Title: Explaining the "Bandwagon" and "underdog" effects
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00102734/00001
 Material Information
Title: Explaining the "Bandwagon" and "underdog" effects
Physical Description: Book
Language: English
Creator: Hollander, Barry A
Copyright Date: 1991
 Record Information
Bibliographic ID: UF00102734
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: oclc - 25604963
ltuf - AJC6967

Full Text








There are many people who deserve recognition for

their assistance in my completing my graduate studies,

which culminate in this dissertation.

First, I would like to thank my committee: Drs. Mary

Ann Ferguson, Michael Martinez, Leonard Tipton, Michael

Weigold, and John Wright. It was a Leonard Tipton

research seminar that first sparked my interest in public

opinion and perceptions of opinion. No doubt this

fascination will continue for some time.

I would also like to thank Kim Walsh-Childers for

acting as a substitute on my final dissertation. Thanks

also to Michael Weigold and Julie Dodd for use of their

students in the dissertation experiments.

Mary Ann Ferguson deserves special note. She hired

me as her research assistant at the Communication Research

center when I was a lowly and hungry master's student

barely able to find a computer's on-off switch. My three-

year apprenticeship with her at the Center is the reason I

continued my graduate studies, and her research seminars

provided me a theoretical direction. Her enthusiasm for

research and teaching was contagious. If ever a graduate

student bears the stamp of his major professor, I suspect

I do. With luck and hard work, I may live up to this


A number of fellow graduate students deserve

acknowledgement, more than I have the time or space to

mention. Geetu Melwani from the period of my work on a

master's degree and fellow doctoral students Erika

Engstrom, Art Emig, and Alan Fried come immediately to

mind. There are many others.

Finally, and most important, I thank my wife Edith

for her patience, love, and support. She worked as I

continued my graduate studies, making it possible for me

to write this dissertation. I look forward to spending

time with her again.




ABSTRACT ............................................


1 INTRODUCTION.................................

A Different Perspective.......................
Research Hypothees.e ..........................
Summary of Dissertation Contents..............

2 PUBLIC OPINION AND POLLS.....................

Polls and Mass Communication.................
Bandwagon Research...........................
Conformity Research...........................
The Poll Bandwagon--A Summary.................

3 PERSUASION RESEARCH..........................

Cognitive Response Analysis..................
Persuasion and Individual Differences.........

4 UNCERTAINTY ORIENTATION......................

Theoretical Discussion.......................
Findings in Uncertainty Orientation..........
Other Personality Theories....................
On the Study of Personality, Politics,
and Communication..........................

5 SYNTHESIS AND HYPOTHESES.....................

Bandwagon Research and Persuasion.............
Consensus as a Heuristic Cue.................
Summary and Hypothees........................
Overview of the Studies.......................
















Methdology................................... 47
Results ....................................... 52
Summary...................................... 60
Notes ......................................... 64

7 STUDY 2 METHODOLOGY.......................... 66

The News Articles ........................... 66
Measurement of Variables..................... 71
Uncertainty Orientation...................... 72

8 STUDY 2 RESULTS .............................. 77

Assumptions.................................. 77
Manipulation Checks.......................... 78
Tests of Hypotheses.......................... 80
Other Significant Findings................... 94
Notes ......................................... 95

9 DISCUSSION................................... 96

Summary of Study Findings.................... 96
Limitations of the Studies.................... 100
Conclusions.................................. 105
Implications for Bandwagon Research.......... 107
Future Research............................... 110


A NEWS ARTICLE EXAMPLES........................ 113


ORIENTATION QUESTIONS...................... 130

REFERENCES.......................................... 132

BIOGRAPHICAL SKETCH................................. 147

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Barry A. Hollander

December 1991

Chairman: Leonard Tipton
Major Department: Mass Communication

Concerns about the influence of public opinion polls

have existed almost as long as polls themselves.

"Bandwagon" research results have been inconsistent,

however, even with an underdog effect being found. The

conceptual framework of persuasion research from social

psychology was used to review the bandwagon literature and

to offer hypotheses. The individual difference variable

of uncertainty orientation was presented as one

explanation as to why some people are influenced in the

direction of a poll while others are influenced in the

opposite direction.

Three hypotheses were presented. First, it was

hypothesized that subjects would be more susceptible to

poll influence in a low relevance situation as opposed to

one of high relevance. Second, it was hypothesized that

certainty-oriented persons would be more likely to move in

a "bandwagon" direction, or with the poll influence, while

uncertainty-oriented persons would be more likely to move

in an "underdog" direction, or opposite the poll

influence. Finally, it was hypothesized that when

relevance was high, certainty-oriented persons would be

susceptible to poll influence.

A 2 (relevance) X 2 (uncertainty orientation) X 3

(type of poll) factorial design experiment was conducted

using news articles with bogus poll results. Mixed

support was found for all three hypotheses, primarily in

situations where relevance was manipulated rather than

between issues thought to differ in relevance. It was

suggested that future research into public opinion poll

influence take into consideration the relevance of the

campaign or issue under consideration as well as the

personality traits of those hearing such poll results.



In 1935, U.S. Rep. Walter Pierce of Oregon introduced

legislation that would have prohibited use of the mail for

gathering political straw ballots. If enacted, pollsters

convicted under the law faced a maximum penalty of one

year in prison or a $1,000 fine or both. Such polls are

"powerful, subsidized propaganda," Pierce warned in a

later Public Opinion Quarterly article, and should be

recognized as "a potent, if not the most powerful, agency

now used to influence public opinion" (Pierce, 1940, p.

241 and p. 243, respectively).

While Pierce's bill failed, the potential influence

of public opinion polls has been a focus of debate for

more than 50 years. Pollsters assert that their efforts

reflect rather than affect public opinion, arguing no

"bandwagon" effect exists (Gallup & Rae, 1940; Robinson,

1937). Other early commentators such as Pierce (1940) and

Lewis (1940), however, doubted that polls were so innocent

and incapable of persuasion.

Despite arguments, evidence from survey and

experimental studies remains inconclusive. Reviews by

Marsh (1983) and Merkle (1991) found that while some

studies do favor a bandwagon interpretation, many do not.

Laboratory studies of artificial elections coupled with

bogus poll results produce an "underdog" effect, of people

shifting to the minority rather than majority view. In

general, there appears little support for the bandwagon

thesis, particularly in controlled settings, which seem to

favor an underdog effect. Possible reasons for such

inconsistent conclusions range from the kinds of issues or

candidates studied to how the "majority" and "minority"

are identified. Few attempts have been made, however, to

explain and systematically test why people may choose the

bandwagon or underdog route or the process that takes

place. Almost exclusively devoted to elections at the

expense of issues, the bandwagon literature typically

assumes an undifferentiated public, though some scholars

have begun to investigate the influence of predispositions

on likelihood to move with or against a poll (e.g.,

Lavrakas, Holley, & Miller, 1991).

A Different Perspective

Some public opinion scholars have hinted at the

possibility that polls are influential in situations of

less importance to the public. McBride (1991) suggests

such influence is more likely in political races below the

presidential level, while Hickman (1991) points to

presidential primaries rather than elections as better

opportunities to find the bandwagon effect.

Another perspective is offered, this one from social

psychology. Theorists such as Petty and Cacioppo (1986a,

1986b) outline two distinct modes of processing which bear

directly on how and why persuasion takes place. The key

to these persuasion models is whether people have the

motivation and ability to process a message. Where

motivation is low, people are thought to be more likely to

use shortcuts in making decisions. An issue or political

campaign of little personal relevance, then, is unlikely

to provide the motivation necessary to carefully

scrutinize all the questions involved. In such a case, a

credible source or the presence of some other source

characteristic can be of greater influence. In situations

of high relevance, such source characteristics are thought

to have little influence. With this conceptual framework

in mind, the research in poll influence is reviewed and an

attempt is made to organize the results along the lines

seen in these persuasive models.

As noted above, one criticism of bandwagon research

is its focus on an undifferentiated public. One

explanation for both the bandwagon and underdog effects

may also be in people themselves. That is, some people

may be more persuaded by polls, while others may be more

open to the arguments presented by a minority. The

individual difference variable uncertainty orientation is

offered as one possible explanation (Sorrentino, Short, &

Raynor, 1984). Briefly, Sorrentino and his colleagues

posit that some people -- certainty-oriented persons --

seek to maintain clarity about themselves and their

environment. That is, they will avoid self-diagnostic

situations if they feel that information is threatening,

and, in fact, when facing a situation of personal

importance they will rely on simple decision-making rules

rather than consider all the information possible. Their

polar opposites, uncertainty-oriented persons, are open to

such information and in fact seek it out.

Research Hypotheses

Three research hypotheses can be generated from this

discussion, all of which are discussed in more detail in

later chapters. First, it is expected that issues of

little personal relevance will be more influenced by poll

results than issues of high personal relevance. When

considering uncertainty orientation, it is hypothesized

that certainty-oriented persons will be more likely to be

persuaded by a poll than uncertainty-oriented persons.

Finally, building on other work in uncertainty

orientation, it is expected this relationship between

uncertainty orientation and poll results will be more

pronounced in situations of high relevance, which

certainty-oriented persons would find threatening. That

is, certainty-oriented persons would be even more

persuaded by a poll in personally relevant situations

rather than in low-relevant situations.

Summary of Dissertation Contents

Chapter 2 reviews the research on the influence of

polls, most of which has been conducted by public opinion

and mass communication scholars, while Chapter 3 discusses

the persuasion literature from social psychology. Chapter

4 reviews the work on uncertainty orientation, and Chapter

5 provides a synthesis and the resulting hypotheses.

Study 1 (Chapter 6) is designed mainly to select

issues for use in Study 2. The second study tests the

research hypotheses discussed above. Chapter 7 presents

the methodology of Study 2 and Chapter 8 the results. The

final chapter provides an overall discussion of the



Separating the poll from public opinion is difficult,

and public opinion sometimes defined as that which a

public opinion poll measures. Public opinion is often

viewed as the aggregation and distribution of individual

opinions, simply "any collection of individual opinions"

(Childs, 1939, p. 331). Others regard public opinion as a

communication process (e.g., Cooley, 1909; Price 1989).

Price and Roberts (1988) distinguish between the

pollster's view of public in public opinion as a noun and

their conceptualization of public as an adjective

describing the opinion process. An important feature of

this conceptualization is F. D. Allport's (1937) view of

public opinion formation occurring through an individual's

observation of his social environment, a theoretical

orientation extended by others, notably Noelle-Neumann's

"spiral of silence" theory (Noelle-Neumann, 1974, 1977).

Noelle-Neumann (1974, p. 51) suggests individuals

possess a "quasi-statistical organ" that allows them to

assess the opinion environment, largely through

information provided by the mass media. However, it

appears from research such as the false consensus effect

(Ross, Greene, & House, 1977; Marks & Miller, 1987) and

pluralistic ignorance (Fields & Schuman, 1976; O'Gorman &

Garry, 1976) that if such an "organ" exists, it is indeed

a faulty one. False consensus research has demonstrated

that people tend to believe others think as they do, even

when those individuals are obviously in a minority.

Pluralistic ignorance has to do with misunderstanding of

true opinion distribution.

Polls and Mass Communication

Key to Noelle-Neumann's spiral of silence theory is

how people learn about the distribution of opinion. A

likely source is public opinion polls, a popular staple of

the news media since early in the century. More than one-

third of newspapers surveyed reported the use of polls to

gather information (Rippey, 1980), and about 15 percent of

newspaper election stories focused on polls (Stovall &

Solomon, 1984). While there has been some improvement in

how the media conduct or report on polls (Salwen, 1985),

others charge news organizations have fail to provide

necessary explanatory information on polls and, when they

do so, the reliance on such polls places a "horse-race

metaphor" emphasis on elections (Broh, 1980). Content

analyses of poll stories found newspapers often fail to

provide the information deemed necessary by polling

experts (Miller & Hurd, 1982; Sanders, Rollberg, &

Buffalo, 1989), despite findings that inclusion of this

information may improve polls' credibility (Paletz, et

al., 1980; Salwen, 1987).

People are aware of public opinion and polls. Gollin

(1980) reported that 84 percent of a 1976 national sample

had heard of polls. Kohut (1986) and Roper (1986) both

report results suggesting the public regards polls and

pollsters as generally credible.

While polls are a popular means of people learning

about the distribution of opinion on public issues, many

question whether, rather than reflecting public opinion,

polls can actually influence public opinion (see, for

example, Bogart, 1972; Gollin, 1980; Roll & Cantril,


Bandwagon Research

Early scholars and pundits termed the influence of

actual or perceived majority opinion on election results

as the "bandwagon" effect. A bandwagon effect occurs when

the majority receives additional support from the

publication of a poll; an underdog effect is just the

opposite, in which people respond to an underdog's status

and go against the majority or with the minority. This

notion of a bandwagon effect has existed almost as long as

public opinion polling, although much of the literature is

simply speculative.

Though a number of elegant theoretical models have

been published arguing for the effect (Gartner, 1976;

Henshel & Johnston, 1987; H. A. Simon, 1954; Straffin,

1977; Zech, 1975), the literature is sprinkled generously

with reports of null findings (Beniger, 1976; Dizney &

Roskens, 1962; Fleitas, 1971; Lang & Lang, 1984). Adding

to the confusion is the discovery of an "underdog" effect

(Ceci & Kain, 1982; Laponce, 1966).

As discussed in the previous chapter, early in this

century congressional efforts to control straw ballots

failed. Even the news media, a major user of polling

today, argued for stricter regulation. A 1936 New York

Times editorial, for example, cautioned that polls often

"develop a bandwagon rush" and "in cases where public

opinion is less definitely rooted, and the allegiance of

large groups of voters is less definitely won, a

'bandwagon' tendency may play a decisive part in

determining the result of an election" ("Straw Ballots,"

p. 22). Similar arguments were made by others (F. D.

Allport, 1940; Perry, 1968; Stoler, 1986). Early

pollsters brought forth more systematic though problematic

evidence. Directly following Pierce's 1940 condemnation

of polls in Public Opinion Quarterly, for example, Gallup

and Rae (1940) presented sequential survey evidence

finding no bandwagon effect.

Research in public opinion poll influence typically

takes one of two methodological directions. The first is

the use of survey, panel, or trend survey data to uncover

opinion movement; the second is the use of experiments or

quasi-experimental methods in laboratory or field

settings. First we review the survey evidence, then that

from experiments.

Survey Evidence

One popular survey method is to ask respondents what

influenced their decisions and whether, more specifically,

polls had some influence (Hastings & Hastings, 1990;

Roshwalb & Resnicoff, 1971)--a self-report method not

without criticism as to whether individuals are aware or

can accurately describe what influenced their opinions

(Nisbett and Ross, 1977). Berelson, Lazarsfeld and McPhee

(1954) found only about 10 percent mentioned polls as a

source of information about changes in who they expected

to win the presidential campaign. In asking directly

about poll influence, Hastings and Hastings (1990) found

between 3 and 11 percent of those polled willing to admit

such influence. Roshwald and Resnicoff's (1971) survey of

1,000 respondents found no voluntary mentions of polls as

being influential in a vote decision and, when asked, 66

percent of the respondents said polls were not at all

important in their decision-making.

Other survey research has attempted to track effects

of polls through panel or trend studies. One of the

earliest discussions of the effect on voting by

Lazarsfeld, Berelson, & Gaudet (1948) found many voters

attempting to sense the direction of public opinion and

the outcome in order to vote "with the winner." Follow-up

work (Berelson, Lazarsfeld, and McPhee, 1954) revealed

that a bandwagon effect (perception guiding preference)

and a projection explanation (preference guiding

perception) carried about equal weight in presidential

election voting.

In another use of survey methodology, an extensive

study by Beniger (1976) analyzed data from the date of

gathering and publication of Gallup candidate preference

polls and state primary elections from 1936 to 1972.

Essentially, Beniger's sequential method was looking for

whether a poll influenced the primary, or vice versa. In

83 of 183 preference polls followed by a subsequent poll,

the ratings of the leading candidate increased from the

first to the second poll, while in 81 cases it fell. Like

many other sequential poll studies (Kavanagh, 1981; Roll &

Cantril, 1980; McBride, 1991), Beniger found no

consistent bandwagon effect and even hints of an equally

inconsistent underdog effect, though others have

successfully demonstrated more indirect poll effects such

as those on campaign contributions, volunteer work, or

endorsements by key political actors (Henshel & Johnston,


Non-experimental cross-sectional evidence cannot

resolve the crucial question of causality. Trend studies

provide a slightly clearer picture, but the results do not

support the intuitive appeal and theoretical justification

of a bandwagon effect of poll influence. Other

methodologies have not improved matters.

Experimental and Quasi-Experimental Evidence

Given the weaknesses noted above, Campbell (1958)

asserted "the question of a bandwagon effect resulting

from publication of poll results can only be settled by

experimentation" (p. 252). Even controlled settings have

not been kind to the bandwagon thesis, however, with

findings as mixed as those from survey research.

To investigate the question, Campbell suggested a

national manipulation of poll results, a difficult and

costly task in the 1950s and one far too expensive--if

even possible -- today. A similar but less ambitious

study reported by Gallup (1972) compared cities with and

without newspapers carrying the Gallup Poll and found no

differences in candidate preferences. No further details

were provided.

An early experiment by Cook and Welch (1940) used

college students and as a stimulus the infamous 1936

Literary Digest poll showing Alf Landon leading Roosevelt

in the presidential election. Students informed about the

Digest poll were more likely to favor Landon than those in

the control group not told about the poll. A second

experiment conducted over a longer time period produced

results in the direction of a bandwagon prediction, though

the results were not statistically significant. These

results led Cook and Welch to suggest any bandwagon effect

may be a small one, reminiscent of Robinson's (1937)

cautioning that the bandwagon thesis contained "more an

element of fancy than fact" (p. 49).

Katz (1972) argued that it was not sheer numbers from

polls that resulted in influence, "but the character of

the majority in relation to the individual's psychological

group membership" (p. 24). That is, a poll of similar

others was thought to have more influence than a poll of a

more general public. Atkin (1969) found some support by

using "student preferences" on one issue, though no effect

was found on two other issues.

Two British studies favored an underdog effect on

likelihood to turn out to vote, indirectly affecting

results. The first, cited in Teer and Spence (1973),

varied information provided to three treatment groups of

400 voters before an election. There was no effect on

preference, though supporters of the party they were told

was behind at the polls said they were more likely to turn

out to vote. In the second study, Gaskell (1974) varied

information given about who was ahead in the polls, but

rather than a no-information control group, he told

respondents the polls were close. Fewer supporters said

they would vote if a clear-cut victory was predicted, but

this finding was stronger for supporters of the party

predicted to win. However, deBock's (1972) investigation

of the one-sided 1972 presidential election with a field

experiment suggested the trailing candidate may suffer

loss of support intensity and turnout motivation after

supporters have been exposed to election poll results.

Navazio (1977) divided a sample of 500 persons into

two groups, one a control and the other an experimental

group that was told of bogus national poll results. The

mail survey found no poll effect. In a similar vein,

Roper, in a study reported by Cantril (1980), conducted a

split-ballot trial with presidential approval ratings by

introducing half of the questions with "As you know, all

the polls have been showing support for Carter going down.

We'd like to get your opinion about him." Few technical

details were provided by the author, though it was

reported that the experimental manipulation resulted in

only a 1 percent movement from Carter.

Ceci and Kain (1982) asked students their preferences

for Carter and Reagan in the 1980 presidential election

during morning classes. Before being asked, some sections

of the class were told a survey of "college-educated

persons" showed support for Carter (first section of the

class), Reagan (second section), or no result (third

section). Later that night, confederates posing as

national pollsters telephoned students and requested a

presidential preference. Prior to this, students were

again provided bogus poll data. Despite the

methodological problems of this study's use of intact

groups, the results point to movement toward the underdog.

A quasi-experimental panel study by Cloutier, Nadeau,

and Guay (1989) found 75 percent of their 1,005 subjects

had the same attitude over a one-month period. However,

14 percent changed in the direction of the poll

manipulation, while 11 percent went with the underdog or

minority view.

Fewer studies combine experimental control with a

representative sample. A two-wave national panel survey

conducted during the 1988 Bush/Dukakis presidential

campaign by Lavrakas, Holley and Miller (1991) included an

experimental manipulation of informing respondents about

actual public opinion poll results. Some interesting

results were obtained. Persons who did not.graduate from

high school and who were told that Bush was leading in the

polls showed a significant increase in uncertainty about

which candidate they would vote for compared to those with

similar education who were not told Bush was leading. No

effect was seen on those with higher education. The

authors conclude better-educated persons were more

resistant to the external source of influence.

Summarizing the experimental research, Marsh (1983)

concluded no experimental evidence exists for a bandwagon

hypothesis, though there is some support that a perception

of growing strength may aid in the bandwagon process.

Conformity Research

In a related field of study, research on conformity

has long documented the influence of the majority on the

minority in small-group settings (Allen, 1965; Kiesler &

Kiesler, 1969). The majority's superior size, power, and

status provides a firm base for establishing social

reality and a position in which to reward those who agree

(Festinger, 1950). Recent findings have led many to

question whether the minority must be willing and helpless

recipients or whether a minority cam instead bring about

modification and change (Moscovici & Personnaz, 1980),

speculation not unlike the underdog thesis. It has even

been suggested that majority and minority influence

processes differ (Moscovici, 1980; Nemeth, 1987), though

many others maintain a single process is in effect (Doms,

1983; Latane & Wolf, 1981; Wolf, 1987). Latane's (1981)

social impact theory is one example of single process

argument, proposing that influence by either a majority or

minority is a multiplicative function of the strength,

immediacy, and number of its members.

Typically these studies rely on a single confederate

who disagrees with the majority on some perception task,

the results showing a persistent minority can influence

both the public and private responses of a majority. In

one study, Moscovici & Personnaz (1980) presented subjects

a series of blue slides consistently labeled as green by

the confederate. On each trial, subjects were required to

indicate the color of the slide and the color of the

afterimage perceived on a screen following removal of the

slide. The confederate was presented as either a member

of the majority or minority. The minority viewpoint was

found to influence subjects' judgments of the afterimage.

The authors conclude that the persistent minority causes

the majority to start a validation process by considering

the deviant response and its reasoning. "In other words,

as a result of trying to see or understand what the

minority saw or understood, the majority begins to see and

understand as the minority would" (Moscovici & Personnaz,

1980, p. 272).

It appears the minority, at least in these

experimental settings, can indeed influence the majority,

confirming Asch's original suggestion (Asch, 1951) that

some persons distort their perceptions of an unambiguous

stimuli as a function of influence of others. In

Moscovici's (Moscovici, 1980) theory, a consistently

stated minority opinion can lead to careful scrutiny by

the group, a detailed processing of the minority position

similar to that of systematic or central processing. This

influence is held to be latent, indirect, or covert, an

internalized change of opinion not necessarily revealed

publicly. Influence by the majority is held to bring

about manifest, direct, or overt influence, what

researchers call public compliance.

One study in this paradigm stands out as closest to

the bandwagon question (Kaplowitz, Fink, D'Alessio, &

Armstrong, 1983). Subjects were presented six issue

statements with different bogus poll results about those

issues. Subjects were either told they would either have

to discuss their attitudes publicly or were given a

private response option. No conformity effect was found

for those who reported high commitment to their attitudes

about the issues, even those responding privately.

However, the authors did find the bogus-poll technique

could produce a conformity effect on those with little

commitment to their attitudes. A previous study (Tyson &

Kaplowitz, 1977) concluded that if people respond

privately, results of public opinion polls will not

directly influence responses to issues on which they have

strong commitment to their views.

The Poll Bandwagon--A Summary

Throughout the accumulation of these assorted

findings on poll influence, a number of factors have been

suggested for opinion movements, including perception of

majority trends (Marsh, 1984), the strength or

unexpectedness of these perceptions (Aronson, Turner, &

Carlsmith, 1963), and the commitment to an opinion

(Kaplowitz, Fink, D'Alessio, & Armstrong, 1983). In some

research a bandwagon effect is found, in others an

underdog effect. Some studies find no results, while

others find both effects occurring.

Why the difficulties? Merkle (1991) suggests a

number of methods of uncovering the underlying process:

more powerful experimental designs, care in looking for

both bandwagon and underdog effects, assessing or

controlling for previous perceptions, analysis of the

relevance of reference groups, the kinds of topics used as

a stimulus, investigation of personality characteristics,


and consideration of how subjects interpret the poll


Another way to approach the question is from a single

theoretical perspective that attempts to consolidate

findings in persuasion research in general and to apply

that to poll influence studies in particular. The next

chapter focuses on the work by cognitive response analysis

theorists and two specific and complementary models of



Soon after Gordon Allport declared attitudes to be

"the most distinctive and indispensable concept in

contemporary American social psychology" (Allport, 1935,

p. 798), the construct blossomed as a research focus.

Motivated by World War II, early scholars in social

psychology investigated how persuasion and propaganda

influenced attitudes (Hovland, Lumsdaine, & Sheffield,

1949; Hovland, Janis, & Kelley, 1953).

The framework described by Lasswell (1948a) and

elaborated by Hovland et al. (1953) most closely resembles

the work of cognitive response theorists and is the

theoretical direction taken in this dissertation. It has

enjoyed remarkable growth in the past 10 years. The early

approach focused on "who says what to whom with what

effect," the investigation of the communicator, the

message, the audience, and the effects of attitude change.

Cognitive Response Analysis

Cognitive response analysis delves further into the

audience, arguing that attitude change is primarily a

function of an individual's personal elaboration of

external stimuli (Fiske & Taylor, 1984; Petty, Ostrom, &

Brock, 1981). A stimulus causes a cognitive effect, which

results in a response. Variables in this research include

audience involvement, message repetition, message

comprehension, and communicator credibility.

Critical to the cognitive response perspective is

audience involvement or personal relevance, the variable

most mentioned as a determinant of processing styles.

Research in the social judgment perspective shows high

involvement to be associated with less attitude change,

while the cognitive response to persuasion model shows

just the opposite, a contrast noted by Petty and Cacioppo

(1979a) and explained by differences in conceptions of

involvement (O'Keefe, 1990). Without involvement, social

cognitive theorists argue, an individual will not use

thoughtful, central, systematic information-processing

strategies and instead will rely on peripheral, heuristic,

or shortcut strategies that depend heavily on such

irrelevant cues as communicator characteristics (Chaiken,

1980; Petty & Cacioppo, 1979b). Simply put, less involved

persons are more likely to operate "on automatic," while

highly involved persons will operate in a more

"controlled" mode (Fiske & Taylor, 1984, p. 354).

This research into the two distinct styles of making

social judgments is best exemplified by the elaboration

likelihood model (ELM) of persuasion and the heuristic-

systematic model (HSM) of persuasion (see Petty &

Cacioppo, 1986a or 1986b, for more complete descriptions

of the ELM. For a similar view on HSM, see- Chaiken,

Liberman, & Eagly, 1989).

Elaboration in the ELM is generally described as "the

extent to which a person thinks about the issue-relevant

argument contained in a message" (Petty & Cacioppo, 1986b,

p. 128). In their conceptualization, when elaboration

likelihood is high, more cognitive resources should be

used by the receiver. Use of issue-relevant elaboration

typically results in the new arguments, or a personal

translation of them, being integrated into the underlying

belief structure or schema for the attitude object.

Current social psychological research provides support for

two general modes of processing discussed above, termed by

these theorists as central and peripheral routes to


In this view, the likelihood of elaboration depends

on a person's motivation and ability to evaluate the

communication. Motivation is influenced by such variables

as personal relevance or personal responsibility, while

ability to process can be influenced either by personal

characteristics such as prior knowledge or by such

variables not related to the person as message

comprehension or distraction in the communication

situation. With both motivation and ability present,

elaboration likelihood is high. With only one or neither

present, elaboration likelihood is low.

This model presents a continuum from no thought about

the issue-relevant information to complete elaboration on

the issues. On this theoretical continuum's elaboration

side is the central route to persuasion, on the opposite

end the peripheral route. The former has persuasion

achieved through close scrutiny to message arguments and

consideration of other issue-relevant material, the latter

is conducted through use of some simple decision rule or

heuristic principle to evaluate an advocated position.

Persuasion through the central route is posited to be long

lasting, while persuasion through the peripheral route is

relatively temporary. Examples of heuristics are number

of arguments presented, use of expert sources,

communicator credibility, likability, and attractiveness.

The two routes are not mutually exclusive; rather they

represent the extremes of the elaboration likelihood

continuum. For example, a mixture of central and

peripheral processes are theorized to occur at moderate

levels of elaboration (Petty & Cacioppo, 1986a).

A series of experiments by Petty, Cacioppo, and their

colleagues show that high relevance increases the

likelihood of issue-relevant thinking and invoking the

central route to persuasion (Petty & Cacioppo, 1979a,

1979b, 1984; Petty, Cacioppo, & Goldman, 1981).

The HSM (see Chaiken, 1980, 1986, 1987; Chaiken &

Eagly, 1983) similarly views heuristic processing as

involving more limited cognitive effort and capacity. It

goes further, however, by describing heuristics as learned

knowledge structures that may be used by social

perceivers. Such rules can range from "Experts'

statements can be trusted" to "If most people think so, it

must be right." In general, both models agree that two

modes of processing exist, and that heuristics such as

source expertise or consensus can lead to less systematic

processing of arguments.

Persuasion and Individual Differences

The Petty, Cacioppo, and Goldman (1981) study of the

effects of argument strength and communicator expertise on

persuasive effectiveness provides a clear example of the

cognitive response perspective. Subjects were presented a

message with an issue that was either relatively involving

personally (designed to prompt high elaboration) or not

involving (presumably to cause less elaboration).

Personal relevance or involvement was manipulated by

making receivers believe the recommendations in the

message for comprehensive examinations for senior

undergraduates were being considered for adoption at their

university the following year (high involvement) or 10

years later (low involvement). Also manipulated were

argument quality (strong versus weak) and the expertise of

the communicator (high versus low). For those in the

high-relevance condition, strong arguments produced more

attitude change than weak arguments, but source expertise

did not affect persuasion. In the low-relevance

condition, subjects were more persuaded by an expert than

by a non-expert source, but the quality of the arguments

in the message did not affect subjects' attitudes. It

appears that when issue relevance is high, people are more

persuaded by message content (the central route); but when

relevance is low, they are more persuaded by attributes or

cues (peripheral route).

But is this always the case? Little work has been

conducted on individual differences in the effects of

personal relevance, though Cacioppo, Petty, and Morris

(1983) demonstrated that high need for cognition, defined

as the tendency for people to engage in and enjoy

thinking, resulted in more central processing than for

those low in the need for cognition. Need for cognition

was not found to interact with the manipulations and only

enhanced the likelihood for central processing. Others

questioned whether personal relevance might actually

reduce the likelihood for elaboration for some people


(Sorrentino, Bobocel, Gitta, Olson, & Hewitt, 1988).

Uncertainty orientation, identified by Sorrentino, Short,

and Raynor (1984), is one such possible dimension.


Uncertainty orientation is an individual difference

occupying a niche called the "warm look," an interaction

and interdependence of motivation and cognition to explain

behavior (Sorrentino & Higgins, 1986; Sorrentino & Short,

1986). The construct remains in an early state of

development and researchers have concentrated less on a

definition than on describing those who are certainty and

uncertainty-oriented, conceding "we are still learning

about them ourselves" (Sorrentino & Roney, 1990, p. 242).

This chapter describes the theoretical and

operational rationale behind the construct and sketches

the research to date.

Theoretical Discussion

Definitions proposed for uncertainty orientation (and

as we will see later, the measurement itself) have changed

as well, though Sorrentino maintains the variable is a

cognitive rather than motivational one (Sorrentino &

Short, 1986). For example, need for cognition (Cacioppo &

Petty, 1982) has been described as a variable measuring

motivation to think, while uncertainty orientation is a

measure of when to think (Sorrentino, et al., 1988).

Uncertainty orientation has been described as the

degree to which situations of certainty versus uncertainty

are cognitively relevant (Sorrentino, Short, & Raynor,

1984); as involving the degree of uncertainty surrounding

the outcome of one's activity (Sorrentino & Short, 1986);

as how one orients him or herself toward the uncertainty

concerning the self or the environment (Sorrentino &

Hancock, 1987); as people's relative interest in either

maximizing information gain or maintaining clarity (Roney

& Sorrentino, 1987); and as an individual's approach to

uncertainty or ambiguity (Sorrentino, et al., 1988). In

general, uncertainty orientation is the concern or

interest in finding out new things about the self and the

environment and the tolerance of ambiguity about the self

or environment.

What should interest mass communication scholars is

Sorrentino's contention that uncertainty orientation is

primarily concerned with information value and the

maintenance and preservation of current knowledge.

Interest in the variable arose from Atkinson and Raynor's

(1974) theory of achievement motivation and replies from

cognitive informational theorists (Trope, 1975; Weiner,

1972). In short, it was argued that situations that

attain clarity about the self result in the greatest

arousal of achievement-related motives for uncertainty-

oriented people, and situations that maintain clarity lead

to the greatest arousal of these motives for certainty-

oriented people. One series of experiments showed the

theory of achievement motivation only held for

uncertainty-oriented persons (Sorrentino, Short, & Raynor,

1984), that is, certainty-oriented people demonstrated no

interest in using a diagnostic achievement situation to

learn about themselves.

Generally, it is posited that the uncertainty-

oriented person seeks to attain clarity about the self and

the environment, while the certainty-orientated individual

seeks to maintain present clarity about the self or the

environment. To attain clarity, one is willing or likely

to deal with information that may even threaten the self-

perception or challenge beliefs. Maintaining clarity is

just the opposite, the tendency to hold on to beliefs and

to avoid situations that are self-diagnostic. Attaining

or maintaining clarity is done through use or failure to

use or process information that is available.

Historical Perspective

The prototype for uncertainty orientation is traced

to Rokeach's The Open and Closed Mind (1960), which

presents a continuum of "gestalt types" to "psychoanalytic

types." Rokeach characterized the former as having a

"need for a cognitive framework to know and understand"

and the latter as having the "need to ward off threatening

aspects of reality" (Rokeach, 1960, p. 67). Thus the

open-minded person possesses a belief system oriented

toward the new, the closed-minded a belief system oriented

toward the familiar, a dichotomy similar to the Freudian

notion of basic trust and mistrust of the world. The more

closed the system, according to this argument, the more

difficult it is to distinguish between information

received about the world and information about the

source--a perspective not unlike that of persuasion

theorists discussed in the previous chapter that

emphasizes elaboration of issue matter versus influence by

source characteristics.

Rokeach's work included the famous "Joe Doodlebug"

experiments to explore differences between those Rokeach

called "closed" and "open" persons. The experiments

showed that persons with "closed" systems had more

difficulty with integrating new beliefs and often showed

early closure. These results were not associated with

intelligence or ideology.

Sorrentino also notes the influence of Kagan (1972),

who described uncertainty resolution as one of four

primary determinants of behavior. The motive to resolve

uncertainty, he writes, is the "wish to know" (p. 54).

While not discussing individual differences, Kagan

suggested that people in a state of uncertainty will

either assimilate, remove, or escape from an event that is

discrepant from an established schema. Certainty-oriented

persons would attempt to remove or escape such discrepant

information, while uncertainty-oriented persons would

attempt to assimilate such information.

Cognitive Perspective

The term schema shows up in Sorrentino's work as

defense for uncertainty orientation as a largely cognitive

variable with motivational roots. Building on the work of

construct assessability (Bargh 1984; Higgins & King,

1981), King and Sorrentino (1988) found that uncertainty-

oriented and certainty-oriented persons differ in the

chronic accessibility of positive and negative constructs

reflecting certainty (e.g., "cautious" and "stubborn") and

uncertainty (e.g., "adventurous" and reckless").

In perhaps the most complete explanation of the

construct, the authors write that their model assumes

that uncertainty-oriented persons are those
who have been rewarded for autonomous
exploratory behavior. Such persons develop
general schemas for situations that allow
resolution of uncertainty about the self and
the environment. When placed in these
situations, they are effectively charged to
resolve this uncertainty. Certainty-oriented
persons, on the other hand, have not been
rewarded for autonomous, exploratory behavior
and in fact might have been punished for such
behavior. Consequently, these persons
gravitate toward and develop schemas for safe
and familiar situations or schemas that do

not require dealing with uncertain
situations. Hence, they are effectively
charged by situations involving certainty and
the self and the environment (Sorrentino, et
al., 1988, p. 358)

An "affective charge" closely resembles the language

of motivation scholars who describe motivation as the

drive that energizes, orients, and guides behavior.

Uncertainty-oriented individuals will be motivated in

situations that involve uncertainty about the self or the

environment, while certainty-oriented persons should be

more motivated in situations that involve certainty about

the self or the environment, setting into motion

respective cognitive responses.

Findings in Uncertainty Orientation Research

The initial test of this hypothesis by Sorrentino et

al. (1984) found that for uncertainty-oriented persons,

relevant sources of motivation were most aroused in

performance situations that contained uncertainty about

the self or the environment. The reverse was found for

certainty-oriented persons. The authors argued their

findings integrate existing cognitive and motivational

interpretations of achievement behavior.

This first study was aimed at explaining problems

arising from motivational research. In work more closely

related to social influence, Sorrentino & Hewitt (1984)

tested whether there would be differences in how

uncertainty-oriented and certainty-oriented people would

approach information that was personally relevant. As

predicted, uncertainty-oriented persons chose activities

aimed at resolving uncertainty about a new and potentially

important ability, whereas certainty-oriented persons

actually chose an alternative activity that would tell

them nothing new about this ability. This finding held

even when the information was likely to be positive.

Again studying personal relevance, Sorrentino and

Roney (1986) demonstrated that motivational arousal would

occur in situations relevant to a person's uncertainty

orientation. Achievement-related motives of uncertainty-

oriented persons were most aroused in a diagnostic task

about a new and possibly important ability (mental

flexibility), but those of certainty-oriented persons were

most aroused in a non diagnostic task.


An important aspect of uncertainty orientation is use

of authoritarianism, a construct elaborated by Adorno,

Frenkel-Brunswik, Levinson, & Sanford (1950) and ably

criticized by Shils (1954) as dealing only with

authoritarianism of the "right." The preoccupation is

understandable, however, given the political climate of

the 1940s. Still, most of the early questions measuring

the construct were directed at the political right.

One of the earliest and most influential studies

linking psychological and political concepts, The

Authoritarian Personality, identified a group of

individuals who possessed anti-Semitic attitudes and then

expanded this view to generally ethnocentric aversions for

all people. Theoretical development of the F-scale (for

fascism) was drawn largely from Freudian concepts of

superego, ego, and id, and was traced to childhood

development (Sanford, 1971). Authoritarians were found to

display considerable cognitive rigidity and intolerance

for ambiguity (Block & Block, 1951; Frenkel-Brunswick,

1949; Jones, 1954; Rokeach, 1948; Steiner & Johnson,

1963), to believe others thought the same way they do

(Granberg, 1972; Scodel & Mussen, 1953; Simons, 1966), to

favor conservative political attitudes (Adorno, et al.,

1950), and to vote for conservative or authoritarian

candidates (Higgins, 1965; Milton, 1952; Poley, 1974;

Wrightsman, 1965).

Other Personality Theories

There is a long history of attempts to provide an

exhaustive taxonomy of personality (Cattell, 1943; Digman,

1990; Fiske, 1949; Norman, 1963). Some consensus for a

"Big Five" model appears to be emerging (Digman, 1990).

Of these five factors, one dimension appears closely

related to uncertainty orientation. This dimension goes

by a number of labels, but perhaps can best be described

as an openness to experience (Costa & McCrae, 1985; Coan,


Another similar personality dimension is repression-

sensitization (Byrne, 1964), where persons are thought to

deal with threatening stimuli in one of two ways--a person

who approaches threats is a sensitizer, a person who

avoids threats is a repressor. Repressors focus on

supportative information and avoid information that is not

supportative (Olson & Zanna, 1979).

On the Study of Personality. Politics, and Communication

"To talk about politics without reference to human

beings," Lippmann wrote, "is just the deepest error in our

political thinking" (1913, p. 2). Early work by Lasswell

emphasized the need to bring psychology into political

analysis (Lasswell, 1930, 1948b, 1951). Yet personality

variables, which indicate an organized, stable internal

predisposition a person brings to a situation, have not

enjoyed a great deal of success in political communication

research. Even supporters of the theoretical perspective

concede personality variables are unlikely to account for

much of the variance when looking at such behaviors as

joining a major political party or in voting, though

evidence exists that such variables are "selectively felt

at a number of points in the political process" (Knutson,

1973, p. 44).

Much of this "vast and uneven work" falls under the

externalization hypothesis, which treats political

opinions as externalized versions of the mind's inner

conflicts (Kinder & Sears, 1985, p. 676). Sorrentino

argues for extending the work in individual differences

from the achievement area to that of social influence and

persuasion. Reviewing the work of persuasion theorists,

Sorrentino and others questioned whether uncertainty

orientation may interact with personal relevance in a

persuasive context. As discussed earlier, research has

typically shown that perceivers process information more

thoughtfully when the subject has high rather than low

personal relevance (Petty & Cacioppo, 1981; Fiske &

Taylor, 1984; Nisbett & Ross, 1980). The authors

replicated the work of persuasion theorists, primarily

that of Petty & Cacioppo (1984) and Petty et al. (1981),

to study the effects of personal relevance on attitude

change as a function of one's uncertainty orientation.

Sorrentino et al. (1988) predicted that, unlike

uncertainty-oriented persons, high personal relevance

would make certainty-oriented persons less careful or

systematic in their processing of message arguments and

more dependent on heuristics, or persuasion cues, than

would low personal relevance. The investigators found

that personal relevance does not appear to increase

systematic processing for all persons--that certainty-

oriented persons are more motivated to consider a matter

further when personal relevance decreases. These

individuals do not prefer to reason matters out when

facing a situation with uncertainty about themselves or

their environment, the authors conclude. "When forced

into these situations, they rely on heuristics rather than

use their own judgment" (p. 368). This reversal of a

principle "thought to be universal," they add, highlights

the value of an individual-differences approach in social

influence (p. 369).

The question is not one of merely varying the

strength of a persuasive effect, then, but varying the

direction of the effect. "In other words, variables

assumed to enhance systematic forms of information

processing may work for some, but may actually dampen such

tendencies in others" (Sorrentino & Hancock, 1987, p.



Bandwagon Research and Persuasion

Reviewing the bandwagon research from the perspective

of persuasion theory offers some clues as to why the

bandwagon effect occurs in some cases but not in others.

A key variable in this paradigm is personal

relevance. A look at many bandwagon and poll studies

finds some influence in what we might suspect are low

relevance issues, and little or no persuasive (or

bandwagon) effects on what we might suspect are high

relevance issues. For example, Atkin's (1969) use of a

"student preference" poll found effects only on an income

tax surcharge issue but not on the issues of marijuana or

reopening the Warren Commission investigation

(controversial matters at the time). While no measures of

relevance are provided in the study, it seems likely that

the tax surcharge issue was less important to his college

student subject pool that the use of marijuana or the

Warren Commission issue. Atkin used a three-week period

between measures, thus it may be that the poll had more

influence on the less relevant issue but no effect on the

higher relevance issues.

A look at Table 5-1 provides a look at some selected

poll persuasion studies. Though drawn largely from Merkle

(1991), the table organizes the list chronologically with

bandwagon or bandwagon/underdog findings listed first,

underdog-only effects listed second, and then null studies

listed third.

Most null findings occurred in the 1970s, a time when

belief in persuasive effects of mass communication were in

their decline. Studies conducted in the 1980s have

uncovered greater support for the bandwagon/underdog

thesis, perhaps due to a reliance on experiments and


Looking closely at the table of the thirteen studies

finding a bandwagon or bandwagon/underdog effect, only

five were elections or candidate evaluations. Both of the

studies finding only an underdog effect were elections,

while seven of eight studies finding no poll influence

were of an election or candidate evaluation. It appears

elections offer less an opportunity to demonstrate a

poll's influence. This hardly seems surprising if, as is

suggested by the ELM and other persuasion models, people

are more highly involved in presidential or other

elections or hold more closely to candidate evaluations

than they would an abstract issue, therefore less likely

to be persuaded by a heuristic such as a poll result.

Table 5-1

Chronological Listing of Selected Bandwagon/Underdog and
Consensus Persuasion Studies

Authors Subjects Topic Conclusion

Cook & Welch

Allard (1941)

et al., (1944)
and Berelson,
et al., (1954)

Atkin (1969)

Roshwalb &
Resnicoff (1971)

de Bock (1972)

Meyers, et al.

et al. (1983)

Marsh (1984)

Mackie (1987)

et al. (1989)

et al. (1990)

























Some bandwagon

Support for
bandwagon effect

Support for
bandwagon and
underdog effect

Some support for
bandwagon effect

Indirect support
for bandwagon

Support for
indirect bandwagon

Support for
bandwagon effect

Bandwagon support
on low-commitment

Trend produced
bandwagon effect

Support for

Support for

Support for

Table 5-1--continued

Authors Subjects Topic Conclusion

Laponce (1966)

Ceci & Kain



College Pres.
students election

Support for
underdog effect

Some support for
underdog effect

Dizney & Roskens

Sales Research
Ltd. (1964) Cited
in Teer & Spence

Fleitas (1971)

Gaskell (1974)

Beniger (1976)

Tyson & Kaplowitz

Navazio (1977)

Roper (1980)
cited in
Cantril (1980)















on Nixon

on Carter

No support for

No support for

No support for

No support for

No support for

No support for
bandwagon effect

No support for

No support for

Looking at issues, Kaplowitz, et al. (1983) found no

conformity effect on what subjects described as high-

committment issues, but some conformity on low commitment

issues, a finding that would be anticipated in the

persuasion perspective. Atkin (1969) also found support

for the bandwagon thesis, but only on one of three issues

(one less likely to be relevant to respondents).

Consensus as a Heuristic Cue

A number of different heuristic principles have been

suggested in persuasion studies, but three have received

the most research attention: the credibility, liking, and

consensus heuristics (O'Keefe, 1990). The consensus

heuristic can be broadly defined as "if other people

believe it, then it's probably true" (O'Keefe, 1990, p.

107). Two recent studies using the cognitive response

conceptual framework to investigate consensus effects

arrive at different, though not contradictory,


Axsom, Yates, & Chaiken (1987) studied audience

response as a consensus cue in order to clarify a history

of mixed results, finding these reactions influenced

opinions only under low involvement conditions; in high

involvement conditions only argument quality influenced

opinions, a result similar to that of Kaplowitz, et al.


In a series of experiments more closely aligned to

this dissertation, Mackie (1987) investigated the kind of

processing induced by consensus information. Mackie found

subjects simultaneously exposed to a majority poll result

with which they disagreed and a minority with which they

agreed showed more issue-relevant processing of the

majority message. She concluded that the majority

provides validity to the arguments presented, directing

attention to them and resulting in considerable

processing. That is, disagreeing with a majority results

in more attention being paid to the issue-relevant facts

themselves, leading to systematic processing. Majority

status, then, can in some cases provide the motivation for

recipients to systematically process communications in

certain circumstances.

Summary and Hypotheses

Polls influence people--sometimes. The conceptual

framework of the ELM and HSM provides one hope of

unraveling these tangled findings. As noted by Marsh

(1983), research on the bandwagon effect or poll influence

in general has tended to focus on elections at the expense

of issues. This dissertation will explore issues and poll

influence using the persuasion paradigm, which typically

uses dependent variables other than a "yes" versus "no" or

one candidate versus the other, instead relying on an

attitude measure.

If an issue is personally relevant, then a consensus

heuristic is predicted to have less influence. If an

issue is not relevant, presentation of consensus should

have more chance to sway attitudes. Following this, it is

hypothesized that an issue poll will be more persuasive in

low- rather than high-relevance issues.

Second is the matter of uncertainty orientation. In

a series of speculations about the possible interactive

effects of uncertainty orientation, Sorrentino and Hancock

(1987) argue that uncertainty-oriented persons, exposed to

consensus information, would try to discover why a

minority insisted on its side of an issue in the face of a

majority. Certainty-oriented persons would be more likely

to ignore the inconsistency and "go along with the crowd"

(Sorrentino & Hancock, 1987, p. 256). Here we come to the

bandwagon versus underdog effect. If this theorizing

holds, uncertainty-oriented persons should be less likely

to be influenced by a poll or move in the "underdog"

direction, while certainty-oriented persons should be more

persuaded by poll.

Finally, based on the work of Sorrentino, et al.

(1988), it is hypothesized that in situations of high

relevance, certainty-oriented persons threatened by such

relevance would turn to a heuristic cue such as a poll and

be more influenced by a poll than would uncertainty-

oriented persons.

Overview of the Studies

This dissertation is organized around two studies.

Study 1 largely focuses on the testing of issues for use

in Study 2, though it also includes analysis of two

experimental manipulations helpful in deciding how best to

proceed with the second phase.



Choosing the Issues

Two steps were taken to decide what issues to use.

First, students from four mass communication writing

classes, of about 20 students each, were asked in fall

1990 and spring 1991 to list the three most personally

important social or political issues.

In the first issue-listing phase, the Persian Gulf

crisis received the most mentions (26), followed by

abortion (15), the environment (14), AIDS (11), the

economy (11), education (9), censorship, drugs, elderly

and child abuse, and energy (each with 6). Next, from

this list, issue statements were presented to two sections

of a large introductory mass communications class for

attitude and personal relevance measures.

Two questions accompanied each issue statement, the

first asking for the subjects' attitude toward the issue,

the second for the personal relevance of that issue. Both

were measured on 1-to-5 scales as provided by the machine-

read answer sheets (bubble sheets). An issue attitude

could range from a 1 for "disagree" to a 5 for "agree."

Relevance measures could range from l-to-5 as well, a 1

representing "not relevant" and a 5 "very relevant."

There were two reasons for this phase of the

research. First, two issues rated as low and high

relevance by students themselves would be presented to the

subjects in the later study. Second, those issues had to

be relatively unrelated to uncertainty orientation. An

association between the issues and uncertainty orientation

might have occurred due to a number of factors, the most

likely being ideological differences between certainty-

oriented and uncertainty-oriented persons.

Uncertainty Orientation

Measures of uncertainty orientation resembles those

used to assess achievement-related motives (Atkinson,

1958; Atkinson & Feather, 1966). Frederick and Sorrentino

(1977) developed a projective measure of n-uncertainty

from which to infer to relevance of uncertain situations.

Early attempts to unconfound authoritarianism and

acquiescence response sets (Christie, Havel, & Seidenberg,

1958; Couch & Keniston, 1960) have led many to question

the usefulness of the construct. Sorrentino and his

colleagues adopted the measure of authoritarianism of

Byrne and Lamberth (1971) to infer the relevance of

certain situations. Previous research has shown that

people high in authoritarianism are more intolerant of

ambiguity (Kirscht & Dillehay, 1967) and have less

experience with uncertain situations (Kelman & Barclay,

1963). Interrater reliabilities for the n-uncertainty

projective measure have been found to be above .90 by

Frederick and Sorrentino (1977) and Sorrentino (1977) had

similar results for the authoritarianism measure.

Generally, sentence leads are used for the projective

test, with the open-ended responses coded for certainty

and uncertainty imagery (see Fredrick & Sorrentino, 1977,

for detailed scoring instructions).

The acquiescence-free measure of authoritarianism

results was used with the n-uncertainty measures as

opposing elements of uncertainty orientation. This

resultant measure is used, according to the authors,

because persons scoring low on n-uncertainty may not

necessarily be high in certainty orientation, and vice

versa. The method is not unusual in the motivational

literature, for example, where two uncorrelated measures

are used to measure achievement motivation (Atkinson &

Feather, 1966). Briefly, z-scores are computed on the two

measures for a resultant measure and a tertile split of

the scores are used. The middle third is typically

excluded from further analysis because of inconsistent

results with this middle group.

Projective measures are time consuming and in some

research settings, such as surveys, not feasible. A 35-

item questionnaire, still being validated with the

projective measures, has been produced by Sorrentino and

his colleagues and is used in the research reported here.

This instrument contains two sets of questions. The first

set is designed to measure need for uncertainty, using

questions rather than the previous projective measure.

The second set of questions is the same as the original

instrument and is designed to measure authoritarianism.

Sorrentino and his colleagues believe the new

instrument essentially taps the same construct and initial

testing is positive, though it is also thought that the

pen-and-pencil measure does not have the explanatory power

of the projective method (Sorrentino, R. M., personal

communication, 1989, 1991).1

Independent and Dependent Variables

Following the issues on the Study 1 instrument were

measures of media reliance, attitudes about poll

credibility, importance of opinions of others,

demographics, political ideology, and uncertainty

orientation. Finally, two of the issue statements were

presented again for attitude measures (comprehensive exams

and agricultural pesticide reforms). Bogus poll results

were built into the issue statements and these versions

randomly assigned to subjects.

One manipulation involved the poll sample. Some

subjects received a version that said the poll was of

college students, others received a version that stated

the poll was of Americans. A second manipulation had

subjects either being told a single poll result or a

version that added that a trend was seen in the direction

of the poll result. A third factor was dropped due to

question wording problems, reducing the number of subjects

available for analysis.2 To summarize, subjects received

both issues (senior comprehensive exams and agricultural

reforms in pesticide use) and were randomly assigned to

one of four conditions: (1) Poll of Students; (2) Poll of

Americans; (3) Trend-Poll of Students; and (4) Trend-Poll

of Americans. Uncertainty orientation was added to this,

making it a 2 (uncertainty- or certainty-oriented) X 2

(poll or trend and poll) X 2 (students or Americans as

sample) analysis design. A control group was also

included, these subjects receiving only the issue

statements a second time with no poll results.

An example of the dependent variable question is one

from the poll-student condition, which stated: "A public

opinion poll shows college students favor university

senior comprehensive exams to graduate, with 76 percent of

college students in favor. What do you think?" Subjects

could disagree or agree on the same l-to-5 scale as above.

The second question in this condition stated: "The same

poll shows college students in favor of agricultural

reforms in pesticide use, with 78 percent of college

students in favor. What do you think?" The slight

difference in percentages of the bogus poll results was

used to decrease suspicion. These issues differed

significantly in relevance as reported by the subjects,

with comprehensive exams being slightly higher in

relevance (M = 3.9) than pesticide reform (M = 3.6, t(190)

= 2.6, p<.01).


One hundred and twenty-eight students in the two

sections of an introduction to mass communication class

took part in Study 1 on March 5, 1991. Nearly 63 percent

were female. A plurality listed themselves as advertising

majors (45, or 36.3 percent).

Issue Statements

Students were first provided the issue statements

(Table 6-1 contains the issue statements and Table 6-2 the

attitude and relevance mean scores and the association of

these scores with uncertainty orientation). Subjects were

presented 17 issue statements followed by the attitude and

relevance questions after each issue. A bogus

agricultural act received the lowest relevance score (M =

2.4), while worries about the recession and-job

availability produced the highest score (M = 4.6).

Relevance scores differing by about one-tenth of a point

are significantly different at the .05 level.

The recession/jobs issue met the criteria of both

high relevance and apparently no relationship to

uncertainty orientation. Of the "low relevance" issues,

the best choice appeared to be the capital gains tax (M =

2.9), also not significantly related with uncertainty

Table 6-1

Issue Questions and Mean Relevance and Attitude Scores

Relev. Attit.

1. Congress should pass the recently-
proposed Agricultural Act of 1991. 2.4 3.9

2. The United States can help its own
economy and that of poor nations with
increased use of the International
Monetary Fund. 2.9 4.4

3. The capital gains tax should
be changed. 2.9 4.4

4. Mandatory AIDS testing should be
required for some people. 3.1 3.1

5. Weak federal regulations are to
blame for the U.S. savings and
loan crisis. 3.1 4.1

Table 6-1--continued

Relev. Attit.

6. Reforms are needed in the use
of pesticides in agriculture. 3.7 3.0

7. Taxes are so high that they must
be cut even if it means a reduction
of public services. 3.7 4.4

8. Senior comprehensive exams for
graduating seniors is a good idea. 3.9 3.3

9. Women should have the right to an
abortion if they wish. 4.0 3.2

10. We should support the allied
efforts in the Persian Gulf War. 4.1 3.3

11. U.S. universities should toughen
education requirements to be
competitive in the world. 4.2 3.8

12. President Bush is doing a good job. 4.2 2.4

13. Americans need to sacrifice more
in order to cut our energy use. 4.2 3.0

14. We should do more to protect the
environment, even at the cost
of some businesses. 4.3 4.2

15. Jobs should go to the best person
regardless of race or sex. 4.6 2.5

16. The proposed University of Florida
tuition increase is not necessary. 4.6 4.0

17. The recession means good jobs will be
hard to come by for the next few years. 4.6 4.6

Note Issues in boldface are those selected for use in
Study 2. Mean scores of approximately 0.1
difference are significantly different at the .05

Table 6-2

Attitude and Relevance Scores by Uncertainty Orientation

Attitude Relevance

Issue C-O U-O t C-O U-O t
























































































































Note C-0 is

certainty-oriented, U-0
d. a p<.05; b p<.01; c

is uncertainty-

orientation. These two issues were significantly

different in personal relevance (t(191) = 20.1, E<.001)

and were selected for further use in the later study.

Uncertainty Orientation

A subset of the questions provided by Sorrentino and

his colleagues was used after factor analyses in bothStudy

1 and Study 2 were inconclusive. The subset of six need

for uncertainty and six authoritarianism items were used

(a more detailed discussion of the factor analyses is

provided in Chapter 7).

The need for uncertainty scale had a mean of 4.0 (SD

= 0.7) and the authoritarianism scale a mean of 2.8 (SD =

0.7). As expected, the two had a significant negative

association (r = -.22, p<.001). The need for uncertainty

measure demonstrated fair internal consistency

(standardized Cronbach's Alpha of .77). The

authoritarianism scale was less internally consistent,

with a Cronbach's Alpha of .59.

To create a resultant uncertainty orientation

measure, scores on the authoritarianism scale were

subtracted from the need for uncertainty scale, resulting

in a scale that ranged from -1.5 to 3.5 (M = 1.3, SD =

1.1). Individuals high in this resultant measure are

considered uncertainty-oriented, while those low are

considered more certainty-oriented. As noted in earlier

discussion of the construct, the middle tertile is

typically dropped from further analysis, leaving the

extremes of uncertainty-oriented and certainty-oriented

persons but reducing the number of subjects available.

This procedure was not done in this research. Instead,

the middle tertile is included to keep these subjects in

the analysis.

Uncertainty-oriented subjects were more liberal (M =

3.7) than certainty-oriented subjects (2.7, t(82) = 3.9,

p<.001) However, uncertainty-oriented persons did not

differ from certainty-oriented persons on newspaper or

television reliance, the credibility they placed on polls

reported in newspapers, or in the importance they place in

other people's opinions.


Analysis of variance was used to test for significant

interactions. There are three main assumptions in the use

of analysis of variance: that there be normally

distributed treatment populations, homogeneity of error

variance, and independence of error components (Keppell,

1982). The F-distribution is robust against violations of

the first two conditions, though such violations are more

serious in situations with unequal cell sizes. In

general, Monte Carlo studies have shown large sample sizes

will offset any moderate violation of normality, with

significance usually misinterpreted by the F-test by only

a small amount (for example, a reported significance of

.05 when actually .06).

Various statistical tests are available to compare

variances (e.g., Hartley, Cochran, and Bartlett), though

as Keppell (1982) notes, many share in common a

sensitivity to departures from normality as well as

homogeneity of variance. That is, though experimental

groups may demonstrate homogeneity of variance, a lack of

normality will result in these homogeneity statistical

tests reporting a lack of homogeneity among groups. Use

of the SPSS-X statistical package confines the tests to

the Bartlett-Box F, Cochran's C, and Hartley's F max.

(SPSS-X, 1988). Examination of Study 1 results revealed

no threats to the assumptions of homogeneity of variance

or normality. This allows us to confidently examine the

unequal cell sizes found in this study. The third

assumption, independence of error components, means that

treatment groups are in no way related to any other

observation in the experiment. Random assignment, used in

the experiment reported here, achieves independence.

Tests of statistical significance are at the

traditional probability level of .05, though some results

below the .10 level are reported when they provide insight

into relationships.

Attitudes on Two Issues

As noted earlier, a pair of issues were selected to

help study the effects of two poll story manipulations.

The first manipulation had subjects receiving either a

single-shot poll result or a poll with additional

information that the trend of opinion was also in that

direction. The second manipulation was of the poll

sample, with the poll either being of Americans in general

or of students. A control group with no manipulations was

asked for an opinion.

No significant interactions were found on the

comprehensive exams issue (see Table 6-3 for the F-table,

Table 6-4 for the cell means). Only one near-significant

two-way interaction was found on the agricultural

pesticides issue, that between uncertainty orientation and

the student versus American poll manipulation (F(2,81) =

2.5, p<.09). This result offers little in terms of

interest since it is apparently the "middle tertile" group

causing the relationship.

No main effects were found between the types of poll

manipulations. However, those in the control group showed

less agreement with pesticide reforms (M = 3.6) than those

receiving any of the poll conditions (M = 4.1, t(116) =

2.3, E<.02). However, when looking at the more relevant

comprehensive exams issue, no significant difference was

found between persons in the control group and those

receiving some kind of poll condition. In general,

comparisons with the control group showed poll influence

only on the less relevant issue (see Table 6-7).


Two issues -- jobs during a recession and the capital

gains tax -- were selected for Study 2 based on the

requirements that they differ in relevance and that the

Table 6-3

F-Table of Analysis of Variance of Poll Manipulations on
the Agricultural Pesticides Issue

Sum Of Mean Sig.
Source of Variation Squares DF Square F Of F

Main Effects 3.5 4 0.9 1.0 .39
1 Poll vs. Trend 0.4 1 0.4 0.5 .50
2 Students vs. Americans 0.1 1 0.1 0.1 .75
3 Uncertainty Orientation 3.0 2 1.5 1.8 .18

2-Way Interactions 5.5 5 1.1 1.3 .27
1 By 2 0.4 1 0.4 0.4 .51
1 By 3 0.8 2 0.4 0.5 .62
2 By 3 4.2 2 2.1 2.5 .09

3-Way Interactions 0.8 2 0.4 0.4 .62
1 By 2 By 3 0.8 2 0.4 0.4 .62

Explained 9.9 11 0.9 1.1 .40

Residual 68.3 81 0.8

Total 78.1 92 0.8

attitude and relevance measures not be associated with

uncertainty orientation. Study 1 included a pilot

experiment investigating whether using students as a

sample in the bogus poll would be more influential than

Americans in general, and whether a report of a single

poll or poll with reports of a trend would be more

influential. Few differences were expected given the mild


Table 6-4

Attitude about Agricultural Pesticide Reforms by Poll
Manipulations and Uncertainty Orientation.

Poll Trend

Cert.-Oriented 4.0 4.2
Students Polled N = 10 N = 9

Moderate U-O 3.8 3.2
Students Polled N = 4 N = 5

Uncert.-Oriented 4.4 4.5
Students Polled N = 10 N = 8

Cert.-Oriented 4.8 4.1
Americans Polled N = 4 N = 7

Moderate U-O 4.2 3.8
Americans Polled N = 9 N = 10

Uncert.-Oriented 4.0 4.0
Americans Polled N = 10 N = 7

Note. Interaction F(2,81) = 0.4,



F-Table of Analysis of Variance of Poll Manipulations on
the Comprehensive Exams Issue

Sum Of Mean Sig.
Source of Variation Squares DF Square F Of F

Main Effects
1 Poll vs. Trend
2 Students vs. Americans
3 Uncertainty Orientation

2-Way Interactions
1 By 2
1 By 3
2 By 3

3-Way Interactions
1 By 2 By 3



1.4 2
1.4 2








0.3 .71
0.3 .71



15.9 11 1.4

173.1 81




189.0 92 2.1


Table 6-6

Attitude about Senior Comprehensive Exams by Poll
Manipulation and Uncertainty Orientation.

Poll Poll/Trend

Cert.-Oriented 2.1 2.2
Students Polled N = 10 N = 9

Middle Tertile 2.3 3.0
Students Polled N = 4 N = 5

Uncert.-Oriented 2.6 2.6
Students Polled N = 10 N = 8

Cert.-Oriented 2.0 3.3
Americans Polled N = 4 N = 7

Middle Tertile 2.0 2.6
Americans Polled N = 9 N = 10

Uncert.-Oriented 2.9 3.1
Americans Polled N = 10 N = 7

Note. No significant interaction.

Table 6-7

Mean Attitudes (SD) Toward Two Issues by Poll/Trend and
Student/American Manipulations

Type of Issue

Comp. Exams Pesticides

Poll 2.4 (1.4) 4.2 (0.9)

Trend 2.7 (1.5) 4.0 (1.0)

Students 2.4 (1.4) 4.1 (1.0)

Americans 2.7 (1.4) 4.1 (0.9)

Poll By Students 2.3a (1.4) 4.1 (0.9)

Poll By Americans 2.4 (1.4) 4.2 (0.9)

Trend By Students 2.5 (1.4) 4.1 (1.0)

Trend By Americans 3.0ab(1.5) 4.0c (0.9)

Control 2.4b (1.4) 3.6c (1.1)

Note. Column entries sharing a superscript significantly
different at the .05 level.


Because a particular type of computer-readable answer
sheet was used, the scales had to be a 1-to-5 measure
instead of the usual six-point scale provided for
uncertainty orientation. Both scales were presented in
terms of statements where respondents could either
answer a 1 for "disagree" across to a 5 for "agree."
Responses of 2 or 4 were for "somewhat disagree" or
"somewhat agree," while a 3 was for "neither disagree
or agree." The typical z-score method of previous
uncertainty orientation research was not used since


both scales making up the variable have the same range
(l-to-5) and the reporting of the means is perhaps more
meaningful than that of z-scores.

2 A third factor manipulated the direction of the poll
result. Half of the subjects were to be told the poll
supported senior comprehensive exams and reforms in
agricultural pesticide use, while half were to be told
the poll shows most people against the issues. A
problem with question wording on the "against"
manipulation, making it unclear what was being asked,
was discovered after a first session with students.
These 35 cases were dropped from analysis and a second
session with another set of students did not include
versions with this manipulation, leaving a total number
of subjects at 128.


The News Articles

News Article Preparation

Two short news articles about the capital gains tax

and jobs were prepared with a desktop publishing system to

resemble genuine newspaper articles (see Appendix A for

versions of each article). Both articles contained two-

line headlines and bogus "jumps" informing the reader to

continue reading on another page. In addition, a breakout

quote or piece of information was presented in larger

format to make the bogus poll results more accessible.

The articles were written in a similar fashion as to

structure, the placement and length of quotations, the use

of sources, and placement of manipulated information.

The Manipulation of Relevance

Issues, no matter how careful the presentation, may

differ in ways other than those intended or manipulated.

Despite the relevance differences inherent in the issues

as reported by Study 1 subjects, the first stimulus has to

do with personal relevance.

It is apparent that the capital gains tax and the

possibility of a tight job market are very different, not

only in the relevance people may place in them, but also

in the kind of issues they represent. The capital gains

tax is a largely political matter. The jobs issue can be

viewed politically, but it is likely subjects would

consider it on other, more personal grounds. While this

may seem to fit the "relevance" definition, it is possible

subjects might fit the jobs issue into a different kind of


Persuasion research in the cognitive response

paradigm typically manipulates relevance on the same issue

(see Petty & Cacioppo, 1986a, 1986b, for examples;

O'Keefe, 1990, for a discussion). By manipulating

relevance through telling subjects that the issue is or is

not important to them at the present, researchers are able

to avoid the pitfalls that come with differences between

issues other than relevance itself. The question is one

of theoretical and methodological control versus a more

"real-world" test. This dissertation attempts both by

providing subjects two issues apparently differing in

personal relevance and then manipulating relevance within

each issue. The manipulation occurred in the second

paragraph of each version of the article. In the "low

relevance" condition of the news article on the capital

gains tax, for example, readers were told that "experts

predict the final decision will not affect college

students." In the jobs and recession article with a "high

relevance" manipulation, readers were told that "analysts

predict most college students will be greatly affected."

The Manipulation of Consensus

The second manipulation had to do with perceptions of

consensus through use of poll results. Subjects could

receive one of three kinds of stories: a poll result in

the direction of the article's chief arguments, a poll in

the opposite direction, and a control group with no poll


For example, in the capital gains tax article,

subjects in the supporting poll condition--where poll

results were in the direction of the majority of arguments

presented--the fourth paragraph informed readers that a

"recent nationwide public opinion poll showed 81 percent

of Americans support the reduction of the capital gains

tax, with 15 percent against and 4 percent undecided." In

addition, the "blurb" or "pull out" added that "polls show

most Americans are for a reduction in the capital gains

tax." Subjects could also receive a version with the

polls results contradicting the article's argument. The

same sentence on poll results was provided, but instead

the majority argued either against a capital gains tax or

against the likelihood that jobs would be difficult to

find for college students. A control group received a

slightly shorter article (minus the paragraph on the poll

results) and with a "blurb" pulled from the article


No hypotheses were presented concerning possible

differences between how people process a supporting or

contradictory poll or differences in the persuasive

abilities of either type of poll. On the one hand, we

might expect an contradictory poll to promote central

processing, perhaps making the poll less influential

(similar to the argument of Mackie, 1987). On the other

hand, when the issue is of little relevance, we might

expect the contradictory poll to persuade people as a

consensus heuristic. This dissertation does not directly

probe the different processes, but does use this research

as a guide in explaining results. In general, it is

expected that the contradictory poll will act in a similar

fashion as the supporting poll.

Summary of Manipulations

Volunteers from an introductory advertising class

participated in Study 2 over a two-week period for extra

credit. Subjects could participate on March 19 or 20,

1991, for what will be called from this point Tl. During

this time subjects were measured on attitudes about the

two issues selected for further study, uncertainty

orientation, demographics, and other variables.

The design was a 3 (uncertainty versus middle versus

certainty oriented) X 2 (low and high relevance) X 3 (poll

result for or against or control) factorial for both


Subjects therefore could be assigned to any of six

conditions for each article (three poll manipulations and

two relevance manipulations). This meant a 6 X 6 or 36

different questionnaire versions. In addition, the order

of article presentation was randomized to ensure no order

effects would be present, resulting in 72 different

questionnaire versions. Based on their categorization on

uncertainty orientation, subjects were assigned to any of

these 72 versions. Each subject's computer-readable

bubble sheet from T1 was clipped to the assigned T2

version and subjects picked up these instruments at the T2

stage. To simplify presentation of the results, the

condition where the poll agrees with most of the news

articles will be called the "supporting poll" condition;

the condition in which the poll results are the opposite

of the arguments will be called "contradictory."

Before returning one week later for T2, uncertainty

orientation scores were determined and subjects randomly

assigned to experimental conditions. The T2 instrument

opened with additional media use questions then presented

the first news article followed by a thought-listing task

and attitude questions. A set of filler questions

followed, and then the second article was presented along

with its questions. The order of the two articles was

randomized to ensure no order effects occurred.

The chief independent variable measured at T1 was

uncertainty orientation, handled in the same way as in

Study 1 except for differences in scaling. Rather than a

l-to-5 scale, a different computer-readable answer sheet

version was used, this version ranging from 0 to 9. This

scale was adopted because it allows more movement on the

attitude measures, a greater range for uncertainty

orientation responses, and the possibility of answers

beyond the l-to-5 range constraints such as age or days a

week subjects read a newspaper.

Measurement of Variables

This study views attitudes as the evaluation of an

object and as representing a residue of experience;

different from moods, which are more temporary, attitudes

are relatively enduring though susceptible to change

(O'Keefe, 1990).

Attitude was measured by the same question at T1 and

T2 on both issues. The results of the T1 attitude measure

were used as a covariate and the results at T2 were used

as the dependent variable on each issue in an analysis of

covariance. This allowed the statistical control of

attitudes before the experimental manipulation.

Subjects were asked, on a 0-to-9 scale, to strongly

disagree or strongly agree to the following statements:

"The recession means jobs for college graduates will be

hard to come by for the next several years" and "The

federal government should cut the capital gains tax." The

questions at T1 were listed with a number of other

attitude measures to ease suspicion.

A second approach to this dependent variable, and one

perhaps more closely resembling traditional bandwagon

research, is to look for differences between T1 and T2.

To investigate this possibility, a change score was also

computed by subtracting the T1 attitude from the one at

T2. Therefore, a negative score represents a decrease in

support for the issue statement and a positive score an

increase in support.

Uncertainty Orientation

As noted briefly in the previous chapter, factor

analysis of the full 35-item index provided by Sorrentino

and his colleagues was problematic. In both studies, a

principal components and principal factor analysis using

varimax and oblique rotations arrived at a 10-factor

solution, some factors consisting of only a single item.

This analysis, however, could be interpreted as a three-

or perhaps two-factor solution (Appendix B contains all

questions with means and standard deviations, Appendix C

the 10-factor solution). The first factor appears to tap

the general construct of need for uncertainty, although

one authoritarianism question also loaded on this factor.

The second factor was made up of only three questions

asking about respect for authority, parents, and friends.

The third factor are all authoritarianism questions

generally aimed as beliefs and values.

Forcing a two-factor solution from all these

questions resulted in some authoritarianism items loading

on the need for uncertainty scale, though no uncertainty

items loaded on the authoritarianism factor.

After scrutiny of the factor analysis results and

using the larger Study 2 subject pool, a subset of 12

items was selected and a second set of factor analyses

conducted (factor analysis of the Study 1 pool resulted in

roughly the same solution). The six need for uncertainty

questions asked whether subjects enjoyed spending time

discovering new things or finding out why things happen.

The six selected authoritarianism measures focused on the

need for discipline, obedience, and respect for authority.

A principal factors analysis with a varimax rotation

showed the items explaining 36.5 percent of the variance

with a correlation between the factors of -.24. Both

scales demonstrated moderate internal consistency, with

Cronbach's Alpha of .81 for the need for uncertainty scale

and at .73 for the authoritarianism scale. The mean score

for need for uncertainty was 6.3 (SD = 1.4) and for

authoritarianism was 4.5 (SD = 1.6). As in Study 1, the

authoritarianism score for each subject is subtracted from

the need for uncertainty orientation score and an overall

index constructed. A tertile split was conducted with the

middle third included in analysis (see Table 1 for means

and factor loadings of the 12 questions used in this


The shortened instrument resulted in little movement

of subjects from being categorized as either uncertainty-

oriented or certainty-oriented under the original format

(X2(4) = 238.5, p<.0001). Only six persons found

themselves classified differently, three in each

direction. That is, three originally classified as

certainty-oriented were later classified as uncertainty-

oriented, and vice versa.

Table 7-1

Factor Analysis Results of Study 2 Uncertainty Orientation

Question Mean SD Fl F2

I enjoy spending time
discovering new things.

I like to find out why
things happen.

I like to fool around
with new ideas, even if
they turn out later to be
a total waste of time.

I enjoy thinking about
ideas that challenge my
views of the world.

I usually try to learn
about something I don't
understand, even if I
might end up in trouble
because of it.

I often put myself in
situations in which I
could learn something new.

What the youth needs most
is strict discipline, rugged
determination, and the will
to work and fight for family
and country.

6.5 1.8

6.9 1.8

6.1 2.0

6.9 1.9

5.6 2.2

6.5 1.7

4.6 2.3













-.17 .62

Table 7-1--continued

Question Mean SD Fl F2

There is hardly anything
lower than a person who
does not feel a great
love, gratitude, and
respect for his or her

Young people sometimes
get rebellious ideas, but
as they grow up they ought
to get over them and settle

Obedience and respect
for authority are the most
important virtues children
should learn.

No sane, normal, decent
person could ever think of
hurting a close friend
or relative.

Books and movies ought not
to deal so much with the
unpleasant and seamy side
of life; they ought to
concentrate on themes
that are entertaining
or uplifting.

Eigen Value

Percent Variance Explained

4.9 2.5

4.9 2.3

4.4 2.5

4.8 2.7

3.6 2.5















Note Loadings above are from the rotated structure
matrix of a principal factors analysis with oblique
rotation. The correlation between factors was -.24.


The experiment in Study 2 was conducted over a two-

week period.1 Two-hundred and eighty-nine subjects

participated in the first phase of the experiment (Tl).

Two hundred and sixty-three returned for T2, a loss of

approximately 9 percent. The average age of all subjects

was 20.6 (SD = 2.7). Fifty-nine percent of the subjects

were female and 41 percent male. Nearly half reported

their academic standing as being juniors (137, or 47.4



Analysis of covariance were the primary statistical

tests used in analysis.2 As in Study 1, examination for

assumptions found no violations of normality or

homogeneity of variance.

Study 1 was used to select issues of differing

relevance and with no relationship to uncertainty

orientation. Association of simple issues statements with

uncertainty orientation may differ in news articles as

compared to simple issue statements. T-tests of pretest

attitudes (before the manipulations) revealed no

significant difference in attitudes toward either issue

and uncertainty orientation.

Manipulation Checks

In a manipulation check of the poll condition,

subjects were asked to agree or disagree with a statement

that most Americans agree with the stance argued in the

article. If the manipulation was successful, those

receiving the condition where poll results favored the

article should provide a higher score (measured on a 0-to-

9 scale) than those in the control group, and those

receiving a poll in the opposite direction (a

contradictory poll) should demonstrate less agreement as

compared to the control group. T-tests show significant

differences in all cell means in the expected directions

(see Table 8-1).

Manipulations of relevance were mixed, however. A

question after each article asked subjects to agree or

disagree (on a 0-to-9 scale) as to whether "the article

was very relevant to me personally." As Table 8-2 shows,

the manipulation appears to have worked on the capital

gains tax article, with those receiving the low relevance

paragraph (N = 3.1) reporting less personal relevance than

those receiving the high relevance paragraph (M = 3.7,

t(261) = -2.5, R<.01). However, there was no significant

difference in relevance reported for the highly relevant

jobs issue manipulation, perhaps due to a ceiling effect

of the already high relevance associated with that

Table 8-1

Means (SD) of Manipulation Check Asking How. Likely Most
Americans Agree by Poll Type and Issue

Poll Manipulation Received

Issue Contradictory Poll Control Supporting Poll

Jobs For 4.8 (3.2) 6.4 (2.0) 7.6 (1.6)

Capital 3.7 (2.9) 4.6 (1.7) 7.1 (1.8)
Gains Tax

Note. All row entries are significantly different from
each other at the .01 level by t-tests.

Table 8-2

Means (SD) of Manipulation Check of Effects of
Experimental Manipulation of Personal Relevance on
Relevance Scores by Issue

Relevance Manipulation Received

Issue Low High

Jobs For
Graduates 7.1 (2.0) 7.4 (2.1)

Gains Tax 3.1a (2.1) 3.7a (2.3)

Note. Entries sharing a superscript are significantly
different at the .01 level by t-tests.

issue.3 Relevance scores associated with the jobs issue

are significantly higher than those associated with the

capital gains issue regardless of relevance manipulation.

Tests of Hypotheses

This study has three chief hypotheses. The first,

based on cognitive response models of persuasion (e.g.,

the elaboration likelihood model of Petty & Cacioppo,

1986a, 1986b), predicted that poll results would be more

influential on attitudes about less relevant issues as

compared to more relevant issues. The second hypothesis

had to do with explaining why the bandwagon or underdog

effects are seen in different studies. Following the

reasoning of Sorrentino (Sorrentino & Hancock, 1987), it

was hypothesized that uncertainty-oriented persons are

more likely to wonder why a minority thinks the way it

does, to "go with the underdog," to react against a poll

result as compared to certainty-oriented persons, while

certainty-oriented persons would be more likely to "jump

on the bandwagon," to be influenced by a poll. Third,

following the work of Sorrentino, et al. (1988), it was

hypothesized that in high-relevant situations, certainty-

oriented persons would be even more likely to be

influenced by a poll as compared to uncertainty-oriented

persons. The results section is organized around these

three hypotheses.

In the following results, T2 attitudes are used as

the dependent variable and the Tl attitudes treated as

covariates. For the capital gains tax issue, the Tl

attitude had a raw regression coefficient of 0.18. For

the jobs and recession issue, the raw regression

coefficient was 0.30. Where change scores are reported in

tables (usually in parentheses), these are the result of a

subtraction of T2 attitude from Tl attitude. Statistical

tests are for the attitude measures. Change scores are

provided as a guide for change between Tl and T2.

Finally, before moving to the results, recall that if

the poll is influential, those in the contradictory poll

condition should demonstrate lower attitude scores as

compared to the control group. On the other hand, an

influential supporting poll should result in attitude

scores greater than those seen among those in the control


Hypothesis 1 Poll Effects and Relevance

The first hypothesis predicts that polls are more

influential in situations of low relevance (Petty &

Cacioppi, 1986a, 1986b). However, poll influence was

found only on the more relevant jobs issue (see Table 8-

3). Those receiving the supporting poll showed greater

approval of the article's arguments (M = 6.8) than those

in the control group (M = 6.2, t(175) = -2.0, E<.05) or

those receiving the contradictory poll (M = 6.0, t(170) =

-2.5, p<.01). No significant differences were found on

the capital gains tax issue (F-tables for both issues are

found in Table 8-4 and 7-5).

Capital gains tax issue

Looking within issues, a two-way interaction between

poll manipulation and personal relevance was found on the

capital gains tax issue (E(2,242) = 3.2, p<.05). First,

looking at the contradictory poll, no influence was seen

Table 8-3

Means of T2 Attitudes (Change Scores) by Issue and Poll
Manipulation Received

Poll Manipulation Received

Issue Contradictory Poll Control Supporting Poll

Capital 4.8 (+0.3) 5.2 (+0.3) 5.1 (+0.5)
Tax N = 88 N = 87 N = 85

Jobs For 6.0 (-0.6)a 6.2 (-0.2)b 6.8 (+0.6)ab
Graduates N = 85 N = 90 N = 86

Note. Row entries sharing superscript are significantly
different at the .05 level by t-tests. Cell means
for the jobs issue are significantly higher than
those on the capital gains tax issue.

Table 8-4

F-Table of Analysis of Covariance of Capital Gains Tax
Issue by Poll and Relevance Manipulations and Uncertainty

Sum Of Mean Sig.
Source of Variation Squares DF Square F Of F


T1 Attitude

Main Effects

1 Poll Type
2 Relevance
3 Uncertainty Orientation

2-Way Interactions

1 By 2
1 By 3
2 By 3







1 31.4

1 31.4

5 1.9


6 8.8














3-Way Interactions

1 By 2 By 3






4 0.9

4 0.9

114.9 18 6.4

840.0 242 3.5

954.9 260 3.7

0.3 .90






T1 Attitude

Raw Regression Coefficient



Table 8-5

F-Table of Analysis of Covariance of Jobs Issue by Poll
and Relevance Manipulations and Uncertainty Orientation

Sum Of Mean Sig.
Source of Variation Squares DF Square F Of F


T1 Attitude

Main Effects

1 Poll Type
2 Relevance
3 Uncertainty Orientation

2-Way Interactions

1 By 2
1 By 3
2 By 3







1 114.4 29.3 .00

1 114.4 29.3


5 35.7 9.1 .00




8 4.4 1.1 .35




3-Way Interactions

1 By 2 By 3



4 8.8 2.3 .06

4 8.8 2.3

363.3 18 20.2 5.2



949.0 243

1312.3 261



Tl Attitude

Raw Regression Coefficient






among those receiving the high-relevance manipulation (M =

4.8 for both the control group and contradictory poll

conditions; see Table 8-6). Among persons receiving the

low-relevance version, those in the control group were

less approving of the issue (M = 5.6) than those receiving

the contradictory poll (M = 4.8, t(87) = -1.7, p<.05).4

An effect opposite the one hypothesized is found when

considering the supporting poll. For persons in the low-

relevance condition, those receiving the poll were less

approving (M = 4.8) than those in the control group (M =

5.6, t(85) = 2.2, p<.02). Among those told the issue was

Table 8-6

Means of T2 Attitudes (Change Scores) on Capital Gains Tax
by Poll and Relevance Manipulations

Poll Manipulation Received

Contradictory Poll Control Supporting Poll

Low Relevance 4.8 (+0.1)a 5.6 (+0.8)ab 4.8 (+0.1)b
N = 46 N = 42 N = 44

High Relevance 4.8 (+0.5) 4.8 (-0.1)c 5.4 (+0.9)c
N = 43 N = 45 N = 41

Note. Interaction F(2,254) = 3.2, p<.04. Entries sharing
a superscript are significantly different at the
.05 level.

of high personal relevance, those receiving the poll were

more supportative (M = 5.4) than those in the control

group (M = 4.8, t(84) = -1.8, p<.04).

Jobs and recession issue

There was no significant two-interaction between

relevance and poll found on the jobs issue (E(2.243) =

0.9, p<.41; Table 8-7). No differences were found between

the control group and those receiving the contradictory

poll regardless of relevance condition. Also, no

difference between the control group and the supporting

poll group was found among those in the high-relevance

condition (M = 6.8 and M = 7.0, respectively). However, a

Table 8-7

Means of T2 Attitudes (Change Scores) on Jobs by Poll and
Relevance Manipulations

Poll Manipulation Received

Contradictory Poll Control Supporting Poll

Low Relevance 5.3 (-1.2) 5.6 (-.O0)a 6.7 (0.0)a
N = 45 N = 46 N = 43

High Relevance 6.7 (+0.1) 6.8 (+0.8) 7.0 (+1.0)
N = 40 N = 45 N = 43

Note. No significant interaction. Entries sharing a
superscript are significantly different at the .05
level by t-tests.

t-test of the mean scores among those told the issue was

of little relevance showed those in the control group less

supportive (M = 5.6) than those receiving the supporting

poll (M = 6.7, t(87) = -2.5, p<.01).

Summary of hypothesis 1 results

The poll appeared to influence subjects on the more

relevant jobs issue, but not on the less relevant capital

gains tax issue.

Looking at within-issue relevance, the contradictory

poll influenced only those in the low relevance condition,

while the supporting poll influenced persons in both

relevance conditions, but not in a way that would be

hypothesized -- those in the low relevance condition

demonstrating more approval when shown the supporting

poll, while those told the issue was very relevant showed

less approval.

Hypothesis 2 Polls and Uncertainty Orientation

It was hypothesized that uncertainty-oriented persons

would be less persuaded by a poll than certainty-oriented


Capital gains tax issue

A two-way interaction near the traditional level of

significance was found for poll manipulation and

uncertainty-orientation (F(4,242) = 1.9, p<.10; see Table


Looking first at the contradictory poll, certainty-

oriented persons in the control group were more supportive

(M = 5.5) than those receiving the poll (M = = 4.4, t(54)

= -2.1, p<.03). The mean scores of uncertainty-oriented

persons were in the expected direction, but were not

statistically significant, with those showing greater

approval when in the contradictory poll condition (M =

5.3) than in control group (M = 4.7, t(55) = -1.1, p<.14).

Table 8-8

Means of T2 Attitudes (Chanae Scores) toward Capital Gains
Tax by Poll and Uncertainty Orientation

Poll Manipulation Received

Contradictory Poll Control Supporting Poll

Certainty 4.4 (+0.2)a 5.5 (+0.5)a 5.1 (+0.6)
Oriented N = 29 N = 29 N = 31

Moderate 4.7 (-0.2) 5.3 (+0.2) 5.6 (+1.2)
Uncert-O. N = 26 N = 36 N = 25

Uncertainty 5.3 (+0.7) 4.7 (+0.4) 4.5 (-0.2)
Oriented N = 29 N = 27 N = 29

Note. Interaction F(4,242) = 1.9, R<.10. Entries sharing
a superscript are significantly different at the
.05 level by t-tests.

No significant differences were found between the

control group and those receiving the supporting poll.

Jobs and recession issue

No significant two-way interaction between poll and

uncertainty orientation was found on the jobs issue

((4,243) = 1.8, E<.14; see Table 8-9). T-tests of the

expected differences found no significant poll influence

among certainty-oriented persons. Uncertainty-oriented

persons were influenced by both poll versions as compared

to the control group (M = 5.7), with those receiving the

Table 8-9

Means of T2 Attitudes (Change Scores) toward Jobs and
Recession by Poll and Uncertainty Orientation

Poll Manipulation Received

Contradictory Poll Control Supporting Poll

Certainty 6.3 (-0.7) 5.8 (-0.6) 6.6 (+0.1)
Oriented N = 27 N = 40 N = 18

Moderate 6.4 (+0.2) 7.3 (+1.0) 7.0 (+1.4)
Uncert-O. N = 34 N = 24 N = 33

Uncertainty 4.5 (-2.2)a 5.7 (-0.6)ab 6.8 (+0.1)b
Oriented N = 24 N = 27 N = 35

Note. Interaction F(4,243) = 1.8, ns. Entries sharing a
superscript are significantly different at the .05
level by t-tests.

contradictory poll showing a decrease in support (M = 4.5,

t(49) = -1.7, p<.05) and those receiving the supportive

poll increasing in support (M = 6.8, t(66) = -1.9, p<.02).

Summary of hypothesis 2 results

The capital gains tax issue offered some support for

the second hypothesis. Among those receiving the

contradictory poll, certainty-oriented persons were more

likely to move in the direction of the poll while

uncertainty-oriented persons were more likely to move in

the opposite direction, as compared to the control group.

No support for the hypothesis was found in the jobs issue.

Hypothesis 3 Polls. Relevance, and Uncertainty

The third hypothesis predicted that certainty-

oriented persons would be more sensitive to situations of

high relevance and would be more influenced by polls in

those situations.

Capital gains tax issue

No significant three-way interaction was found

between uncertainty orientation, poll type, and personal

relevance on the capital gains tax ((4,242) = 0.3, p<.90;

see Table 8-10). T-tests to explore the hypothesized

relationship found a pattern in comparing certainty-

oriented and uncertainty-oriented persons in the control

group and contradictory poll condition, though neither of

the findings reaches the traditional level of

significance. In these, certainty-oriented persons told

the issue was of personal relevance showed more approval

in the contradictory poll condition (M = 4.1) than those

in the control group (M = 4.9, t(55) = 1.1, p<.15).

Uncertainty-oriented persons in the high-relevance

condition, however, moved in the opposite

Table 8-10

Means of T2 Attitudes (Chanae Scores) Toward Capital Gains
Tax by Poll. Relevance, and Uncertainty Orientation

Contradictory Poll Control Supporting Poll

Cert.-Oriented 4.7 (+0.6)a 6.0 (+0.5)ab 5.1 (+0.5)b
Low Relevance N = 14 N = 13 N = 15

Mod. Unc-O 4.4 (-0.4) 5.4 (+0.7) 5.0 (+0.2)
Low Relevance. N = 19 N = 20 N = 11

Uncert.-Oriented 5.5 (+0.2) 5.7 (+1.6)c 4.3 (-0.3)c
Low Relevance N = 13 N = 9 N = 18

Cert.-Oriented 4.1 (-0.1) 4.9 (+0.5) 5.1 (+0.6)
High Relevance N = 15 N = 13 N = 14

Mod. Unc-O 5.4 (+0.3) 5.3 (-0.4) 6.1 (+1.8)
High Relevance N = 10 N = 16 N = 16

Uncert.-Oriented 5.1 (+1.1) 4.1 (-0.3) 4.8 (0.0)
High Relevance N = 18 N = 16 N = 11

Note. No significant interaction. Selected entries
sharing superscripts significantly different at the
.05 level by t-tests.

direction, with those receiving the contradictory poll

demonstrating less approval (M = 5.1) than those in the

control group (M = 4.1, t(32) = 1.5, p<.07).

Jobs and recession issue

A significant three-way interaction was found between

uncertainty orientation, poll type, and relevance on the

Table 8-11

Means of T2 Attitudes toward Jobs Issue by Poll.
Relevance, and Uncertainty Orientation

Contradictory Poll Control Supporting Poll

Cert.-Oriented 4.9 (-1.9) 5.4 (-1.4)a 7.1 (+0.2)a
Low Relevance N = 14 N = 19 N = 14

Mod. Unc-O 6.0 (+0.1) 6.7 (+0.4) 6.7 (+0.6)
Low Relevance N = 21 N = 7 N = 14

Uncert.-Oriented 4.4 (-2.9) 5.4 (-1.3) 6.3 (-0.6)
Low Relevance N = 10 N = 20 N = 16

Cert.-Oriented 7.9 (+0.6)b 6.3 (+0.3)b 6.0 ( 0.0)
High Relevance N = 13 N = 15 N = 11

Mod. Unc-O 6.8 (+0.3) 7.5 (+1.2) 7.4 (+2.2)
High Relevance N = 19 N = 17 N = 13

Uncert.-Oriented 4.6 (-1.4) 6.3 (+0.6) 7.3 (+0.7)
High Relevance N = 8 N = 13 N = 19

Note. Interaction F(4,243) = 2.3), p<.07. Entries
sharing a superscript are significantly different
at the .05 level by t-tests.

jobs issue (E(4,243) = 2.3, p<.06; see Table 8-11).

Instead of the hypothesized effect of high relevance on

certainty-oriented persons, those receiving the

contradictory poll were more approving of the jobs issue

(M = 7.9) than those in the control group (M = 6.3, t(26)

= 2.1, p<.03). No difference was found among certainty-

oriented persons in the high relevance condition either in

the control group of receiving the supporting poll. Poll

influence was also seen in situations of high relevance

for uncertainty-oriented persons (rather than certainty-

oriented, as hypothesized). Among uncertainty-oriented

persons told the issue was of high relevance, those in the

control group showed more approval (l = 6.3) than those

receiving the contradictory poll (M = 4.6, t(19) = -1.4,

p<.09) and less approval than those receiving the

supporting poll version (M = 7.3, t(30) = -1.2, p<.13),

though the last finding is not statistically significant.

Summary of hypothesis 3 results

Little support for hypothesis 3 was found on the

capital gains tax issue, but results from the jobs and

recession issue were opposite the direction predicted. On

the less relevant capital gains tax issue, certainty-

oriented persons told the issue was of high relevance and

who received a poll contradictory to the main thrust of

the news article were more influenced in the direction of

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - Version 2.9.7 - mvs