seems
likely that self-enhancement cues may be accorded greater weight in
assessing one's confidence in a rumor.

RUMOR AS SENSE MAKING

In chapter 5 (this volume) we considered rumor sense making at the
level of the individual and the group. At the individual level, we argued
that many rumors convey a stable-cause explanation. Do they indeed?
This hypothesis could be tested by procuring blind judges to evaluate
a sample of rumors from a variety of domains and coding them for the
presence or absence of a stable-cause explanation (x causes y and this
cause is relatively stable over time). Less retrospective approaches could
query participants about recent rumors:
Think of a recent bit of information you heard about which
you [or your group, acquaintances, friends, coworkers, or
family] are unsure. Describe the information and the situation
 Summary, Model, and Research Agenda 251

surrounding it [these descriptions could be coded for stable -

cause explanations]. Rate the extent to which this information
explains or makes sense of a current situation or event. What
current situation or event, if any, does this information
explain? According to the information, what is the cause of the
situation or event? Finally, rate the extent to which this cause
will [or will not] last over time.

We also presented evidence suggesting that rumors lead people

to perceive strong relationships between variables that are weak or
nonexistent. This idea could be applied to increase knowledge of how
racial stereotypes spread. It may be that rumors embedded with stable-
cause explanations teach people to associate race with particular behav-
iors. This idea could be first explored by collecting race-related rumors:
"Think of a recent bit of information you heard—about which you [or
your group, acquaintances, friends, coworkers, or family] are unsure—
that involved a member of another race." Participants could then de-
scribe the information and the situation surrounding it and these de-
scriptions could be coded for stable-cause explanations. Research in
this vein would be particularly interesting among child and school-
age populations.
At the group level, research is needed to replicate the findings—
reviewed in chapter 5 (this volume)—pertaining to the statement con-
tent, communicative postures, and dynamic flow in rumor interaction
episodes over the Internet. These investigations should also explore
the same phenomena in face-to-face interaction episodes through tape
recording and transcription: Are the same patterns evident? Other in-
teresting questions: What would the patterns of content, posture, and
flow look like in a highly anxious group or one that perceived an
imminent threat? The sense-making process might be more rapid and
less deliberative in nature. In a similar vein, how would they look in
a highly identified ingroup in the context of ethnic conflict with an
outgroup? The sense-making process might include a conformity-
demanding posture in which ingroup criticism was not permitted. Fi-
nally, how might these patterns change in the context of a more highly
cohesive set of participants, say those in a very close-knit social unit?
We speculate that the sense-making process may again proceed more
quickly as compared with the Internet rumor discussions we analyzed.

RUMOR ACCURACY
In chapters 6 and 7 (this volume) we reviewed several questions related
to rumor content change, accuracy base rates, and mechanisms related
to how rumors become more and less accurate. We explored the litera-
ture on rumor content change and proposed that leveling (loss of
252 R U M O R P S Y C H O L O G Y

details) occurs mostly in some real-life rumor episodes marked by low

ambiguity and in which a group is primarily engaged in simply passing
along the information, whereas adding (gain of details) occurs mostly
in situations of high ambiguity and importance and the group is quite
interactive and collaborative. This hypothesis should be tested experi-
mentally to put to rest the persistent scientific legend that rumors
always level. Research on this topic dovetails with an emphasis in social
psychology on understanding social processes in the context of the act
of communicating.
With regard to rumor accuracy base rates, we call for a revival of
accuracy studies in naturalistic environments. With the proliferation
of computer messaging archives, such studies ought to be far less labor-
intensive than the bulk of those performed in the 1960s and 1970s.
Rumors communicated via the Internet ought to be collected from a
variety of domains and to sample a variety of situation types, such as
high-involvement collaboration and low-involvement minimal-
discussion serial transmission. (At the same time, efforts can be made
to ascertain the relative frequency of these two situations.) After collec-
tion, the percentage of rumors or rumor elements that are true should
then be ascertained. Of course, if such efforts can measure mechanisms
affecting rumor accuracy discussed in chapter 7 (this volume), all the
better; such data will of course be difficult to collect naturalistically
and retrospectively. A series of diary-type web-based studies may be
the most promising methodological avenue for collecting information
of this type in field settings. Participants could be charged with the
ongoing task of journaling rumors they hear and responding daily to
items aimed at measuring cognitive, motivational, situational, group,
and network data. If accuracy trends are also recorded, it may be
possible to assess how widespread the Matthew accuracy effect is (accu-
racy trends bifurcate).
The set of accuracy mechanisms reviewed in chapter 7 (this vol-
ume) represents a rich source of theoretical propositions to be explored
and tested. We propose several ideas here pertaining to each set. With
regard to cognitive mechanisms, much could be gained by adopting
the teller-listener-extremity effect paradigm—which uses serial
transmission—but also allowing discussion (interaction) at each trans-
mission node (Baron, David, Brunsman, & Inman, 1997, also proposed
this). Because discussion allows for greater precision in transmission,
accuracy should increase. In addition, we proposed that anxiety in-
creases the reliance on extant cognitive structures (e.g., stereotypes)
during transmission. This hypothesis could be tested by assessing each
participant's level of implicit or explicit stereotyping, then manipulating
participant anxiety before performing the tasks involved in the teller-
listener-extremity effect paradigm. Lower anxiety and allowing partici-
 Summary, Model, and Research Agenda 253

pant discussion during transmission should each lead to less extreme

judgments. These effects should be especially pronounced when the
relevant cognitive structures (e.g., the stereotypical football player) are
experimentally and subliminally primed (e.g., with a tachistoscope).
With regard to motivational mechanisms of accuracy, the method-
ology used by Lyons and Kashima (2001) could be extended by allowing
discussion at each transmission node; fact-finding, as compared with
relationship- and self-enhancement motivations, should increase accu-
racy. Second, very little work has been done on the conceptualization,
measurement, and effect of the motivation to be "in the know" or
to transmit rumors "for effect" (Sinha, 1952); we suspect that this
motivation powerfully affects rumor accuracy. Third, the effects of self-
enhancement motivation on accuracy (and rumor content change) is
an area ripe for investigation; we have already outlined these ideas in
the rumor motivations section earlier.
With regard to situational features, to our knowledge, no experi-
mental work has followed up on R. H. Turner and Killian's (1972)
idea that high collective excitement results in less stringent norms
about what information is acceptable, especially in close—versus
diffuse—groups. In addition, a group's corporate capacity to check
rumor verity is an area in need of greater conceptual clarity and
measurement. One goal of this work would be to reliably measure a
group's capacity to check, and to relate this to overall rumor accuracy
levels during any given rumor episode. One interesting question
here: "How does the capacity to check fare against antiaccuracy
motivations?" That is, how accurate are rumors in groups that have
a high capacity to check rumor veracity but are strongly motivated
to believe or disbelieve them? This question seems highly amenable
to experimentation. We speculate that capacity to check trumps moti-
vation when they are in conflict; rumors that are clearly false will
not thrive, even among a group that strongly desires them to be true.
We reason that in situations in which the rumor can be objectively
checked, even relationship- and self-enhancing motivations come to
act in service of accuracy because sending true information is a deeply
engrained social rule. This research question has important practical
implications; as previously discussed, many troubling rumor effects
arise in situations of conflict in which people are strongly motivated
to believe the worst about rival groups.
With regard to group mechanisms, the application of conformity
literature to rumor phenomena seems a promising line of inquiry.
In what circumstances does the transmission of an unverified bit of
information cease to be sense-making activity and become a means of
influence? When is external acceptance of a rumor demanded? In
more specific terms, how does the well-known set of factors affecting
254 R U M O R P S Y C H O L O G Y

conformity (e.g., unanimity, public balloting, group cohesion, low

self-monitoring, group size) affect external acceptance of a rumor?
Research of this sort would explore how, for example, "racialized nar-
ratives" (Maines, 1999) are not only common sets of beliefs circu-
lating within communities, but also mandatory. A second group
mechanism—culture—is witnessing resurgence in social psychology
in general. Indeed, Lyons and Kashima (2001) found that culture
informs rumor content change in serial transmissions of rumor. To
what extent are rumors "culturally anchored" as proposed by Shibutani
(1966)? The question first requires a sharper conceptualization—and
measurement—of what we mean by cultural anchoring. The question
might best be investigated in the context of well-developed cultural
differences such as collectivism and individualism. For example, would
the terminal story of the drunken football player in Baron et al.'s (1997)
teller-listener-extremity effect study highlight more collectivistic than
individualistic themes among Chinese as compared with American
participants?
Finally, the investigation of network mechanisms is surely one of
the most promising current avenues of accuracy research. In general,
these research paths can be fruitfully guided by Buckner's (1965) con-
ceptualizations of group orientation and MI; researchers can begin by
sharpening these constructs, then systematically evaluating Buckner's
(and our) hypotheses about how they act and interact to affect accu-
racy. Perhaps the greatest research fruit can be picked from this woe-
fully underresearched question: "How do network transmission
configurations—the network structure of information flow—affect
rumor accuracy?" In chapter 7 (this volume) we posited first that
different configurations differentially affect the influence of informa-
tion liaisons; serial and cluster transmission grant more influence to
liaisons than do MI configurations. This hypothesis could be evaluated
by experimentally manipulating network transmission configuration
and recording content changes at each node; liaisons should affect
content less in the MI condition.
Second, Buckner posited that MI increased the speed of rumor
content change relative to other configurations, but that group orienta-
tion moderated the direction of this change toward or away from accu-
racy. For example, groups that are accuracy motivated—as compared
with relationship-enhancement motivated—should produce more ac-
curate rumors and produce them more quickly in multiply interactive
than in serial or cluster-transmission configurations. However, groups
that are motivated by relationship enhancement—as compared with
accuracy—should produce less accurate rumors and (also) produce
them more quickly under MI than under serial or cluster transmission.
 Summary, Model, and Research Agenda 255

This idea has not been experimentally tested. The mechanisms that we
propose are at work in these hypotheses can also be investigated: MI,
as compared with serial and cluster transmission configurations, affords
the opportunity for greater checking of information (both cross-
checking across sources—as when Person C hears rumor variations
from both Persons A and B—and checking of precision accuracy—as
when Person C ensures that she heard the rumor correctly from Person
A by discussing it with Person A). MI seems to afford greater opportunity
for group forces, such as conformity, culture, and common perceptual
biases heightened by collective excitement, to operate, especially in
contexts in which the capacity to check is minimal.
We have been discussing how configurations of information flow—
MI, serial transmission, and cluster transmission—affect rumor accu-
racy. We dubbed these configurations network transmission configurations.
It is obvious, however, that these configurations occur in the context
of actual social networks and these social networks can themselves
be configured differently; we call networks social space configurations.
Elsewhere we have reviewed these social space configurations and
how they affect rumor transmission over time and across social space
(DiFonzo & Bordia, in press). Three examples: In the usual representa-
tion of a torus configuration, each individual is relationally connected
to his four closest "neighbors"—north, south, east, and west—in a
two-dimensionally uniform distribution that resembles a grid. A ribbon
configuration occurs when the person is connected to four neighbors
aligned as in a line—two on that person's left and two on the right. A
typical family configuration occurs when most of the people that one
is connected with are also connected with each other (Latane & Bour-
geois, 1996) and therefore form communities. These concepts have
great relevance for accuracy research. Theorists and researchers can
investigate how network transmission configurations are related to
social space configurations, especially as they relate to accuracy. It seems
likely that social space configurations may predispose and possibly
constrain certain network transmission configurations. For example,
we speculate that MI is more likely to occur in family than in ribbon
configurations in view of the fact that family configurations predispose
interaction within communities. Another interesting question for this
research vein is, "How would social space configuration affect the spatial
distribution of rumor accuracy over time?" We speculate that pockets
of accurate and inaccurate rumors are more likely to occur in "clumpier"
types of social space configurations such as the family (see DiFonzo &
Bordia, in press, for a closely related presentation of hypotheses about
the spatial distribution of rumors over time). This area is rich with
potential and virtually untapped.
256 R U M O R P S Y C H O L O G Y

TRUST AND RUMOR TRANSMISSION

In chapter 8 (this volume) we explored the effects of trust on rumor
transmission by reporting results of a longitudinal study of an organiza-
tion undergoing a radical downsizing. One main finding was that dis-
trust of management was a better predictor of rumor transmission than
were anxiety and uncertainty. This finding—the first of its kind—
should be replicated in other organizational settings. The current body
of experimentally based knowledge about rumor merely mentions
trust. In studies of rumor transmission, trust should become a central
and regularly measured antecedent, along with anxiety, uncertainty,
belief, loss of a sense of control, and importance. In addition to the
sense-making (solving a puzzle) context, rumor should also regularly
be considered in the context of intergroup conflict and threat; these of
course involve perceptions of trust.
The broader extensions of this research are ripe for discovery and
practically significant as well: Beyond management and staff, what
roles does trust play in the spread of negative rumors related to any set
of groups, especially those in conflict? The role of rumors in intergroup
relations has only occasionally been systematically studied (e.g., Kakar,
2005; Knapp, 1944); in addition to indicating the underlying fears,
beliefs, and attitudes of each party, future research could directly mea-
sure trust in the rival camp. We expect that distrust is strongly related
to belief and transmission of negative rumors—even after anxiety and
uncertainty are accounted for. (In addition, more general attitudes of
basic mistrust, anomie, and powerlessness should correlate with belief
in and transmission of all sorts of negative rumors.) Longitudinal re-
search could investigate the direction of causality in a fashion similar
to our analysis in chapter 2 (this volume): Does distrust lead to negative
rumor transmission, or does transmitting negative rumors (through a
process of schema activation or cognitive dissonance) lead to distrust?
We suspect that both occur. We suspect that both anxiety and distrust
feed off of one another, although we think that because attributions
of malevolent intent inevitably lead to worry, the distrust —> anxiety
relation is the stronger of the two. Finally in this vein, trust should be
experimentally manipulated in studies investigating rumor interven-
tions. Raising trust should itself reduce negative rumor transmission.
In addition, raising trust should aid rumor management efforts that
try to reduce uncertainty and anxiety.
Broadening this research raises an issue in need of conceptual
clarification. In our longitudinal study we measured trust in manage-
ment. In this case, management was both the source of formal informa-
tion and, in the context of the downsizing, the outgroup (anecdotal
reports and observation of rumor content showed evidence of us-
 Summary, Model, and Research Agenda 257

versus-them thinking). What of situations—quite common in conflict

situations—in which the outgroup and the source of formal information
are different entities? Trust of each entity (e.g., trust in the rival group
and trust in the press) should be measured. What role does each play
in the generation and survival of negative rumors? We speculate that
distrust of the press may be the more serious of the two.
In chapters 2 and 8 (this volume) we introduced several new depen-
dent variables related to rumor transmission: the number of different
rumors heard, the number of different rumors passed, and the likeli-
hood of transmission (the proportion of rumors heard that were
passed). These measures need further testing and validation. We suggest
that the diary method of data collection be used to more carefully record
rumors heard and passed as well as a description of the transmission
episodes. We also suggest that self-reports of these variables be com-
pared with reports from friends and coworkers. Other measures, such
as the number of different occasions on which a rumor was told and the
number of different people to whom it was spread (Pezzo & Beckstead,
2006), should also be considered. Each construct is in need of concep-
tual sharpening: What does each signify? For reasons outlined in chap-
ter 8 (this volume), we think that the likelihood of transmission quo-
tient should become the standard rumor transmission outcome variable
for actual rumors. Results from these studies would then be conceptu-
ally comparable to similar measures of behavioral intention in labora-
tory studies (e.g., "How likely, on a scale of 0% likely to 100% likely,
are you to share this information with a coworker?").

RUMOR MANAGEMENT
The effective control of harmful rumors is of great practical relevance
in a variety of domains, including public health, natural disaster and
crisis management, civil and military administration, organizational
communication, and PR. In chapter 9 (this volume) we reviewed re-
search in this area. It is clear that a great deal more needs to be done.
Although informed advice is readily available in the practitioner litera-
ture, very little systematic empirical research has been conducted on
this topic.
More research is needed on ways to prevent rumors. Intergroup
distrust (i.e., distrust between management and employees or between
rival ethnic groups) plays an important role in the spread of rumors;
when distrust is high, ambiguous events are interpreted as threatening.
How can the likelihood of this happening be reduced? Greater contact
between the rival groups (committees comprising management and
employee representatives or citizen forums with diverse ethnic repre-
sentation) may foster trust, reduce uncertainty about motives, and
258 R U M O R P S Y C H O L O G Y

lower the likelihood that outgroup-derogating rumors would find trac-

tion. Management of anxiety and uncertainty during change is difficult.
Strategies that structure the uncertainty should be useful in preventing
rumors. We recommend that laboratory and field studies with experi-
mental (involving communication strategies that structure uncertainty)
and control conditions be conducted to assess the effectiveness of rumor
prevention strategies.
More research is also needed on the effectiveness of rumor rebut-
tals. The role of the source of the rebuttal is important, but several
questions arise: How does trust in the source affect rebuttal effective-
ness? Higher trust will lead to greater rebuttal effectiveness, but is this
a linear relationship or monotonic? Just how much trust is necessary?
How do perceptions of the source of the rebuttal change after the
rebuttal? Ingroup sources are perceived as credible and are likely to be
more effective. However, what happens when an ingroup source rebuts
a rumor that was derogating to the outgroup? Does the source garner
respect for standing up for truth or lose authority and respect for selling
out? How does the motivational context of the rebuttal source influence
effectiveness? If the rebutter has something to gain, the rebuttal would
be less effective. In a similar vein, how do the motivations of the target
(or the rumor public) affect their acceptance of the rebuttal? We would
expect that people motivated by fact-finding would be more persuaded
by credible arguments, whereas those motivated by self-enhancement
would be persuaded by strategies that assuage hurt esteem. Studies
that manipulate motivations and then test the effectiveness of different
rebuttals will help test these ideas. Finally, what are the long-term
effects of rebuttals? Is there a sleeper effect (in which levels of belief
in the rumor may dip soon after the rebuttal, but resurface after the
passage of time)? Longitudinal designs will help explore the role of
memory processes in the long-term effectiveness of rebuttals.
Much of the research on rebuttals has been conducted in laboratory
settings; field studies are needed. Moreover, research on rumor man-
agement needs an interdisciplinary approach. Theoretical perspectives
from cognitive psychology, social psychology, and rhetoric and commu-
nication need to combine with approaches in marketing, crisis manage-
ment, and PR to arrive at context-relevant understanding of, and strate-
gies for, rumor management.

Conclusion

Rumors continue to be a firm feature of social and organizational

landscapes. We hope that we have put forth a clearer set of conceptual-
 Summary, Model, and Research Agenda 259

izations about rumor, its effects, motivations behind its transmission,

its sense-making functions, its accuracy, the role of trust in rumor
transmission, and how rumor may be managed. We have proposed an
integrated model of rumor phenomena. We have outlined a compre-
hensive agenda for rumor research. We hope that this volume has
served to highlight old and new questions surrounding rumor. We
expect that it has served to put useful new frames—drawn from social
and organizational literature—around the central aspects of rumor phe-
nomena. We trust that this more integrated psychological and sociologi-
cal approach to the study of rumor, the social and organizational frames
offered, and the set of current research methods and questions put
forth will be helpful to rumor researchers in the future.
 References

10 die in Haiti as a false rumor is met by of Abnormal and Social Psychology, 40,
violent street protests. (1991, January 3-36.
28). The New York Times, p. A3. Allport, G. W., & Postman, L. J. (1947a). An
Abalakina-Paap, M., & Stephan, W. G. analysis of rumor. Public Opinion Quar-
(1999). Beliefs in conspiracies. Journal of terly, 10, 501-517.
Political Psychology, 20, 637-647. Allport, G. W., & Postman, L. J. (1947b). The
Abelson, R. P., & Lalljee, M. (1988). Knowl- psychology of rumor. New York: Holt, Rine-
edge structures and causal explanation. hart & Winston.
In D. J. Hilton (Ed.), Contemporary science Ambrosini, P. J. (1983). Clinical assessment
and natural explanation: Commonsense con- of group and defensive aspects of rumor.
ceptions of causality (pp. 175-203). New International Journal of Group Psychother-
York: New York University Press. apy, 33, 69-83.
Ackerstrom, M. (1988). The social con- Anderson, C. A. (1983). Abstract and con-
struction of snitches. Deviant Behavior, crete data in the perseverance of social
9, 155-167. theories: When weak data lead to un-
Adams, J. S. (1965). Inequity in social ex- shakable beliefs. Journal of Experimental
change. In L. Berkowitz (Ed.), Advances in Social Psychology, 19, 93-108.
experimental social psychology (Vol. 2, pp. Anderson, C. A. (1985). Argument avail-
267-296). New York: Academic Press. ability as a mediator of social theory per-
Agnes, M. (Ed.). (1996). Webster's new world severance. Social Cognition, 3, 235-249.
dictionary and thesaurus. New York: Simon Anderson, C. A., Krull, D. S., & Weiner, B.
& Schuster. (1996). Explanations: Processes and con-
Aiken, L. S., & West, S. G. (1991). Multiple sequences. In E. T. Higgins & A. W.
regression: Testing and interpreting interac- Kruglanski (Eds.), Social psychology hand-
tions. Newbury Park, CA: Sage. book of basic principles (pp. 271-296). New
Ajzen, I. (1977). Intuitive theories of events York: Guilford Press.
and the effects of base-rate information Anderson, C. A., Lepper, M. R., & Ross, L.
on prediction. Journal of Personality and (1980). Perseverance of social theories:
Social Psychology, 35, 303-314. The role of explanation in the persistence
Allport, F. H., & Lepkin, M. (1945). Wartime of discredited information. Journal of Per-
rumors of waste and special privilege: sonality and Social Psychology, 39, 1037-
Why some people believe them. Journal 1049.

261
262 R E F E R E N C E S

Anderson, C. A., & Slusher, M. P. (1986). Baron, R. M., & Kenny, D. A. (1986). The
Relocating motivational effects: A syn- moderator-mediator variable distinction
thesis of cognitive and motivational ef- in social psychological research: Concep-
fects on attributions for success and fail- tual, strategic, and statistical considera-
ure. Social Cognition, 4, 270-292. tions. Journal of Personality and Social Psy-
Andreassen, P. B. (1987). On the social psy- chology, 51, 1173-1182.
chology of the stock market: Aggregate Baron, R. S., David, J. P., Brunsman, B. M.,
attributional effects and the regressive- & Inman, M. (1997). Why listeners hear
ness of prediction. Journal of Personality less than they are told: Attentional load
and Social Psychology, 53, 490-496. and the Teller-Listener Extremity effect.
Antaki, C. (1988). Explanations, communi- Journal of Personality and Social Psychology,
cation and social cognition. In C. Antaki 72, 826-838.
(Ed.), Analysing everyday explanation: A Bartlett, F. C. (1932). Remembering. Cam-
casebook of methods (pp. 1-14). London: bridge, England: Cambridge University
Sage. Press.
Antaki, C., & Fielding, G. (1981). Research Bauer, R. A., & Gleicher, D. B. (1953).
on ordinary explanations. In C. Antaki Word-of-mouth communication in the
(Ed.), The psychology of ordinary explana- Soviet Union. Public Opinion Quarterly,
tions of social behaviour: Vol. 23. European 17, 297-310.
Monographs in Social Psychology (pp. 27- Baumeister, R. F., Bratslavsky, E., Finken-
55). London: Academic Press. auer, C., & Vohs, K. D. (2001). Bad is
Anthony, S. (1973). Anxiety and rumour. stronger than good. Review of General Psy-
Journal of Social Psychology, 89, 91-98. chology, 5, 323-370.
Anthony, S. (1992). The influence of per- Baumeister, R. F., Campbell, J. D., Krueger,
sonal characteristics on rumor knowl- J. L, & Vohs, K. D. (2003). Does high self-
edge and transmission among the deaf. esteem cause better performance, inter-
American Annals of the Deaf, 137, 44-47. personal success, happiness, or healthier
Arndt, J. (1967). Word of mouth advertising: lifestyles? Psychological Science in the Public
A review of the literature. New York: Adver- Interest, 4(1), 1-44.
tising Research Foundation. Baumeister, R. F., & Leary, M. R. (1995).
Asch, S. E. (1955). Opinions and social pres- The need to belong: Desire for interper-
sure. Scientific American, 193, 31-35. sonal attachments as a fundamental hu-
Ashford, S., & Black, J. (1996). Proactivity man motivation. Psychological Bulletin,
during organizational entry: The role of 117,497-529.
desire for control. Journal of Applied Psy- Baumeister, R. P., Zhang, L., & Vohs, K. D.
chology, 81, 199-214. (2004). Gossip as cultural learning. Re-
Austin, M. J., & Brumfield, L. (1991). P&G's view of General Psychology, 8, 111-121.
run-in with the devil. Business and Society Seal, D. J., Ruscher, J. B., &• Schnake, S. B.
Review, 7<S(Summer), 16-19. (2001). No benefit of the doubt: Inter-
Back, K., Festinger, L., Hymovitch, B., Kel- group bias in understanding causal expla-
ley, H., Schachter, S., & Thibaut, J. nation. British Journal of Social Psychology,
(1950). The methodology of studying ru- 40, 531-543.
mor transmission. Human Relations, 3, Begg, I. M., Anas, A., & Farinacci, S. (1992).
307-312. Dissociation of processes in belief: Source
Bacon, F. T. (1979). Credibility of repeated recollection, statement familiarity, and
statements: Memory for trivia. Journal of the illusion of truth. Journal of Experimen-
Experimental Psychology: Human Learning tal Psychology: General, 121, 446-458.
and Memory, 5, 241-252. Belgion, M. (1939). The vogue of rumour.
Bandura, A. (2001). Social cognitive theory: Quarterly Review, 273, 1-18.
An agentic perspective. Annual Review of Bennett, G. (1985). What's modern about
Psychology, 52, 1-26. the modern legend? Fabula, 26, 219-229.
 References 263

Berger, C. R. (1987). Communicating under Fine, C. Heath, &• V. Campion-Vincent

uncertainty. In M. E. Roloff & G. R. (Eds.), Rumor mills: The social impact of ru-
Miller (Eds.), Interpersonal processes: New mor and legend (pp. 87-101). New York:
directions in communication research (pp. Aldine Press.
39-62). Newbury Park, CA: Sage. Bordia, P., DiFonzo, N., & Chang, A. (1999).
Berger, C. R., & Bradac, J. J. (1982). Lan- Rumor as group problem-solving: Devel-
guage and social knowledge: Uncertainty in opment patterns in informal computer-
interpersonal relations. London: Edward mediated groups. Small Group Research,
Arnold. 30(1), 8-28.
Bird, D. A. (1979). Rumor as folklore: An inter- Bordia, P., DiFonzo, N., Haines, R., & Chase-
pretation and inventory. Unpublished doc- ling, L. (2005). Rumor denials as per-
toral dissertation, Indiana University, suasive messages: Effects of personal
Bloomington, IN. relevance, source, and message charac-
Black beliefs on AIDS tallied. (1995, Novem- teristics. Journal of Applied Social Psychol-
ber 2). Rochester Democrat and Chronicle, ogy, 35, 1301-1331.
p. Bl. Bordia, P., DiFonzo, N., Irmer, B. E., Gal-
Blake, R. H., McFaul, T. R., & Porter, W. H. lois, C., & Bourne, M. (2005). Consumer
(1974, November). Authority and mass me- reactions to corporate rumor refutations. Un-
dia as variables in rumor transmission. Paper published manuscript.
presented at the annual meeting of the Bordia, P., DiFonzo, N., & Schulz, C. A.
Western Speech Communication Associ- (2000). Source characteristics in denying
ation, Newport Beach, CA. rumors of organizational closure: Hon-
Blake, R. R., & Mouton, J. S. (1983). The esty is the best policy. Journal of Applied
urge to merge: Tying the knot success- Social Psychology, 11, 2301-2309.
fully. Training and Development Journal, Bordia, P., DiFonzo, N., & Travers, V. (1998).
37, 41-46. Denying rumors of organizational
Blumenfeld, L. (1991, July 15). Procter & change: A higher source is not always bet-
Gamble's devil of a problem: Anti- ter. Communications Research Reports, 15,
Satanism watchdogs turn up the heat. 189-198.
Washington Post, pp. Bl, B6. Bordia, P., Hobman, E., Jones, E., Gallois, C.,
Bobo, L., & Kluegel, J. R. (1993). Opposition & Callan, V. J. (2004). Uncertainty dur-
to race-targeting: Self-interest, stratifica- ing organizational change: Types, conse-
tion ideology, or racial attitudes. American quences, and management strategies.
Sociological Review, 58, 443—464. Journal of Business and Psychology, 18,
Boehm, L. E. (1994). The validity effect—a 507-532.
search for mediating variables. Personality Bordia, P., Hunt, L., Paulsen, N., Tourish, D.,
and Social Psychology Bulletin, 20, 285-293. & DiFonzo, N. (2004). Communication
Bordia, P. (1996). Studying verbal interac- and uncertainty during organizational
tion on the Internet: The case of rumor change: Is it all about control? European
transmission research. Behavior Research Journal of Work and Organizational Psychol-
Methods, Instruments, & Computers, 28,149- ogy, 13, 345-365.
151. Bordia, P., Jones, E., Gallois, C., Callan, V. J.,
Bordia, P., & DiFonzo, N. (2002). When so- & DiFonzo, N. (in press). Management
cial psychology became less social: Prasad are aliens! Rumors and stress during or-
and the history of rumor research. Asian ganizational change. Group & Organiza-
Journal of Social Psychology, 5, 49-61. tion Management.
Bordia, P., & DiFonzo, N. (2004). Problem Bordia, P., & Rosnow, R. L. (1998). Rumor
solving in social interactions on the In- rest stops on the information superhigh-
ternet: Rumor as social cognition. Social way: A naturalistic study of transmission
Psychology Quarterly, 67, 33-49. patterns in a computer-mediated rumor
Bordia, P., & DiFonzo, N. (2005). Psycholog- chain. Human Communication Research,
ical motivations in rumor spread. In G. A. 25, 163-179.
264 R E F E R E N C E S

Brock, T. C. (1968). Implications of com- Cialdini, R. B., & Trost, M. R. (1998). Social
modity theory for value change. In influence: Social norms, conformity, and
A. Greenwald, T. C. Brock, & T. M. Os- compliance. In D. T. Gilbert, S. T. Fiske,
trom (Eds.), Psychological foundations of at- & G. Lindzey (Eds.), The handbook of social
titudes (pp. 243-276). New York: Aca- psychology (4th ed., Vol. II, pp. 151-192).
demic Press. Boston: McGraw-Hill.
Brunswik, E. (1952). The conceptual frame- Clampitt, P. G., DeKoch, R. J., &• Cashman,
work of psychology. International encyclo- T. (2000). A strategy for communicating
pedia of unified science. Chicago: University about uncertainty. Academy of Manage-
of Chicago Press. ment Executive, 14, 41-57.
Brunvand, J. H. (1981). The vanishing hitch- Cohen, G. L., Aronson, J., & Steele, C. M.
hiker. New York: Norton. (2000). When beliefs yield to evidence:
Brunvand, J. H. (1984). The choking Dober- Reducing biased evaluation by affirming
man. New York: Norton. the self. Personality and Social Psychology
Buckner, H. T. (1965). A theory of rumor Bulletin, 26, 1151-1164.
transmission. Public Opinion Quarterly, Cohen, J. (1988). Statistical power analysis for
29, 54-70. the behavioral sciences (2nd ed.). Hillsdale,
Burlew, L. D., Pederson, J. E., & Bradley, B. NJ: Erlbaum.
(1994). The reaction of managers to the Cornwell, D., & Hobbs, S. (1992). Rumour
pre-acquisition stage of a corporate mer- and legend: Irregular interactions be-
ger: A qualitative study. Journal of Career tween social psychology and folklorists.
Development, 21, 11-22. Canadian Psychology, 33, 609-613.
Burt, R., & Knez, M. (1996). Third-party Creed, W. E. D., & Miles, R. E. (1996). Trust
gossip and trust. In R. M. Kramer & T. R. in organizations: A conceptual frame-
Tyler (Eds.), Trust in organizations: Fron- work linking organizational forms, man-
tiers of theory and research (pp. 68-89). agerial philosophies, and the opportunity
Thousand Oaks, CA: Sage. costs of controls. In R. E. Kramer & T. R.
Calvo, M. G., & Castillo, M. D. (1997). Tyler (Eds.), Trust in organizations: Fron-
Mood-congruent bias in interpretation of tiers of theory and research (pp. 16-38).
ambiguity: Strategic process and tempo- Thousand Oaks, CA: Sage.
rary activation. The Quarterly Journal of Crick, N. R., Nelson, D. A., Morales, J. R.,
Experimental Psychology, 50, 163-182. Cullerton-Sen, C., Casas, J. F., & Hick-
Cantera, K. (2002, January 3). Vigilant man, S. (2001). Relational victimization
Utahns. The Salt Lake Tribune, p. Al. in childhood and adolescence: I hurt you
Caplow, T. (1947). Rumors in war. Social through the grapevine. In J. Juvonen &
Forces, 25, 298-302. S. Graham (Eds.), School-based peer harass-
Cato, F. W. (1982). Procter & Gamble and ment: The plight of the vulnerable and victim-
the devil. Public Relations Quarterly, 27, ized (pp. 196-214). New York: Guilford
16-21. Press.
Chapman, L. J., & Chapman, J. P. (1969). Davis, K. (1972). Human behavior at work.
Illusory correlation as an obstacle to the San Francisco: McGraw-Hill.
use of valid psychodiagnostic signs. Jour- Davis, K. (1975, June). Cut those rumors
nal of Abnormal Psychology, 74, 271-280. down to size. Supervisory Management,
Choe, S., Hanley, C. J., & Mendoza, M. 2-6.
(1999, October 17). GIs admit murdering Davis, W. L., & O'Connor, J. R. (1977). Se-
civilians in S. Korea. Rochester Democrat rial transmission of information: A study
and Chronicle, pp. 1A, 3A. of the grapevine. Journal of Applied Com-
Chulov, M., Warne-Smith, D., & Colman, E. munication Research, 5, 61-72.
(2004, February 17). Rumour the spark Day, R. S. (1986). Overconfidence as a result
that fired racial tinderbox. The Australian, of incomplete and wrong knowledge. In
pp. 1, 6. R. W. Scholz (Ed.), Current issues in West
 References 265

German decision research (pp. 13-30). DiFonzo, N., & Bordia, P. (in press). Rumors
Frankfurt am Main, Germany: Lang. influence: Toward a dynamic social im-
DeClerque, J., Tsui, A. O., Abul-Ata, M. P., pact theory of rumor. In A. R. Pratkanis
& Barcelona, D. (1986). Rumor, misin- (Ed.), The science of social influence. Phila-
formation, and oral contraceptive use in delphia: Psychology Press.
Egypt. Social Science and Medicine, 23, DiFonzo, N., Bordia, P., & Rosnow, R. L.
83-92. (1994). Reining in rumors. Organizational
Deener, B. (2001, September 20). Rumors Dynamics, 23(1), 47-62.
rattle market. The Dallas Morning News. DiFonzo, N., Bordia, P., & Winterkorn, R.
De Fleur, M. L. (1962). Mass communica- (2003, January). Distrust is a key ingredient
tion and the study of rumor. Sociological in negative rumor transmission. Paper pre-
Inquiry, 32, 51-70. sented at the 4th Annual Meeting of the
DePaulo, B. M., & Kashy, D. A. (1998). Ev- Society for Personality and Social Psy-
eryday lies in close and casual relation- chologists, Los Angeles, CA.
ships. Journal of Personality and Social Psy- DiFonzo, N., Hantula, D. A., & Bordia, P.
chology 74, 63-79. (1998). Microworlds for experimental re-
Dewey, J. (1925). Experience and nature. Chi- search: Having your (control and collec-
cago: Open Court. tion) cake, and realism too. Behavior Re-
Dietz-Uhler, B. (1999). Defensive reactions search Methods, Instruments, & Computers,
to group-relevant information. Group 30, 278-286.
Processes and Intergroup Relations, 2, 17-29. Dirks, K. T., & Ferrin, D. L. (2001). The role
DiFonzo, N. (1994). Piggy-backed syllogisms of trust in organizational settings. Organi-
for investor behavior: Probabilistic mental zational Science, 12, 450-467.
modeling in rumor-based stock market trad- Dirks, K. T., & Ferrin, D. L. (2002). Trust
ing. Unpublished doctoral thesis, Temple in leadership: Meta-analytic findings and
University, Philadelphia. implications for research and practice.
DiFonzo, N., & Bordia, P. (1997). Rumor Journal of Applied Psychology, 87, 611-628.
and prediction: Making sense (but losing Donnelly, F. K. (1983, Spring). People's Al-
dollars) in the stock market. Organiza- manac predictions: Retrospective check
tional Behavior and Human Decision Pro- of accuracy. Skeptical Inquirer, 48-52.
cesses, 71, 329-353. Dunbar, R.I. M. (1996). Grooming, gossip,
DiFonzo, N., & Bordia, P. (1998). A tale of and the evolution of language. Cambridge,
two corporations: Managing uncertainty MA: Harvard University Press.
during organizational change. Human Re- Dunbar, R. I. M. (2004). Gossip in evolu-
source Management, 37, 295-303. tionary perspective. Review of General Psy-
DiFonzo, N., & Bordia, P. (2000). How top chology, 8, 100-110.
PR professionals handle hearsay: Corpo- Dwyer, J. C., & Drew, C. (2005, September
rate rumors, their effects, and strategies 29). Fear exceeded crime's reality in New
to manage them. Public Relations Review, Orleans. The New York Times, pp. Al, A22.
26, 173-190. Eder, D., & Enke, J. L. (1991). The structure
DiFonzo, N., & Bordia, P. (2002a). Corporate of gossip: Opportunities and constraints
rumor activity, belief, and accuracy. Pub- on collective expression among adoles-
lic Relations Review, 150, 1-19. cents. American Sociological Review, 56,
DiFonzo, N., & Bordia, P. (2002b). Rumor 494-508.
and stable-cause attribution in prediction Edwards, O. (1989, April). Leak soup. GQ
and behavior. Organizational Behavior and Magazine, p. 228.
Human Decision Processes, 88, 329-353. Einhorn, H. J., & Hogarth, R. M. (1986).
DiFonzo, N., & Bordia, P. (2006). Rumor in Judging probable cause. Psychological Bul-
organizational contexts. In D. A. Hantula letin, 99, 3-19.
(Ed.), Advances in psychology: A tribute to Ellis, R. J., & Zanna, M. P. (1990). Arousal
Ralph L. Rosnow (pp. 249-274). Mahwah, and causal attribution. Canadian Journal
NJ: Erlbaum. of Behavioural Science, 22, 1-12.
266 R E F E R E N C E S

Emory, D. (n.d.) Does Osama bin Laden Firth, R. (1956). Rumor in a primitive soci-
own Snapple? Retrieved December 13, ety. Journal of Abnormal and Social Psychol-
2005, from http://urbanlegends.about, ogy, 53, 122-132.
com/library/blsnapple.htm Fisher, D. R. (1998). Rumoring theory and
Esposito, J. L. (1986/1987). Subjective fac- the Internet: A framework for analyzing
tors and rumor transmission: A field in- the grass roots. Social Science Computer Re-
vestigation of the influence of anxiety, view, 16, 158-168.
importance, and belief on rumormonger- Fischle, M. (2000). Mass response to the
ing (Doctoral dissertation, Temple Uni- Lewinsky scandal: Motivated reasoning
versity, 1986). Dissertation Abstracts Inter- or Bayesian updating? Political Psychology,
national, 48, 596B. 21, 135-159.
Esposito, J. L., & Rosnow, R. L. (1983, Fiske, S. T. (2003). Five core social motives,
April). Corporate rumors: How they start plus or minus five. In S. J. Spencer & S.
and how to stop them. Management Re- Fein (Eds.), Ontario Symposium on Person-
view, 44-49. ality and Social Psychology: Vol. 9. Motivated
Fama, E. F., Fisher, L., Jensen, M. C, & social perception (pp. 233-246). Mahwah,
Roll, R. (1969). The adjustment of stock NJ: Erlbaum.
prices to new information. International Fiske, S. T. (2004). Social beings: A core motives
Economic Review, 10, 1-21. approach to social psychology. Hoboken,
Fearn-Banks, K. (2002). Crisis communica- NJ: Wiley.
tions: A casebook approach. Mahwah, NJ: Fiske, S. T., Lin, M., & Neuberg, S. L. (1999).
Erlbaum. The continuum model: Ten years later.
Fein, S., & Spencer, S. J. (1997). Prejudice In S. Chaiken & Y. Trope (Eds.), Dual
as self-image maintenance: Affirming the process theories in social psychology (pp. 231-
self through derogating others. Journal of 254). New York: Guilford Press.
Personality and Social Psychology, 73, 31-44. Fiske, S. T., &Taylor, S. E. (1991). Socialcog-
Festinger, L. (1957). A theory of cognitive disso- nition (2nd ed.). New York: Random
nance. Evanston, IL: Row, Peterson. House.
Festinger, L., Cartwright, D., Barber, K., Flanagan, J. C. (1954). The critical inci-
Fleischl, J., Gottsdanker, J., Keysen, A., dent technique. Psychological Bulletin,
et al. (1948). A study of rumor: Its origin 51, 327-358.
and spread. Human Relations, I, 464-485. Foster, E. K. (2004). Research on gossip:
Fine, G. A. (1985). Rumors and gossiping. Taxonomy, methods, and future direc-
In T. Van Dijk (Ed.), Handbook of discourse tions. Review of General Psychology, 8,
analysis (Vol. 3, pp. 223-237). London: 78-99.
Academic Press. Foster, E. K., & Rosnow, R. L. (2006). Gossip
Fine, G. A. (1992). Manufacturing tales: Sex and network relationships: The pro-
and money in contemporary legends. Knox- cesses of constructing and managing dif-
ville: University of Tennessee Press. ficult interaction. In D. C. Kirkpatrick,
Fine, G. A. (2005, April 1-3). Does rumor lie: S. W. Duck, & M. K. Foley (Eds.), Relating
Narrators and the framing of unsecured infor- difficulty (pp. 161-201). Mahwah, NJ:
mation. Paper presented to the Sante Fe Erlbaum.
Institute's Conference on Deception: Freedman, A.M. (1991, May 10). Rumor
Methods, Motives, Contexts, and Conse- turns fantasy into bad dream. The Wall
quences, Santa Fe, NM. Street Journal, pp. Bl, B5.
Fine, G. A., Heath, C., & Campion-Vincent, Frith, B. (2001, August 29). AMP's silence
V. (Eds.) (2005). Rumor mills: The social on NAB merger rumours spoke volumes.
impact of rumor and legend. Chicago: The Australian, p. Ml.
Aldine. Fromkin, H. L. (1972). Feelings of interper-
Fine, G. A., & Turner, P. A. (2001). Whispers sonal undistinctiveness: An unpleasant
on the color line: Rumor and race in America. affective state. Journal of Experimental Re-
Berkeley: University of California Press. search in Personality, 6, 178-185.
 References 267

Gigerenzer, G., Hoffrage, U., & Kleinbolt- work for the social sciences. Review of Gen-
ing, H. (1991). Probabilistic mental mod- eral Psychology, 7, 251-298.
els: A Brunswikian theory of confidence. Harcourt, J., Richerson, V., & Wattier, M. J.
Psychological Review, 98, 506-528. (1991). A national study of middle man-
Gillin, B. (2005, September 28). Tales of agers' assessment of organization com-
mass murder, rape proving false. Rochester munication quality. Journal of Business
Democrat and Chronicle, p. 7A. Communication, 28, 348-365.
Gilovich, T., Vallone, R., & Tversky, A. Hardin, C. D., & Higgins, E. T. (1996).
(1985). The hot hand in basketball: On Shared reality: How social verification
the misperception of random sequences. makes the subjective objective. In R. M.
Journal of Personality and Social Psychology, Sorrentino & E. T. Higgins (Vol. Eds.),
17, 295-314. Handbook of motivation and cognition: Vol. 3.
Gluckman, M. (1963). Gossip and scandal. The interpersonal context (pp. 28-84). New
Current Anthropology, 4, 307-316. York: Guilford Press.
Goggins, S. M. (1979). The wormburger scare: Hari, J. (2002, December 31). Well, they
A case study of the McDonald's corporation's would say that, wouldn't they? Australian
public relations campaign to stop a damag- Financial Review, p. 42.
ing rumor. Unpublished master's thesis, Harris, B., & Harvey, J. H. (1981). Attribu-
Georgia State University, Athens. tion theory: From phenomenal causality
Goleman, D. (1991, June 4). Anatomy of to the intuitive social scientist and be-
a rumor: It flies on fear. The New York yond. In C. Antaki (Ed.), The psychology
Times, p. C5. of ordinary explanations of social behaviour:
Goodwin, S. A., Operario, D., & Fiske S. T. Vol. 23. European Monographs in Social Psy-
(1998). Situational power and interper- chology (pp. 57-95). London: Academic
sonal dominance facilitate bias and in- Press.
equality. Journal of Social Issues, 54, Harris, S. G. (1994). Organizational culture
677-698. and individual sensemaking: A schema-
Goswamy, M., & Kumar, A. (1990). Sto- based perspective. Organizational Science,
chastic model for spread of rumour sup- 5, 309-321.
ported by a leader resulting in collective Hasher, L., Goldstein, D., & Toppino T.
violence and planning of control mea- (1977). Frequency and the conference of
sures. Mathematical Social Sciences, 19, referential validity. Journal of Verbal
23-36. Learning and Verbal Behavior, 16,107-112.
Green, D. F. (1984). Rumor control strategies Heath, C., Bell, C., & Sternberg, E. (2001).
for corporations. Unpublished master's Emotional selection in memes: The case
thesis, University of Texas at Austin. of urban legends. Journal of Personality
Grice, H. P. (1975). Logic and conversation. and Social Psychology, 81, 1028-1041.
The William James Lectures. In P. Cole & Heider, F. (1958). The psychology of interper-
J. L. Morgan (Eds.), Syntax and semantics: sonal relations. New York: Wiley.
Vol. 3. Speech acts (pp. 41-58). New York: Hellweg, S. A. (1987). Organizational grape-
Academic Press. vines. In B. Dervin & M. J. Voigt, Progress
Gross, A. E. (1990, October 22). Crisis man- in communication sciences (Vol. 8, pp. 213-
agement: How Popeyes and Reebok con- 230). Norwood, NJ: Ablex Publishing.
fronted product rumors. Adweek 's Market- Hershey, R. (1956). Heed rumors for their
ing Week, p. 27. meaning. Personnel Journal, 34, 299-301.
Gudykunst, W. B. (1995). Anxiety/uncer- Hicks, R. D. (1990). Police pursuit of Sa-
tainty management (AUM) theory: Cur- tanic crime: Part 2: The Satanic conspir-
rent status. In R. L. Wiseman (Ed.), Inter- acy and urban legends. Skeptical Inquirer,
cultural communication theory (pp. 8-57). 14, 378-389.
Thousand Oaks, CA: Sage. Higgins, E. T. (1981). The "communication
Guerin, B. (2003). Language use as a social game": Implications of social cognition.
strategy: A review and an analytic frame- In E. T. Higgins, C. P. Herman, & M. P.
268 R E F E R E N C E S

Zanna (Eds.), Ontario Symposium on Per- JDBGMGR.EXE. (2002, September). Re-

sonality and Social Psychology: Vol. 1. Social trieved June 7, 2004, from http://
cognition (pp. 343-392). Hillsdale, NJ: www.snopes.com/computer/virus/
Erlbaum. jdbgmgr.htm
Higham, T. M. (1951). The experimental Jennings, D. L., Amabile, T. M., & Ross, L.
study of the transmission of rumour. Brit- (1982). Information covariation assess-
ish Journal of Psychology, 42, 42-55. ment: Data-based versus theory-based
Hilton, D. J., & Slugoski, B. R. (1986). judgments. In D. Kahneman, P. Slovic,
Knowledge-based causal attribution: The & A. Tversky (Eds.), Judgment under uncer-
abnormal conditions focus model. Psycho- tainty: Heuristics and biases (pp. 211-230).
logical Review, 93, 75-88. New York: Cambridge University Press.
Hirschhorn, L. (1983). Cutting back: Retrench- Jones, E. E., & Davis, K. E. (1965). From acts
ment and redevelopment of human and com- to dispositions: The attribution process in
munity services. San Francisco: Jossey- person perception. In L. Berkowitz (Ed.),
Bass. Advances in experimental social psychology
Hogg, M., & Abrams, D. (1988). Social iden- (Vol. 2, pp. 220-226), New York: Aca-
tifications: A social psychology of intergroup demic Press.
relations and group processes. London: Jung, C. G. (1916). Bin beitrag zur psycho-
Routledge. Igie des gertichtes [A contribution in the
Holtgraves, T., & Grayer, A. R. (1994). I am psychology of rumor]. In C. E. Long
not a crook: Effects of denials on percep- (Trans.), Collected papers on analytical psy-
tions of a defendant's guilt, personality, chology (pp. 176-190). New York: Wiley.
and motives. Journal of Applied Social Psy- (Original work published 1910)
chology, 24, 2132-2150. Jung, C. G. (1959). A visionary rumour.
Horn, H., & Haidt, J. (2002, January). Psst, Journal of Analytical Psychology, 4, 5-19.
Did you hear? Exploring the gossip phenome- Jungermann, H., & Thiiring, M. (1993).
non. Poster presented at the Annual Causal knowledge and the expression of
Meeting of the Society of Personality and uncertainty. In G. Strube & K. F. Wender
Social Psychologists, Savannah, GA. (Eds.), The cognitive psychology of knowl-
Horowitz, D. L. (2001). The deadly ethnic riot. edge (pp. 53-73). Amsterdam: Elsevier
Berkeley: University of California Press. Science.
Houmanfar, R., & Johnson, R. (2003). Orga- Kahneman, D., & Tversky, A. (1973). On
nizational implications of gossip and ru- the psychology of prediction. Psychological
mor. Journal of Organizational Behavior Review, 80, 237-351.
Management, 23, 117-138. Kahneman, D., & Tversky, A. (1979). Pros-
Hovland, C, & Weiss, W. (1951). The influ- pect theory: An analysis of decision under
ence of source credibility on communica- risk. Econometrica, 47, 263-291.
tion effectiveness. Public Opinion Quar- Kahneman, D., & Tversky, A. (1982). Sub-
terly, 75, 635-650. jective probability: A judgment of repre-
Hunsaker, P. L., & Coombs, M. W. (1988). sentativeness. In D. Kahneman, P. Slovic,
Mergers and acquisitions: Managing the & A. Tversky (Eds.), Judgment under uncer-
emotional issues. Personnel Journal, 67, tainty: Heuristics and biases (pp. 32-47).
56-78. New York: Cambridge University Press.
Iyer, E. S., & Debevec, K. (1991). Origin of Kakar, S. (2005). Rumors and religious riots.
rumor and tone of message in rumor In G. A. Fine, V. Campion-Vincent, &
quelling strategies. Psychology and Market- C. Heath (Eds.), Rumor mills: The social im-
ing, 8, 161-175. pact of rumor and legend (pp. 53-59). New
Jaeger, M. E., Anthony, S., & Rosnow, R. L. York: Aldine.
(1980). Who hears what from whom and Kamins, M. A., Folkes, V. S., & Perner, L.
with what effect: A study of rumor. Per- (1997). Consumer responses to rumors:
sonality and Social Psychology Bulletin, 6, Good news, bad news. Journal of Con-
473-478. sumer Psychology, 6, 165-187.
 References 269

Kapferer, J.-N. (1989). A mass poisoning ru- Knobloch, L. K., & Carpenter-Theune, K. E.
mor in Europe. Public Opinion Quarterly, (2004). Topic avoidance in developing
53, 467-481. romantic relationships: Associations with
Kapferer, J.-N. (1990). Rumor in the stock intimacy and relational uncertainty. Com-
exchange. Communications, 52, 61-84. munication Research, 31, 173-205.
Kapferer, J.-N. (1990). Rumors: Uses, interpre- Knopf, T. A. (1975). Rumor, race and
tations, and images (B. Fink, Trans.). New riots. New Brunswick, NJ: Transaction
Brunswick, NJ: Transaction Publishers. Publishers.
(Original work published 1987) Koenig, F. W. (1985). Rumor in the market-
Kelley, H. H. (1973). The processes of causal place: The social psychology of commercial
attribution. American Psychologist, 28,107- hearsay. Dover, MA: Auburn House.
128. Roller, M. (1992). Rumor rebuttal in the
Kelley, S. R. (2004). Rumors in Iraq: A guide marketplace. Journal of Economic Psychol-
to winning hearts and minds. Unpub- ogy, 13, 167-186.
lished master's thesis, Naval Postgraduate Roller, M. (1993). Rebutting accusations:
School, Monterey, CA. Retrieved No- When does it work, when does it fail.
vember 16, 2004, from http://tb.eses. European Journal of Social Psychology, 23,
nps.navy.mil/04Sep_Kelley.pdf 373-389.
Kenrick, D. T., Maner, J. K., Butner, J., Komarnicki, M., & Walker, C. J. (1980,
Li, N. P., Becker, D. V., & Schaller, M. March). Reliable and valid hearsay: Conver-
(2002). Dynamical evolutionary psychol- gent and divergent rumor transmission. Pa-
ogy: Mapping the domains of the new per presented at the Eastern Psychologi-
interactionist paradigm. Personality and cal Association meeting, Hartford, CT.
Social Psychology Review, 6, 347-356. Kramer, R. M. (1999). Trust and distrust
Kerner, O., Lindsay, J. V., Harris, F. R., Abel, in organizations: Emerging perspectives,
I. W., Brooke, E. W., Thornton, C. B., et enduring questions. Annual Review of Psy-
al. (1968). Report of the National Advisory chology, 50, 569-598.
Commission on Civil Disorders (Report No. Krull, D. S., & Anderson, C. A. (1997). The
1968 O - 291-729). Washington, DC: U.S. process of explanation. Current Directions
Government Printing Office. in Psychological Science, 6, 1-5.
Kimmel, A. J. (2004a). Rumors and rumor Ruhn, T. S. (1996). The structure of scientific
control: A manager's guide to understanding revolutions (3rd ed.). Chicago: University
and combating rumors. Mahwah, NJ: of Chicago Press.
Erlbaum. Runda, Z. (1987). Motivated inference: Self-
Kimmel, A. J. (2004b). Rumors and the fi- serving generation and evaluation of
nancial marketplace. Journal of Behavioral causal theories. Journal of Personality and
Finance, 5, 134-141. Social Psychology, 53, 636-647.
Kimmel, A. J., & Reefer, R. (1991). Psycho- Kunda, Z. (1990). The case for motivated
logical correlates of the transmission and reasoning. Psychological Bulletin, 108, 480-
acceptance of rumors about AIDS. Jour- 498.
nal of Applied Social Psychology, 21, 1608- Kunda, Z. (1999). Social cognition: Making
1628. sense of people. Cambridge, MA: MIT Press.
Kirkpatrick, C. (1932). A tentative study in Kurland, N. B., & Felled, L. H. (2000). Pass-
experimental social psychology. American ing the word: Toward a model of gossip
Journal of Sociology, 38, 194-206. and power in the workplace. Academy of
Knapp, R. H. (1944). A psychology of ru- Management Review, 25, 428-438.
mor. Public Opinion Quarterly, 8, 22-27. Latane, B., & Bourgeois, M. J. (1996). Ex-
Knapp, S. D. (Ed.). (1993). The contemporary perimental evidence for dynamic social
thesaurus of social science terms and syn- impact: The formations of subcultures in
onyms: A guide for natural language com- electronic groups. Journal of Communica-
puter searching. Phoenix, AZ: Oryx. tion, 46, 35-47.
270 R E F E R E N C E S

Lazar, R. J. (1973). Stock market price nation, and world in a historical per-
movements as collective behavior. Inter- spective. Family Community Health, 12(2),
national Journal of Contemporary Sociology, 52-59.
10, 133-147. Lynn, M. (1991). Scarcity effects on desir-
Leary, M. R. (1995). Self-presentation:Impres- ability: A quantitative review of the com-
sion management and interpersonal behav- modity theory literature. Psychology and
ior. Boulder, CO: Westview Press. Marketing, 8, 43-57.
Leavitt, H. J., & Mueller, R. A. (1951). Some Lyons, A., & Kashima, Y. (2001). The repro-
effects of feedback on communication. duction of culture: Communication pro-
Human Relations, 4, 401-410. cesses tend to maintain cultural stereo-
Lerbinger, O. (1997). The crisis manager: Fac- types. Social Cognition, 19, 372-394.
ing risk and responsibility. Mahwah, NJ: MacLeod, C., & Cohen, I. L. (1993). Anxiety
Erlbaum. and the interpretation of ambiguity: A
Lev, M. (1991, February 6). Carter stock text comprehension study. Journal of Ab-
drops again on rumors. The New York normal Psychology, 102, 238-247.
Times, p. D4. Maines, D. R. (1999). Information pools and
Levin, J., & Arluke, A. (1987). Gossip: The racialized narrative structures. The Socio-
inside scoop. New York: Plenum Press. logical Quarterly, 40, 317-326.
Litman, J. A., & Pezzo, M. V. (2005). Indi- Malkiel, B. G. (1985). A random walk down
vidual differences in attitudes towards Wall Street (4th ed.). New York: Norton.
gossip. Personality and Individual Differ- Marks, A. (2001, October 23). From survival
ences, 38, 963-980. tales to attack predictions, rumors fly. The
Litwin, M. L. (1979, January). Key commu- Christian Science Monitor, p. 2.
nicators—They lock out rumors. National Marting, B. (1969). A study of grapevine com-
Association of Secondary School Principals munication patterns in a manufacturing or-
Bulletin, 17-22. ganization. Unpublished doctoral disserta-
London, I. D., &• London, M. B. (1975). Ru- tion, Arizona State University, Tempe.
mor as a footnote to Chinese national Marty, E. M. (1982). Satanism: No soap.
character. Psychological Reports, 37, 343- Across the Board, 19(11), 8-14.
349. Matthews, L., & Sanders, W. (1984). Effects
Lord, C. G., Lepper, M. R., & Ross, L. (1979). of causal and noncausal sequences of in-
Biased assimilation and attitude polariza- formation on subjective prediction. Psy-
tion: The effects of prior theory on subse- chological Reports, 54, 211-215.
quently considered information. Journal Mausner, J., & Gezon, H. (1967). Report on
of Personality and Social Psychology, 46, a phantom epidemic of gonorrhea. Ameri-
1254-1266. can Journal of Epidemiology, 85, 320-331.
Lorenzi-Cioldi, F., & Clemence, A. (2001). McAdam, J. R. (1962). The effect of verbal in-
Group processes and the construction of teraction on the serial reproduction of rumor.
social representations. In M. A. Hogg & Unpublished doctoral dissertation, Indi-
S. Tindale (Eds.), Group processes (pp. 311- ana University, Bloomington, IN.
333). Maiden, MA: Blackwell. McEvily, B., Perrone, V., & Zaheer, A.
Lott, B. E., & Lott, A. J. (1985). Learning (2003). Introduction to the special issue
theory in contemporary social psychol- on trust in an organizational context. Or-
ogy. In G. Lindzey & E. Aronson (Eds.), ganization Science, 14, 1-4.
The handbook of social psychology (3rd ed., McGee, G. W., & Ford, R. C. (1987). Two
Vol. 1, pp. 109-135). New York: Ran- (or more?) dimensions of organizational
dom House. commitment: Reexamination of the af-
Lowenberg, R. D. (1943). Rumors of mass fective and continuance commitment
poisoning in times of crisis. Journal of scales. Journal of Applied Psychology, 72,
Criminal Pathology, 5, 131-142. 638-642.
Lynch, R. D. (1989). Psychological impact of McMillan, S. (1991). Squelching the rumor
AIDS on individual, family, community, mill. Personnel Journal, 70(10), 95-99.
 References 271

Merton, R. K. (1968, January 5). The Mat- users' private files. The Wall Street Journal,
thew effect in science. Science, 159, 56-63. p. Bl.
Michelson, G., & Mouly, S. (2000). Rumour Mirvis, P. H. (1985). Negotiations after the
and gossip in organizations: A conceptual sale: The roots and ramifications of con-
study. Management Decision, 38, 339- flict in an acquisition. Journal of Occupa-
346. tional Behaviour, 6, 65-84.
Michelson, G., & Mouly, V. S. (2004). Do Modic, S. J. (1989, May 15). Grapevine
loose lips sink ships? The meaning, ante- rated most believable. Industry Week,
cedents and consequences of rumour and 235(10), 11.
gossip in organizations. Corporate Com- Monday, Monday. (2002, October 1). Re-
munications: An International Journal, 9, trieved December 16, 2005, from http://
189-201. www.snopes.com/rumors/fema.htm
Mihanovic, M., Jukic, V., & Milas, M. Monge, P. R., & Contractor, N. S. (2000).
(1994). Rumours in psychological war- Emergence of communication networks.
fare. Socijalna Psihijatrija, 22, 75-82. In F. M. Jablin & L. L. Putnam (Eds.), The
Mikkelson, B. (1999, November 19). The un- new handbook of organizational communica-
kindest cut. Retrieved November 2, 2004, tion: Advances in theory, research, and meth-
from http://www.snopes.com/horrors/ ods (pp. 440-502). Thousand Oaks, CA:
robbery/slasher.asp Sage.
Mikkelson, B. (2001, November 24). Citi- Morin, E. (1971). Rumour in Orleans. New
bank rumor. Retrieved December 2, 2005, York: Pantheon Books.
from http: / / www.snopes.com / rumors / Mullen, B. (1989). Advanced BASK meta-
citibank.htm analysis. Hillsdale, NJ: Erlbaum.
Mullen, P. B. (1972). Modern legend and
Mikkelson, B. (2002, April 28). You've got to
rumor theory. Journal of the Folklore Insti-
be kidneying. Retrieved June 7, 2004,
tute, 9, 95-109.
from http://www.snopes.com/horrors/
Newman, M. E. J. (2003). The structure and
robbery/kidney.htm
function of complex networks. SIAM Re-
Mikkelson, B. (2003, January 12). Trade-
view, 45, 167-256.
mark of the devil. Retrieved November
Newstrom, J. W., Monczka, R. E., & Reif,
22, 2004, from http://www.snopes.com/
W. E. (1974). Perceptions of the grape-
business/alliance/procter.asp
vine: Its value and influence. Journal of
Mikkelson, B. (2004a, July 8). Deja 'TOO. Re-
Business Communication, 11, 12-20.
trieved November 9, 2004, from http://
Nisbett, R., & Ross, L. (1980). Human
www.snopes.com/critters/malice/ inference: Strategies and shortcomings of
kangaroo.htm social judgment. Englewood Cliffs, NJ:
Mikkelson, B. (2004b, September 23). Lights Prentice-Hall.
out! Retrieved November 2, 2004, from Nkpa, N. K. U. (1977). Rumors of mass poi-
http://www.snopes.com/horrors/ soning in Biafra. Public Opinion Quarterly,
madmen/lightout.asp 41, 332-346.
Mikkelson, B. (2005, September 27). Killer Noon, M., & Delbridge, R. (1993). News
dolphins. Retrieved November 22, 2005, from behind my hand: Gossip in organi-
from http://www.snopes.com/Katrina/ zations. Organization Studies, 14, 23-36.
rumor/dolphins.asp Ojha, A. B. (1973). Rumour research: An
Mikkelson, B., & Mikkelson, D. P. (2004, overview. Journal of the Indian Academy of
August 23). Verses, foiled again. Retrieved Applied Psychology, 10, 56-64.
November 25, 2005, from http://www. Pendleton, S. C. (1998). Rumor research re-
snopes.com/politics/bush/bibleverse.asp visited and expanded. Language & Com-
Miller, D. L. (1985). Introduction to collective munication, 18, 69-86.
behavior. Belmont, CA: Wadsworth. Peters, H. P., Albrecht, G., Hennen, L., &
Miller, M. W. (1991, May 1). Computers: Stegelmann, H. U. (1990). "Chernobyl"
'Prodigy' headquarters offered peeks into and the nuclear power issue in West
272 R E F E R E N C E S

German public opinion. Journal of Envi- abuse of persuasion (Rev. ed.). New York:
ronmental Psychology, 10, 121-134. Freeman.
Peterson, W. A., & Gist, N. P. (19 51). Rumor Pratkanis, A. R., & Greenwald, A. G. (1989).
and public opinion. American Journal of A socio-cognitive model of attitude struc-
Sociology, 57, 159-167. ture and function. In L. Berkowitz (Ed.),
Pettigrew, T. F. (1979). The ultimate attribu- Advances in experimental social psychology
tion error: Extending Allport's cognitive (Vol. 22, pp. 245-285). New York: Aca-
analysis of prejudice. Personality and Social demic Press.
Psychology Bulletin, 5, 461-476. Priester, J. R., & Petty, R. E. (1995). Source
Petty, R. E., & Cacioppo, J. T. (1981). Atti- attributions and persuasion: Perceived
tudes and persuasion: Classic and contempo- honesty as a determinant of message
rary approaches. Dubuque, IA: Brown. scrutiny. Personality and Social Psychology
Petty, R. E., & Wegener, D.T. (1998). Atti- Bulletin, 21, 637-654.
tude change: Multiple roles for persua- Quist, R. M., & Resendez, M. G. (2002). So-
sion variables. In D. T. Gilbert, S. T. Fiske, cial dominance threat: Examining social
& G. Lindzey (Eds.), The handbook of social dominance theory's explanation of prej-
psychology (4th ed., Vol. I, pp. 323-390). udice as legitimizing myths. Basic and Ap-
Boston: McGraw-Hill. plied Social Psychology, 24, 287-293.
Pezzo, M. V., & Beckstead, J. (2006). A Rajecki, D. W. (1990). Attitudes (2nd ed.).
multi-level analysis of rumor transmis- Sunderland, MA: Sinauer Associates.
sion: Effects of anxiety and belief in two Rawlins, W. (1983). Openness as problem-
field experiments. Basic and Applied Social atic in ongoing friendships: Two conver-
Psychology, 28, 91-100. sational dilemmas. Communication Mono-
Pinsdorf, M. K. (1987). Communicating when graphs, 50, 1-13.
your company is under siege: Surviving public Robinson, S. L. (1996). Trust and breach of
crisis. Lexington, MA: Lexington Books. the psychological contract. Administrative
Ponting, J. R. (1973). Rumor control cen- Science Quarterly, 41, 574-599.
ters: Their emergence and operations. Rose, A. M. (19 51). Rumor in the stock mar-
The American Behavioral Scientist, 16, ket. Public Opinion Quarterly, 15, 461-486.
391-401. Rosenberg, L. A. (1967). On talking to a
Popper, K. R. (1962). Conjectures and refuta- newspaper reporter: A study of selective
tions: The growth of scientific knowledge. perception, distortion through rumor,
New York: Basic Books. professional gullibility, or how to ride the
Porter, E. G. (1984). Birth control discon- Zeitgeist for all it is worth. American Psy-
tinuance as a diffusion process. Studies in chologist, 22, 239-240.
Family Planning, 15, 20-29. Rosenthal, M. (1971). Where rumor raged.
Pound, J., & Zeckhauser, R. (1990). Clearly Trans-Action, 8(4), 34-43.
heard on the street: The effect of takeover Rosenthal, R. (1979). The "file drawer prob-
rumors on stock prices. Journal of Busi- lem" and tolerance for null results. Psy-
ness, 63, 291-308. chological Bulletin, 86, 638-641.
Prasad, J. (1935). The psychology of ru- Rosenthal, R. (1991). Meta-analyticprocedures
mour: A study relating to the great Indian for social research (Rev. ed.). Newbury
earthquake of 1934. British Journal of Psy- Park, CA: Sage.
chology, 26, 1-15. Rosenthal, R., & Rosnow, R. L. (1991). Essen-
Prasad, J. (1950). A comparative study of tials of behavioral research: Methods and data
rumours and reports in earthquakes. Brit- analysis (2nd ed.). New York: McGraw-
ish Journal of Psychology, 41, 129-144. Hill.
Pratkanis, A. R., & Aronson, E. (1991). Age Rosnow, R. L. (1974). On rumor. Journal of
of propaganda: The everyday use and abuse Communication, 24(3), 26-38.
of persuasion. New York: Freeman. Rosnow, R. L. (1980). Psychology of rumor
Pratkanis, A. R., & Aronson, E. (2001). Age reconsidered. Psychological Bulletin, 87,
of propaganda: The everyday use and 578-591.
 References 273

Rosnow, R. L. (1988). Rumor as communi- Rothbaum, F., Weisz, J. R., & Snyder S. S.
cation: A contextualist approach. Journal (1982). Changing the world and chang-
of Communication, 38, 12-28. ing the self: A two-process model of per-
Rosnow, R. L. (1991). Inside rumor: A per- ceived control. Journal of Personality and
sonal journey. American Psychologist, 46, Social Psychology, 42, 5-37.
484-496. Rousseau, D. M., Sitkin, S. B., Burt, R. S., &
Rosnow, R. L. (2001). Rumor and gossip in Camerer, C. (1998). Not different after
interpersonal interaction and beyond: A all: A cross-discipline view of trust. Acad-
social exchange perspective. In R. M. Ko- emy of Management Review, 23, 393-404.
walski (Ed.), Behaving badly: Aversive be- Rousseau, D. M., & Tijoriwala, S. A. (1999).
haviors in interpersonal relationships (pp. What's a good reason to change? Moti-
203-232). Washington, DC: American vated reasoning and social accounts in
Psychological Association. promoting organizational change. Jour-
Rosnow, R. L., Esposito, J. L., & Gibney, L. nal of Applied Psychology, 84, 514-528.
(1988). Factors influencing rumor Roux-Dufort, C., & Pauchant, T. C. (1993).
spreading: Replication and extension. Rumors and crisis: A case study in the
Language & Communication, 8, 29-42. banking industry. Industrial and Environ-
Rosnow, R. L., & Fine, G. A. (1976). Rumor mental Crisis Quarterly, 7, 231-251.
and gossip: The social psychology of hearsay. Rudolph, E. (1971). A study of informal com-
New York: Elsevier. munication patterns within a multi-shift pub-
Rosnow, R. L., & Foster, E. K. (2005, April). lic utility organizational unit. Unpublished
Rumor and gossip. Psychological Science doctoral dissertation, University of Den-
Agenda, 79(4). Retrieved April 21, 2005, ver, Denver, Colorado.
from http://www.apa.org/science/psa/ Rudolph, E. (1973). Informal human com-
apr05gossip.html munication systems in a large organiza-
Rosnow, R. L., & Georgoudi, M. (1985). tion. Journal of Applied Communication Re-
"Killed by idle gossip": The psychology
search, 1,7-23.
of small talk. In B. Rubin (Ed.), When
Ruscher, J. B. (2001). Prejudiced communica-
information counts: Grading the media
tion: A social psychological perspective. New
(pp. 59-74). Lexington, MA: Lexington
York: Guilford Press.
Books.
Sabini, J., & Silver, M. (1982). Moralities of
Rosnow, R. L., & Kimmel, A. J. (2000). Ru-
mor. In A. E. Kazdin (Ed.), Encyclopedia everyday life. New York: Oxford Univer-
of psychology (Vol. 7, pp. 122-123). New sity Press.
York: Oxford University Press & Ameri- Salancik, G. R., & Pfeffer, J. (1977). A social
can Psychological Association. information processing approach to job
Rosnow, R. L., Yost, J. H., & Esposito, J. L. attitudes and task design. Administrative
(1986). Belief in rumor and likelihood of Science Quarterly, 23, 224-253.
rumor transmission. Language & Commu- Scanlon, T. J. (1977). Post-disaster rumor
nication, 6, 189-194. chains: A case study. Mass Emergencies,
Ross, L., Lepper, M. R., & Hubbard, M. 2, 121-126.
(1975). Perseverance in self-perception Schachter, S., & Burdick, H. (1955). A field
and social perception: Biased attribu- experiment on rumor transmission and
tional processes in the debriefing para- distortion. Journal of Abnormal and Social
digm. Journal of Personality and Social Psy- Psychology, 50, 363-371.
chology, 32, 880-892. Scheper-Hughes, N. (1990). Theft of life. So-
Ross, L., Lepper, M. R., Strack, F., & Stein- ciety, 27(6), 57-62.
metz, J. (1977). Social explanation and Schweiger, D. M., & DeNisi, A. S. (1991).
social expectation: Effects of real and hy- The effects of communication with em-
pothetical explanations on subjective ployees following a merger: A longitudi-
likelihood. Journal of Personality and Social nal field experiment. Academy of Manage-
Psychology, 35, 817-829. ment Journal, 34, 110-135.
274 R E F E R E N C E S

Sedikides, C., & Anderson, C. A. (1992). Smith, E. R. (1994). Social cognition contri-
Causal explanations of defection: A butions to attributional theory and re-
knowledge structure approach. Person- search. In P. G. Devine, D. L. Hamilton,
ality and Social Psychology Bulletin, 18, &T. M. Ostrom (Eds.), Social cognition: Im-
420-429. pact on social psychology (pp. 77-108). San
Sedikides, C., & Skowronski, J. J. (1991). Diego, CA: Academic Press.
The law of cognitive structure activation. Smith, G. H. (1947). Beliefs in statements
Psychological Inquiry, 2, 169-184. labeled fact and rumor. Journal of Abnor-
Sedivec, D. J. (1987). Network analysis of the mal and Social Psychology, 42, 80-90.
accuracy process within the grapevine. Un- Smith, L. C., Lucas, K. C., & Latkin, C.
published master's thesis, North Dakota (1999). Rumor and gossip: Social dis-
State University, Fargo. course on HIV and AIDS. Anthropology &
Seligman, M. E. P., Abramson, L. Y., Sem- Medicine, 6, 121-131.
mel, A., & von Baeyer, C. (1979). De- Steele, C. M. (1988). The psychology of self-
pressive attributional style. Journal of Ab- affirmation: Sustaining the integrity of
normal Psychology, 88, 242-247. the self. In L. Berkowitz (Ed.), Advances
Shadish, W. R., & Haddock, C. K. (1994). in experimental social psychology (Vol. 21,
Combining estimates of effect size. In pp. 261-302). San Diego, CA: Academ-
H. Cooper & L. V. Hedges (Eds.), The ic Press.
handbook of research synthesis (pp. 261- Stevens, L. E., & Fiske, S. T. (1995). Motiva-
282). New York: Russell Sage Foun- tion and cognition in social life: A social
dation. survival perspective. Social Cognition, 13,
Shanker, T. (2004, March 23). U.S. team in 189-214.
Baghdad fights a persistent enemy: Ru- Struthers, C. W., Menec, V. H., Schonwet-
mors. The New York Times, p. Al. ter, D. J., & Perry, R. P. (1996). The ef-
Sherif, M. (1936). The psychology of social fects of perceived attributions, action
norms. Oxford, England: HarperCollins. control, and creativity on college stu-
Shibutani, T. (1966). Improvised news: A socio- dents' motivation and performance: A
logical study of rumor. Indianapolis, IN: field study. Learning and Individual Differ-
Bobbs-Merrill. ences, 8, 121-139.
Sinha, D. (1952). Behaviour in a cata- Sugiyama, M. S. (1996). On the origins of
strophic situation: A psychological study narrative: Storyteller bias as a fitness-
of reports and rumours. British Journal of enhancing strategy. Human Nature, 7,
Psychology, 43, 200-209. 403^25.
Sinha, D. (1955). Rumours as a factor in Suls, J. M., & Goodkin, F. (1994). Medical
public opinion during elections. The East- gossip and rumor: Their role in the lay
ern Anthropologist, 8, 63-73. referral system. In R. F. Goodman &
Skarlicki, D. P., & Folger, R. (1997). Retalia- A. Ben-Ze'ev (Eds.), Good gossip (pp.
tion in the workplace: The roles of distrib- 169-179). Lawrence: University Press
utive, procedural, and interactional jus- of Kansas.
tice. Journal of Applied Psychology, 82, Tabachnick, E.G., & Fidell, L. S. (1996/
434-443. 2001). Using multivariate statistics (4th
Slackman, M. (2003, June 14). A tale of ed.). Boston: Allyn & Bacon.
two cities. The Gazette (Montreal, Quebec, Teenager arrested after cyber hoax causes chaos.
Canada), p. Fl. (2003, April 3). Retrieved May 22, 2003,
Smeltzer, L. R. (1991). An analysis of strate- from http: / / www.thestandard.com.hk /
gies for announcing organization-wide thestandard/txtarticle_v.cfm?articleid=
change. Group and Organization Studies, 38028
16, 5-24. Terry, D. J., Tonge, L., & Callan, V. J. (1995).
Smeltzer, L. R., & Zener, M. F. (1992). De- Employee adjustment to stress: The role
velopment of a model for announcing of coping resources, situational factors
major layoffs. Group and Organization and coping responses. Anxiety, Stress, and
Studies, 17, 446-472. Coping, 8, 1-24.
 References 275

lesser, A., & Rosen, S. (1975). The reluc- Van der Linden, P., & Chan, T. (2003). What
tance to transmit bad news. In L. Berkow- is an urban legend? Retrieved August 2003
itz (Ed.), Advances in experimental social from http://www.urbanlegends.com/afu.
psychology (Vol. 18, pp. 193-232). New faq/index.htm
York: Academic Press. VanDijk, T. A. (1987). Communicating racism:
Tommy rumor: The truth behind the rumor. Ethnic prejudice in thought and talk. New-
(1999, January 11). Retrieved April 15, bury Park, CA: Sage.
2005, from http://www.tommy.com/ Van Groezen, B., Leers, T., & Meijdam, L.
help/rumor/rumorOprah.jsp (2002). The vulnerability of social secu-
Trope, Y., & Liberman, A. (1996). Social hy- rity when fertility is endogenous. Journal
pothesis testing: Cognitive and motiva- of Institutional and Theoretical Economics,
tional mechanisms. In E. T. Higgins & 158, 715-730.
A. W. Kruglanski (Eds.), Social psychology: Verma, S. K. (2003, February 21). I would
Handbook of basic principles (pp. 239-270). rather die than eat beef, says PM. The
New York: Guilford Press. Statesman (India). Retrieved March 3,
Trope, Y., & Thompson, E. P. (1997). Look- 2003, from LexisNexis database.
ing for truth in all the wrong places? Victor, J. S. (1989). A rumor-panic about a
Asymmetric search of individuating in- dangerous Satanic cult in western New
formation about stereotyped group York. New York Folklore, 15, 23-49.
members. Journal of Personality and Social Vigoda, R. (1993, November 5). Heard about
Psychology, 73, 229-241. the headlights? The big lie comes sweep-
Turner, P. A. (1993). I heard it through the ing into town. Philadelphia Inquirer, pp.
grapevine: Rumor in African-American cul- Bl, B8.
ture. Berkeley: University of California Walker, C. J. (1996, March). Perceived control
Press. in wish and dread rumors. Poster presented
Turner, R. H. (1964). Collective behavior. In at the Eastern Psychological Association
R. E. L. Paris (Ed.), Handbook of modern Meeting, Washington, DC.
sociology (pp. 382-425). Chicago: Rand Walker, C. J. (2003, January). If you can't say
McNally. something good, say something bad. Paper
Turner, R. H. (1994). Rumor as intensified presented at the 4th annual meeting of
information seeking: Earthquake rumors the Society for Personality and Social Psy-
in China and the United States. In R. R. chologists, Los Angeles, CA.
Dynes & K. J. Tierney (Eds.), Disasters, col- Walker, C. J., & Beckerle, C. A. (1987). The
lective behavior, and social organization effect of anxiety on rumor transmission.
(pp. 244-256). Newark: University of Journal of Social Behavior and Personal-
Delaware Press. ity, 2, 353-360.
Turner, R. H., & Killian, L. M. (1972). Collec- Walker, C. J., & Elaine, B. (1991). The viru-
tive behavior (2nd ed.). Englewood Cliffs, lence of dread rumors: A field experi-
NJ: Prentice-Hall. ment. Language & Communication, 11,
Tybout, A. M., Calder, B. J., & Sternthal, B. 291-297.
(1981). Using information processing Walker, C. J., & Struzyk, D. (1998, June).
theory to design marketing strategies. Evidence for a social conduct moderating func-
Journal of Marketing Research, 18, 73-79. tion of common gossip. Paper presented to
Unger, H. (1979, June). Psst—heard about the International Society for the Study of
Pop Rocks? Business rumors and how Close Relationships, Saratoga Springs,
to counteract them. Canadian Business, NY.
p. 39. Walton, E. (1961). How efficient is the
U.S. Department of State Bureau of Inter- grapevine? Personnel, 38, 45-49.
national Information Programs. (2005, Weenig, M. W. H., Groenenboom, A. C.
January 14). The 4000 Jews rumor. Re- W. J., & Wilke, H. A. M. (2001). Bad
trieved December 18, 2005, from http:// news transmission as a function of the
usinfo.state.gov/media/Archive/2005/ definitiveness of consequences and the
Jan/14-260933.html relationship between communicator and
276 R E F E R E N C E S

recipient. Journal of Personality and Social System: Origins and applications. Philadel-
Psychology, 80, 449-461. phia: PEP Press.
Wegner, D. M., Coulton, G. F., & Wenz- White, R. W. (1959). Motivation reconsid-
laff, R. (1985). The transparency of de- ered: The concept of competence. Psycho-
nial: Briefing in the debriefing paradigm. logical Review, 66, 297-333.
Journal of Personality and Social Psychology, Wilke, J. R. (1986). Rumor as a social phenom-
49, 338-346. enon: An analysis of three crisis rumors of
Wegner, D. M., Wenzlaff, R., Kerker, R. M., the 1970's. Unpublished master's thesis,
& Beattie, A. E. (1981). Incrimination Auburn University, Auburn, AL.
through innuendo: Can media questions Wood, W. (1999). Motives and modes of
become public answers? Journal of Person- processing in the social influence of
ality and Social Psychology, 40, 822-832. groups. In S. Chaiken & Y. Trope (Eds.),
Weinberg, S. B., &Eich, R. K. (1978). Fight- Dual process theories in social psychology (pp.
ing fire with fire: Establishment of a ru- 547-570). New York: Guilford Press.
mor control center. Communication Quar- Wood, W. (2000). Attitude change: Persua-
terly, 26, 26-31.
sion and social influence. Annual Review
Weinberg, S. B., Regan, E. A., Weiman, L.,
of Psychology, 51, 539-570.
Thon, L. J., Kuehn, B., Mond, C. J., et
Worth, R. F. (2005, September 1). Stampede:
al. (1980). Anatomy of a rumor: A field
950 die as Iraqi crowd panics. Retrieved
study of rumor dissemination in a univer-
September 1, 2005, from http://www.
sity setting. Journal of Applied Communica-
sltrib.com
tion Research, 8, 156-160.
Weiner, B. (1985). An attributional theory Yandell, B. (1979). Those who protest too
of achievement motivation and emotion. much are seen as guilty. Personality and
Psychological Review, 92, 548-573. Social Psychology Bulletin, 5, 44-47.
Weiss, W. H. (1982). The supervisor's problem Zaremba, A. (1988). Working with the orga-
solver. New York: American Manage- nizational grapevine. Personnel Journal,
ment Association. 67, 38-42.
Werner, W. P. (1976). The distortion of rumor Zaremba, A. (1989, September/October).
as related to prejudice and stereotypes. Un- Management in a new key: Communica-
published thesis, Montclair State College, tion networks. Industrial Management,
Montclair, NJ. 31, 6-11.
Wert, S. R., & Salovey, P. (2004). A social Zingales, F. (1998, February). What's a com-
comparison account of gossip. Review of pany's reputation worth? Global Finance,
General Psychology, 8, 122-137. p. 17.
Wheelan, S. A., Verdi, A. F., & McKeage, R.
(1994). The Group Development Observation
Author Index

B
Abelson, R. P., 115 Back, K., 4, 178
Abrams, D., 78 Bacon, F. T., 102
Abramson, L. Y., 117 Bandura, A., 71
Abul-Ata, M. P., 42 Barcelona, D., 42
Ackerstrom, M., 173 Baron, R. M., 84n
Adams, J. S., 52 Baron, R. S., 4, 160, 163, 252, 254
Agnes, M., 17 Bartlett, F. C., 137, 163, 174
Aiken, L. S., 181n Bauer, R. A., 13, 186
Ajzen, I., 119 Baumeister, R. F., 20, 43, 75, 247
Albrecht, G., 15 Beal, D. J., 78
Allport, F. H., 4, 93, 96, 97, 102, 102n, 103, Beattie, A. E., 218
118, 168, 234 Beckerle, C. A., 74
Allport, G. W., 4, 5, 13, 14, 15, 17, 18, 25, Beckstead, J., 76, 257
26, 36, 37, 38, 41, 72, 77, 78, 79, 80, Begg, I. M., 102
93, 117, 118, 119, 135, 136, 137, Belgion, M., 72
137n, 138, 139, 141, 143, 144, 145, Bell, C., 4, 16
160, 163, 164, 165, 166, 168, 172, Bennett, G., 24, 25
174, 175, 178, 237 Berger, C. R., 72
Amabile, T. M., 118 Bird, D. A., 26, 36, 101, 102, 164
Ambrosini, P. J., 22, 168 Black, J., 72, 73
Anas, A., 102 "Black Beliefs," 90, 249
Anderson, C. A., 114, 115, 116, 117, 126, Elaine, B., 69, 73
222 Blake, R. H., 71, 101
Andreassen, P. B., 45 Blumenfeld, L., 90
Antaki, C., 14, 116n Bobo, L., 78
Anthony, S., 4, 73, 74, 92n Boehm, L. E., 102
Arluke, A., 19 Bordia, P., 3, 4, 14, 37, 38, 39, 40, 44, 46,
Arndt, J., 166, 167 50, 69, 71, 72, 73, 78, 79, 89, 103,
Aronson, E., 4, 5, 79 115, 118, 119, 120, 121, 125, 134,
Aronson, J., 78 138, 145, 165, 172, 174, 176, 185n,
Asch, S. E., 13 187, 190, 191, 203, 206, 210, 214,
Ashford, S., 72, 73 215, 217, 219, 225, 231, 232, 235,
Austin, M. J., 90, 206, 210 244, 246, 247, 255

277
278 A U T H O R I N D E X

Bourne, M., 219 Davis, K. E., 116n

Bradac, J. J., 72 Davis, W. L., 140
Bradley, B., 133 Day, R. S., 90
Bratslavsky, E., 43 Debevec, K., 103, 213
Brock, T. C., 77 DeClerque, J., 42, 97, 100
Brumfield, L., 90, 206, 210 Deener, B., 11
Brunsman, B. M., 4, 160, 252 De Fleur, M. L., 139, 140
Brunswik, E., 91 DeKoch, R. J., 225
Brunvand, J. H., 24, 25 Delbridge, R., 21
Buckner, H. T., 8, 130, 136, 138, 160, 172, DePaulo, B. M., 75
173, 174, 175, 176, 177, 178, 179, Dewey, J., 16
180, 192, 241, 254 Dietz-Uhler, B., 78
Burdick, H., 4, 72, 140, 141, 143, 144, 169, DiFonzo, N., 3, 4, 14, 16, 37, 38, 39, 40, 44,
178, 191 46, 52, 69, 71, 72, 73, 78, 79, 89, 103,
Burlew, L. D., 133 106, 115, 118, 119, 120, 121, 125,
Burt, R., 186, 202 134, 138, 145, 165, 172, 174, 176,
185n, 187, 190, 191, 203, 206, 210,
214, 215, 217, 219, 225, 226, 231,
232, 235, 244, 246, 247, 255
Cacioppo, J. T., 100
Dirks, K. T., 8, 186, 187, 188, 201, 241
Callan, V. J., 38, 50, 73, 206, 231
Donnelly, F. K., 101
Calvo, M. G., 74
Drew, C., 159, 171
Camerer, C., 186
Dunbar, R. I. M., 19, 20, 21
Campbell, J. D., 247
Dwyer, J. C., 159, 171
Campion-Vincent, V., 4
Cantera, K., 11
Caplow, T., 18, 26, 39, 72, 76, 140, 144,
165, 170, 171, 172, 178, 179 Edwards, O., 212
Carpenter-Theune, K. E., 75 Eich, R. K., 171
Cartwright, D., 4 Einhorn, H. J., 116
Cashman, T., 225 Ellis, R. J., 163
Castillo, M. D., 74 Emory, D., 11
Cato, F. W., 90 Esposito, J. L., 4, 72, 73, 90, 92n, 191
Chan, T., 24
Chang, A., 118
Chapman, J. P., 119, 154
Fama, E. F., 44, 46, 120
Chapman, L. J., 119, 154
Farinacci, S., 102
Chaseling, L., 103, 217
Fearn-Banks, K., 226
Choe, S., 172
Fein, S., 78
Chulov, M., 185, 206 Ferrin, D. L., 8, 186, 187, 188, 201, 241
Cialdini, R. B., 70, 71, 75
Festinger, L., 4, 72, 75, 93, 117, 164, 165,
Clampitt, P. G., 225 169, 173, 178
Clemence, A., 174
Fielding, G., 116n
Cohen, G. L., 78 Fine, G. A., 4, 20, 22, 23, 25, 26, 36, 41, 79,
Cohen, I. L., 74
205
Cohen, J., lOOn
Finkenauer, C., 43
Colman, E., 185, 206
Firth, R., 136, 173
Contractor, N. S., 175 Fischle, M., 92
Coombs, M. W., 71 Fisher, L., 44, 120
Comwell, D., 23 Fiske, S. T., 4, 5, 13, 14, 15, 19, 71, 75, 76,
Coulton, G. F., 223
116
Creed, W. E. D., 186 Flanagan, J. C., 106, 147
Folger, R., 52
D Folkes, V. S., 76, 166
David, J. P., 4, 160, 252 Foster, E. K., 5, 20, 21, 22
Davis, K., 4, 52, 133, 134, 144, 175, 176, Freedman, A. M., 43, 115, 202
191, 192, 210 Fromkin, H. L., 77
 Author Index 279

Hunsaker, P. L., 71
Gallois, C, 38, 73, 206, 219, 231 Hunt, L., 73
Georgoudi, M., 4, 12, 13, 19, 20, 21, 22
Gezon, H., 170
Gibney, L., 4, 72, 191 Inman, M., 4, 160, 252
Gigerenzer, G., 90, 91 Irmer, B. E., 219
Gillin, B., 159 Iyer, E. S., 103, 213
Gilovich, T., 120
Gist, N. P., 15, 18, 135, 136, 138, 140, 167,
172, 175, 178 Jaeger, M. E., 4, 73-74, 92n
Gleicher, D. B., 13, 186 "JDBGMGR.EXE," 4, 171
Gluckman, M., 20 Jennings, D. L., 118
Goggins, S. M., 3, 92n Jensen, M. C., 44, 120
Goldstein, D., 101 Johnson, R., 4
Goleman, D., 213 Jones, E., 38, 73, 116n, 206, 231
Goodkin, F., 42 Jukic, V., 79
Goswamy, M., 210 Jung, C. G., 96, 168
Grayer, A. R., 219 Jungermann, H., 115
Green, D. F., 172, 205
Greenwald, A. G., 92 K
Grice, H. P., 165, 219 Kahneman, D., 43, 120, 232
Groenenboom, A. C. W. J., 76, 171 Kakar, S., 256
Gudykunst, W. B., 165 Kamins, M. A., 76, 81, 166, 169
Guerin, B., 75, 76 Kapferer, J.-N., 4, 22, 23, 24, 25, 26, 36, 78,
79, 174, 187, 210
H Kashima, Y., 144, 165, 253, 254
Haddock, C. K., 92 Kashy, D. A., 75
Haidt, J., 20, 21, 39 Keefer, R., 76, 92n, 103
Haines, R., 103, 217 Kelley, H. H., 115, 116
Hanley, C. J., 172 Kelley, S. R., 37, 38, 74
Hantula, D. A., 44, 138 Kenny, D. A., 84n
Harcourt, J., 133, 134 Kenrick, D. T., 75
Hardin, C. D., 71 Kerker, R. M., 218
Harris, B., 116n Kerner, O., 42
Harris, S. G., 174 Killian, L. M., 4, 7, 13, 125, 136, 138, 142,
Harvey, J. H., 116n 164, 167, 169, 170, 172, 173, 175,
Hasher, L., 101 176, 178, 236, 253
Heath, C., 4, 16 Kimmel, A. J., 4, 13, 16, 22, 76, 92n, 103,
Heider, F., 116n, 117 212
Hellweg, S. A., 37, 39, 146, 172-173, Kirkpatrick, C., 135, 136, 137
176 Kleinbolting, H., 90
Hennen, L., 15 Kluegel, J. R., 78
Hershey, R., 37 Knapp, R. H., 37, 38, 74, 79, 90, 93, 101,
Higgins, E. T., 71, 76n, 165 102, 119, 164, 210, 231, 256
Higham, T. M., 137, 144, 166 Knapp, S. D., 92
Hilton, D. J., 114, 116 Knez, M., 202
Hobbs, S., 23 Knobloch, L. K., 75
Hobman, E., 73 Knopf, T. A., 4, 36, 41, 119, 174
Hoffrage, U., 90 Koenig, F. W., 4, 17, 36, 44, 115, 172, 205,
Hogarth, R. M., 116 210, 224, 227
Hogg, M., 78 Koller, M., 103, 217
Holtgraves, T., 219 Komarnicki, M., 177
Horn, H., 20, 21, 39 Kramer, R. M., 186, 201
Horowitz, D. L., 41, 79 Krueger, J. I., 247
Houmanfar, R., 4 Krull, D. S., 114
Hovland, C., 100 Kuhn, T. S., 167
Hubbard, M., 222 Kumar, A., 210
280 A U T H O R INDEX

Kunda, Z., 77, 80 Mirvis, P. H., 71

Kurland, N. B., 20 Modic, S. J., 134
Monczka, R. E., 133
"Monday, Monday," 11
Lalljee, M., 115 Monge, P. R., 175
Latkin, C., 19 Morin, E., 4
Lazar, R. J., 44, 120 Mouly, V. S., 4, 21, 22
Leary, M. R., 75, 79 Mouton, J. S., 71
Leavitt, H. J., 139, 144, 175 Mueller, R. A., 139, 144, 175
Leers, T., 42 Mullen, B., 92
Lepkin, M., 4, 93, 96, 97, 102, 102n, 103, Mullen, P. B., 23, 25, 26
118, 168, 234
Lepper, M. R., 92, 222, 223 N
Lerbinger, O., 202 Neuberg, S. L., 71
Lev, M., 16 Newman, M. E. J., 155
Levin, J., 19 Newstrom, J. W., 133
Liberman, A., 164, 165 Nisbett, R., 176, 223
Lin, M., 71 Nkpa, N. K. U., 140
Lindsay, J. V. Noon, M., 21
Litman, J. A., 20
O
London, I. D., 93
O'Connor, J. R., 140
London, M. B., 93
Ojha, A. B., 12
Lord, C. G., 92
Lorenzi-Cioldi, F., 174
Lott, A. J., 51 Pauchant, T. C., 141
Lott, B. E., 51 Paulsen, N., 73
Lowenberg, R. D., 168 Pederson, J. E., 133
Lucas, K. C., 19 Felled, L. H., 20
Lynch, R. D., 4 Pendleton, S. C., 4
Lynn, M., 77 Perner, L., 76, 166
Lyons, A., 144, 165, 253, 254 Perrone, V., 186
Perry, R. P., 117
M Peters, H. P., 15
MacLeod, C., 74 Peterson, W. A., 15, 18, 135, 136, 138, 140,
Maines, D. R., 164, 254 167, 172, 175, 178
Malkiel, B. G., 44, 120 Pettigrew, T. F., 78
Marks, A., 11 Petty, R. E., 100, 213, 215
Marting, B., 144 Pezzo, M. V., 20, 76, 257
Marty, E. M., 90, 115, 205 Pfeffer, J., 51
Matthews, L., 120 Pinsdorf, M. K., 205
Mausner, J., 170 Popper, K. R., 17
McAdam, J. R., 137, 139, 144, 175 Porter, E. G., 101
McEvily, B., 186 Porter, W. H., 101
McFaul, T. R., 101 Postman, L. J., 4, 5, 13, 14, 15, 17, 18, 25,
McKeage, R., 122 26, 36, 37, 38, 41, 72, 77, 78, 79, 80,
Meijdam, L., 42 93, 117, 118, 119, 135, 136, 137,
Mendoza, M., 172 137n, 138, 139, 141, 143, 144, 145,
Menec, V. H., 117 160, 163, 164, 165, 166, 168, 172,
Merton, R. K., 154 174, 175, 178, 237
Michelson, G., 4, 21, 22 Pound, J., 44, 120, 139, 144
Mihanovic, M., 79 Prasad, J., 4, 15, 41, 42, 72, 93, 169, 171,
Mikkelson, B., 11, 24, 36, 42, 69, 205, 229 173
Mikkelson, D. P., 69 Pratkanis, A. R., 4, 5, 79, 92
Milas, M., 79 Priester, J. R., 215
Miles, R. E., 186
Miller, D. L., 26, 36, 142, 175 Q
Miller, M. W., 113 Quist, R. M., 78
 Author Index 281

Smith, E. R., 116, 116n

Rajecki, D. W., 91, 92, 93 Smith, G. H., 145
Rawlins, W., 75 Smith, L. C., 19, 20, 21
Reif, W. E., 133 Snyder, S. S., 73
Resendez, M. G., 78 Spencer, S. J., 78
Richerson, V., 133 Steele, C. M., 77, 78
Robinson, S. L., 186, 202 Stegelmann, H. U., 15
Roll, R., 44, 120 Steinmetz, J., 223
Rose, A. M., 44, 120, 235 Sternberg, E., 4, 16
Rosen, S., 75, 76, 166 Stevens, L. E., 75, 76
Rosenberg, L. A., 171 Strack, P., 223
Rosenthal, M., 117, 164 Struthers, C. W., 117
Rosenthal, R., 92, 100, lOOn, 102, 103n Struzyk, D., 22
Rosnow, R. L., 3, 4, 5, 6, 7, 12, 13, 16, 17, Suls, J. M., 42
18, 19, 20, 21, 22, 25, 26, 37, 69, 72,
73, 74, 76, 77, 78, 80, 89, 90, 92, 92n,
100, lOOn, 102, 103n, 115, 118, 121, Taylor, S. E., 116
134, 135, 136, 138, 164, 191, 195, 206 "Teenager Arrested," 42
Ross, L., 92, 119, 176, 222, 223 "10 Die in Haiti," 3, 16
Rothbaum, F., 73 Terry, D. J., 50
Rousseau, D. M., 186, 187 Tesser, A., 75, 76, 166
Roux-Dufort, C, 141 Thompson, E. P., 116
Rudolph, E., 144 Thuring, M., 115
Ruscher, J. B., 78, 167 Tijoriwala, S. A., 186, 187
"Tommy Rumor," 226
Tonge, L., 50
Sabini, J., 20, 22 Toppino, T., 101
Salancik, G. R., 51 Tourish, D., 73
Salovey, P., 19, 21 Travers, V., 103, 214
Sanders, W., 120 Trope, Y., 116, 164, 165
Scanlon, T. J., 140 Trost, M. R., 70, 71, 75
Schachter, S., 4, 72, 140, 141, 143, 144, 169, Tsui, A. O., 42
178, 191 Turner, P. A., 4, 26, 36, 41, 78, 79, 119, 164,
Schnake, S. B., 78 165, 169, 202
Schonwetter, D. J., 117 Turner, R. H., 4, 7, 13, 14, 41, 125, 136, 138,
Schulz, C. A., 4, 103, 203, 215 139, 142, 164, 167, 169, 170, 171, 172,
Sedikides, C., 52, 115 173, 175, 176, 178, 210, 236, 253
Sedivec, D. J., 140, 144, 178 Tversky, A., 43, 120, 232
Seligman, M. E. P., 117 Tybout, A. M., 224
Semmel, A., 117
Shadish, W. R., 92 V
Shanker, T., 80 Vallone, R., 120
Sherif, M., 13 Van der Linden, P., 24
Shibutani, T., 4, 7, 13, 14-15, 17, 36, 41, 72, Van Dijk, T. A., 80
125, 136, 137, 138, 139, 165, 166, Van Groezen, B., 42
167, 168, 170, 171, 173, 174, 179, Verdi, A. F., 122
186, 236, 254 Verma, S. K., 79
Silver, M., 20, 22 Vigoda, R., 118
Sinha, D., 78, 79, 135, 142, 143, 144, 167, Vohs, K. D., 20, 43, 247
170, 253 von Baeyer, C., 117
Sitkin, S. B., 186
Skarlicki, D. P., 52 W
Skowronski, J. J., 52, 115 Walker, C. J., 15, 22, 69, 73, 74, 177
Slackman, M., 89 Walton, E., 133, 144, 172
Slugoski, B. R., 114, 116 Warne-Smith, D., 185, 206
Slusher, M. P., 115 Wattier, M. J., 133
Smeltzer, L. R., 44, 133 Weenig, M. W. H., 76, 171
282 A U T H O R INDEX

Wegener, D. T., 213 Wood, W., 70

Wegner, D. M., 218, 223, 224 Worth, R. F., 35
Weinberg, S. B., 102, 144, 171
Weiner, B., 114, 117
Weiss, W., 52, 100 Yandell, B., 218
Weisz, J. R., 73 Yost, J. H., 4, 90, 191
Wenzlaff, R., 218, 223
Werner, W. P., 144
Wert, S. R., 19, 21 Zaheer, A., 186
West, S. G., 181n Zanna, M. P., 163
Wheelan, S. A., 122 Zaremba, A., 133, 134
White, R. W., 71 Zeckhauser, R., 44, 120, 139, 144
Wilke, H. A. M., 76, 171 Zener, M. F., 44, 133
Wilke, J. R., 24, 36 Zhang, L., 20
Winterkorn, R., 4, 185n Zingales, P., 44
Subject Index

Attitude, 92-93
Accuracy, 145, 159. See also Rumor accuracy and rumor evaluation, 92-100
as motivation, 165-166 Attitude change research, 213
Accusation, absence of, 217-218 Attitude effects, 44, 50-61
Adding, 134-138, 140-141, 237, 252 Attribution theory, 116n, 219
Affirmative rumoring, 139
Allport, Floyd H., 4, 96-97, 102, 102n6, 103, B
168 Back, Kurt, 4, 178-179
Allport, G. W., 4-5, 14, 18, 36-37, 72, 77, Baron, R. S., 160, 163, 254
80, 93, 117, 137n, 138-139, 141, 143- Bartlett, F. C., 174
145, 160-164, 166, 168, 172, 174, Base-rate information, 119-120
178, 237 Beattie, A. E., 218-219
Alpac Bottling Company, 226 Beckerle, C. A., 74
Ambiguity, 13-14, 52, 139-140, 188, 237, Behavior effects, 44-46
243 Being in the know, 167, 253
Ambrosini, P. J., 22, 168 Belief, 134, 214-217, 234, 244, 249-250. See
American Psychological Association, Psyc- also Rumor evaluation
INFO Thesaurus, 12, 22 and rumor transmission, 69, 76-77, 248
Anderson, C. A., 114, 126 Belief perseverance, 222-224, 243
Anthony, S., 74, 92n Belief research, 90-91
Antiregressive prediction, 119-120, 235 Binomial Effect Size Display, 100, 103n7
Anxiety, 74, 163, 168-170, 256 Bird, Donald Allport, 101
and organizational change rumors, 51-61 Blake, R. H., 101
reducing, 210, 214-217, 224-225, 247, Bordia, P., 185n, 215, 232
256 Brunswik, Egon, 7, 91, 234, 250
and rumor transmission, 69, 71, 73-75, Bubble Yum (bubble gum), 43-44
242, 244, 248, 252-253 Buckner, H. Taylor, 8, 130, 138, 173-174,
and trust, 189-190, 194-195, 198-201, 176-180, 241
203 Burdick, H., 72, 141, 143-144, 178
Appropriateness, in rumor rebuttal, 216
Arndt, Johan, 166
Arthur W. Page Society, 39-40 Callan, V. J., 50
Assimilation, 134, 136-138, 141, 164, Capacity to check, 170-172, 240, 253
237 Caplow, T., 18, 39, 72, 140, 144, 165, 170,
Atrocity rumors, 36 172, 178-179
Attentional narrowing, 160-163, 239 Cartwright, Dorwin, 4

283
284 S U B J E C T I N D E X

Causal inference making, 223 Cultural anchoring, 173-174, 254

Causal mental models, 115 Cultural information, gossip as, 20
Causal schemes, 115 Culture
Celebrity gossip, 11-12 and rumor accuracy, 173-174
Channel age, 179, 240-241 and rumor evaluation, 93-96, 240, 243
Chunking, 164 and rumor transmission, 254
Classification issues, 36-39, 245 Curiosity rumors, 37
Close groups, 178, 192 Current events, 16, 25
Cognitive dissonance, 169
Cognitive mechanisms, for rumor accuracy, D
160-165, 176, 239-240, 252-253 Davis, Keith L., 176, 210
Cognitive structure activation, 52, 115 Davis, W. L., 140
Cognitive structures, 115, 235 Debevec, K., 103, 213
Coherence, 167-168 Debriefing, 223
Collective unconscious, 96 DeCIerque, J., 97-100
College students, 39, 148-153, 181-182 Denial transparency, 223-224
Commercial rumors, 36 Dewey, J., 16
Communication, formal, 51-61, 179, 212 Diffuse groups, 178
Communication channels, 171, 179, 181 DiFonzo, N., 16, 185n
age of, 179, 240-241 Dinkins, David, 202
Communication networks, 175 Dirks, Kurt, 8, 186-188, 201
Communication roles, 125 Disaster rumors, 36, 142, 159, 229
Communication styles, 125 Distortion, 159. See also Rumor distortion
Communicative postures, 122-126, 236 Distrust, 186-190, 195, 201-203, 241, 256
Compounding, 135 Dixon, Jeane, 101
Computer messaging archives, 252 Downsizing, corporate, 51, 186, 190-200,
Computer rumors, 217. See also Internet; 232, 241, 256. See also Corporate
Urban legends rumors; Organizational change
Condensation, 135 Dramatization, 167
Confidence judgments, 90-91 Dread rumors, 37-38, 41, 51-61, 74, 225,
Confidence ratings, 249 231
Confirmation bias, 164, 223 Dummy coding, 149
Conformity, 173, 240, 253-254
Conscientiousness, 188
Consistency, 92-100, 110 Earthquake rumors, 41-42, 173-174, 210
Consistency theory, 93 Education, and rumor evaluation, 103
Contagion, behavioral, 170 Edwards, Owen, 212
Context Efficient market theory, 46
and rumor rebuttal, 217-222, 226-227 Elaboration, 135, 137n
and rumor transmission, 79-86, 243 "Eleanor Club" rumors, 15, 93
Continuous scales, use of, 33 Ellis, R. J., 163
Control, 217 Embedding, 137, 137n, 138
lack of, 69, 71, 73, 225, 244 Embroidering, 135
Corporate rumors, 106-111, 113, 115, 133, Employee attitudes, 50-61
147, 154, 180-181, 190-200, 202, Employee productivity, 44, 50-61
205-206, 224. See also Organizational Employees' intention to stay, 50-61
rumors; Rumor management Employee stress, 49-50
effects, 46-49 Entertainment
Coulton, G. F., 223-224 gossip as, 20
Cover-up, perceived, 221-222 urban legends as, 24
Crime rumors, 72-73, 140-141, 167 Equity theory, 52
Crisis model, 139 Esposito, J. L., 72-73
Crisis rumors, 36 Evaluative talk, gossip as, 20-22
Critical ability, diminishing, 170 Evidence, 17. See also Rumor research
Critical incident, 147 agenda; Rumor research studies
Critical set orientation, 176, 180, 241 anecdotal, 39
Cues, 91, 106-111, 234, 249-250 Exaggeration, 136, 143, 167
 Subject Index 285

Excitement, collective, 170, 240, 253 Hershey, R., 210

Exclusion, gossip and, 20 Higham, T. M., 166
Expert consensus, 110-111 Hilfiger, Tommy, 226
Explanation, 116 Hoffrage, U., 90
ready-made, 116, 223, 235 Holtgraves, T., 219
Explanation kernels, 115 Horn, H., 21, 39
Explanation process, 114-116 Honesty, and rumor rebuttal, 216, 226
Explanation theory, 114-115, 126, 235 Hostility rumors, 38
External ramifications, 48-49
Eyewitness perceptions, false, 171 I
Illusory correlation, 118-119, 154, 235, 251
Illusory truth effect, 102
Fabrication, 135 Importance, 237
Fact-finding motivation, 70-75, 80, 82-83, and accuracy motivation, 166
165, 233-234, 244, 246-249, 253 and rumor evaluation, 103
Falsifiability principle, 17-18, 245 and rumor transmission, 69, 71-73
Ferrin, Donald, 8, 186-188, 201 Inequity, perceptions of, 52
Festinger, Leon, 4, 72, 93 Informal networks, 72, 170
Fine, G. A., 36, 79 Information, lack of, 172
Fiske, S. T., 116 Information Dimensions Scale (IDS), 27,
Folkes, V. S., 76 230, 245
Former Soviet Union, 186 Information statements, rumors as, 16-17,
Free feedback, 175 117, 243
Free riders, 21 Ingroup, 78, 169
Initial impressions, persistence of, 222-224
Institute for Public Relations, Gainesville,
Gestalt memory studies, 138 Florida, 39-40
Gezon, H., 170 Instrumental relevance, 16
Gibney, L., 72 Integrative model, 243-244
Gigerenzer, G., 90 Interaction, 175, 240, 252-253. See also
Gist, N. P., 18, 135, 140, 167, 178 Rumor discussion
Goldstein, D., 101-102 Intergroup relations, rumor in, 256
"Good Times" computer virus, 217 Internal attitudes, 48-49
Gossip, 12-13, 19-23, 230. See also Rumor Internal behaviors, 48-49
research Internet, 72, 121-130, 178, 217, 236, 251-
Grapevine research, 39, 133-134, 144, 146, 252. See also Urban legends
176, 239 Invention, 135, 137n, 139-141, 173
Grayer, A. R., 219 Iraq, war in, 37
Grice, H. P., 219 Israel, 39
Groenenboom, A. C. W. J., 76 Iyer, E. S., 103, 213
Group biases, 173-174
Group cohesion, 19-20
Group epistemic norms, 21, 174, 240 Jaeger, M. E., 73-74, 92n
Group mechanisms, for rumor accuracy, Judgment biases, 118-120, 235
173-175, 239-240 Jung, C. G., 96, 168
Group membership, 20, 78, 81-86, 169
Group orientation, 176, 179, 241, 254 K
Group structure, 178 Kamins, M. A., 76, 81, 169
Group thinking, 13 Kapferer, J.-N., 24, 36-37, 174, 187
Gudykunst, W. B., 165 Kashima, Y., 253-254
Keefer, R., 76, 103
H Kelley, S. R., 37-39
Haidt, J., 21, 39 Kerker, R. M., 218-219
Hale, Lorraine, 202 Killian, L. M., 142, 167, 178, 253
Hasher, L., 101-102 Kimmel, A. J., 76, 103
Heider, F., 117 Kirkpatrick, C., 135
Hellweg, S. A., 39, 146 Kleinbolting, H., 90
286 S U B J E C T INDEX

Knapp, R. H., 37, 79, 93, 101-102, 210 Network mechanisms, for rumor accuracy,
Knopf, T. A., 36, 174 175-179, 239-240, 254
Knowledge structures, 114-115 Neutral outside sources, and rumor rebuttal,
Koenig, F. W., 36, 44, 212-213 226
Koller, M., 103, 217 News, 16-17, 122, 230. See also Rumor
Komarnicki, M., 177-178 research
Krull, D. S., 114 improvised, 122
Nisbett, R., 176
Nkpa, N. K. U., 140
Legend, 18-19. See also Urban legends No-comments strategy, 219-222, 225, 243
modern, 23 Nonregressive predictions, 120
Lens model of judgment, 91, 234, 250
Lepkin, Milton, 96-97, 102, 102n6, 103, 168 O
Leveling, 134-141, 163, 166, 237, 251 O'Connor, J. R., 140
Liaisons, 175-176, 192, 254 Ojha, A. B., 12
Liberman, A., 164 One-liners, rumor as, 18
Life cycle of rumors, 206-210 Organizational change, 38, 40-41, 44, 180-
Likelihood of transmission (LOT), 191-192 181, 187-190, 231, 243, 256-257. See
Litigation, 205, 210 also Organizational rumors; Trust
London, I. D., 93 Organizational citizenship behavior (OCB),
London, M. B., 93 188
Lowenberg, R. D., 168 Organizational research, 201-203
Lyons, A., 253-254 Organizational rumors, 37-38, 133, 232. See
also Rumor management
consumer concern, 37-38
M costly error, 37-38
Main effects model of trust, 186-188, downsizing, 51, 186, 190-200, 206, 232,
195-198 241, 256
Mass poisoning rumors, 168 and employee stress, 49-50
Matthew accuracy effect, 154-155, 239, 252 frequency, 39-41
Mausner, J., 170 internal/external, 40^1
Maxims of conversation, 219 job quality, 37-38
McCartney, Paul, 135, 164 job security, 37—38
McDonald's Corporation, 205-206, 224 pecking order, 37-38
McKeage, R., 122 and rumor accuracy, 182-183
Medical rumors, 42 turnover, 37-38
Memes, 16 Organized groups, 170
Memory limits, 138, 163, 239 Outcome relevance, 16
Merton, Robert K., 154 Outcome-relevant involvement, 69, 72-73,
Miller, D. L., 36 178, 248
Misinformation, 17 Outgroup, 78, 169
Moderator model of trust, 187-190
Morality, urban legends and, 24
Motivation, for rumor transmission, 71-80 Perceptual biases, 164-165, 240
competing, 81, 85-86 Perner, L., 76
contextual determinants, 79-86 Persuasion research, 100-101, 213, 215
Motivational approach, 70, 231, 234, 244, Peterson, W. A., 18, 135, 140, 167, 178
246-249 Plausibility, 14
Motivational mechanisms, for rumor accu- Police investigations, 210
racy, 165-169, 239-240, 253 Political rumors, 79, 164
MUM effect (minimize unpleasant mes- Popper, Karl R., 17
sages), 75-76, 84, 166 Pop Rocks (candy), 44
Porter, E. G., 101
N Postman, Leo J., 4-5, 14, 18, 36-37, 72, 77,
Nebulous forms, 33-34 80, 93, 117, 137n, 138-139, 141, 143-
Network analysis, 175 145, 160-164, 166, 168, 172, 174,
Network characteristics, 243-244 178, 237
 Subject Index 287

Power relationships, gossip and, 20-21 and legend, 18-19

Prasad, Jamuna, 4, 15, 42, 93, 173-174 and news, 16-17, 122, 230
Press campaign, 205 as ready-made explanation, 116-118,
Primary control, 73 223, 235
Principal components analysis, 48—49 as threat management, 15, 38, 231, 243
Probabilistic mental models (PMMs), 90-91 as unverified statement, 17-18
Problem solving, 72, 122. See also Sense and urban legends, 12, 25-26, 230
making Rumor accuracy, 142-144, 153-155, 238-
Procter & Gamble Corporation, 17, 115, 205 239, 251-255
Prodigy (Internet service provider), 113, 115 Matthew accuracy effect, 154-155, 239,
Product rumors, 36, 43^14, 79, 166, 219- 252
222. See also Corporate rumors; Rumor Rumor accuracy mechanisms, 160, 239-241,
management 251-255
Projection, 168 cognitive mechanisms, 160-165, 176,
Prospect theory, 42-43 239-240, 252-253
Proximity, and capacity to check, 172 group mechanisms, 173-175, 239-240
Public relations professionals, 39-41, 46-49, motivational mechanisms, 165-169, 239-
147, 176-178, 210-212, 232, 242 240, 253
network mechanisms, 175-179, 239-240,
R 254
Race rumors, 15, 36, 41^3, 78, 80, 93, 115, situational features, 169-173, 239-240,
118-119, 164-165, 202, 249 253
Reality testing. See Rumor evaluation Rumor accuracy studies, 145-147
Recall bias, 153-154 field interviews (1996), 147
Reciprocity, 175 student rumor survey no. 1, 148-149
Relationship-enhancement motivation, 70, student rumor survey no. 2, 149-153
75-77, 81-82, 84, 90, 233-234, 244, Rumor accuracy trend, 148
246-249 Rumor activity, 176. See also Rumor discus-
and rumor accuracy, 166-168, 240 sion; Rumor evaluation; Rumor
Reliability (psychometrics), 143 transmission
Repetition effect, 225. See also Rumor Rumor attribution, 100-101, 235. See also Ex-
repetition planation process
Report of the American National Advisory Com- Rumor circulation, 16
mission on Civil Disorders, 41-42 Rumor comparison, 172
Reputation, and accuracy motivation, 166 Rumor confirmation, 212, 225-226
Risk management, rumor as, 187 Rumor construction, deliberative or extempo-
Robinson, S. L., 202 raneous, 170
Rochester Institute of Technology, 28, Rumor content, 121-122, 173-174
81-82 change patterns, 134-142, 237
Roosevelt, Eleanor, 15 Rumor denial, 212. See also Rumor rebuttal
Rosenberg, L. A., 171 Rumor discussion, 13, 36, 72, 121-130, 240
Rosnow, Ralph L., 4, 7, 17-18, 37, 72, 74, communicative postures, 122-126, 236
77, 92n, 135 content statements, 121-122, 126-130,
Ross, L., 176 236
Rousseau, D. M., 187 Rumor distortion, 142-143, 176, 238
Rumor, 13-19, 145, 229-230, 243 Rumor diversity, 141
attitude effects, 44, 50-61 Rumor effects, 232-233, 245-246. See also At-
behavioral effects, 43-44 titude effects; Behavior effects
content, 16-19 Rumor evaluation, 80, 89-90, 206
contexts, 13 Rumor frequency, 39-41, 231
fallout, 41-44 Rumor generation, 89, 206
forms, 36-39 Rumor Interaction Analysis System (RIAS),
functions, 14-15, 22 121-122, 236
and gossip, 12-13, 22-23, 33, 230 Rumor interaction episodes (RIEs), 121, 130,
as information statement, 16-17, 117, 236, 251
243 Rumor management, 206-212, 242, 257-258
internal/external, 231 recommendations, 224—227
288 S U B J E C T I N D E X

Rumor persistence, 18, 172, 222-224 serial transmission (ST), 134, 137-140,
Rumor precision, 142-145, 238 160-163, 176-178, 237, 240
Rumor prevention, 210, 257 Rumor transmission configuration, 175-179,
Rumor propaganda, 78-79, 246-247 240, 255
Rumor public, 37-38 Rumor verification, 17-18. See also Rumor
Rumor rebuttal, 102-103, 203, 210, 212- evaluation
222, 226, 242-243, 258 Rumor verity, 142-145, 154, 238
context, 217-222, 226-227
source characteristics, 213-216, 226
Rumor repetition, 101-102, 106, 110 Sabini, J., 22
Rumor research Salovey, P., 19
cluster analysis, 125-126 Satanism rumor, 205-206. See also Corporate
content analysis, 121-122 rumors
critical-incident methodology, 147 Scanlon, T. J., 140
future agenda, 245-248 Schachter, Stanley, 4, 72, 141, 143-144, 178
history, 4-5 Secondary control, 15, 73
information dimensions, 27-32 Sedivec, D. J., 144-145, 178
meta-analytic methods, 91-92 Self-enhancement motivation, 70, 77-82,
microworld simulation, 44-46 167, 233-234, 244, 246-249, 253
motivational approach, 70, 231, 234, 244, and rumor accuracy, 168-169, 240
246-249 Self-esteem research, 247
principal components analysis, 212 Sense making, 14-15, 38, 41, 72, 138, 231,
use of exemplars, 33 243, 250-251
Rumor research agenda, 245-248 collective, 113, 121-130, 236
Rumor research studies individual, 113-120, 234-235
field interviews (1996), 147, 180-181 Sharpening, 134-138, 141, 237
organizational change study, 191-200 Shibutani, Tamotsu, 4, 7, 13, 36, 167, 170,
public relations professionals, 39-41, 46- 173-174, 179, 186
49, 147, 176-178, 210-212, 232 Silver, M., 22
stockbrokers, 106-111, 118 Sinha, D., 142-144, 170
student rumor survey no. 1, 148-149 Situational features, and rumor accuracy,
student rumor survey no. 2, 149-153, 169-173, 239-240, 253
181-182 Skepticism, 172, 174, 177, 181, 210
Rumors, false, 90, 135, 142, 154, 239 Snowballing, 135, 141, 167
counterfeits, 143, 238 Social bonding, gossip and, 20
fallen stars, 143, 238 Social costs, of rumor transmission, 247
hopefuls, 143, 238 Social desirabililty bias, 154
Rumors, negative, 76, 232-233, 241, Social identity, 70, 79-80
256 Social interaction, 70
Rumors, planted, 139-140, 247 Social learning, 51-52
Rumors, positive, 202 Social space configurations, 255
Rumors, true, 142, 154, 172, 225-226, family configuration, 255
238-239 ribbon configuration, 255
converts, 143, 238 torus configuration, 255
grainies, 143, 238 Social status, 77
stars, 143, 238 Source credibility, 172, 213
Rumor transmission, 18, 51, 89, 206, 210, and rumor evaluation, 100-101, 106, 110
233-234, 246-249, 257. See also Ru- and rumor rebuttal, 213-216, 226
mor content; Trust Spread of rumors. See Rumor transmission
cluster transmission, 175, 240 Stable-cause attribution, 46, 116-119, 235
collaborative approach (COL), 134, 137- Stakeholder status, 110
138, 144, 237 Stereotyping, 78, 154, 164-165, 252-253
factors in, 69-70 Stockbrokers, 106-111, 118, 234
measures of, 191-192 Stock market rumors, 38, 44-46, 118-120,
motivations for, 71-80 139, 234-235
multiple interaction (MI), 130, 175-179, Storytelling, 24-26, 75-76
240-241, 254-255 Stroh Brewery Company, 224, 227
 Subject Index 289

Structure, 18-19 Uncritical set orientation, 176

Suggestibility, 170 University of Rochester (UofR), 82
Symbolization, 164 Unstable-cause attribution, 117-118
Unverified statements, rumor as, 17-18
Urban legends, 12, 23-26, 230. See also Ru-
Tall tales, 24 mor research
Taylor, S. E., 116 car theft at mall, 36
Taylor Manifest Anxiety Scale, 74 headlights hoax, 42-43, 171-172
Teller-listener-extremity effect, 160-163, kangaroo, 23-24
252-254 K-Mart snake, 26
Terror attacks of September 11, 2001, 11 stolen kidney, 24
Terry, D. J., 50 "The Hook," 24
Thibaut, John, 4 vanishing hitchhiker, 24
Threat management, 15, 38, 231, 243 Urgency of action, 171-172
Tijoriwala, S. A., 187
Time, as situational feature, 172-173, 240
Tonge, L., 50 Validity (psychometrics), 142
Toppino, T., 101-102 Verdi, A. P., 122
Tracking strategy, 45-46
Training, and stable-cause attribution, 46 W
Trends, perceived, 110 Walker, C. J., 73-74, 177-178
Trope, Y., 164 Wartime rumors, 37, 39, 78-79, 93, 96-97,
Tropical Fantasy (soft drink), 43, 115, 202 101, 118, 138, 140, 168
Trust, 51, 185-186, 201-203, 216, 225, 241- Wedge-driving rumors, 15, 37, 51-61, 80,
242, 244, 256-257 97, 169, 189, 231, 247
main effects, 186-188, 195-198, 201 Weenig, M. W. H., 76
moderating effects, 188-190, 198-201 Wegner, D. M., 218-219, 223-224
Trust building, 225 Weinberg, S. B., 102
Truth kits, 172, 205 Weiner, B., 114
Turner, P. A., 36, 169 Wenzlaff, R., 218-219, 223-224
Turner, Ralph H., 7, 139, 142, 167, 178, 210, Wert, S. R., 19
253 Wheelan, S. A., 122
Tybout, A. M., 224 Whispering campaigns, 79
Wilke, H. A. M., 76
U Wilke, J. R., 24, 36
Uncertainty Winterkorn, Rob, 51, 185n
and organizational change rumors, 51—61 Wish rumors, 37, 41, 74, 231
reducing, 210, 224-225, 256
and rumor transmission, 69, 71-72, 234-
235, 244, 248 Yandell, B., 218
structuring, 212, 225, 242, 258
and trust, 189-190, 194-195, 199-201,
203 Zanna, M. P., 163
About the Authors

Nicholas DiFonzo earned his PhD in social and organiza-

tional psychology from Temple University in 1994, where
he was awarded the Marianthi Georgoudi Dissertation
Award for philosophical and theoretical contributions to the
field of psychology. He is currently professor of psychology
at the Rochester Institute of Technology. He is a member of
numerous professional associations including the American
Psychological Association, the American Psychological Soci-
ety, and the Society for Personality and Social Psychologists.
Dr. DiFonzo currently serves as consulting editor for The
Journal of Social Psychology.
He has published approximately 30 articles, book chap-
ters, and technical reports pertaining to the topic of rumor.
He has studied how rumors propagate through networks in
social space and across time, the mechanisms by which ru-
mors become accurate or distorted, motivations involved in
rumor transmission, processes involved in believing a rumor,
how rumor is differentiated from other forms of communica-
tion, how rumor processes are affected by organizational
trust, and how rumors influence social and economic behav-
ior. He has also pursued practical applications of rumor
theory including how harmful rumors may be most effec-
tively refuted. Dr. DiFonzo has also given approximately 30
presentations and invited addresses at academic conferences

291
292 A B O U T THE A U T H O R S

on these topics, as well as several seminars to business audiences on

the topic of managing rumors.
Dr. DiFonzo recently received a grant from the National Science
Foundation to investigate how rumors propagate through social net-
works over time. He has also received grants from the Institute for
Public Relations to study corporate rumors, their effects, and how top
corporate public relations officers handle them. He has investigated
rumor attributional processes and boomerang effects in rumor refuta-
tion in projects funded by the Australian Research Council. Dr. DiFonzo
organized and maintains an Internet discussion group, Rumor-Gossip
[email protected], for scholars interested in the topics of ru-
mor and gossip. Dr. DiFonzo is currently serving as expert witness for
the Procter & Gamble Corporation on the topic of the famous (and
false) Satanism rumor and product rumors.

Prashant Bordia is an associate professor in the School of Management

at the University of South Australia. He obtained his PhD in social and
organizational psychology from Temple University in 1996, where he
was awarded the Vanderveer Fellowship in 1995. He is a member of
the U.S. Academy of Management and the Australian Psychological
Society and an international affiliate of the American Psychological
Association. He is a consulting editor for The Journal of Social Psychology
and on the editorial boards of Group & Organization Management,
the Journal of Business and Psychology, and the Journal of Business
Communication.
Dr. Bordia has studied rumor as collective sense making on the
Internet, effectiveness of rumor rebuttal strategies, rumors during orga-
nizational change and, more recently, the role of motivations in rumor
spread. He was awarded the University of Queensland Foundation
Early Career Research Excellence Award in 2003. Dr. Bordia has pub-
lished over 40 articles in refereed journals and book chapters and
has presented numerous papers in international conferences. He has
received best paper awards at the Academy of Management and the
Australian I/O Psychology conferences. He has received competitive
research grant funding from the Australian Research Council and the
U.S. National Science Foundation to study rumors and has conducted
research and consulting projects with several public and private sector
organizations. Dr. Bordia is currently serving as expert witness for the
Procter & Gamble Corporation on the topic of the famous (and false)
Satanism rumor and product rumors.