.meawquwwsaé .

.3:

1.... . 1:71;... ..
is V ‘ .

a v.3“...

3H3. .
, V 3?.
{Li}; 1 .81“.
4...}. :3 Kim .2 u. ‘
V ﬁTihé: .
V 1 “ﬁg." .159.
u. an... . xu 5.:
V . a ,
.mmmﬁu‘iu‘ﬁﬂa V
. V . 331?... 3 g V xi
{mm c: .
‘ . . V . "ﬁx? U'ﬁ1qCQ. «I L.
.twvvn. u .V
V 2.3.! V
‘ 1 I
.438. 3 A... V V
a

I 3.3

an. sgﬁr ¢. ‘
.ﬁﬁmﬁnﬁnﬁu
‘ n’v’..". .li'
3. .ﬁnzhkﬁw.
twig .

. i

)5: ...........

Ls kw...

 

II. I!’

n
. biz:
.silbcaif...
V iii-3‘. 15:35. .5;
. luxldu:3‘v.§ixvv .2111: 14.9 I...
5 if. ..

:15.

in. 5.3. 3..ch
2.2.3ratyu513 r.
59 3.. 5-..,

. th3 AX!

«3 list» ”:6.

: ¢ .. $3 a . 53....

4.3... 3.5:}..2Lu?
2.3.93? nip

:3:

all. 1‘ 11-.

 

 

 

.l. :3 « 71.1.}
V";

 

 

 

 

i‘i-lvli-
-. 39.3... g . V, La... ﬁkgﬁhﬁnﬂmﬁgg
.1 +._§.VV...( , é .?

.3. w I
ILIIAII! .v
ti... 4-
., . s... 5.2.: 1h.

This is to certify that the
dissertation entitled

EXEMPLIFICATION OF HEALTH MESSAGES IN THE MEDIA

presented by
TRACY R. WORRELL

has been accepted towards fulﬁllment
of the requirements for the

Doctoral degree in Communication

 

 

Major Professor's Signature

 

4&1qu f .181 3.006

Date

MSU is an Afﬁrmative Action/Equal Opportunity Institution

 

LIBRARY
Michigan State
University

 

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/CIRC/Date0ue.indd-p.1

 

EXEMPLIFICATION OF HEALTH MESSAGES IN THE MEDIA
By

Tracy R. Worrell

A DISSERTATION
Submitted to
Michigan State University

in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Communication

2006

ABSTRACT
EXEMPLIFICATION OF HEALTH MESSAGES IN THE MEDIA
By
Tracy R. Worrell

Research on exempliﬁcation demonstrates that exemplars are superior to base-rate
information in Shaping the perceptions that result from message exposure. Though the
extensive body of work in this area suggests that this effect is robust, Krupat et a1. (1997)
found that there are conditions in which an abstract base-rate may be more inﬂuential
than a concrete exemplar. The present study attempts to replicate and extend Krupat et
al.’s research in a health communication context. It examines the ability of information
utility to moderate the relative inﬂuence of exemplars versus base-rates under conditions
of high and low issue importance.

The paper begins by introducing exempliﬁcation theory and explicating concepts
dealing with issue importance, information utility, exemplars and base-rates. Logic from
this discussion is use to predict that issue importance and information utility will interact
to moderate the inﬂuence of exemplar discrepancy on perceptions of problem severity
and population estimates related to the issue of smoking related anxiety among college
students. A uniquely weak exempliﬁcation effect was predicted for conditions in which
issue importance and information utility were high. An experiment was conducted to test
this prediction.

A 2 X 2 X 2 mixed factorial design varied the exemplar discrepancy (extreme,
low) and information utility (high, low) in a message on smoking related anxiety

presented to respondents who varied on issue importance (high, low). After exposure,

respondents were asked to report their perceptions of the problems severity, to estimate
its pervasiveness in the population, and to indicate their behavior intentions. Analysis
produced a signiﬁcant three-way interaction, but the pattern of means associated with the
interaction was not consistent with predictions. Instead of the uniquely weak
exempliﬁcation effect predicted for when issue importance and information utility were
both high, the effect was unmistakably apparent in this condition. Means associated with
the interaction showed a different pattern.

In addition to the reported effect when issue importance and information utility
were high, the effect was also apparent when issue importance and information utility
were both low, the combination of conditions best suited theoretically for
exempliﬁcation. Unexpectedly, exempliﬁcation effects were absent under conditions
where issue importance was low while information utility was high, and where issue
importance was high while information utility was low. Thus, the exemplar effect was
absent when either issue importance or information utility was high. Discussion makes
note of the fact that the results are consistent with earlier evidence indicating that
exempliﬁcation occurred under both high and low conditions of issue importance;

however, the effect was not stable when information utility varied.

To my parents: for giving me the support and freedom to accomplish my goals. I never
would have had the courage to try if I didn’t always know that you’d be there to catch me
if I failed.

iv

ACKNOWLEDGEMENTS

I would like to acknowledge several people for helping me during my doctoral
work. First and foremost, thank you to my advisor, Dr. Ron Tarnborini, without your
generous time and commitment I know that I would not be where I am today. I know that
due to your guidance I have become, and will continue to become, a better researcher and
writer. To my “other” advisor, Dr. Charles Atkin, thank you for all of your help and
guidance for these past several years. I am also very grateful for having an exceptional
doctoral committee and wish to thank Dr. John Sherry and Dr. Frank Biocca for their
continual support and encouragement.

Thank yOu to the student assistants, both graduate and undergraduate, that helped
me in gathering data. I’m also extremely grateful to the staff for their help not only with
the copying and mailing of dissertation items but also the continued support and help
throughout my years at Michigan State.

Finally, I’d like to thank my family and ﬁiends. To my parents, Roger and Linda,
I swear I’m done with college now, thank you for all of your support and cheering. To
my brother Travis and sister Kim, thank you for always having my back. Harlan, thank
you for being ‘my favorite little boy.’ James, sorry this took me away from meeting you
right away. My friends, Tracy, Michelle, and Angie, without knowing that you would
always be there I don’t think I could have survived, thank you for being there and helping

me keep life in perspective.

TABLE OF CONTENTS

Introduction ................................................................................................ 1
Deﬁning Exemplars ........................................................................... 4
Deﬁning Base-rates ........................................................................... 6
Information Utility ............................................................................. 7
Issue Importance ............................................................................. 10
Comparing the Inﬂuence of Exemplar

and Base-Rate Information ...................................................... 12
Information Utility as a Moderator of

Base-rate lnfonnation’s Inﬂuence .............................................. 16
Practical Applications of Information Utility

as a Moderator Message Inﬂuence ............................................. 19
Issue Importance and Information Utility

in the Context of Health Communication ...................................... 20
Formal Hypotheses ........................................................................ 25

Method ................................................................................................ 28
Overview .................................................................................... 28
Pilot Study .................................................................................. 29

Participants ......................................................................... 29
Procedure ........................................................................... 30
Measures ........................................................................... 31
Main Study .................................................................................. 34
Participants ......................................................................... 34
Procedure ........................................................................... 35
Stimulus Materials ................................................................ 36
Measures ........................................................................... 38

Results ................................................................................................ 41
Pilot Study Phase One ..................................................................... 41
Pilot Study Phase Two ..................................................................... 47
Main Study .................................................................................. 52

Discussion ............................................................................................ 68
The Inﬂuence of Message Features on Problem Perceptions ........................ 68

Issue Importance .................................................................. 68
Information Utility ................................................................ 70
Exemplar Discrepancy ............................................................ 72
Interactions of Issue Importance, Information Utility,
and Exemplar Discrepancy ........................................................ 74
The Inﬂuence of Message Features on Behavior Intentions .......................... 81
Implications for Health Communication Research .................................... 82
Limitations .................................................................................. 86
Future Research ............................................................................ 91
Contributions to Exempliﬁcation Theory and Health Communication ............. 92

vi

Appendices ........................................................................................... 94

Pilot Study — Phase One Questionnaire ................................................. 95
Pilot Study - Phase Two Questionnaire ............................................. ["105
News Stories .............................................................................. 115
Main Study —Questionnaire ............................................................. 120
Perceptions of Credibility ............................................................... 132
Questions Measuring Respondent Knowledge of Base-Rates ...................... 134
Information Utility Items ................................................................ 136
Perception of Problem Severity ......................................................... 138
Behavior Outcomes ....................................................................... 140
Issue Importance — Measures Factor Loadings ....................................... 142
Information Utility - Measures Factor Loadings .................................... 144
Perception of Problem Severity - Items Used ....................................... 146
Behavior Intentions Scale ............................................................... 148

References .......................................................................................... 149

vii

LIST OF TABLES

Table 1: Percent Scoring High and Low on Issue Importance for Six Pilot Issues. . . . .....43
Table 2: Descriptive Statistics on Issue Importance for Six Pilot Issues ..................... 44
Table 3: Descriptive Statistics for Issue Importance Scale .................................... 53
Table 4: Results of t-tests on Information Utility Subscales .................................. 55
Table 5: Means and Standard Deviations for Study Variables ................................ 56
Table 6: Zero-Order Correlations between Study Variables .................................. 57
Table 7: Descriptive Statistics for High and Low Issue Importance ......................... 59
Table 8: Means Associated with Perceptions of Problem Severity Main Effects .......... 60

Table 9: Means Associated with Perceptions of Problem Severity Interaction Effects. . .60
Table 10: Means Associated with Population Estimates Main Effects ...................... 61

Table 11: Means Associated with Population Estimates Interaction Effects ............... 61

viii

LIST OF FIGURES

Figure 1: Perceptions of Problem Severity for Respondents

High and Low on Issue Importance ..................................................... 64
Figure 2: Interaction effect of information utility and issue importance

on perceptions of problem severity ...................................................... 66
Figure 3: Interaction effect of information utility and issue importance

on population estimates .................................................................... 67

ix

Introduction

Which would be more compelling; being told that 68% of college students have
anxiety disorders that are linked to smoking or hearing a story of one student’s mental
difﬁculties brought on by nicotine? Would one stand out more than the other? One is
statistical information; the other is a story from one person’s life. Exempliﬁcation theory
(Zillmann, 2002) examines the role of these two factors, base-rates and exemplars, in
communication messages. Much of what currently exists in exempliﬁcation research is
examining the viewer’s consumption and absorption of information put forth by the news
media (e.g. Aust & Zillmann, 1996; Berger, 2000; Berger, 2002; Brosius & Bathelt,
1994; Daschmann, 2000; Gibson & Zillmann, 1994; Gibson & Zillmann, 1998; Gunther
& Christen, 1999; Newhagen & Reeves, 1992; Perry & Gonzenbach, 1997; Perry &
Gonzenbach, 2000; Zillmann, Gibson, Sundar, & Perkins, 1996). These studies
predominantly focus on viewer differences attending to, believing, and retaining base-rate
versus exemplar information. This study brings exempliﬁcation into the health
communication message arena examining the role of base-rates and exemplars in anti-
smoking messages targeting young adults.

The premise of exempliﬁcation is that individuals not only pay more attention to
but are inﬂuenced more by exemplars than by base-rate data (Gibson & Zillmann, 1998).
News reports often use exempliﬁcation to illustrate news issues of large scope through
the use of individual examples that are to some degree representative of the larger issue,
and through this draw an audience to their stories (Gibson & Zillmann, 1998).

Much of exempliﬁcation theory is based on logic arguing that there are

evolutionary advantages that stem from the ability to group events together. Humans do

this by siﬁing through experiences to group those that seem to go together. When they do
this, they code the occurrence of events in some basal quantitative manner that allows
them to make judgments (most often nonconsciously) of how ﬁequently different events
occur (Zillmann, 2000). These judgments are said to be based on two cognitive devices:
the availability heuristic and the representativeness heuristic. The availability heuristic
tells us that judgments of social phenomena are greatly inﬂuenced by the ease with which
information comes to mind (Gibson & Zillmann, 1998). According to Rothbart, Fulero,
Jensen, Howard, and Birrell (1978) availability heuristics can be a useful tool for
assessing frequency or probability of an event. Individuals that employ the availability
heuristic evaluate the frequency of events based on the quickness with which pertinent
instances come to mind (Zillmann, et al., 1996). The representativeness heuristic is a
special case of availability. It stipulates that abstract base-rate information plays little role
in quantitative judgments about event populations. Instead, these judgments are based on
the sample of more concrete exemplars that are available to the individual at the time of
decision making. Exempliﬁcation theory is a simple combination of these heuristics. It
posits that since exemplars come to mind more easily than base-rates when accessing
information, available exemplars will dominate base-rate information when making
judgments of event populations.

The two components of exempliﬁcation theory, exemplars and base-rates, have
been examined both separately and comparatively for decades (e. g. Bar—Hillel, 1980; Bar-
Hillel & Fischhoff, 1981; Gilbert & Jones, 1986; Hayes-Roth & Hayes-Roth, 1977).
Separately scholars have examined things such as exemplars encoded in memory (Hayes-

Roth & Hayes-Roth, 1977), exemplar retrieval from category membership (Castelli, et

al., 2004), and the base-rate fallacy (Bar-Hillel, 1980). Comparatively scholars have
focused more on determining if either base-rates or exemplars can alter apprehension
(Berger, 2000), affect recall of information (Berger, 2002; Heuer & Reisberg, 1990), and
inﬂuence the effect of news information (e. g. Aust & Zillmann, 1996; Gunther &
Christen, 1999; Zillmann, et al., 1996). In all, empirical evidence from research on
exempliﬁcation shows that exemplar effects are robust, having been observed to
inﬂuence a variety of different perceptual and attitudinal outcomes across an assortment
of issues in news and political media research examining several different media (for
review see Brosious, 2003).

Most research on the inﬂuence of exemplars has focused on message effects in a
news or political context. The ﬁndings from this work generally show that exemplars
outweigh base-rate information in terms of their inﬂuence on receivers (e. g. Aust &
Zillmann, 1996; Brosius & Bathelt, 1994; Gibson & Zillmann, 1998). Yet, Zillmann
(2002) contends that it would be premature to conclude that base-rate information is
always inconsequential. In fact, one study by Krupat, Smith, Leach, and Jackson (1997)
shows a situation in which base-rate information was more inﬂuential than an exemplar.
In this instance, however, the investigation was set in a product purchase context.
Zillmann (2002) interprets this research to suggest that reliable quantitative information
can have a greater inﬂuence on judgments than an exemplar in situations where the
information has great diagnostic signiﬁcance (in this case for purchasing a car). If
Zillmann is correct, understanding the extent to which information utility can act as a
moderator of an exemplar’s inﬂuence would have important theoretical and practical

implications for research on message effects. The present study attempts to replicate and

extend Krupat et al.’s (1997) research to examine the importance of information utility in
determining the relative inﬂuence of exemplars versus base-rate information in
persuasive media messages. The Krupat et al. study looked only at information use in
evaluating automobiles; the present study attempts to examine the extent to which
information utility can moderate the relative inﬂuence of exemplars versus base-rate
information in other circumstances. The paper begins by explicating concepts and
processes related to exemplars, base-rates, and information utility. Based on this
discussion, it presents a rationale relating base-rates, exemplars, and information utility
and offers several hypotheses to test this logic. These hypotheses are then tested with an
experiment examining the effect of information utility on exempliﬁcation theory.
Deﬁning Exemplars

An exemplar is an instance of an event population that shares essential features
with all other instances ﬁ'om the group of events that is deﬁned by those features. In a
sense, exemplars are case reports used to represent characteristics typical of a group of
events (Zillmann & Brosius, 2000). For example, a news story reporting on increased
anxiety problems caused by smoking might try to make a point about the larger issue of
health problems related to smoking by presenting an exemplar. If the story is about
smoking the report might describe in detail the story about one particular individual that
has anxiety brought on by smoking. The exemplar may be presented in order to convey
drama and emotion in ways that verbal or written description of the larger event
population or social issue cannot (Aust & Zillmann, 1996). At the same time, questions
persist about the extent to which the drama and emotion associated with the particular

story correspond to the subject at hand. Even more to the point, one must question the

extent to which objective features of the particular example accurately and completely
represent the larger event population.

Commonly, exemplars are illustrative representations of information. This means
that rather then give speciﬁc information that details an issue, an exemplar will provide a
demonstration of an event. They may be looked at as testimonials or qualitative evidence
of an event (e. g. using a victim of anxiety or another health related issue brought on by
smoking) that adds to more objective information that may be given to quantify an
occurrence. Still, in order to function as an exemplar, the particular example must share
essential attributes of the larger issue to make it classiﬁable as a member of the same
population of events. In this regard, not all examples function as exemplars. Examples
that have no features in common with other events cannot be considered a member of that
population of events, and cannot serve as an exemplar for that population (Zillmann,
2002). As such, only events that share several deﬁning features with other events from
the population can serve as exemplars. Yet, even when exemplars share some of the
primary features of an event population, variation in secondary attributes can have
substantial consequence.

Exemplars do not always reliably exemplify phenomena (Zillmann, 2002). This is
the case when an exemplar that is being used is atypical of the phenomenon being
illustrated along primary or secondary attributes of the event population. Exemplars have
been referred to as illustrating a “case in point,” which is only beneﬁcial “for the
exempliﬁcation of a homogeneous event group” (Zillmann & Brosius, 2000, p. 5). In
event populations where cases contain considerable variation in secondary features,

certain secondary attributes presented in an exemplar have the potential to distort

perception. For example, in the aforementioned smoking illustration, if the particular
example used to illustrate the increased anxiety problems brought on by smoking
happened to end with the death of a victim, the presence of this atypical but highly salient
attribute of anxiety is likely to have a strong inﬂuence on receiver perceptions and
behavioral intentions. In this case, the inﬂuence of the exemplar is likely to be quite
different than what we would expect from exposure to messages featuring only the base-
rate information.

Exemplars are ﬁ'equently used in conjunction with base-rates as case illustrations
of event populations (Brosius & Bathelt, 1994). In these instances exemplars have been
found to enhance or detract from given base-rate information as a function of the extent
to which the salient features of the exemplar are representative of the event population.
As with news reports, exemplars are often added to provide a description of an event
being discussed within a report and are chosen for their entertaining qualities (Gibson &
Zillmann, 1994). When the collection of included exemplars is selected in a manner that
will accurately represent variation in secondary event attributes, we should expect their
inclusion to augment the base-rate information. When the included exemplars are not
selected in such a manner, this increases the likelihood that exemplars will detract from
outcomes consistent with what we might expect from base-rate information.

Deﬁning Base-rates

Whereas exemplars rely on speciﬁc case reports to describe and illustrate
essential features of an event group that may or may not accurately represent secondary
features, base-rates present descriptions limited to certain general features of the event

population. Notably, these more limited descriptions are ascertained using systematic

methods expected to produce more reliable representations of the population. Base-rate
information often comes in the form of a numerical representation such as the frequency
with which an event occurs. Typically it is thought of as quantitative information about
population events (Zillmann & Brosius, 2000) or a general description “of the number of
people or things in a given social phenomenon” (Zillmann, et al., 1996, p. 427). In the
smoking example, base-rate information might quantify the number of smokers
diagnosed with anxiety. Such base-rates can be precise (speciﬁc percentages) or vague
(e.g. in a majority of instances), but in either case they will not include illustrations of the
events. Generally speaking, this type of objective base-rate information is less partial and
more truthful than exemplars (Zillmann, 2002).

Base-rates are often said to be less illustrative or vivid than exemplar information.
The fact that base-rates are less illustrative than exemplars often leads to researchers
examining the base-rate fallacy, which is an individual’s tendency to ignore objective but
abstract base-rate information in favor of less relevant but more concrete exemplars that
are readily available in memory (Scheider, 2001). These more entertaining exemplars
draw the attention of many individuals despite the fact that they provide representations
that are potentially atypical and might actually produce misperceptions of the event
population under consideration.
Information utility

Information utility can be understood as the extent to which certain facts or
knowledge help people comprehend “happenings of interest” (Festinger, 1957).
Knobloch, Carpentier, and Zillmann (2003) suggests that information utility is knowledge

that “affords individuals a better orientation in, as well as more effective means of acting

on, their physical and social environments” (p. 92). In other words information utility
refers to how useful certain facts are for helping us make decisions in response to
different environmental challenges. For example, in consumer decision making research,
information that helps a consumer make a difﬁcult product choice could be thought of as
having high information utility (Huh, DeLorma, & Reid, 2004). Choices on issues such
as which prescription drug to use can pose both challenges that are large or small to
different people. In either case, information that relates directly to the health disorder or
the drugs under consideration are likely to be seen as particularly useful in the potential
purchase decision, whereas information unrelated to the drug or its effects would have
little value. Information utility research examines how expectations of message use are
formed and modiﬁed by attempts to organize available information in order to achieve a
speciﬁc objective (Wilton & Myers, 1986). Empirical studies of information have found
that individuals will assess information on the basis of three features leading to increased
utility: (l) the perceived relevance of information to an assigned task, (2) the amount of
new information, and (3) the perceived credibility of the information (Moenaert &
Souder, 1996; Wilton & Myers, 1986). Wilton and Myers (1986) describe perceived
relevance in terms of information that provides an individual with signals that are strong
and pertinent to their long-term memory; something particularly important when
information is sought to address environmental challenges originating from internal
needs. Credibility refers to how “undistorted” information is thought to be (Thayer,
1968), or the believability of stimuli from a source (Wilton & Myer, 1986). If the
information provided is from a source that is trustworthy, users will ﬁnd that information

more credible than information from other unknown or unbelievable sources. Finally,

amount of new information refers simply to whether or not an individual has been
exposed to the same knowledge previously, and how much of the information is novel.

The present study starts with the assumption that information utility will increase
message effectiveness in a health communication context. This expectation is consistent
with research from several ﬁelds showing that the credibility, relevance and new
information in a message govern its inﬂuence on attitude and behavior. Persuasion
literature shows that the attributes that increase message credibility are likely to reduce
counter arguing (Buda & Chamov, 2003) and increase message effect (Thompson, 1975).
Investigations on the beneﬁts of new information show that its inclusion makes a
message more appealing, and increases the likelihood that the message will be processed
and produce persuasion (Trenholm, 1989). Finally, research on relevance not only shows
that a message’s ability to generate relevant thoughts increases persuasive effects
(Trenholm, 1989), but the role of relevance is central to the health context studied here.
According to the health belief model, relevance will increase message effects,
particularly with regard to health recommendations (Perloff, 2001). People are more
persuaded by messages that have costs and beneﬁts that relate to them directly.

The inﬂuence of information utility can also be seen in the selective exposure
behavior of individuals, where the increased perceived utility of information is thought to
result in a greater drive to seek that information (Sears & Freedman, 1967). When
physical or social environments confront them with uncertainty, individuals may attempt
to seek out information that helps to explain the phenomenon (Berlyne, 1954). In this

sense, information utility has been found to be an essential determinant in the selective

exposure of information, and an important factor in attempts to understand the inﬂuence
of exemplars and base-rate information (Zillmann, 2000).
Issue Importance

Since exemplar discrepancy is thought to result from careless information
processing, we might expect that factors promoting more careﬁrl scrutiny of messages
would eliminate exemplar discrepancy’s effect. One factor that comes immediately to
mind in this regard is issue involvement. People who are highly involved with an issue
should be expected to pay more careful attention to the information contained in
messages related to the issue. Issue involvement (what we’re calling issue importance is
similar to what others call issue involvement) has been found to be an important factor in
several areas of media research, particularly in a public safety or health context. For
example Miller and Millar’s (2000) work on the persuasive inﬂuences of message
ﬁaming shows that issue involvement increases intentions to drive more cautiously after
exposure to a safe driving message. Skumanich and Kintsfather’s (1996) research on
organ donation shows that issue involvement is the strongest determinant of compliance
to a persuasive message asking individual’s to Sign an organ donor card.

Research on issue importance is consistently shown to distinguish people’s
perceptions and reactions to persuasive messages. This has been found to be the case in
the health communication context (Salmon & Atkin, 2003) and many other areas of
media inﬂuence. By and large, though not without exception, issue importance is found
to minimize message effects, with high issue involvement respondents less affected by
message factors than those low on involvement. The Elaboration Likelihood Model

(Petty & Cacioppo, 1986) explain this as a function of receiver’s motivation and ability to

10

process messages more closely, suggesting that those more involved in a topic will
inspect the message more closely. Close scrutiny to a message should increase the
receiver’s ability to ﬁnd ﬂaws in evidence and arguments used to persuade. Under these
conditions, only a very strong persuasive message should succeed. As such, high issue
involvement should minimize message effects in most circumstances. When we apply
this reasoning to exempliﬁcation logic, we might expect that those less involved would
process heuristically and be more affected by exempliﬁcation than individuals high on
issue involvement. Whereas the careless attention to the type of discrepant exemplars
often found in media that fail to accurately represent base-rate data, what Zillmann
(2002) calls “misexempliﬁcation,” might strengthen exempliﬁcation, close attention by
those highly involved should negate this effect.

It is clear from the logic above that issue involvement alone might be expected to
inﬂuence behavioral and perceptual reactions to the information presented in a message.
Moreover, there is reason to believe that it might moderate the effect that different
message factors have on these outcome responses. However, the limited evidence in
research on exempliﬁcation has failed to show that issue involvement can moderate the
effect of exemplar discrepancy. Research by Daschmann (2001 , as reported in Brosius,
2003) shows that exempliﬁcation effects occur regardless of a recipient’s level of
involvement with an issue. In fact, they occur even apart from one’s prior opinions on an
issue. In light of this, we might expect the effects of exempliﬁcation are robust. Certainly,
we should expect issue involvement to have an independent effect on issue perceptions.
However, in light of Daschmann’s ﬁndings, should we expect it to act as a moderator of

exempliﬁcation? By itself it seems that we should not expect issue importance to

11

moderate exempliﬁcation, but perhaps this might change when it is considered in
combination with variables such as information utility.

The extent to which information utility inﬂuences selective exposure can be seen
as a ﬁrnction of three message variables related to the concept of issue importance.
Although scholars use several different terms to label these variables they are essentially:
(l) the magnitude of threat or opportunity, (2) the likelihood that the threat or opportunity
will actually occur, and (3) the immediacy of occurrence (Festinger, 1957; Knobloch,
Carpentier, & Zillmann, 2003). Prior research suggests that people pay closer attention to
stories that are high on these attributes (Knobloch, Carpentier, & Zillmann, 2003). Atkin
(1985) suggests that this type of increased attention can be found in selective exposure to
media, where issue importance related to needs for guidance and reinforcement promotes
selection of media content suited just for this purpose.

If issue importance and information utility can moderate exposure and attention to
message features, questions arise concerning the role they play in determining the relative
inﬂuence of features such as exemplars and base-rate information. Although the general
superiority of exemplars over base-rate information might be expected, modiﬁcations in
reception processes suggest the potential for conditional change in outcome expectations.
Before examining the role of base-rates and exemplars, with this in mind, the
comparative advantages and disadvantages of each type of message feature needs to be
examined.

Comparing the Inﬂuence of Exemplar and Base-Rate Information.
Very few studies look at base-rates in isolation when attempting to examine

outcomes associated with their use. This is likely due to the fact that receivers often

12

perceive base-rates as boring and dry, particularly without exemplar information to
supplement the simple facts. Yet, the presence of exemplars to supplement base-rates
seems capable of reducing the inﬂuence of base-rate information. Some research shows
that when exemplars are present base-rate information may be underweighted (e.g. Adler,
1997). Other studies have found that even when no exemplars are present, base-rates are
often neglected if an individual has direct experience with the issue at hand (e. g. Goodie,
1997).

In general, base-rates have not been examined as often as exemplars, perhaps
because of the belief by some scholars that base-rates are virtually ignored when
exemplar information is present (e. g. Koehler, 1996). For example, in research by
Zillmann et al. (1996) the inﬂuence of base-rate information clearly pales compared to
exemplars. If researchers expect this pattern to be consistent regardless of other concerns,
then it would be of little consequence to examine the inﬂuence of base-rates when
exemplar inﬂuence is present. However, not all scholars agree that this is the case.
Although studies show that base-rates are overwhelmed by exemplars in many
circumstances (Berger, 2000), it does not appear that they are always ignored. Prior
investigations suggest that base-rates are perceived by decision makers as more reliable
and can inﬂuence decisional conﬁdence more than exemplars (Adler, 1997). Base-rates
are perceived as more truthful (Zillmann, 2002), precise (Zillmann et al., 1996), and
accurate (Zillmann, Perkins, & Sundar, 1992 (as cited in Zillmann & Brosius, 2000)) than
exemplars. Moreover, a variety of factors seem capable of altering the attention people

pay to base-rates and thus the inﬂuence that could potentially result.

13

Unlike base-rates, the inﬂuence of exemplars is often studied separately, as well
as in comparison with base-rates, to determine not only why they may be better than
base-rates but also what types of exemplars may have the strongest affect on individuals.
Several factors have been identiﬁed as potential sources of exemplar’s superior strength
(Brosius & Bathelt, 1994). For example, some scholars believe the strength of exemplar
inﬂuence results from the fact that exemplars are often more entertaining and sensational
than base-rates (Zillmann et al., 1996) and can convey more human drama than base-rates
(Aust & Zillmann, 1996). Zillmann (2002) claimed that exemplar effects would increase
over time; however, evidence does not support this. Several studies found that although
the effect of exemplars remained stable over time they did not become stronger (e.g.
Gibson & Zillmann 1994 & Zillmamr, Gibson, Sundar, & Perkins, 1996).

Factors such as these are often held capable of activating mental processes that
strengthen an exemplar’s affect. For example, some claim that exemplars are more likely
to evoke emotions than base-rates (Aust & Zillmann, 1996). Aside from emotional
components, exemplars are also more likely to invoke an automatic response in
individuals (Castelli et al., 2004) by coming to mind easily. Exemplars can activate and
inﬂuence an individual’s judgments through involuntary cognitive processes without
being accessible to conscious recall (Sia et al., 1999). Zillmann et al. (1996) asserts that
exemplars operate like ready-made shortcuts that make complex information more
understandable. Moreover, these shortcuts are made even more accessible by exemplars
that are vivid. Vivid exemplars tend to be more memorable than pallid base-rate data, and
this may increase exemplar availability in a manner that makes them more inﬂuential

(Berger, 2000).

14

Although most studies seem to focus on the comparative advantages of exemplars
over base-rate information, some research suggests several disadvantages to the use of
exemplars, particularly in relation to perceptual bias that might result from their use.
Frequently, and usually inadvertently, exemplars can present distorted versions of truth
(Rodgers, 2003). It has been found that exemplars are at best only somewhat
representative of their population and in some cases entirely nonrepresentative (Zillmann,
2002). The ease of their cognitive activation can lead to misjudgrnents of the frequency
with which different elements are present in a population (Zillmann et al., 1996). Such
misjudgrnents are proffered to have far reaching consequences, such as those observed by
cultivation researchers who claim that TV images lead people to believe that the world is
scarier than it really is and to over exaggerate their own chances to become victims of
crime (Berger, 2000). Exemplars are also said to play a role in perpetuating negative
stereotypes, and in isolation are frequently found to be inefficient sources of information
(see Sherman, 1996).

For the most part, exemplar information has been found to be less reliable than
base-rate information. In part, due to the fact that it is seldom collected or presented in a
systematic manner (Gibson & Zillmann, 1994). Exemplars are often presented to
supplement base-rates without making sure that the exemplar information is
representative of the population as a whole. In some cases the exemplar may be only one
case out of virtually thousands. As such, though most research has focused on how the
advantages of exemplars have outweighed the advantages of base-rate information, the
advantages of base-rate information should not be overlooked. Almost by deﬁnition,

base-rate information is expected to provide more reliable representations of population

15

characteristics and thought less likely to produce distorted perceptions. As such, we
might expect more functional outcomes to result from exposure to base-rate information
under conditions where its inﬂuence on receivers is strongest. One moderating factor
thought to strengthen the inﬂuence of base-rate information is information utility.
Information Utility as a Moderator of Base-rate Information ’3 Inﬂuence.

In their research comparing the inﬂuence of exemplars and base-rate information
on decisions to purchase a car, Krupat et al. (1997) found that whereas base-rate
information was regarded as less interesting than exemplars it was rated as signiﬁcantly
more informative. As already mentioned, Krupat et al. also found that this base-rate
information was more inﬂuential than an exemplar in this purchase decision context; an
observation in contrast to most research on exempliﬁcation. Can we surmise that the two
are interrelated? Is it the case that perception of the base-rate infonnation’s informative
value was responsible for the observed anomaly? If so, what process might account for
this? One possible explanation can be found in an understanding of the relationship
between information utility and attentional vigilance.

Knobloch et al. (2003) argue that when information is found to be useful it can
activate curiosity and exposure. Such an outcome might alter the perceived level of
interest in and subsequent attention given to the information provided. Previous research
shows that information utility can affect attention to messages, speciﬁcally messages
containing the type of objective facts characteristic of base-rate information (Wilton &
Myers, 1986). If one argues that the information utility of message components can
moderate attention to those components and, subsequently, the inﬂuence of those

components on outcomes from message exposure, then a case can be made that the

16

inﬂuence of base-rate information should supersede the inﬂuence of exemplars in
situations where the base-rate information has greater information utility than exemplars.
If information is perceived as useful then message receivers would more likely attend to
that information than to exemplars.

The reasoning offered here suggests the presence of an important moderator in
our understanding of exempliﬁcation processes. Instead of envisioning a simple process
in which message effects are governed by the inﬂuence of exemplars, we might propose a
more complex process in which issue importance interacts with information utility to
determine the inﬂuence strength of both exemplar and base-rate information. The unique
predictions offered here are limited to conditions of great issue importance. In cases
where issue importance is low, we should expect the inﬂuence of exemplars to exceed
that of base-rate information regardless of the informational utility found in either of
these message features. This prediction stems from the lirrrited receiver attentional
vigilance to all message features expected when issue importance is low. Under these
circumstances the greater accessibility of exemplars over base-rate information
recognized by Zillmann and Brosius (2000) should be constant. However, our predictions
might differ when issue importance is high. Under these conditions we expect that
message variables facilitating information utility will produce outcomes that contrast with
the commonly observed pattern in which the inﬂuence of exemplars supersedes that of
base rate information. Given conditions of great issue importance, when base-rate
information is thought to contain greater information utility than an exemplar, its

inﬂuence on receiver outcomes should exceed exemplar effects. By contrast, when the

17

information utility of an exemplar is thought to exceed base-rate information, exemplar
inﬂuence should dominate.

Iftlris understanding is correct, when issue importance is high the ability of
messages to inﬂuence receivers Should vary as a function of the information utility found
in different message features. In other words, information utility should moderate the
inﬂuence of message features (such as exemplars and base-rate information) on message
receivers. As the information utility of message features increases, their inﬂuence on
receiver outcomes should increase. Consequently, factors that determine the information
utility of a message should govern the inﬂuence of these different message components.
Based on prior information utility research, we should expect that the attention to and
inﬂuence of message features will vary as a function of the perceived relevance,
credibility, and amount of new information contained in a message feature. Individuals
will assess information as having greater utility as their perception of the relevance, the
amount of new information, and the credibility information increases. As such, these
features will moderate the inﬂuence of message features such as exemplars and base-rate
information. These expectations can be represented conceptually as the following set of
predictions.

1. Issue importance will moderate overall attention to exemplar and base-rate
message features. When issue importance is low, attentional vigilance to both
message features will be low. When issue importance is high, attentional
vigilance to message features will be moderated by informational utility.

2. When overall attentional vigilance to exemplar and base-rate message features is

low, selection of information used to make judgments about event populations

18

will be governed by message features affecting the ease with which information is
available in memory. In this regard, exemplars of concrete events will be
generally more available for access and retrieval from memory than more abstract
base-rate information.

. When overall attentional vigilance to exemplar and base-rate message features is
high, attentional vigilance to message features will rise as a function of increased
informational utility. In this regard, the perceived relevance, credibility, and
amount of new information contained in message features will moderate the
selection of information used to make judgments about event populations in the
following manner. If the perceived relevance, credibility, and amount of new
information is high, attentional vigilance to message features will increase. If the
perceived relevance, credibility, and amount of new information is low,
attentional vigilance to message features will decrease. Generally in this regard,
the perceived relevance, credibility, and amount of new information contained in
base-rates will be greater than that for exemplars. As such, base-rates will be
selected for use in making judgments about event populations more often than

exemplars.

Practical Applications of Information Utility as a Moderator Message Inﬂuence.

The logic described above outlines the general process by which information

utility should moderate the relative inﬂuence of exemplars versus base-rate information.

It argues that when issue importance is high, message factors that strengthen the

information utility of base-rate information will increase inﬂuence of base-rate

information compared to that of exemplars. Whereas the logic does not limit itself to any

19

one communication context, the present study is concerned particularly with the use of
health information; an area in which we might expect issue importance and information
utility to play substantial roles in message exposure. Although it is unlikely to be true in
all cases, it is not hard to imagine a health communication exchange where issue
importance is elevated by the magnitude of some health threat, the likelihood that the
threat will actually occur, or the immediacy of its occurrence. In these situations, the
perceived relevance, credibility, and amount of new information contained in a message
should moderate the relative strength of exemplar versus base-rate information on
receiver outcomes in the manner advocated by the above propositions. When issue
importance is pronounced, information utility should moderate the outcome of message
exposure such that when the perceived relevance, credibility, or amount of new
information contained in a message is high, the inﬂuence of base-rate information will
supersede the inﬂuence of exemplars. On the other hand, when issue importance is
nominal, we should expect the often observed pattern in which exemplars have a greater
inﬂuence than base-rate information.
Issue Importance and Information Utility in the Context of Health Communication

Issue importance and information utility have been issues of central concern in the
persuasive health communication literature. Concern over some threat to public health is
the essence of many persuasive health communication messages. In similar manner, the
importance of information utility can be seen in theoretical research such as the health
belief model (HBM) (Rosenstock, 1990) and the extended parallel processing model
(EPPM) (Witte, 1992) that focuses on the roles of efficacy and ability. However, the

centrality of information importance in theory may seem at odds with strategy apparent in

20

many public health campaigns that seem to focus more on the use of highly emotional
exemplars. While research in this area might suggest that the use of such exemplars is
effective with many audiences, the logic presented in this study raises questions
concerning the effectiveness of reliance on exemplars under circumstances of acute
attentional vigilance. In some health situatibns, the type of people likely to pay close
attention might be the exact audience you are concerned with. Could it be that base-rates
play a more important role with these critical receivers? If so, the practical implications
for health communication are large, making our understanding of the roles of issue
importance and information utility critical.

Issue Importance. As previously mentioned, issue importance is made up of three
factors dealing with perceived threat: the magnitude of the threat, the likelihood of
occurrence, and the immediacy of occurrence. In health communication research
persuasive messages related to health threats have often examined the extent to which
fear and efﬁcacy induced by messages promote behavior change in receivers. Much
health communication work, particularly work with fear appeals, is concerned with the
extent to which the target of a message perceives a particular threat to be severe
(magnitude) and perceives himself/herself to be susceptible (likelihood) to the threat (see
Witte, 1992). The higher the perceived severity and susceptibility, the greater threat an
individual may feel (Witte & Morrison, 2000).

Although the immediacy of the threat is not examined directly in most health
communication research, it is often examined indirectly when examining behavior
change as part of overall perceived susceptibility. Whether or not an individual believes

that a threat is imminent is not studied as much as whether they believe the threat can

21

happen at all. The present study focuses more speciﬁcally on immediacy as a factor of
issue importance based on the belief that this aspect of overall perceived susceptibility is
an essential determinant of attentional vigilance. If the viewer does not believe that the
threat will occur in the near future it is less likely that threat will be perceived as high at
the moment of message reception. As such, the moderating inﬂuence of threat on
selective exposure to message content might be lost.

When a health threat is seen as severe and probable an individual is more
motivated to seek information that can help reduce the anxiety created. Under these
circumstances, we might expect people to process messages content more closely.
Though Daschmann (2001, as reported in Brosius, 2003) shows that exempliﬁcation
effects are robust, and occur regardless of the recipient’s level of involvement, it seems
plausible that base-rate information would play a stronger role under some conditions,
especially if message content is perceived as high on information utility.

Information Utility. Information utility has been examined on the basis of three
factors; the perceived relevance of information, the amount of new information, and the
perceived credibility of the information. When examining health communication
messages each of these factors have been shown to enhance selection of and/or attention
to health communication messages. In research on alcohol intervention, Neighbors,
Palmer, and Larimer (2004) found that when individuals believed that the information
contained in a persuasive message was relevant they were more likely to attend to
message content and be open to the behavioral change it suggested. One possible
explanation for this is based on the manner in which information is processed cognitively

when its utility is high. It has been found that individuals process health-related

22

information systematically when the health issue is perceived relevant but process it
heuristically when it is not (Rotlrman & Schwarz, 1998).

Previous research on health communication also shows that credibility can
enhance the inﬂuence of persuasive messages (McGuire, 2001). Within the health
communication literature credibility usually refers to both the source of the message and
the evidence provided within it. Messages are perceived as more credible when it appears
that the source knows the facts on the relevant health topic and is reporting them honestly
(McGuire, 2001). Interestingly in this regard, statistical information such as that often
found in base-rate reports has been found to enhance the credibility of evidence within
health messages (Atkin, 2001 ). However, since the present study looks at the inﬂuence of
base-rates independent from its role here, we focus on other aspects of credibility. There
are a variety of factors that can detract from perceived credibility such as evidence of
nrisinfonnation, inaccurate data, and an untrustworthy source.

The ﬁnal determinant of information utility deals with amount of new
information. This factor is likely to be particularly relevant to many persuasive health
messages where campaigns might suffer from the perception that a message presents
nothing new. If the logic here related to base-rate inﬂuence is correct, health
communicators would need to focus on presenting base-rate information in forms that
appear to add to the receiver’s knowledge base. While at ﬁrst this might seem like a
considerable challenge overcome only by the constant gathering of new information, the
variety of methods that can be used to present base-rate information offer message
alternatives that should create the perception that a message contains new knowledge.

The use of alternative methods for presenting base-rate information should not only make

23

the amount of new information in a message appear greater, the presentation of abstract
information in several different forms should also increase the likelihood that receivers
with different learning styles will see the information in a form easier for them to
understand. In this manner, the use of different presentation styles might also facilitate
the understanding of abstract base-rate information often difficult for some to grasp.

The application of issue importance and information utility to the health
communication context allows us to represent the aforementioned predictions as a set of
speciﬁc hypotheses offered for examination. The ﬁrst three hypotheses simply posit
direct effects of issue importance, information utility, and exemplar discrepancy on
outcome variables. The fourth hypothesis considers the interaction among these variables,
and how information utility and issue importance combine to moderate the inﬂuence of
exemplar versus base-rate information. The logic offered before proffers that when issue
importance is low, the perceptual reaction of respondents (e.g., population estimates, and
perceptions of problem severity/personal vulnerability) concerning a health disorder will
be inﬂuenced more strongly by the presentation of concrete exemplars of individual
experiences with the health disorder than by the presentation of abstract base-rate
information about the proportion of people who are diagnosed with the disorder. This
would be the case regardless of the information utility of a message. By contrast, when
issue importance is high, the effect of base-rates and exemplars on these reactions will be
moderated by the information utility (i.e., perceived relevance, perceived credibility and
amount of new information) in exemplars and base-rates. The moderation is such that
when information utility is high, perceptions will be inﬂuenced more strongly by the

presentation of abstract base-rate information regarding the proportion of people who are

24

diagnosed with the disorder than by the presentation of concrete exemplars of individual
experiences with the health disorder; whereas when information utility is low, once again,
perceptions will be inﬂuenced more strongly by concrete exemplars than by abstract
base-rate information. Past exempliﬁcation research (Gibson & Zillmann, 1994) has
observed the strength of exemplar inﬂuence by comparing responses to stories with
minimal versus extreme exemplar discrepancy (i.e., the extent to which the exemplar and
base-rate information differ). The current research uses this approach to examine a set of
hypotheses designed to examine the moderation of exempliﬁcation effects resulting from
the combination of three variables.
Formal Hypotheses

The research and theory above suggest a set of four hypotheses that encapsulate
the issues investigated in this study. Hypotheses one to three posit direct effects for the
three key variables combined in hypothesis four. Hypothesis four details the interaction
predicted in this study, the examination of which is the main focus of this investigation.

The ﬁrst hypothesis attempts to replicate previous work examining the role of
issue involvement on message effects in the ﬁeld of public safety and health
communication (e.g. Millar & Millar, 2000; Skumanich & Kintsfather, 1996).

H1: When exposed to a news story about a health issue problem, respondents who

ascribe high importance to the issue (in this case the issue of smoking related

anxiety) will give higher ratings of the problem’s perceived severity and higher

estimates of the percent of the population affected by the problem than

respondents who ascribe low importance to the issue.

25

The second hypothesis attempts to replicate the ﬁndings of previous research on the
effects of information utility in the ﬁeld of health communication (e. g., Perloff, 2001).
H2: Exposure to a news story about a health issue problem that contains high
levels of information utility (in this case the issue of smoking related anxiety) will
lead to higher ratings of the problem’s perceived severity and higher estimates of
the population affected by the problem than exposure to a news story that contains
low levels of information utility.
The third hypothesis extends the extensive research on the effects of exemplar
discrepancy in news and political media (cf. Brosious, 2003) to the ﬁeld of health
communication.
H3: Exposure to a news story about a health issue problem that contains high
exemplar discrepancy will bias (in the direction of the exemplar) respondents’
ratings of the problem’s perceived severity, and estimates of the percent of the
population affected by the problem more than exposure to a news story that
contains low exemplar discrepancy.
The fourth hypothesis uses the logic stated above to predict the manner in which issue
importance and information utility combine to moderate the inﬂuence of exemplar
discrepancy.
H4: Information utility and issue importance will moderate the effect of exemplar
discrepancy such that the following will be observed:
a) When issue importance is low, perceptual responses (i.e., population
estimates, perceptions of problem severity) to a message containing base-

rate information with highly discrepant exemplars will show a stronger

26

bias toward the exemplar than the base-rate information. This will be
equally true under conditions in which the information utility of a message
is high or low.
b) When issue importance is high, responses (i.e., population estimates,
perceptions of problem severity/personal vulnerability) to a message
containing base-rate information with highly discrepant exemplars will
show a stronger bias toward the base-rate under conditions where the
message’s information utility is high than under conditions where
information utility is low.

Research Question — Will issue importance, information utility, and exemplar

discrepancy, individually or in combination, have an affect on behavioral intentions?

27

Method
Overview

The present study was conducted in two parts: a pilot study followed by the main
study. In the main experiment a 2 X 2 X 2 mixed-model design varied issue importance
(high, low), information utility (high, low) and the degree to which an exemplar
representation differs from base-rate information; hereon referred to as exemplar
discrepancy (low, extreme).

The pilot study was conducted in two phases. In the ﬁrst phase, survey items
examined the existing variance inherent in the self-evaluated importance that individuals
place on smoking and several other health issues in order to determine if smoking was a
viable issue for use in this study or, if it was not, to identify another issue for use in the
study. A viable issue is one where we can identify that distinct groups of individuals
exists in the population who are high and low on perceived issue importance. In the
second phase, once a usable issue had been identiﬁed, four different print versions of a
health information story were created to vary in terms of information utility (high, low)
and exemplar discrepancy (low, extreme). Survey items examined respondent’s
perceptions of these stories as a manipulation check designed to determine the extent to
which the story versions successfully manipulate information utility and exemplar
discrepancy.

In the main study, participants took part in an experiment involving a story
manipulation and a survey. Respondents read one of four versions of an online story
containing information about the dangers of smoking. The four versions of the story were

manipulated to represent all combinations of information utility (high, low) and exemplar

28

discrepancy (low, extreme) factors. Outcome measures examined the inﬂuence of
exemplar discrepancy under varying conditions of issue importance and information
utility on two respondent perceptions: 1) population estimates, the percent of the
population suffering from smoking related disorders, and 2) problem severity/personal
vulnerability, an individual’s beliefs about the social threat worsening and their chances
of actually experiencing a threat (Gibson & Zillmann, 1994; Witte, 1992) to smoking
related disorders.
Pilot Study
Participants

All participants within this study were recruited from a communication course at a
large Midwestern University. For the ﬁrst phase of the pilot 52 participants were
surveyed. As this study was only used to determine topic viability, little demographic
information was collected at this time. Respondents received course credit or extra credit
for participation. A student sample was selected because students are members of the
target audience for several health issues including that of cigarette smoking. Although
college students are a limited sample as far as age is concerned successful health
campaigns do target select audiences in order to best present their messages. This
experiment used the same technique as successful health campaigns by focusing the
messages presented to a target audience. This provides a more realistic test of the
assumptions and hypotheses presented than using unspeciﬁc, broader groups of

individuals.

29

Procedure

In the ﬁrst phase of the pilot study, respondents were recruited in class and asked
to ﬁll out a survey. Subjects were told that they would be completing a survey about
various health issues in order to identify which issues are important to them as college
students and to determine their current knowledge of those subjects. A consent form
provided repeated this information and explained that participation was voluntary.

Subjects were asked to answer questions regarding a variety of health issues
including, sexually transmitted infections, pregnancy, cigarette smoking, binge drinking,
drug abuse, and domestic violence. Once the ﬁrst phase was completed it was determined
that cigarette smoking was a viable issue for use. If this had not been the case then
another more viable issue would have been selected based on the results of the survey.

The second phase of the pilot study recruited a separate sample of 207 students
ﬁom another communication course within the same Midwestern University. During this
phase respondents were given one of four story versions (that vary information utility and
exemplar discrepancy) to read and respond to in a questionnaire. These story versions
included; high information utility, extreme discrepancy, n= 46; high information utility,
low discrepancy, n= 49; low information utility, extreme discrepancy, n= 56; and low
information utility, low discrepancy, n= 56.

Along with the stories on smoking, respondents read a story on terrorism and
genetic screening. These stories are to be used in the main study to help mask the true
nature of the experiment. Once the stories had been read the students were asked
questions pertaining to each story’s perceived relevance, perceived credibility, and the

amount of new information. They were then asked to determine how similar or different

30

the exemplar is to the base-rate information. Finally, they responded to items intended to
pilot test the outcome variables in this study (population estimates of those suffering from
smoking related issues and perceptions of problem severity/personal vulnerability to
smoking related issues).

Measures

First phase of the pilot stuay. Items were designed to measure the issue
importance respondents ascribe to various health issues, behavioral patterns related to
these issues, the perceived credibility of possible information sources, and knowledge of
facts related to these issues for possible use as the base-rate for the main study.

Issue importance was measured with a total of 54 items (9 for each issue)
designed to assess the perceived magnitude of several health related issues (e.g., sexually
transmitted infections, cigarette smoking, drug abuse) along with the likelihood and
immediacy of the issue’s occurrence. The perceived magnitude of the threat was assessed
by responses to items such as “Do you feel personally threatened by the health risks
associated with smoking?” Likelihood and immediacy measures included items such as
“Do you feel that a college student like you is likely to be affected by smoking” and “I
will be affected by smoking soon.” Responses to the questions were assessed on a 7-point
Likert-type scale from 1 = “not at all” to 7 = “extremely.” All the items were scored so
higher agreement corresponds with higher issue importance.

Respondent behavior relevant to the health issues was measured by a single item
with four response choices repeated for each of the six health issues. For example, “I
consider myself a smoker,” is followed by response options including “yes,” “no,” “I

don’t consider myself a smoker, but I do smoke socially,” and “I used to smoke but I quit

31

years _ months ago.” It is expected that non-smokers will constitute the low
issue importance group and smokers will be placed in the high issue importance group.
Those individuals that do not currently smoke are expected to not feel a threat ﬁom
smoking (excluding concerns related to second hand smoke), whereas those that are
smokers should score higher in this regard.

Source credibility was assessed by giving subjects a list of 30 possible sources of
health information and asking them to rate credibility on a 7-point Likert-type response
scale ranging from 1 = “not at all” to 7 = “extremely.” Possible sources included the
CDC, Surgeon General, The National Enquirer, Newsweek, and 26 other sources that
may be used as channels for health information.

Baseline data related to the amount of new information a potential message might
contain was gathered through 27 open-ended items designed to determine respondent
knowledge of certain facts pertaining to each health issue. These items include, “What
percent of all sexually transmitted infections occur in people 25 or younger?” and “About
what percent of smokers under the age of 25 have anxiety problems associated with
cigarette smoking?” All items can be found in Appendix A.

Second phase of the pilot study. Respondents within this phase also received
course credit or extra credit for participation. Students that participated in the ﬁrst phase
of the pilot study did not participate in this phase. Participants responded to items
examining the perceived information utility and exemplar discrepancy contained in
stories in order to determine the extent to which the four versions of the smoking story
successfully manipulated these factors. The same items were used to assess perceptions

of the two additional stories on terrorism and genetic screening. In addition to measures

32

of information utility and exemplar discrepancy, participants responded to items designed
to pilot test the outcome measures in the main experiment. These included measures of
population estimates, perceptions of problem severity, and personal vulnerability
regarding a health disorder.

Conﬁrrnatory factor analysis (CFA) was used to test the content validity of all
multiple (three or more) item measures. Scale items were retained if they passed an
internal consistency test, involving (a) a check of face validity and (b) an examination of
factor loadings and errors. Items with poor face validity and factor loadings of less than
.60 and/or greater errors in association with other items than what would be expected by
sampling error were dropped. All items failing to meet these criteria were dropped from
scale construction. The reliability of each scale was assessed using Chronbach’s Alpha
((1).

Perceptions of high and low levels of information utility were measured with (14)
items included to assess the perceived relevance, credibility, and the amount of new
information contained within each story. Perceived relevance (ﬁve items) and credibility
(six items) were measured with questions like, “The story on smoking and anxiety applies
directly to my life,” “I believe that cigarette smoking can cause anxiety.” Again
responses were collected on a 7-point Likert scale from 1 = “not at all” to 7 =
“extremely.” Amount of new information (three items) was assessed by asking subjects
to report whether they knew the information provided in the story prior to reading the
story and if the story added to their knowledge.

Exemplar discrepancy was judged by (three) items asking the subjects to rate how

similar the information in the exemplar is to the base-rate. These items included questions

33

such as, “On a scale of l to 7 rate how different you think the example about anxiety is
from the statistical information.” One question on behavior was also included to assess if
there is a difference between smokers and non-smokers.

The outcome measure included ﬁve items. The ﬁrst item was designed to assess
population estimates concerning the percent of the population suffering ﬁ'om smoking
related ailments: “What percent of college student smokers have anxiety problems?

%” Responses to the four items assessing problem severity/personal vulnerability
were collected on a 7-point Likert scale from 1 = “not at all” to 7 = “extremely.” These
include items such as: “How severe is the problem of smoking related anxiety among
college students,” and “How likely do you think it is that smoking related anxiety might
become a problem for you?” Behavioral intentions and outcomes regarding smoking
behavior were measured with four items using the same 7-point Likert scale. These
include items such as; “How likely is this information about anxiety to affect your current
smoking behavior,” and “How likely is it that you will seek more information on health
problems associated with smoking.” All items can be found in Appendix B.

Main Study
Participants

The main study had a total of 220 participants recnrited from the same
Midwestern University through an online database. None of the individuals that
participated in the main study participated in either of the earlier pilot studies.
Respondents received course credit or extra credit for participation. Participants were
assigned to one of four groups; high information utility/low exemplar discrepancy, n =

54, high information utility/extreme exemplar discrepancy, n = 54, low information

34

utility/low exemplar discrepancy, n = 55, and low information utility/extreme exemplar
discrepancy, n = 57. The majority of participants were female (52%) and were Caucasian
(75%). Fifteen percent were African American, 4% Chicano-Latino, 1% Asian/Paciﬁc
Islander, .5% Middle Eastern, .5% Native American and the remaining 4% recorded
themselves as Other. The mean age was 19.67 (SD = 1.54) and the majority of
participants were either freshman or sophomores (76%).

Procedure

Subjects were told that they were participating in a study examining the health
behaviors of college students. A consent form explained the purpose of the study to the
participants informing them that participation is voluntary and they may discontinue the
study at any time without penalty. Students were asked to not place their name on the
questionnaire.

Subjects were told that they were going to read a series of news articles about
social issues. Subjects were told little about the study so they were not sensitized to the
true purpose of the study. They were also told that after the stories they would be asked
various questions concerning their views on the topics addressed within the stories.
Respondents were recruited using an online database (Experimetrix) and were asked to
access a website (www.survevmonk_ev.com) in order to participate in the study. There
was a consent form for participants to sign (by typing their name) prior to beginning the
experiment that again reminds them that participation is voluntary.

Participants were randomly assigned to read one of the four different versions of
the smoking story induction that varies information utility and exemplar discrepancy. The

last two stories dealt with two relevant social issues selected to be one high and one low

35

on issue importance; terrorism and genetic screening. After reading the stories
participants were directed to ﬁll out a brief questionnaire designed to measure the
outcome variables in this study (population estimates of those suffering from smoking
related issues, and perceptions of problem severity/personal vulnerability to smoking
related issues) and to act as a manipulation check on information utility. After subjects
ﬁnished the survey they were directed to a page containing information debriefrng them
on the exact nature of the study. This debrieﬁng informed each subject of the facts
pertaining to cigarette smoking and made them aware that they were exposed to an
extremely discrepant exemplar that might have been misleading. Subjects were also given
a website link to access (www.anti-smoking.org) which contains ﬁrrther information
about the dangers of smoking as well as steps on how to effectively quit.
Stimulus Materials

Four versions of the online message induction were created to manipulate an
informational message about the dangers of smoking. All stories were centered on the
presentation of the same base-rate statistic; however, message features were altered to
vary the perceived information utility and exemplar discrepancy of the message. Story
one represented the high utility/low discrepancy condition, the story was constructed to
make both the base-rate information and exemplars appear high on information utility in
terms of perceived credibility, relevance and amount of new information. For example, in
the high information utility version the source of the information was clearly labeled as
one that was established as credible from the pilot study (e. g. the CDC, and being
reported in The US. News & World Report). In addition, the story included information

that scored high in the pilot study for relevance and newness. Exemplars were

36

constructed to appear consistent with the base-rate information presented. For example,
the exemplar described a group of college students where the individuals that are smokers
are having anxiety issues.

Story two represented the high utility/extreme discrepancy condition. This story
contains the same message structure used to make the base-rate information and
exemplars appear high on information utility; however, in this version the exemplar
included with the story contains attributes that are extremely uncharacteristic of the
norms suggested in the base-rate information. For example, this story is about a college
student smoker that has no apparent anxiety issues.

Story three represents the low utility/low discrepancy condition. This version
maintains the message discrepancy presentation used in story one, however, it was
constructed to make both the base-rate information and exemplars appear to have little
information utility. In the low information utility conditions each of the three factors was
manipulated to reduce credibility, relevance. Credibility was reduced by choosing one of
the sources from the pilot study (e. g. The National Enquirer, as well as using an
“unnamed source”) that scored low in credibility. Low relevance was manipulated by
using information that is not pertinent to college students. For example, the base-rate and
exemplars in the report were altered so that they were based on a population of people
who are senior citizens. Amount of new information was not altered as part of the
information utility manipulation. The decision to leave amount of new information
unchanged was made in order to avoid the validity threats associated with potential
confounds that might be introduced by this type of change. Altering the amount of new

information would require changing some of the basic facts used in each story. This was

37

thought a likely source for confounds. The addition of a manipulation check following
the outcome measure in the main study was included in order to guarantee that
information utility was successfully manipulated in spite of the fact that amount of new
information remained unchanged. Finally, story four represented low utility/extreme
discrepancy. This version maintained the information utility presentation from story three
and the message discrepancy presentation used in story one.

The two stories being used to shield the true nature of the experiment deal with
terrorism and genetic screening. Both of these are issues that are relevant to society and
are plausible as news stories that would be read with a story associated with cigarette
smoking. The stories were the same for each participant regardless of condition. The
information given in each of these stories also contained base-rate and exemplar
information so the format of each of the three stories is similar. All stories can be found
in Appendix C.

Measures

Experiment. Following the reading of the stories participants completed a
questionnaire that included: 1) measures of the two main outcomes variables in this study
(population estimates, and perceptions of problem severity/personal vulnerability,), 2)
items designed to act as a manipulation check on information utility, 3) measures of the
issue importance respondents ascribe to smoking, and 4) items assessing issue
perceptions relevant to the other stories in the experiment. These last items were included
to disguise the questionnaires true purpose.

The outcome measures included two sets of questions. Population estimates and

problem severity/personal vulnerability were assessed with the same ﬁve items used to

38

measure these variables in phase two of the pilot study. Behavioral outcomes were
measured using four items. The ﬁrst two items replicate the items used in phase two of
the pilot study. Two additional items using the same response scale included questions
asking: “How likely are you to quit smoking or not start based on the information about
anxiety?” and “How likely do you think others would be to quit smoking based on this
information about anxiety?” These measures were used to examine if there were
differences between smokers and non-smokers.

The manipulation check on information utility replicated the 14 items used in
phase two of the pilot study to measure the credibility (six items), relevance (four items)
and amount of new information (four items) for each of the three stories read by
participants. Assessment of perceptions relevant to the other stories was made by
modifying the 14 items used to measure credibility, relevance, and amount of new
information. Two sets of 14 items were adapted so that they pertained to the issues found
in the two stories. A manipulation check of issue importance was assessed for each of the
stories that respondents read. The issue importance questions included nine items for
each health issue: three each for magnitude, likelihood and immediacy. The form of these
items matched those used in the pilot test asking participants to respond to items such as
“Do you feel personally threatened by the health risks associated with smoking”
(magnitude), “Do you feel that a college student like you is likely to be affected by
smoking” (likelihood), and “I will be affected by smoking soon” (immediacy).

The questions for the two additional stories on terrorism and genetic screening
were modiﬁed for the different topics. The additional items measured the demographics

of the sample population. These included asking questions in regards to gender, current

39

GPA, and household income. The ﬁnal two questions of the questionnaire assessed
whether or not respondents discovered the true nature of the story. The question asked
was, “While completing this experiment I believe that I have detected the purpose of the
study.” If the answer given was “Yes” participants were asked to name what they

believed the purpose to be. All items can be found in Appendix D.

40

Results
Pilot Study Phase One

The ﬁrst part of the pilot study was conducted with several goals in mind
including: 1) to determine if there are two distinct groups (high, low) in the perceived
issue importance associated with the Six health topics in order to identify one for use in
the main experiment, 2) to measure the perceived credibility of 30 information sources,
3) to measure knowledge of facts related to the six health issues in order to determine if
the target population will already be familiar with base-rates for the issues that would be
used in the main study, and 4) to identify behavioral patterns related to these health topics
as possible markers of individuals scoring high or low on the issue importance of each
topic.

High versus Low Issue Importance. The main purpose of the ﬁrst phase was to
identify a single topic for use in the main study. The experiment in the main study called
for an issue on which there existed distinct groups of individuals in the population who
scored high and low on perceived issue importance. Once identiﬁed as such, respondents
could be assigned one of the two issue importance conditions (high, low). For the
purposes of this study, issue importance was deﬁned as the extent to which respondents
perceived a health topic to be threatening along the three dimensions of issue importance
identiﬁed earlier, the perceived magnitude of threat along with perceptions of the
likelihood and immediacy of the threat affecting them personally. In order to identify an
issue for use in the study, respondent scores on questions measuring the perceived
magnitude, likelihood and immediacy associated with each of the six pilot tested issues

were examined.

41

For an issue to satisfy the requirements of the study, participant responses would
need to show a pattern which demonstrated that two distinct groups (high and low) of
people existed in the population. Evidence of two distinct groups was deﬁned
operationally here as a situation where there existed a sufﬁcient number of respondents
who scored high (5 or above on the 7-point scale) or low (3 or below on the scale) on all
items used to measure each of the three dimensions of issue importance.

Respondents answering inconsistently or giving a neutral response of 4 to any
item were considered neither distinctly high nor low. As such, an issue was considered
high on importance for the respondent if he/she scored high for each separate attribute of
issue importance (magnitude, likelihood, and immediacy) as well as on overall issue
importance (i.e., a score of between 15 and 21) and was considered low on importance if
the respondent scored in low for each separate attribute as well as on overall issue
importance (i.e., a score of between 3 and 9). The list of potential topics in the
questionnaire included: sexually transmitted infections, unplanned pregnancy, smoking,
drug use, drinking alcohol, and domestic violence.

As a ﬁrst step in examining the six potential issues; scores on the individual
dimensions and the overall measure of issue importance were inspected to determine the
percentage of respondents that fell into the categories established as low and high. As
shown in Table 1, although there is no single issue for which a large percent of
respondents fall within both the high and low groups across all indicators of issue
importance, smoking and STI Show a substantial number of respondents falling within
both high and low groups on two of the three individual dimensions (magnitude and

likelihood). Moreover, on the third dimension (immediacy) as well as the overall issue

42

importance, smoking and STI show a large percent of respondents falling within the low
issue importance group and at least some respondents, though smaller in number, falling
within the high issue importance group. These distributions appear considerably superior
to other issues such as pregnancy and domestic violence where no respondents fall in the
high category on the immediacy dimension of issue importance, or drug use and alcohol
where only 2 % of respondents fall in the high category on both the immediacy and
overall dimensions of issue importance.

Table 1

Percent Scoring High and Low on Issue Importance for Six Pilot Issues

 

Health Magnitude Likelihood Irnmediacy Overall
Issue Issue
Importance
High Low High Low High Low High Low

 

 

 

 

 

 

 

 

 

 

 

 

STI 25% 37 15 44 4 88 8 55
Pregnancy 29 35 21 52 0 92 2 47
Smoking 29 35 17 58 4 85 8 57
Drug Use 12 62 4 79 2 94 2 80
Alcohol 13 37 13 44 2 85 2 56
Domestic 8 79 4 83 0 94 0 85
Violence

 

 

 

 

 

Upon initial inspection, smoking and STI seem somewhat indistinguishable in
their suitability for use in this study. However, additional examination of the means and
standard deviations of participant responses to these issues led to the selection of one
issue. Based on the possibility that the unusually high standard deviation observed for
STI scores on the likelihood dimension of issue importance might indicate the presence
of outliers that could distort the results of the main experiment, the decision was made to
use smoking for the current investigation. Notably, the percent of respondents

categorized as high on overall issue importance was smaller than hoped. This appears to

43

result from the low scores on immediacy observed for smoking as well as all other health
issues examined. In order to address potential problems associated with this issue, the
manipulation of news stories in the main study was modiﬁed speciﬁcally to facilitate the
sense of immediacy in respondents by focusing on how smoking affects college students
suddenly and unexpectedly while they were enrolled as students. The means and standard
deviations computed on measures of magnitude, likelihood and immediacy for all issues
can be seen in Table 2.

Table 2

Descriptive Statistics on Issue Importance for Six Pilot Issues

 

Health Magnitude Likelihood Irnmediacy Overall
Issue Issue
Importance
Mean SD Mean SD Mean SD Mean SD

 

STI 3.82 1.55 4.58 6.31 1.91 1.16 10.33 7.15

 

Pregnancy 3.71 1.78 3.31 1.39 1.96 .91 8.92 3.22

 

Smoking 3.91 1.73 3.30 1.51 2.08 1.32 9.20 3.49

 

DrugUse 3.72 9.09 2.50 1.19 1.59 1.04 6.59 2.89

 

Alcohol 3.26 1.36 3.37 1.11 2.22 1.02 8.85 2.70

 

 

Domestic 2.11 1.46 2.12 1.17 2.24 4.55 6.45 2.81
Violence

 

 

 

 

 

 

 

 

 

 

Determining Perceptions of Credibility. After identifying smoking as the topic for
use in the main study, descriptive statistics were computed on the perceived credibility
scores of 30 pilot tested information sources (see Appendix E). This was done in order to
identify their potential for use in the main experiment’s manipulation of information
utility. Information utility is expected to vary from high to low (in part) as a function of
credibility. Of the nine news media included in the pilot test the US. News & World
Report ranked the highest on credibility with 60% of respondents stating that the

magazine was credible (i.e., choosing 5 or higher on the 7-point Likert scale) as a news

44

source for health information; M = 4.46, SD = 1.4. The most credible source for health
information among the 21 individuals and organizations pilot tested was the Center for
Disease Control with 98% of respondents stating that the organization was credible as a
source; M = 6.5, SD = .75. The news media scoring lowest in credibility for health
information was The National Enquirer, with 94% of people ranking them low (i.e.,
choosing 3 or lower on the 7-point Likert scale) on credibility and no respondents
choosing higher than the neutral point; M = 1.2, SD = .75. The least credible individual
or organization was the “unidentiﬁed source” with 90% of respondents ranking this
source as low on credibility, M = 1.59, SD = 1.05. Tests on credibility scores for The US.
News & World Report and The National Enquirer showed that these media were
signiﬁcantly different from each other; t (51) = 15.42, p < .001, two-tailed. Similar tests
on credibility scores for the Center for Disease Control and an unnamed source showed
that these sources were also signiﬁcantly different, t (51) = 27.31, p < .01, two-tailed.
Based on these results, it was decided to use a statement by the Center for Disease
Control appearing in the US. News & World Report for the high credible source, and a
statement by an unnamed source appearing The National Enquirer for the low credible
source in the main study.

Determining Knowledge for Manipulation of Base-Rates. Participant responses to
questions pertaining to information about smoking were exanrined to identify facts with
which respondents were unfamiliar. This was done to aid in the development of base-rate
manipulations used in the main study. Descriptive statistics for questions measuring
respondent knowledge of base rates associated with all pilot tested issues can be found in

Appendix F. When examining only the question pertaining to smoking, response to the

45

question “About what percent of smokers under the age of 25 have anxiety problems
associated with cigarette smoking?” indicates that knowledge about this issue is generally
ﬂawed. No respondent correctly identiﬁed the number of college student smokers that
suffer from anxiety as 30% (National Institute on Drug Abuse, 2001). By and large,
responses varied greatly from each other as well as from the actual ﬁgure, M = 40.00, SD
= 25.08. This observation not only reinforced the notion that sample knowledge of this
issue is poor, but also that the error in these perceptions is unsystematic. Important here is
the implication that respondents as a group had no real idea about the base-rate for this
issue. If respondents had shown considerable knowledge of base-rates associated with the
issue, or if the same erroneous belief was strongly shared among respondents, the ability
to manipulate perceptions of this issue in the main study (and thus, its suitability for use)
would have been greatly damaged. Knowing this wasn’t the case not only reinforced the
choice of smoking related anxiety as the issue for the main study, but also gave the
researcher great freedom in choosing the best base-rate to use in the main study.

A ﬁgure of 68% was selected for use as the base-rate in the main study. Selection
of this ﬁgure was based on the following logic. The goal of the main experiment was to
examine the extent to which exemplar discrepancy would override the inﬂuence of base-
rate information. As such, the ability to unambiguously manipulate exemplar discrepancy
was an important consideration. Since pilot-study students, on average, already perceived
the base-rate for smoking-related anxiety to be somewhat low (M = 40%), setting a low
base-rate for the main experiment (such as the correct 30% ﬁgure) might impede efforts
to manipulate exemplar discrepancy. The 68% ﬁgure was decided on because it was

thought to be distant enough from the average ﬁgure observed, but not so high as to seem

46

improbable to respondents. As a side note, no respondent in the pilot study selected 68%
as the number of college student smokers that suffer ﬁ'om anxiety.

Smoking Behaviors Related with Issue Importance. Data were inspected to
determine whether or not smoking behaviors were related to perceptions of issues
importance. If so, these behaviors might be useﬁrl markers of individuals scoring high or
low on these issues. Since only two of the respondents within the sample were smokers, it
was difﬁcult to determine the extent to which smoking predicted these perceptions.

Pilot Study Phase Two

Analyses in phase two of the pilot study functioned as: 1) a manipulation check to
determine if the four experimental versions of the smoking story successfully
manipulated perceived information utility (high, low) and exemplar discrepancy (low,
extreme), and 2) a reliability check on outcome measures including population estimates
(percent of the population suffering from smoking related disorders), problem
severity/personal vulnerability (beliefs that social threat is worsening and chances of
experiencing smoking related disorders), and behavioral intentions regarding smoking
behavior.

Information Utility Induction Check. The manipulation check on information
utility began by examining the reliability of the 14 items used to measure the perceived
relevance, amount of new information, and perceived credibility of the four manipulated
news stories (i.e., perceived information utility). Two of the four items measuring
perceived relevance failed the initial CFA test and were dropped. The remaining two
items were averaged to create a measure of perceived relevance ((1 = .87). Of the six

credibility items, only one failed the initial CFA test and was dropped. The ﬁve items

47

were averaged to create a measure of perceived credibility (a = .83). All three items for
new information met the criteria for factor loadings and passed the tests for error. These
items were averaged to create a measure of new information (a = .90). The complete set
of questions with factor loadings and excluded items for all three scales are shown in
Appendix G.

An induction check was performed by conducting 2X2 AN OVA on the items
measuring the perceived relevance, credibility, and new information found in the
stimulus messages in order to examine the strength of the information utility
manipulation (low, high), and to determine if the second manipulated message variable,
exemplar discrepancy (low, extreme), unintentionally inﬂuenced these responses either
alone or through its interaction with the information utility manipulation. The ﬁrst
AN OVA was conducted on responses to the items measuring relevance. The analysis
produced a signiﬁcant main effect for information utility, F (1,207) = 28.55, p < .01, if =
.12. The means associated with this effect Show that individuals in the high information
utility condition (M = 3.98) scored higher than those in the low information utility
condition (M = 2.89). The higher score indicates that perceived relevance was higher for
those in the high information utility group. The main effect for exemplar discrepancy was
also signiﬁcant, F (1,207) = 6.64, p < .05, 1'13 = .03. The pattern of means shows that
information in the low discrepancy condition (M = 3.65) was perceived as more relevant
than information in the extreme discrepancy condition (M = 3.12). The interaction
between information utility and exemplar discrepancy was trivial, F < 1.

The second ANOVA was performed on the items measuring perceived credibility.

This analysis produced a signiﬁcant main effect for information utility, F (1,207) =

48

186.87, p < .01 , if = .48. The means associated with this effect Show that individuals in
the high information utility condition (M = 5.02) scored higher than those in the low 1
information utility (M = 3.13). The higher score indicates that credibility of information
was higher for those in the high information utility group. The main effect for exemplar
discrepancy and the interaction between information utility and exemplar discrepancy
were trivial, F < 1.

A third 2X2 was conducted examining new information. The analysis produced a
signiﬁcant main effect for information utility, F (1,207) = 44.26, p < .01, 1'12 = .18. The
means associated with this effect show that individuals in the high information utility
condition (M = 3.79) scored higher than those in the low information utility condition (M
= 2.67). The higher score indicates that new information was higher for those in the high
information utility group. The main effect for exemplar discrepancy was non-sigrriﬁcant,
F (1,207) = 2.79, p = .10, 1'12 = .01, and the interaction between information utility and
exemplar discrepancy was trivial, F <1.

Exemplar Discrepancy Induction Check. The strength of the exemplar
discrepancy induction was measured with two items. Inspection to determine whether or
not the two items formed a single scale indicated that combining the items would be
unreliable (or = .40). As such, the induction check was performed by conducting a 2X2
AN OVA on the two separate items to examine the strength of the exemplar discrepancy
manipulation (low, extreme) on responses to these items and, in addition, to determine if
the second message variable manipulation, information utility (low, high), unintentionally
inﬂuenced these responses either alone or through its interaction with the exemplar

discrepancy manipulation. The ﬁrst AN OVA was conducted on responses to the item

49

measuring whether the example given was “representative of the statistical information
presented.” The analysis produced a signiﬁcant main effect for exemplar discrepancy, F
(1,207) = 42.44, p < .01, 1'12 = .17. The means associated with this effect show that
individuals in the extreme discrepancy condition (M = 3.53) scored higher than those in
the low discrepancy condition (M = 2.72). The higher score indicates disagreement with
the statement, and showed that respondents in the extreme discrepancy condition felt the
example was less representative of the base-rate (i.e., more discrepant) than those in the
no discrepancy condition. The main effect for information utility and the interaction
between information utility and exemplar discrepancy were trivial, F<1.

The second AN OVA was performed on the item measuring the extent to which
respondents felt that the example about anxiety differed from the statistical information.
The analysis produced a signiﬁcant main effect for exemplar discrepancy, F (1 ,207) =
32.25, p < .01, 1'12 = .14. The means associated with this effect show that individuals in the
extreme discrepancy condition (M = 4.98) scored higher than those in the low
discrepancy condition (M = 3.82). This observation corroborated ﬁndings from the ﬁrst
analysis indicating that the examples in the extreme discrepancy condition were
perceived as more discrepant from the base-rate than those in the no discrepancy
condition. The main effect for information utility was again, non-signiﬁcant, F (1,207) =
1.11, p =.29, and the interaction between information utility and exemplar discrepancy
were trivial, F <1

Overall, the ﬁndings associated with the induction checks point to a successful
manipulation of information utility and exemplar discrepancy. The main effects for the

information utility manipulation on measures of perceived relevance, credibility and

50

amount of new information, and for the exemplar discrepancy manipulation on the two
items measuring perceived discrepancy indicate that the four news stories are capable of
creating the conditions needed to represent the populations desired for testing in the main
experiment. Although the main effect of the exemplar discrepancy manipulation on
perceived relevance showed that the extreme exemplar discrepancy reduced the perceived
relevance of a message, this ﬁnding is nether surprising or of great concern. It can be
understood simply as an indication that exemplars inconsistent with the rest of a story are
understood as less relevant to the story than are exemplars that are consistent with the
story. More important are the observations that a) the information utility induction
successfully inﬂuenced perceived information utility (i.e., perceived relevance, credibility
and amount of new information), and that b) the exemplar induction did not interact with
the information utility induction to alter perceived information utility. As such, the four
stories were deemed acceptable for use in this study.

Reliability of Outcome Measures. Responses to items used to measure the two
outcome variables (perceptual and behavioral outcomes) were inspected to determine the
quality of the data collected and to identify any abnormalities that might suggest
problems. Perceptual outcomes were assessed with two measures. The ﬁrst, population
estimates, was measured with a single item asking respondents to state the “percent of
college student smokers that have anxiety problems.” Examination of scores on this item
revealed no irregularities. The second, problem severity, was measured by a four-item
scale. Behavioral outcomes were measured with four items. Analysis of the multi-items
measures for problem severity and behavioral outcomes began by conducting

conﬁrmatory factor analysis (CFA) on each set of items and computing reliabilities on

51

the resulting scales. One of the four items measuring problem severity failed to meet the
established factor loadings criteria and was dropped. When this item was removed the
reliability for the three-item scale was a = .87. The complete set of questions with factor
loadings and excluded items for the scale are shown in Appendix H. Analysis on the
behavioral outcome measures revealed that the items were not reliable when combined to
a single scale (CF A showed each item meeting the selected criteria of .60, however, a =
.50). As such, the four items were retained as single-item indicators for analysis. Each
item was a 7-point scale with 1= “not at all” and 7 = “extremely.” The complete set of
questions and items for the behavior outcomes are shown in Appendix I along with factor
loadings and descriptive statistics of each item.
Main Study

Prior to the hypothesis tests, measures and manipulations used in the main study
were examined to inspect for abnormalities and determine data quality. Confrrrnatory
factor analysis (CFA) and reliabilities were computed to examine the measurement
quality of scales used to quantify issue importance, information utility, and outcome
variables. Additionally, t-tests and AN OVA were conducted to examine the strength of
the information utility induction.1 The results of these analyses are reported along with
accompanying descriptive statistics.2

Measurement Quality. Inspection of data quality began by conducting

conﬁrmatory factor analysis (CFA) on items measuring the issue importance respondents

 

' No induction check on the exemplar discrepancy manipulation was conducted as part of the main study.
This was done based on consideration of the facts that 1) the pilot study provided clear evidence that the
induction successfully manipulated exemplar discrepancy, and 2) any attempt to measure the strength of the
exemplar discrepancy induction as part of the main experiment would be biased by the experimental

ure.

No analyses were conducted on participant responses to questions related to terrorism and genetic

screening. These items were included only as a ruse to hide the studies focus on smoking related beliefs and
behaviors.

52

ascribe to smoking (the perceived magnitude, likelihood, and immediacy of threat) and
computing reliabilities on the three resulting scales. The ﬁndings were inspected to
determine the quality of the data collected and to identify any abnormalities that might
suggest problems. All three items for both the magnitude scale (a = .82) and the
likelihood scale (a = .72) met the established factor loadings criteria and passed the tests
for error. These items were averaged to create measures of magnitude and likelihood for
each respondent. One of the three items measuring immediacy failed the initial CFA test
and was dropped. The remaining two items were averaged to create a measure of
immediacy (a = .84). Descriptive statistics for the three scales can be seen in Table 3. The
complete set of questions with factor loadings and excluded items for all three scales are
shown in Appendix J.

Table 3

Descriptive Statistics for Issue Importance Scale

 

 

 

 

 

 

 

 

 

 

 

Variable N Mean SD
Low High

Magnitude 48 68 4.05 1.51

Likelihood 74 44 3 .59 l .35

Irnmediacy 172 23 1.88 1.15

Overall 98 52 9.50 3.29

 

 

The quality of items in the three subscales used to measure the perceived
information utility of the news stories on smoking-related anxiety (i.e., relevance, amount
of new information, and perceived credibility) were again examined for internal
consistency using CFA. The results for credibility and new information replicated the
ﬁndings from phase two of the pilot study. The three items measuring new information

once again met established criteria and were averaged to create a scale with reliability of

53

o. = .89. Analysis on the six items measuring perceived credibility also matched those of
the pilot study. The identical item failed the initial CFA test and was again dropped. The
remaining ﬁve items were averaged to create a measure of perceived credibility (a = .79).
The results for perceived relevance produced a slightly different outcome from the pilot
study. Whereas one of the four items fell just short of meeting established criteria for
scale inclusion in the pilot study, all four items met these established criteria here and
were aggregated to create scale reliable at a = .77. The complete set of questions with
factor loadings and excluded items for all three scales are shown in Appendix K.
Finally, CFA and reliability analyses were also conducted on scales used to
measure the outcome variables. One of the four items measuring perceived problem
severity failed the initial CFA test and was dropped. The remaining three items were
averaged to create a measure of perceived problem severity (u = .88). The complete set of
questions with factor loadings and excluded items for problem severity are shown in
Appendix L. Population estimates were measured with a single item, and no reliability
estimates were computed. One of the four items measuring behavioral intentions failed
the CF A test and was dropped. The remaining three items were averaged to create a
measure of behavioral intention (a = .77). The complete set of questions with factor
loadings and the single excluded item for behavior outcomes are shown in Appendix M.
Information Utility Induction Check. Following the approach used in the pilot
study, the strength of the information utility induction was examined with independent
samples t-tests conducted on each of the three information utility subscales.
Corroborating ﬁndings in the pilot study, the results show that respondents assigned to

the high information utility conditions scored higher on all the information utility

54

subscales including perceived relevance, credibility, and new information (see Table 4).
Once again, though the information utility of the news stories was not unusually high in
either condition, the signiﬁcant differences observed across all three dimensions of
information utility increases conﬁdence that the induction varied information utility as
intended.

Table 4

Results of t-tests on Information Utility Subscales.

 

 

 

 

 

 

 

 

 

N Mean SD t- value
Perceived HIGH 103 3.84 1.34 3.38"
Relevance LOW 1 12 2.88 1 .46
Perceived HIGH 103 4.61 .96 9.69M
Credibility LOW 1 1 1 3.28 1.03
New HIGH 108 3.86 1.06 3.29"
Information LOW 1 10 3.3 1 1 .39

 

 

 

 

 

 

"' p < .01
Descriptive Statistics. The descriptive statistics for all measures are shown in

Table 5. Table 6 shows the correlations among these variables and inductions.

55

Table 5

Means and Standard Deviations for Study Variables

 

 

N Mean SD Minimum Maximum

Age 219 19.66 1.54 18 28
Sex' 219 1.53 .50 1
Grade“ 216 1.88 .99 1
Race’ 219 1.59 1.42 l 7
GPA 214 2.95 .59 .06 4.0
lncome‘ 210 3.69 2.16 1 7
Issue Importance5 218 1.88 .87 1 3
Information Utility“ 220 1.51 .50 1 2
Exemplar Discrepancy 7 220 1.50 .50 1 2
Population Estimate 214 55.09 20.71 0 100
Problem Severity8 218 3.94 1.26 1 7
Smoking Behavior9 169 1.19 .39 1 2
Behavior Outcome (1 )'° 206 7.78 3.52 1 19

 

' Coded as a dichotomous variable with 1 = male and 2 = female.

’ Coded as a discrete variable with 1 = Freshman, 2 = Sophomore, 3 = Junior, and 4 =
Senior

3 Coded as a discrete variable with 1 = Caucasian, 2 = Aﬁican-American, 3 =
Asian/Paciﬁc Islander, 4 = Chicano-Latino, 5 = Middle Eastern, 6 = Native American,
and 7 = Other

‘Coded as a discrete variable with 1 = Below $25,000, 2 = $25-50,000, 3 = $50-75,000, 4
= S75-100,000, 5 = $100-150,000, 6 = $150-200,000, and 7 = $200 and above

’Issue importance, (l=low, 2 = high),

“Information utility (l=low, 2 = high)

7Exemplar Discrepancy (l=low, 2=extreme)

8Coded as 1 = “not at all” and 7 = “extreme”

9Coded as a dichotomous variable with 1=non smoker and 2 = smoker

loMerged behavioral intentions measure behavior items coded as 1= “not at all” and 21 =
‘6emmc9’

56

Table 6

Zero-Order Correlations between Study Variables

 

1 2 3 4 5 6 7 8 9 10 ll 12 13
1. Age 1.00

2, Sex' -.08 1.00

3. Grade’ .74b -.02 1.00

5. GPA ..09 .07 .14 ..17* 1.00

6. lncomc‘ -23" -.07 -.14 -.11 .04 1.00

7. Issue -.09 -.02 -.01 -.01 -.09 .03 1.00
lmportance’

8. Information -.00 -.0l .03 .10 -.09 .13 -.I4 1.00
Utility“

9. Exemplar ..16' .05 45' .00 .01 .01 .02 .01 1.00
Discrepancy’

10. Population .03 -.00 .05 ..12 ..01 .01 .04 ,3." -.11 1.00
Estimate

11. Problem -.02 ,17' -.01 .04 .08 -.07 .09 -.15 ,20" .35” 1.00
Severity'

12. Smoking .05 .04 .02 .00 -.12 .02 .13 -.01 .10 -.01 .mb 1.00
Behavior9

13. Behavior .02 .06 -.03 -.04 '-.08 45' .11 -.05 ,15' .07 .438 .03 1.00
Outcome (1)'°

 

' Signiﬁcant at p < .05 for two-tailed t-test.

b Signiﬁcant at p < .01 for two-tailed t-test.

' Coded as a dichotomous variable with 1 = male and 2 = female.

’ Coded as a discrete variable with 1 = Freshman, 2 = Sophomore, 3 = Junior, and 4 =
Senior

3 Coded as a discrete variable with l = Caucasian, 2 = African-American, 3 =
Asian/Paciﬁc Islander, 4 = Chicano-Latino, 5 = Middle Eastern, 6 = Native American,
and 7 = Other

4Coded as a discrete variable with 1 = Below $25,000, 2 = $25-50,000, 3 = $50-75,000, 4
= $75-100,000, 5 = $100-150,000, 6 = $150-200,000, and 7 = $200 and above

5Issue importance, (l=low, 2 = high),

6Information utility (l=low, 2 = high)

7Exemplar Discrepancy (l=low, 2=extreme)

8Coded as 1 = “not at all” and 7 = “extreme”

9Coded as a dichotomous variable with 1=non smoker and 2 = smoker

loMerged behavioral intentions measure, behavior items coded as 1= “not at all” and 21 =

“extreme”

57

Placement of Respondents into Conditions of High and Low Issue Importance.
After inspecting the data for quality assurances, the ﬁrst step in analysis of the main study
was to determine the placement of respondents into categories of smoking-related anxiety
issue importance (high, low). Based on this outcome, the second step in analysis was to
test hypotheses explicating the comparative inﬂuence of exemplar versus base-rate
information and the manner in which information utility is predicted to moderate this
inﬂuence. No analyses were conducted on participant responses to questions related to
terrorism and genetic screening since these items were included only to hide the study’s
focus on smoking related beliefs and behaviors.

Once the three scales were computed, the distribution of respondents was
examined to determine whether or not there existed distinct groups of individuals in the
population who scored high and low on the perceived importance of smoking-related
anxiety issues. Duplicating the pilot study, evidence of two distinct groups was deﬁned
operationally as the existence of a sufﬁcient number of respondents who scored high (5
or above on the 7-point scale) or low (3 or below on the scale) for each separate attribute
of issue importance (magnitude, likelihood, and immediacy) and a sufficient number who
scored high (between 15 and 21) or low (between 3 and 9) on overall issue importance.
As shown in Table 7, although the number of respondents ﬁtting the categorization of
high on immediacy is again somewhat limited, the number of respondents falling into the
high and low dimensions of all other categories is substantial. Most importantly, a
substantial number of respondents fall into both the low and high categories of overall
issue importance, with a noticeable improvement over the pilot study in the high issue

importance category.

58

Table 7

Descriptive Statistics for High and Low Issue Importance

 

 

 

 

 

 

 

 

 

 

 

Variable N Mean SD

Low High
Magnitude 48 68 4.05 1.51
Likelihood 74 44 3.59 1 .35
Irnmediacy 172 23 1.88 1.15
Overall Importance 98 52 9.50 3.29

 

 

Tests on Four Hypotheses. The four hypotheses were tested using a 2X2X2
AN OVA to examine the effect of issue importance (high, low), information utility (high,
low), and exemplar discrepancy (extreme, low) on perceptions of problem severity scores
and population estimates. Results for effects on problem severity Show main effects for
issue importance [F (1,147) = 29.18, p < .01, partial n2 = .17 1, information utility [F
(1,147) = 4.33, p < .05, partial r72 = .03 ], and exemplar discrepancy [F (1,147) = 6.00, p
< .05, partial if = .04] along with signiﬁcant effects for the interaction between
information utility and issue importance [F (1,147) = 7.67, p < .01, partial if = .05 ] as
well as between information utility, issue importance, and exemplar discrepancy [F
(1,147) = 3.97, p < .05, partial r72 = .03 ]. All other effects for problem severity were
trivial [F < 1]. The means and standard deviation associated with these analyses are
reported in Tables 8 and 9. Results for the effect on population estimates show a main
effect for information utility [F (1,143) = 6.36, p < .05, partial if = .05]. The main effect
for exemplar discrepancy [F (1,143) = 3.80, p =.054, partial if = .03] as well as the
interaction between information utility and issue importance [F (1,143) = 3.88, p = .051,
partial r72 = .03] were substantial, though falling right at instead of below the .05 level of

significance. The main effects for issue importance [F (1 ,143) = 2.17, p = .14, partial if =

59

.02] as well as the interaction between issue importance and exemplar discrepancy [F
(1,143) = 2.46, p = .12, partial if = .02] were notable though failing to meet the .05
standard. All other effects were trivial [F < 1]. The means and standard deviation
associated with these analyses are reported in Table 10 and 11.

Table 8

Means Associated with Perceptions of Problem Severity Main Eﬂects

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Mean SD 95%
Conﬁdence
Intervals
Information Utility Hig 4.35 1.06 4.1 1-4.65
Low 3.68 1.36 3.63-4.25
Issue Importance High 4.75 1.08 4.39-5.00
Low 3.61 1.19 3.46-3.89
Exemplar Extreme 3.72 1.28 3.68-4.23
Discrepancy Low 4.25 1.22 4.16-4.68
Table 9

Means Associated with Perceptions of Problem Severity Interaction Eﬂects

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Information Issue Exemplar Mean SD 95%

Utility Importance Discrepancy Conﬁdence

Intervals

High Low Low 4.21 .88 3.74-4.67

Extreme 4.06 1.13 3.61-4.51

Total 4.13 1.06 3.81-4.46

High Low 5.13 .87 4.59-5.68

Extreme 4.13 1.06 3.46-4.81

Total 4.73 1.05 4.20-5.20

Low Low Low 3.56 1.22 3.15-3.97

Extreme 2.87 l .08 2.47-3.27

Total 3.21 1.18 2.93-3.50

High Low 4.78 1.26 4.16-5.39

Extreme 4.75 1.05 4.14-5.36

Total 4.76 1.13 4.33-5.20

 

 

 

 

 

 

 

60

Table 10

Means Associated with Population Estimates Main Eﬁ’ects

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Mean SD 95% Conﬁdence
Intervals
Information 3131‘ 61.84 14.57 55.90-65.97
Utility Low 49.84 23.71 47.28-56.86
Issue Importance High 60.02 17.35 53.41-64.79
Low 53.16 22.04 49.92-57.91
Exemplar Extreme 52.72 21.95 47.97-58.20
Discrepancy Low 58.01 19.42 55.22-64.64
Table 11

Means Associated with Population Estimates Interaction Eﬁ’ects

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Information Issue Exemplar Mean SD 95%
Utility Importance Discrepancy Conﬁdence
Intervals
High Low Low 62.86 12.80 54.40-71.31
Extreme 60.77 13.55 52.51-69.04
Total 61.79 13.08 55.90-67.73
High Low 67.47 10.65 57.46-77.48
Extreme 52.67 22.38 39.74-65.59
Total 61.92 17.23 51.89-68.42
Low Low Low 46.31 24.82 38.70-53.91
Extreme 45.73 26.23 38.13-53.34
Total 49.84 23.71 40.64-51.40
High Low 63.08 1 1.94 51.89-74.28
Extreme 53.17 21.30 41.97-64.36
Total 58.13 17.63 50.21-66.04

 

 

 

 

 

 

 

 

Hypothesis One: The Eﬂect of Issue Importance. The ﬁrst hypothesis predicted
that there would be a main effect of issue importance on problem severity and population
estimates. The importance respondents ascribed to the issue of smoking related anxiety
was predicted to increase scores on both outcome variables. Support for this hypothesis

was found in the results for problem severity. Consistent with predictions, smoking

61

related anxiety was rated as a more severe problem by respondents scoring high on issue
importance (M = 4.75) than those scoring low (M = 3.61). Notably, though falling short
of accepted standards, the pattern of means for population estimates was also in the
predicted direction. Estimates by those who judged the issue as more important exceeded
estimates by those who judged the issue less important (M = 60.02% and 53.16%
respectively).

Hypothesis Two: The Eﬂect of Information Utility. The second hypothesis
predicted that there would be a main effect of information utility on problem severity and
population estimates. The level of information utility contained in each news article was
predicted to increase scores on these variables. Support for this hypothesis was found in
outcomes for both variables. The main effect on problem severity showed that
respondents exposed to the high information utility message (M = 4.3 5) considered the
problem to be more severe than those in the low information utility group (M = 3.68).
Similarly, the effect on population estimates shows that respondents in the high
information utility group (M = 61.84) gave higher estimates of the percent of college
student smokers with anxiety problems than those in the low information utility group (M
= 49.84).

Hypothesis Three: The Eﬂect of Exemplar Discrepancy. The third hypothesis
predicted that there would be a main effect of exemplar discrepancy on problem severity
and population estimates. Since the extreme discrepancy message in this study used
exemplars that underrepresented the extent of the problem stipulated by the base-rate
(and epitomized in the low-discrepancy message), exemplar discrepancy was predicted to

reduce scores on both outcome variables. Support for hypothesis three can be seen in the

62

results for problem severity. Consistent with this hypothesis, the problem was rated as
less severe by respondents in the extreme exemplar discrepancy group (M = 3.72) than
those in the low discrepancy group (M = 4.25). Similarly, the pattern of means for
population estimates was also in the predicted direction. Respondents exposed to the
extremely discrepant exemplar gave lower estimates of the percent of college student
smokers with anxiety problems (M = 52.72%) than those exposed to the low-discrepant
exemplar (M = 58.01%).

Hypothesis Four: Issue Importance, Information Utility, and Exemplar
Discrepancy. The ﬁnal hypothesis predicted an interaction effect of issue importance,
information utility, and exemplar discrepancy on perceptions of problem severity and
population estimates. Issue importance and information utility were predicted to
moderate the effect of exemplar discrepancy on both variables such that the effect of
exemplar discrepancy would be weakened when both issue importance and information
utility were high. Since the exemplar discrepancy in this study underrepresented the
extent of smoking anxiety, the weakened effect of exemplar discrepancy should be
observed by the unique failure of the exemplar discrepancy induction to reduce scores on
outcome variables when both issue importance and information utility were high.
Although the three-way interaction predicted by this hypothesis was found signiﬁcant for
measures of problem severity, the pattern of results was not consistent with the failure to
observe an exemplar discrepancy effect under the unique conditions predicted.
Nevertheless, the patterns observed in this three way interaction for problem severity as
well as the two-way interactions of information utility and issue importance for measures

of problem severity and population estimates are informative.

63

The Three- Way Interaction of Information Utility, Issue Importance and
Exemplar Discrepancy. The means associated with the three-way interaction of issue
importance, information utility, and exemplar discrepancy on perceptions of problem
severity are diagramed in Figure 1. This ﬁgure shows distinct patterns for respondents
scoring high and low on issue importance.

Figure 1

Perceptions of Problem Severity for Respondents High and Low on Issue Importance

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5.5
5 O\ _
4.5 >
/ ----Lowlssuelmportance
4 .: ' 'e . . — High Issue Importance
~ ‘ . ‘ I ’ ~ . A Extreme Discrepancy
3.5 k . ’
. ‘ . ‘ . Low Discrepancy
3 - g
A
2.5
2 ﬂ
High Low
Information Utlllty

 

 

 

 

Among respondents low on issue importance, evidence of the exemplar
discrepancy effect is apparent only for those exposed to the low information utility
message (see Figure 1). This is indicated by the observation that problem severity scores
in the extreme discrepancy group are noticeably lower (M = 2.87) than those in the low-

discrepancy group (M = 3.56). By comparison, for those exposed to the high information

utility message, problem severity scores in the extreme discrepancy group (M = 4.06)
differ little if at all from those in the low-discrepancy group (M = 4.20).

Conversely, among respondents high on issue importance, the pattern of means
shows that the effect of exemplar discrepancy is strongest when information utility is
high (see Figure 1). Here again, exempliﬁcation is indicated by the fact that problem
severity scores in the extreme discrepancy group (M = 4.13) are noticeably lower than
those in the low-discrepancy group (M = 5.13). By comparison, for those exposed to the
low information utility message, problem severity scores in the extreme discrepancy
group (M = 4.75) differ little if at all ﬁ'om those in the low-discrepancy group (M = 4.78).
Evidence that the effect of exemplar discrepancy is stronger when both issue importance
and information utility are high is in sharp contrast to hypothesis four’s prediction that
exempliﬁcation would be weakened by these conditions, and the expectation that the
unique failure of exemplar discrepancy to reduce perceptions of problem severity would
be observed under these conditions.

The Two- Way Interaction of Information Utility and Issue Importance. The
pattern of means associated with the signiﬁcant interaction of issue importance and
information utility on perceptions of problem severity can be seen in Figure 2. Inspection
of these means shows that the respondents in the low issue importance, low information
utility group (M = 3.21) considered the problem to be less severe than those in all other
conditions including low issue importance, high information utility (M = 4.13), high issue
importance, high information utility (M = 4.73), and high issue importance, low

information utility (M = 4.76).

65

Figure 2

Interaction eﬁ’ect of information utility and issue importance on perceptions of problem
severity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5.5
5
. I
4.5
4 I . - ~ High Issue Importance
3.5 Nu” “mtgwlssuelmportanca
’I
3 .a
2.5 .-
2 ,
High Low
Information Utility

 

 

 

A similar pattern can be seen in the means seen in estimates of the percent of
college student smokers with anxiety problems (see Figure 3). Inspection of these means
shows that respondents in the low issue importance, low information utility group (M =
46.02%) gave much lower estimates of the pervasiveness of smoking related anxiety than
those in the low issue importance, high information utility group (M = 61.79%), the high
issue importance, high information utility group (M = 61.92), and high issue importance,

. low information utility (M = 58.13).

66

Figure 3

Interaction eﬂ'ect of information utility and issue importance on population estimates

 

 

 

 

 

 

 

 

 

 

 

 

 

 

85
55 ‘t ‘ High Issue Importance

“. ----Lowlssuelmportance
50 “

‘-
45
4o .
High Low
information Utility

 

 

 

 

Research Question: The inﬂuence of Exemplar Discrepancy on Behavior
Intentions. The last analysis examined the research question asking if message factors
could inﬂuence behavioral intentions. A 2X2X2 AN OVA was conducted on the
combined behavioral intentions scale to determine if the three message factors
(information utility, issue importance, and exemplar discrepancy) had any separate or
combined inﬂuence on behavioral intentions. The analysis produced a main effect for
issue importance, F (1,140) = 38.50, p < .01, 1'12 = .23. Respondents in the low issue
importance group (M = 6.58) were less likely to believe their behavior would be affected
by the information than those in the high issue importance group (M = 10.43). No other
effects were signiﬁcant. The main effect for exemplar discrepancy produced, F (1 ,140) =
2.00, p = .16. The interaction between issue importance and exemplar discrepancy

produced, F (1,140) = 1.52, p = .22. All other effects were trivial, F < 1.

67

Discussion

Not surprisingly, the investigation found that both issue importance and
information utility were important determinants of perception and intention following
exposure to the report on smoking related anxiety. Similarly, and more notably for the
study at hand, the results show an effect for exemplar discrepancy consistent with the
body of literature from previous research on news and political media. This observation
should be of particular interest to scholars who study the use of media to inform and
shape behavior in the ﬁeld of public health. Most notable, however are the ﬁndings
associated with the two-way and three-way interactions observed for issue importance,
information utility, and exemplar discrepancy’s effect on population estimates of
smoking related anxiety and perceptions of problem severity. Though not consistent with
the predictions set forth in hypothesis four, the pattern of ﬁndings associated with these
interactions suggest insights and raise questions about exempliﬁcation that have
important consequences for theory and practice. This section elaborates on these ﬁndings
and examines their implications for understanding the effect of base-rates versus
exemplars and the conditions that moderate their inﬂuence in a health communication
context. It ends with a discussion of the limitations of this research and suggestions for
future directions.
The Inﬂuence of Message Features on Problem Perceptions
Issue Importance

Hypothesis one predicted that when individuals were exposed to a news story
about a health issue problem, respondents who ascribed high importance to the issue

(here smoking related anxiety) would give higher ratings of the problem’s perceived

68

severity and higher estimates of the percent of the population affected by the problem
than respondents who ascribed low importance to the issue. The analysis on the inﬂuence
of issue importance produced a signiﬁcant main effect on perceptions of problem severity
and a notable main effect on population estimates.

The patterns of ﬁndings associated with these results are consistent with
hypothesis one. Individuals ascribing higher importance to the issue of smoking related
anxiety gave signiﬁcantly higher ratings of the problem’s perceived severity than
respondents who ascribe low importance to the issue. Although the ﬁndings for
population estimates fell short of the criteria set for rejecting the null hypothesis, the data
patterns were in line with predictions. Those within the high issue importance group gave
higher estimates of the percent of smokers affected by anxiety

These ﬁndings are consistent with the notion that individuals high on issue
importance will respond differently to messages then those that are low. As Salmon and
Atkin (2003) stated, a message must be involving to an individual in order for there to be
a message effect. When the information was presented to those high on issue importance,
individuals were more inclined to perceive the problem of smoking related anxiety as
more severe than those low on issue importance. This pattern was also seen in the means
for population estimates, though the differences did not reach levels of signiﬁcance.

These ﬁndings indicate that issue importance alone can inﬂuence perceptual
reactions to the information presented in a message containing exemplars. If we interpret
this in line with logic from the Elaboration Likelihood Model suggesting that the more an
issue is perceived as important the more likely individuals are to attend to a message, the

heightened perceived severity observed here might indicate the presence of a strong

69

message. Otherwise, closer scrutiny might have lead to the type of counter arguing that
should reduce perceptions of severity. This rationale calls attention to the ELM’s ability
to help explain how issue importance can inﬂuence message effects in the health
communication domain in general, and speciﬁcally with anxiety caused by smoking.

The results replicate prior work showing that issue importance can be used to
forecast people’s perceptions and reactions to persuasive messages (Millar & Millar,
2000). Moreover, by reproducing this ﬁnding on perceptions of smoking related anxiety,
they strengthen the view that issue importance has critical implications for research in
health communication. In line with Skumanich and Kintsfather’s (1996) research on
organ donation showing that issue importance predicts the persuasive effect of a message,
these ﬁndings suggest that factors strengthening involvement can facilitate the intended
message effect. Though the increased perceptions of problem severity and population
estimates found among those within high issue importance comes as no surprise in this
instance, we should not overlook the importance here of its independent effect or its
potential to shape the effect of other forces.
Information Utility

Hypothesis two predicted that exposure to the news story on smoking related
anxiety would lead to higher ratings of the problem’s perceived severity and higher
estimates of the population affected by the problem when the story contained high levels
of information utility. The analysis on the inﬂuence of information utility produced
signiﬁcant main effects both on perceptions of problem severity and population
estimates. The means associated with both effects were in line with hypothesized

directions. Exposure to a message high on information utility produced higher estimates

70

of the percent of the population effected by smoking-related anxiety problems, and
increased perceptions the problem’s severity. These outcomes support the notion that the
information utility contained in a news story can increase perceptions of a problem’s
severity and its pervasiveness throughout a population. The results replicate and extend
the ﬁndings of researchers such as Thompson (1975), Trenholm (1989), and Perloff
(2001) to the area of health communication, and show that individuals are affected more
strongly by credible and relevant messages that contain new information than by
messages low on these attributes.

In line with the counter arguing explanation above, the ﬁndings are consistent
here with research suggesting that increased information utility stemming from source
credibility can minimize counter arguing with the evidence and assertions found in a
message (Buda & Chamov, 2003). Perhaps respondents in the low information utility
group discounted information given within the news story. In this case, we should expect
the population estimates and perceptions of problem severity they report to be unaffected
by evidence in the story, and to locate closer to the 40% average estimate observed in the
pilot study among respondents never exposed to the message. The outcomes are also
consistent with research suggesting that information utility results in longer exposure and
attention to messages (Knobloch et al 2003; Sears & Freedman, 1967; Wilton & Myers,
1986). Along similar lines of reasoning as the Elaboration Likelihood logic arguing that
increased attention should limit the effect of weak messages, longer exposure may have
promoted greater cognitive processing of the information presented within the news

story. In this case, the strong evidence presented in the message may have resulted in

71

higher estimations of college student smokers suffering from anxiety and higher
perceptions of problem severity.

According to Knobloch et al. (2003), information utility inﬂuences individual
responses to stories such that when information is found to be useful it can activate
curiosity and exposure. This type of outcome might alter the perceived level of interest in
and subsequent attention given to the information provided. If this is the case,
information utility might alter the level of issue importance that individuals bring to an
exposure setting. The importance of this possibility should not be overlooked. It suggests
the potential for information utility, a message factor, to interact with issue importance
during an exposure setting in a manner that would change this critical recipient attribute.
For health communication professionals, this ﬁnding suggests that messages should be
constructed with information utility in mind. Effectiveness can be maximized by
designing messages containing credible information that the target audience considers
relevant and new.

Exemplar Discrepancy

Hypothesis three predicted that exposure to a news story about a health issue
problem that contains high exemplar discrepancy will bias (in the direction of the
exemplar) respondents’ ratings of the problem’s perceived severity and estimates of the
percent of the population affected by the problem. These ratings were predicted to exceed
those of respondents exposed to a news story that contains low exemplar discrepancy.
The analysis on the inﬂuence of exemplar discrepancy produced one signiﬁcant main
effect on perceptions of problem severity and a substantial main effect on population

estimates.

72

The pattern of ﬁndings associated with these results is consistent with hypothesis
three. Exposure to a message with extremely discrepant exemplars produced lower
estimates of the percent of the population effected by smoking-related anxiety problems
and decreased perceptions the problem’s severity. Since the discrepant exemplars used in
the news story showed cases in which the problem was less severe than the summary
base-rate information reported, the lowered estimates and reduced perceptions show
predicted change in the direction of the exemplar.

These ﬁndings are consistent with the exempliﬁcation theory and the notions that
individuals are more likely to pay attention to exemplars than base-rate information. The
premise of exempliﬁcation is that individuals not only pay more attention to but are
inﬂuenced more by exemplars than by base-rate data (Gibson & Zillmann, 1998). It
appears that when the exemplar and base-rate information presented were extremely
different participants were more likely to attend to the exemplar information than the
base-rate, especially for perceptions of problem severity. When the base-rate and
exemplar were different (in this case meaning that the accompanying exemplars
portrayed cases without anxiety) perceptions of problem severity declined (as did
population estimates).

The results replicate earlier exempliﬁcation research demonstrating that the
perceptions of respondents exposed to news stories are more strongly inﬂuenced by
concrete exemplars than abstract base-rate information, and extends this research to areas
of health communication to show that whereas individuals might consciously believe that
base-rate information is more reliable, perceptions are likely to be swayed by a few

examples that deviate from statistical facts. The consequence of extending

73

exempliﬁcation theory to the ﬁeld of health communication should not be overlooked.
Exempliﬁcation research on news and political media has shown the importance of this
phenomenon to everyday life. It demonstrates that individuals often base their assessment
of social reality more strongly on attributes of the exemplars selected for inclusion in
media reports than statistical representations of information. This has been found
problematic when the exemplar is atypical and chosen for its entertaining or sensational
qualities, a common occurrence within news stories. The same can be expected in health
communications, where the ramiﬁcations might be of equal if not greater consequence.
Evidence here suggests that exemplars used in messages designed to educate or inﬂuence
perceptions of public health issues will play an important role in recipient response.
Decisions to include exemplars selected for their typicality versus their entertaining and
sensational nature should be made with this understanding in mind. Health
communicators should be made aware of how exemplars affect the responses and use this
knowledge when considering message construction. Their decisions should vary as a
function of their goals. Are exaggerated perceptions consistent with the message
designer’s objectives? In this case the entertaining and attention-grabbing qualities of
extremely discrepant exemplars may be desirable. At the same time, health
communicators should be aware that the simple inclusion of statistical base-rate
information is unlikely to correct errant perceptions that can result ﬁ'om the use of
distorted exemplars.
Interactions of Issue Importance, Information Utility, and Exemplar Discrepancy

As in all cases, the manner in which main effects are interpreted and understood is

qualiﬁed by the observation of an interaction. In the present study, evidence of main

74

effects was observed for all three of the focal predictor variables. The question then
becomes one of understanding the conditions under which we might expect the effect of
each predictor to vary as a function of conditions created by the presence of the other
factors. The main effects tell us that ascribing greater importance to an issue increases
perceptions of problem severity, population estimates, and behavior intent, and that both
the presence of exemplars and the information utility contained in a message can have an
effect on perceptions of problem severity and population estimates. Do the moderating
conditions created by different combinations of issue importance, information utility, and
exempliﬁcation strengthen, weaken or even invert the independent effect observed for
each variable. The two-way and three-way interaction observed for perceptions of
problem severity suggests that complex processes may be at work, and calls for close
inspection. Perhaps the best way to understand the manner in which these factors interact
is to consider the patterns associated with these interactions.

The two way interaction between issue importance and information utility on both
perceptions of problem severity (see Figure 2) and population estimates (see Figure 3) are
informative here. Mean scores associated with the effect on problem severity show that
whereas issue importance increases perceived severity, the effect of information utility on
these perceptions is seen mostly (if not exclusively) among those low on issue
importance. When issue importance was high, increased information utility failed to
heighten perceptions of problem severity. The ﬁnding is even more striking on estimates
of the percent of the population affected by smoking anxiety. Once again, the effect of
information utility is seen, for all intents and purposes, only when issue importance is

low. Moreover, the effect of high information utility on low issue importance respondents

75

is so strong that it all but eliminates any inﬂuence of issue importance on this measure.
This can be seen in the population estimates for those high and low on issue importance,
where means are essentially equal when information utility is high (M = 61.92 and M =
61.79 respectively). Notably, the population estimate here for low issue importance
respondents exposed to the high information utility message (M = 61.79) actually comes
up to and slightly exceeds estimates for high issue importance respondents exposed to the
low information utility message (M = 58.13). This shows clearly that the effect of issue
importance is all but eliminated by a high information utility message, an observation
whose understanding might have important implications for health communicators. Why
is it the case that a message with high information utility can alter the way low issue
importance respondents react to a health message? Did the information utility in the
message change the respondent’s perception of issue importance? Did it affect
respondent reactions through some other mechanism without changing perceived issue
importance? Is something unique to the issue at hand? The data at hand do not allow us to
examine these questions, but their answers seem pertinent to health communication
specialists.

Of course, before we become too focused on the two-way interaction, we need to
consider how the three-way interaction between issue importance, information utility, and
exemplar discrepancy might inﬂuence our understanding of the processes at work. As
already stated, the three-way interaction did not produce results consistent with
hypothesis four’s prediction that we would see a uniquely weak exempliﬁcation effect

when issue importance and information utility were high. How can this be understood?

76

Both issue importance and information utility show their expected separate effects
on perceived severity under all conditions except one. Perceived severity scores were
lower than expected for respondents high on issue importance who were exposed to a
message containing high information utility only in the extreme exemplar discrepancy
condition. All other means were consistent with predictions that issue importance and
information utility would increase perceived severity. However, these factors failed to
increase perceived severity when the message contained extremely discrepant exemplars.
This aberration, it seems, is responsible not only for the three-way interaction between
issue importance, information utility, and exemplar discrepancy, but also the two-way
interaction between issue importance and information utility. As such, it deserves greater
attention. What does this outcome tell us? How can it be understood? Hypothesis four
predicted that exempliﬁcation would be uniquely weak when issue importance and
information utility are high, but the data suggest an exemplar effect occurred precisely
under these conditions.

Previous research shows that exempliﬁcation effects are robust. Daschmann’s
(2001, as reported in Brosius, 2003) report suggests that these effects are even strong
enough to overcome any modifying inﬂuence from prior opinions and recipient
involvement. Brosious (2003,) maintains that message recipients in media settings
seldom attend closely enough to messages or think carefully enough about the
information in them to comprehend base-rate information and how it differs from an
exemplar. As a result, judgments are biased by use of a representativeness heuristic
which accounts for the occurrence of exempliﬁcation. If Brosious is correct, we might try

to interpret patterns showing when exempliﬁcation effects did or did not occur in the

77

present data as indicators of attentional vigilance. In this case, the absence of
exempliﬁcation would denote conditions in which awareness and deliberation were high
enough to overcome heuristic bias. Given the presence of the three-way interaction, it
seems that the conditions governing this bias can result from a complex combination of
factors.

The observations in this study suggest that exemplar effects may depend on the
information utility in a message. Notably, our results are consistent with Daschmann’s
(2001, as reported in Brosius, 2003) claim that exempliﬁcation can occur in respondents
who are both high and low on issue involvement. However, the stability of
exempliﬁcation across levels of issue involvement changes when information utility is
considered. Unexpectedly, information utility alters the exempliﬁcation effect for those
high and low on issue involvement. Exempliﬁcation effects are absent among low issue
involvement respondents when information utility is high. At the same time,
exempliﬁcation is not observed for those high on issue involvement when information
utility is low. What does all this mean?

In line with Daschmann, the ﬁndings here show evidence of exempliﬁcation
effects among low issue involvement respondents, though only when information utility
was low. Clearly, these conditions seem best suited to promote the type of effortless
mental processing said to cause exempliﬁcation, and the absence of exempliﬁcation
effects here might have called our procedure into question. Showing evidence of the
effect where its occurrence is most certain while at the same time seeing it absent under
other conditions is consistent with an interpretation that hypothesized moderation

occurred.

78

Although inconsistent with both Daschmann’s report (and the outcome predicted
here) it is not unfathomable to ﬁnd that exempliﬁcation effects were absent among those
high on issue importance when information utility was low. Surely, this ﬁnding would
not have been surprising to Daschmann, whose observations are more notable for
showing that exempliﬁcation effects were robust enough to overcome the type of careful
information processing expected among those highly involved. If Daschmann’s research
showed that exempliﬁcation processes are robust enough to bias message processing
even among those high on issue involvement, perhaps our ﬁndings suggest that this is
limited to situations where information utility is low or where other factors might curtail
the type of close attention to message information expected from the highly involved.

Similarly, it is not surprising to ﬁnd exempliﬁcation effects absent among low
issue importance respondents exposed to a high information utility message. Though we
began by predicting that exempliﬁcation would only be absent (or uniquely weakened)
when those high on issue importance were exposed to a high information utility message,
it is not hard to understand how a high information utility message could foster the type
of close attention and scrutiny that would negate the inﬂuence of discrepant exemplars.

The failure to ﬁnd evidence of exempliﬁcation effects either in the high
information utility, low issue importance or the low information utility, high issue
importance conditions is not perplexingly. It can be understood as indicative of the
capacity for both information utility and issue involvement to elevate levels of attention
and inhibit exempliﬁcation. However, observing that exempliﬁcation effects among high
issue importance respondents occurred only when information utility was high is

undeniably a surprise. Evidence of strong exemplar discrepancy effects appeared

79

precisely under those conditions in which hypothesis four predicted uniquely weakened
exempliﬁcation. Why should we ﬁnd exempliﬁcation in conditions that should promote
the highest levels of attention to the message while, at the same time, ﬁnding it absent in
conditions less well suited to do this? One possibility is that the exemplar discrepancy
comes so unexpectedly in this situation that it resonates with the recipient.

A high informational utility message delivered to respondents with high issue
importance might be thought of as akin to a priest preaching to the choir. The high utility
message is similar to a priest (credible and relevant) whereas those ascribing high
importance to the issue are like the choir. When the priest uses a discrepant exemplar the
choir thinks it must be meaningful. Among respondents who think the issue is important,
a message with high utility might be considered particularly noteworthy. Discrepant
exemplars in this situation should be looked at very closely and given considerable
weight. When the discrepancy is an example that underrates the problem, it might push
perceptions down below what they would be when no discrepancy occurs. This is the
pattern we see here when comparing low and extreme exemplar discrepancy for those in
the high informational utility, high issue importance conditions.

No doubt, this observation was so unexpected that to some it might seem almost
inexplicable. Moreover, given that this study used a single message topic with a relatively
small sample, caution is called for when considering the reliability and scope of this
outcome observed in this study. Nevertheless, the ﬁndings observed here have important
implications for exempliﬁcation theory, as well as the application of this theory to health

communication and other ﬁelds of practice.

80

The Influence of Message Features on Behavior Intentions

The research question explored the potential for issue importance, information
utility, and exemplar discrepancy to affect behavior intentions individually or in
combination. Analysis on the separate and combined inﬂuence of these factors on
measures of behavioral intentions produced a signiﬁcant effect for the inﬂuence of issue
importance and a notable effect for exemplar discrepancy.

The effect for issue importance shows that respondents in the high issue
importance group were more likely to believe that their own smoking behaviors will be
affected by the news stories, more likely to think the news stories will lead others to quit
smoking, and more likely to seek out information regarding anxiety caused by smoking.
By contrast, respondents low on issue importance were less likely to believe their
behaviors would be affected. This observation, that issue importance affects behavioral
intentions, is most notable in view of the fact that one of the items measuring behavior
intention speciﬁcally looks at the likelihood of seeking information about anxiety. This
connects the ﬁndings here on issue importance to existing research on information
processing and selective exposure.

This study was consistent with prior research on selective exposure which shows
that individuals who consider an issue to be high in importance are more likely to seek
out and pay attention to information pertaining to the issue. The higher level of social
signiﬁcance and personal value one attributes to an issue the more important that issue
becomes. As a result, individuals are more likely to actively gather information on the
topic (Aﬁﬁ, Dillow, & Morse, 2004). Evidence of this effect is apparent in studies

investigating exposure to online news showing that people pay closer attention to stories

81

that are high on issue importance (Knobloch, Carpentier, & Zillmann, 2003) and as well
as research on the selection of entertaining media content (Atkin, 1985).

The inﬂuence of issue importance on selective exposure can be found speciﬁcally
with regard to public safety and health in research showing that individuals with a higher
likelihood of falling prey to a threat are more careful when processing information about
an issue than their counterparts (Knobloch, et al. 2003). This notion is central to parallel
processing models which argue that when perceived threat is low individuals are less
likely to carry out further processing of a message (W itte, 1992). The results of the
present study are consistent with these notions, showing that the behavior intentions of
those within the low issue importance condition seemed unaffected by the information
presented. This would be the expected outcome in situations where the information
processing ceased among the low involvement individuals.

Implications for Health Communication Research

The ﬁndings of this study have important implications for increasing attentional
vigilance to health messages. The data show that when issue importance is high the intent
of respondents to seek out information associated with relevant health issues increases.
Since individuals reporting high levels of issue importance are more likely to attend to
the message more closely, we might expect the response of these individuals to differ. If
high levels of magnitude, likelihood, and immediacy of an issue induce greater attention
among the target population, we should expect the strength of evidence and arguments to
increase the effectiveness of exposure, and the effect of other message elements to suffer.
On the other hand, the same cannot be said for low importance individuals. Not only

should we doubt that strong evidence and arguments will increase message effectiveness,

82

in these situations the effect of ﬂawed evidence may play a stronger role. Although the
effect of exemplar discrepancy failed to meet the .05 standard on measures of behavior
intention, the logic here is consistent with expectations that a potentially strong effect can
result from discrepant exemplars that misrepresent population characteristics. Unlike the
situation with high issue importance, among low issue importance respondents
exaggerated exemplars might be successful in producing the desired behavioral intent.
With this in mind, health communicators should pay close attention to the issue
importance level of an intended target, as various message variables are likely to fare
quite differently for high and low issue importance recipients. For example, research on
threatening messages shows that high fear inducing messages (commonly induced using
powerful exemplars) can cause people to focus more on ﬁnding ways to deal with the
emotional trauma associated with the threat than to address the cause of the threat directly
(cf., the extended parallel process model, Witte, 1992). Potentially, in cases where the
emotional trauma induced by a threatening exemplar becomes too severe for the
individual to endure, strong arguments in the message addressing the cause of the threat
may be ignored and result in diminished message effect. One can imagine how preteens
ﬁrst told of the dangers from smoking or addictive drugs might experience extreme levels
of fear from vivid exemplar displays, or how adolescents newly introduced to the threat
of sexually transmitted diseases might respond to repulsive portrayals. If levels of fear
provoked by extreme exemplars reached heights interfering with close attention to
abstract solutions presented in a sound health message, it is little wonder that message
retention in this situation would be limited to simple notions. Perhaps this explains why

simple slogans like “just say no” are easily and well learned when adolescents are ﬁrst

83

introduced to such risks, and why they quickly lose their effect. Attentive audiences age
and may learn ﬁ'om experience that the simplicity of such slogans are simply a case of
misexempliﬁcation, just like the extreme exemplars that ﬁrst provoke their fear.

In this sense, the outcomes observed here have important implications for health
message construction. The ﬁndings show that information utility and issue importance of
messages play a role in perceptions of problem severity and population estimates. Again,
using the example of fear appeals, if recipients exposed to a credible, relevant message
containing new information perceive moderate levels of problem severity, we might
expect the audience to respond to the danger rather than the fear (Witte, 1992). If a
threatening health risk message focuses the recipient’s attention on sound solutions
presented in the message, it might prompt the type of attentional vigilance needed to
extract even those solutions that are somewhat abstract andOnot so easily grasped at a
glance. The ﬁndings here suggest that high issue importance and information utility
produced population estimates much closer to the abstract base-rate reported in the news
story (68%) than the baseline established in the pre-experiment survey (40%). This
ﬁnding might indicate that abstract base-rate information can play a part in shaping
perceptions of problem severity for high-involvement individuals. Moreover, it is
possible that messages containing high information utility can have a similar effect,
whether through the information utility’s capacity to engender high issue involvement or
through some other cognitive mechanism. This adds to existing work on the Health Belief
Model by identifying ways to increase perceived severity and the likelihood of
subsequent behavior change. When a health threat is seen as likely, immediate, and

severe individuals are more prone to develop behavior intentions consistent with the

84

message intent. As such, health communicators would be well advised to construct
credible and relevant messages that contain new information for respondents.

Notably in this regard, in some situations the use of stories or examples may be
considerably less important than making sure that information utility and issue
importance are elevated to sufﬁcient levels. When this is the case individuals should pay
closer attention to strong arguments and statistical evidence. The most promising
conditions for message effectiveness are presented when base-rates and exemplars are
non-discrepant while information utility and issue importance are high for the targeted
population.

Though not speciﬁcally addressed in the research at hand, it is not difﬁcult to
imagine a situation in which a message’s attributes could be altered to increase the
importance of an issue along with the utility of the information it contains; even when the
target population’s initial perception is that the issue is relatively unimportant. We might
even imagine that this is the norm when new health threats are introduced to the public
through information campaigns. Evidence of this is apparent in the observation that
though few individuals actually engaged in smoking, news story attributes manipulated in
this study effectively altered levels of high and low issue importance. Perceptions of issue
importance increased among groups of respondents even though few of them smoked.
From this we might reason that stories about health threats can alter the behavior
intentions of people even when the story deals with an issue unrelated to their own
behavior. Given these circumstances, health communicators concerned with issue
importance must identify not only those trait variables that distinguish groups inherently

high and low on issue importance, but also those message variables that can be used

85

within a story to increase state issue importance. The only requirement is that state
importance persists long enough for message processing to be completed. Unlike the
typical media setting identiﬁed by Brosious (2003) under this scenario message recipients
should attend closely enough to understand how base-rate information differs from
exemplars, and use that knowledge information when forming perceptions based on
exposure.
Limitations

Several limitations in the design and implementation of this study must be
considered when appraising the level of conﬁdence we can have in the outcomes of this
research. These include issues associated with the representative nature of the sample, the
representative nature of a single issue (smoking related anxiety) selected from the
population of all issues, the reliability of the single-item measure of population estimates,
the cross-sectional nature of the design, and the unsupervised nature of the online survey.

The ﬁrst issue deals with several limitations associated with the sample selected
for study. Most conspicuously here, use of a student sample in this study raises questions
about the ability of the observations here to generalize to other populations. In this
instance a student sample is reasonable for several reasons. To begin with, often overused
criticisms directed at the use of student samples should be applied sparingly and only
when there is reason to suspect problems in this regard. Research examining the
outcomes from studies on these different populations shows no signiﬁcant differences
between student and non-student samples across several major disciplines conducting
research on human behavior (Locke, 1986). Considering the case of exempliﬁcation in

particular, expectations that responses from these two populations should differ is even

86

less likely since this research examines experimentally a type of unconscious reaction
that stems from the phylogenetic roots of the human information processing system. If we
consider the context of the particular experience encountered by respondents in this
study, the media exposure setting used, is a common real-world occurrence for college
students. As such, threats associated with mundane realism are negligible. Moreover,
because college students represent an age group that is frequently the target of health
messages (e. g. reducing binge drinking on college campuses, safer sex, and HIV
prevention), in this instance they represent the potential target audience for the practical
applications of this work. As with many health messages, properly targeting an audience
or segment of the population is important to ensure campaign success, giving good cause
for the use of the small and relatively homogenous sample here.

Another issue of note in this sample is that only 52 subjects rated anxiety from
cigarette smoking as high on issue importance. Two questions come immediately to mind
when considering this concern. The ﬁrst question we must ask deals with the extent to
which the percent of respondents in this sample ascribing high issue importance to
smoking is representative of other populations. With no baseline from the general
population is it hard to answer this question. One observation related to this issue deals
with the percent of smokers found in our sample compared to the percent in the
population at large. The present study had 14.5% smokers compared to 23.6 % of
smokers 18-24 in the population at large (Center for Disease Control, 2005), a relatively
small percent in both cases. Though this comparison does not answer the question of
whether or not the issue involvement with smoking in our sample was representative of

the population at large, it does suggest this sample is similar to the population at large in

87

terms of having a vast majority of nonsmokers, though the sample here had slightly fewer
smokers. The second question we must ask deals with the fact that having 52 subjects in
the high issue importance condition limits the power available for analyses, a particular
concern in tests of the three-way interactions.

The second issue deals with limitations associated with the representative nature
of single issue selected for use in this study. One again, two issues come quickly to mind.
First, since smoking is only one issue for a large population of health and non-health
related issues, we must ask about the extent to which unique characteristics of this issue
are responsible for some of the patterns observed. The fact that this research successfully
replicated the main effects observed in previous research on issue involvement,
information utility, and exempliﬁcation supports the contentions that the ﬁndings
observed here are not limited to the single issue. At the same time, the most important
ﬁndings observed in this study, those dealing with two-way and three-way interaction
among these variables are unique to this investigation. As such, questions related to these
ﬁndings are in need of additional reﬂection. Second, we must ask about the extent to
which smoking related anxiety provided the types of distinct populations (in terms of
high and low issue importance) needed to provide a reliable test of the interactions
predicted in this study. This issue starts with concern over the fact that only 52
respondents scored high on issue importance, and touches on discussion already
addressed in the ﬁrst limitation. In addition to this, however, are questions concerning
whether or not this issue provided the type of truly high and/or low population grouping
needed to provide an effective test of the hypothesized interactions. Although the ﬁrst

pilot study showed this to be a viable health issue to use there may have been one not

88

presented within the six issues given in the pilot. Another health behavior, particularly
one for which a sub-sample of the population scoring truly high on issue importance
exists, may provide a better test of hypothesis four. More than anything, the ﬁrst two
limitations identiﬁed here, the representative nature of the sample and the issue of
smoking-related anxiety; signify the need to replicate the ﬁndings of this study in
research using other samples and different issues. This will help determine the extent to
which these ﬁndings generalize to populations and settings.

The third issue deals with the limitations associated with the reliability for three
single-item measures: one measuring population estimates, and two measuring the
strength of the exemplar discrepancy induction. In both cases, the single-item measures
make assessments of reliability impossible and draw into question study claims that
variables failed to represent the measure and manipulation speciﬁed. Population estimate
is one of the two main outcomes variables in this study. As such, any question concerning
its reliability needs to be considered closely. At the same time, the single item measure of
population estimate used in this study follows common practice. Respondents were
simply asked to estimate the percent of college student smokers who have anxiety
problems. The long practice of measuring population estimates in this manner leads us to
conclude that the reliability threat here is not considered severe. The two items measuring
the strength of the exemplar discrepancy induction were used only as a manipulation
check. Since the items did not form a single index, separate analyses were conducted on
both items. Although this leaves some concern that the reliability of the exemplar
discrepancy measure may have been problematic, the fact that analysis on both separate

measures produced consistent results helps ease this concern. ‘

89

The fourth issue deals with the limitations associated with the decision to use a
post-test only research design with a post-test measure at only one point in time. The
post-test only design was selected in this case to reduce threats to internal validity. In
particular, there was concern that a pre-test on perceptions of problem severity and
population estimates might sensitize participants and bias measures on these two main
outcome variables. This threat was considered particularly inauspicious and strongly
inﬂuenced design choice. At the same time, however, the post-test design introduced
other limitations. First, measures used to assign respondents to conditions of issue
importance were gathered aﬁer exposure to the story on smoking. The potential for the
story to bias these responses cannot be overlooked. At the same time, one could argue
that the impact of story exposure should attenuate existing differences and, as a result, the
post-test measure was a conservative approach to distinguishing the two groups. Second,
the absence of a pre-test meant that no baseline measure of population estimates was
available. Though having a baseline on the participants in the main study would have
been valuable, obtaining population estimates in the pilot test from a separate student
provided something in this regard. Not to be overlooked, the fact that the post-test
measure was taken at only one point in time eliminated the ability to assess how long the
observed effects persisted.

The ﬁnal issue deals with the limitations associated with the unsupervised nature
of the online survey. Using the Internet to conduct the experiment allowed participants to
take as long as they wanted to read each news story, it also allowed participants to go
back to the story if they so desired (even though instructions asked them not to). This

ability to review the stimulus messages may have allowed message recipients to more

90

closely review information in the messages related to the questions asked in the survey. If
respondents behaved in this manner, it may have skewed answers of population
estimates. However, the nature of this study minimizes concern with this threat for two
reasons. First, if the bias introduced by the unsupervised nature of the online survey
increased attention across all conditions in this study, the outcome should have attenuated
message effects. The fact that signiﬁcant differences associated with all three predictors
were observed challenges this interpretation. Second, if the bias introduced increased
attention across only under conditions of increased issue importance and/or information
utility, the increased attention would be consistent with the logic said to explain the
outcomes observed.
Future Research

Based on the ﬁndings of this study as well as previous research by Krupat, et al.
(1997), it appears that the superiority of exemplars over base-rate information is limited
by moderating factors. Yet this study is the ﬁrst of its kind to demonstrate this effect and
only begins to identify the circumstances in which this might be true. Future research is
needed not only to replicate this ﬁnding but also to examine the extent to which it will
generalize beyond the contexts of purchasing decisions or health issues perceptions
examined to date. Attempts to replicate the ﬁndings from this study in research that
replicates some of the original work establishing the exemplar effect in the context of
news and political media would help corroborate observations that information utility and
issue importance can moderate exempliﬁcation processes.

In addition to work aimed at increasing conﬁdence in the ﬁndings of this study

through replication of the interactions observed, research is needed to examine the

91

underlying processes that might be responsible for these ﬁndings. The logic for the
pattern of ﬁndings observed here argues that the moderating inﬂuence of information
utility and issue importance results from increased message attention. Research
examining the mediating role of attention in this development is critical to understanding
the processes at hand and the outcomes that should be expected. Similar research
examining how the retention and recall of exemplars and base-rates is predicted by
varying levels of information utility and issue importance would also add greatly to this
understanding. Although research on exempliﬁcation has ﬁrmly established the
phenomenon as an important outcome of media exposure, our understanding of the
underlying processes is still in its infancy. Efforts to explicate the theoretical mechanisms
at work promise considerable applied beneﬁt.
Contributions to Exempliﬁcation Theory and Health Communication

The value of this research can be seen both conceptually in terms of what it adds
to our understanding of exempliﬁcation theory and pragrnatically in terms of its
applications in health communication campaigns. At the conceptual level, the ﬁndings
' here demonstrate that the robust nature of exempliﬁcation effects is not boundless. It can
be moderated by issue importance and information utility. Undeniably, questions about
the stability of ﬁndings regarding those high on issue importance are subject to concerns
about the low number of high issue importance respondents in this study. However,
information utility moderated the exemplar effect among those low on issue importance,
where the number of respondents was ample. This provides more convincing evidence of
the potential for exempliﬁcation to be moderated. The explanation offered here suggests

that moderation might result from increased attention to the information contained in a

92

message. If this is the case, the theoretical implications for exempliﬁcation theory are
considerable, and efforts to identify other variables that too might increase attention and
moderate exempliﬁcation should be pursued.

In terms of the practical implications of this research, the ﬁndings here might
prove useful for efforts by health communication practitioners in efforts to create more
effective campaigns. Knowing under which conditions (particularly conditions of high
information utility and high issue importance) target audiences are more likely to be
affected by base-rates and/or exemplars can help in message design. Although more
research is needed to determine if the ﬁndings here will generalize to other health issues
and populations, it suggests that message attributes capable of increasing information
utility may heighten attention to, if not at least the inﬂuence of base-rate information. To
the extent that a focus on base-rates might promote healthy behaviors, this should be an
effective strategy on some target populations. By contrast, in cases where base-rates
might not be expected to promote desired behavior, efforts to moderate exempliﬁcation
effects might be less desirable. In such cases, message designers might focus simply on

taking advantage of the generally robust nature of exemplars.

93

Appendix A

94

Pilot Study — Phase One Questionnaire

Issue Immrtance
Not at all Extremely

Magnitude

1. Do you feel personally threatened by the
risk of sexually transmitted
infections (diseases). 1 2 3 4 5 6 7

2. In your judgment how great is the threat
presented to you by the risk of
sexually transmitted infections] 2 3 4 5 6 7

3. Compared to other threats to your health,
how big a problem is the risk of
sexually transmitted infections] 2 3 4 5 6 7

4. Do you feel personally threatened by
the risk of unplanned pregnancy.l 2 3 4 5 6 7

5. In your judgment how great is the threat
presented to you by the risk of
unplanned pregnancy. 1 2 3 4 5 6 7

6. Compared to other threats to your health,
how big a problem is the risk of an
unplanned pregnancy. 1 2 3 4 5 6 7

7. Do you feel personally threatened by
the health risks associated with
smoking. 1 2 3 4 5 6 7

8. In your judgment how great is the threat
presented to you by the risk of
smoking. . 1 2 3 4 5 6 7

9. Compared to other threats to your health,
how big a problem is the risk
associated with smoking. 1 2 3 4 5 6 7

10. Do you feel personally threatened by

the health risks associated with
drug use. 1 2 3 4 5 6 7

95

11. In your judgment how great is the threat
presented to you by the risk of
drug use . 1 2

12. Compared to other threats to your health,
how big a problem is the risk

associated with drug use. 1 2

13. Do you feel personally threatened by risks
associated with drinking alcohol.1 2

14. In your judgment how great is the threat
presented to you by the risk of
drinking alcohol. 1 2

15. Compared to other threats to your health,
how big a problem is the risk
associated with drinking alcohol.1 2

16. Do you feel personally threatened by the
problems associated with
domestic violence. 1 2

17. In your judgment how great is the threat
presented to you by the risk of
domestic violence. ] 2

18. Compared to other threats to your health,
how big a problem is the risk
associated with domestic violence] 2

Likelihood

19. Do you feel that a college student
like you is likely to be affected
by a sexually transmitted infection] 2

20. Do you feel that you personally
are likely to contract a' sexually
transmitted infection. 1 2

2] . How oﬁen do you think about the

likelihood of getting a sexually
transmitted infection 1 2

96

22. Do you feel that a college student

like you is likely to be affected
by an unplanned pregnancy. 1 2

23. Do you feel that you personally

are likely to become pregnant

or get a partner pregnant. 1 2
24. How often do you think about the

likelihood of having an unplanned

pregnancy. 1 2

25. Do you feel that a college student
like you is likely to be affected
by smoking. 1 2

26. Do you feel that you personally
are likely to contract a health
problem associated with smoking]

N

27. How often do you think about the
likelihood of having a health
problem due to smoking. 1 2

28. Do you feel that a college student
like you is likely to be affected
by drug use. 1 2

29. Do you feel that you personally
are likely to contract a health
problem associated with drug use.1 2

30. How often do you think about the
likelihood of having a health
problem due to drug use. 1 2

31. Do you feel that a college student
like you is likely to be affected
by drinking alcohol. 1 2

32. Do you feel that you personally
are likely to contract a health problem
associated with drinking alcohol.1 2

97

33. How often do you think about the
likelihood of having a health
problem due to drinking alcohol.1 2

34. Do you feel that a college student
like you is likely to be affected

by domestic violence. 1 2

35. Do you feel that you personally
are likely to contract a health problem
due to domestic violence. 1 2

36. How often do you think about the
likelihood of being impacted by
domestic violence. 2

Irnmediacy

37. I think that I will be affected by a
sexually transmitted infection
within the next year. 1 2

38. I will be affected by a sexually
transmitted infection soon. 1 2

39. I will not be affected by a sexually
transmitted infection any time

soon. ® 1 2
40.]thinkthatlwillbeaffectedbyan

unplanned pregnancy within the

next year. 1 2

41. I will be affected by an unplanned
pregnancy soon. 1 2

42. I will not be affected by an unplanned
pregnancy any time soon. ® 1 2

43. I think that I will be affected by health
problems due to smoking within
the next year. 1 2

44. I will be affected by smoking soon.] 2

98

45. I will not be affected by smoking any

time soon. ® 1 2 3 4 5 6 7
46. I think that I will be affected by health

problems due to drug use within

the next year. 1 2 3 4 5 6 7

47. I will be affected by drug use soon] 2 3 4 5 6 7

48. I will not be affected by drug use
any time soon. ® 1 2 3 4 5 6 7

49. I think that I will be affected by health
problems due to drinking alcohol
within the next year. 1 2 3 4 5 6 7

50. I will be affected by drinking alcohol
soon. 1 2 3 4 5 6 7

51. I will not be affected by drinking
alcohol any time soon. ® 1 2 3 4 5 6 7

52. I think that I will be affected by
domestic violence within the next
year. 1 2 3 4 5 6 7

53. I will be affected by domestic
violence soon. 1 2 3 4 5 6 7

54. I will not be affected by domestic

violence any time soon. ® 1 2 3 4 5 6 7

Here is a list of sources that may present you with a health message. Please rate each
source as to how credible (believable) you think that source would be for health
information (1 = “not at all credible” and 7 = “extremely credible”).

Not at all Extremely
Credible Credible

l. The Center for Disease Control 1 2 3 4 5 6 7
2. The American Cancer Society 1 2 3 4 5 6 7

3. The Lansing State Journal 1 2 3 4 5 6 7

99

4. WebMd on the Internet 1 2
5. A doctor 1 2

6. Someone who has suffered from
the health issue in question 1 2

7. A parent 1 2
8. A college student 1 2
9. A professor 1 2
10. Oprah Winfrey 1 2
11. Tom Cruise 1 2

12. US. News & World Report 1 2

13. Jessica Simpson 1 2
14. LeBron James 1 2
15. Planned Parenthood l 2
l6. MSU president Simon 1 2
17. An unfamiliar adult 1 2

18. National Institute of Health 1 2

19. An unidentiﬁed government ofﬁcial] 2

20. An unidentiﬁed source 1 2
21. Hilary Rodham Clinton 1 2
22. A tobacco company 1 2

23. Journal of Health Communication] 2

24. Surgeon General 1 2
25. Pharmaceutical Company 1 2
26. The National Enquirer l 2

100

27. Time 1 2 3 4 5 6 7

28. Newsweek l 2 3 4 S 6 7
29. Journal of Public Health 1 2 3 4 5 6 7
30. Junk Mail/Spam l 2 3 4 S 6 7

New Information for Possible use as base-rate information.
1. What percent of all sexually transmitted infections occur in people 25 or younger?

%

 

2. What percent of sexually active people will have contracted a sexually transmitted
infection by the age of 24?
%

 

3. Among women, at what age is the rate of gonorrhea is highest?
years-old

4. Among men, at what age is the rate of gonorrhea is highest?
years-old
5. How many million sexually active young adults acquire a sexually transmitted
infection every year?
million
6. Acquired immunodeﬁciency syndrome (AIDS) is one of the leading causes of death
among people of what age group?
years-old
7. About how many million women face unplanned pregnancy each year?
million
8. In women aged 20-24—years old about what percentage of pregnancies occur outside

of marriage?
%

 

9. What percent of college women attending four-year institutions have experienced

pregnancy?
%

 

101

10. About what percent of smokers under the age of 25 have anxiety problems associated
with cigarette smoking?
%

 

l 1. Smoking is known to be related to lower academic achievement in what percent of
college student smokers?
%

 

12. The maximum lung functioning level for college student smokers is what percentage
lower than non-smokers?
%

 

13. The average resting heart rate of young adult smokers is how many beats per minute
faster than nonsmokers.

beats per minute

14. About what percent of college student smokers have increased cough and/or severe

respiratory illnesses due to smoking?
%

 

15. Approximately how many deaths are there per year are smoking associated?
deaths
16. Approximately what percentage of Americans are smokers?

%

 

17. Smoking has been found to accelerate the aging process by how many years?

years

18. Approximately, how many college students between the ages of 18 and 24 die each
year from unintentional alcohol-related injuries, including motor vehicle crashes?

students

 

19. Approximately, how many students between the ages of 18 and 24 are unintentionally
injured each year while under the inﬂuence of alcohol?
students

 

20. How many students between the ages of 18 and 24 are assaulted each year by another

student who has been drinking?
students

 

102

21. How many students between the ages of 18 and 24 are victims of alcohol-related
sexual assault or date rape each year?
students

22. How many students develop an alcohol-related health problem?
students
23. What percentage of adult women experience at least one physical assault by a

partner?
%

 

24. Approximately, what percentage of all annual violence against women is perpetrated
by intimate partners?
%

 

25. The rate of reported violence by intimate partners is highest among women at what
age?

years-old

26. Smoking marijuana is known to cause memory loss in what percent of college student

marijuana smokers?
%

 

27. Smoking marijuana has been directly related to a shortened attention span in what

percentage of college students?
%

 

®These questions were reverse coded.

103

Appendix B

104

Pilot Study — Phase Two Questionnaire
Smoking Stog

Perceived Credibility
1. The Director of the CDC is a credible source

for giving out information that pertains to

health risks associated with smoking. 1 2 3 4 5 6 7
2. The Director of the CDC does not have enough

information on the risks of smoking to

make these claims. ® 1 2 3 4 5 6 7

3. I trust the Director of the CDC to tell the truth
about smoking. 1 2 3 4 5 6 7

4. I have conﬁdence in the accuracy of the
information about smoking and anxiety. 1 2 3 4 5 6 7

5. Ibelieve that cigarette smoking can cause anxiety] 2 3 4 5 6 7

6. I don’t believe that cigarette smoking has an effect
on college students. ® 1 2 3 4 5 6 7

Relevance

7. The story on smoking and anxiety provides
valuable information for college students. 1 2 3 4 5 6 7

8. The smoking related anxiety story provided
information that I ﬁnd very useful. ] 2 3 4 5 6 7

9. The information on smoking and anxiety is

personally relevant to me. 1 2 3 4 5 6 7
10. The story on smoking and anxiety

applies directly to me life. 1 2 3 4 5 6 7
New Information

1]. If (before you read the report in US. News & World Report today) somebody had
asked you to estimate the percent of college student smokers that have anxiety, what
would your answer have been?

%

 

105

Not at all Extremely

11. How informative was the US. News &
World Report on smoking on
anxiety? 1 2 3 4 5 6 7

12. How well educated do you now feel about
the effects of smoking on anxiety after
reading the article in US. News & World
Report? 1 2 3 4 5 6 7

13. To what extent do you think that the report
you read in US. News & World Report
adds to your knowledge about smoking

and anxiety? 1 2 3 4 5 6 7

Behavior
14. I consider myself to be a smoker:

Yes

No

I don’t consider myself a smoker, but I do smoke socially

I used to smoke but I quit years months ago.
Exemplar Discrepancy

15. The information about anxiety in the example given within the story is representative
of the statistical information presented.

Strongly Agree Agree Neutral Disagree Strongly
Disagree

16. Looking at the information provided about anxiety which do you believe to be more
accurate the statistical information or the example given? (circle one)
a. Statistics
b. Example
c. They are the same

17. On a scale of 1 to 7 (with 1 = “not at all” and 7 = “extremely”) rate how different you
think the example about anxiety is from the statistical information.

Population Estimates

19. What percent of college student smokers have anxiety problems? %

106

Perception of Problem Severity

Not at all Extremely
20. How severe is the problem of smoking
related anxiety among college students? 1 2 3 4 5 6 7

21. How likely do you think it is that
smoking related anxiety among
college students will get worse? 1 2 3 4 5 6 7

22. How likely do you think it is that smoking
related anxiety will become a
real problem on the MSU campus? 1 2 3 4 5 6 7

23. How likely do you think it is that smoking
related anxiety might become a
problem for you? 1 2 3 4 5 6 7

Behavioral Intentions

24. How likely is this information about anxiety
to affect your current smoking behavior? 1 2 3 4 5 6 7

25. How likely are you to seek more information on
health problems associated with smoking?1 2 3 4 5 6 7

26. How likely are you to quit smoking or not start
based on the information about anxiety? 1 2 3 4 5 6 7

27. How likely do you think others would be to quit
smoking based on this information about

anxiety? 1 2 3 4 5 6 7
Terrorism Story
Perceived Credibility
Not at all Extremely

1. US. News & World Report is a credible
source for giving out information that
pertains to terrorism. 1 2 3 4 5 6 7

2. US. News & World Report does not have

enough information on terrorism to make
these claims. 1 2 3 4 5 6 7

107

S"

I trust US. News & World Report to tell the

truth about terrorism. 1 2 3 4 5 6 7
4. I have conﬁdence in the accuracy of the

information about terrorism. 1 2 3 4 5 6 7
5. I believe that there are a number of terrorist

attacks every year. ] 2 3 4 5 6 7
6. I don’t believe that there are terrorist attacks. ®1 2 3 4 5 6 7
Relevance

7. The story on terrorism provides valuable
information for college students 1 2 3 4 5 6 7

8. The terrorism story provided information that
I ﬁnd very useful 1 2 3 4 5 6 7

9. The information on terrorist attacks is personally
relevant to me. 1 2 3 4 5 6 7

10. The story on terrorist attacks applies directly to
me life. 1 2 3 4 5 6 7

New Information

1]. If (before you reaﬂe report in US. News & World Report today) somebody had
asked you to estimate the number of terrorist attacks each year, what would your
answer have been?

attacks each year

Not at all Extremely
12. How informative was the US. News &
World Report on terrorist attacks? 1 2 3 4 5 6 7

13. How well educated do you now feel about
terrorist attacks after reading the article
in US. News & World Report? 1 2 3 4 5 6 7

14. To what extent do you think that the report

you read in US. News & World Report adds
to your knowledge about terrorism? 1 2 3 4 5 6 7

108

Behavior
15. I consider myself to be dreadfully worried about terrorism:

__Yes

_ No

_ I don’t consider myself dreadfully worried about terrorism, but I am a bit
worried

__ I used be dreadfully worried but this stopped about years
months ago.

Exemplar Discrepancy

16. The information about terrorism in the example given within the story is
representative of the statistical information presented.

Strongly Agree Agree Neutral Disagree Strongly
Disagree

17. Looking at the information provided about terrorism, which do you believe to be
more accurate the statistical information or the example given?
a. Statistics
b. Example
c. They are the same

18. On a scale of 1 to 7 (with 1 = “not at all” and 7 = “extremely”) rate how different you
think the example about terrorism is from the statistical information.

Population Estimate

19. How many terrorist attacks occurred in 2004?

 

Perception of Problem Severity

Not at all Extremely
20. How severe is the problem of terrorist attacks
for college students? 1 2 3 4 5 6 7
21. How likely do you think it is that
terrorist attacks will get worse? 1 2 3 4 5 6 7

22. How likely do you think it is that terrorist attacks
will become a real problem on the MSU
campus? 1 2 3 4 5 6 7

109

23. How likely do you think it is that terrorist attacks
might become a problem for you? 1 2

Behavioral Intentions

24. How likely is this information about terrorist
attacks to affect your behavior? ‘ 1 2

25. How likely are you to seek more information on
terrorist attacks? 1 2

26. How likely are you to take preventative action based
on this information about terrorist attacks?1 2

27. How likely do you think others would be to take
take preventative action on this information
about terrorist attacks? 1 2

Genetic Screening Story

Not at all
Perceived Credibility

1. The National Institute of Health is a credible
source for giving out information that
pertains to genetic screening. 1 2

2. The National Institute of Health does not have
enough information on genetic screening to
make these claims. ® 1 2

3. I trust the National Institute of Health to tell the
truth about genetic screening. 1 2

4. I have conﬁdence in the accuracy of the
information about genetic screening. 1 2

5. I believe that genetic screening can detect fetal
abnormalities. 1 2

6. I don’t believe that genetic screening has an effect
on abnormality detection. ® 1 2

110

6 7
6 7
6 7

6 7
6 7
Extremely

6 7

6 7

6 7

6 7
6 7
6 7

Relevance

7. The story on genetic screening provides
valuable information for college students. 1 2 3 4 5 6 7

8. The genetic screening story provided
information that I ﬁnd very useful. 1 2 3 4 5 6 7

9. The information on genetic screening is
personally relevant to me. 1 2 3 4 5 6 7

10. The story on genetic screening
applies directly to me life. 1 2 3 4 5 6 7

New information

11. If (mfore you read the remrt in US. News & World Report today) somebody had

asked you to estimate the percent of fetal abnormalities detected by genetic screening,
what would your answer have been?

%

 

Not at all Extremely
12. How informative was the US. News &
World Report on genetic screening? 1 2 3 4 5 6 7

13. How well educated do you now feel about

genetic screening after reading the

article in US. News & World Report? 1 2 3 4 5 6 7
14. To what extent do you think that the report

you read in US. News & World Report adds

to your knowledge about genetic screening?1 ' 2 3 4 5 6 7
Population Estimates
15. What percent of fetal abnormalities are detected by genetic screening? %
Behavior

16. I consider myself to be dreadfully worried about genetic screening:

____Yes

_ No

__ I am not dreadﬁrlly worried about genetic screening, but I am a bit worried
_ I used be very worried but this stopped about years months

ago.

111

Exemplar Discrepancy

17. The information about genetic screening in the example given within the story is
representative of the statistical information presented.

Strongly Agree Agree Neutral Disagree Strongly
Disagree
18. Looking at the information provided about genetic screening, which do you believe to
be more accurate the statistical information or the example given?
a. Statistics

b. Example
c. They are the same

19. On a scale of 1 to 7 (with l = “not at all” and 7 = “extremely”) rate how different you
think the example about genetic screening is from the statistical information.

Perception of Problem Severity

Not at all Extremely
20. How severe is the problem of fear of fetal
abnormalities among college students? 1 2 3 4 5 6 7

21. How likely do you think it is that fears of fetal
abnormalities college students will get worse?1 2 3 4 5 6 7

22. How likely do you think it is that fears of fetal
abnormalities will become a real problem on
the MSU campus? 1 2 3 4 5 6 7

23. How likely do you think it is that fetal
abnormalities might become a problem for you?l 2 3 4 5 6 7

Behavioral Intentions

24. How likely is this information about genetic
screening to affect your behavior? 1 2 3 4 5 6 7

25. How likely are you to seek more information on
genetic screening? 1 2 3 4 5 6 7

26. How likely would you be to take use genetic
screening based on this information? 1 2 3 4 5 6 7

112

27. How likely do you think others would be to use
genetic screening based on this information?1 2 3

®These questions were reverse coded.

113

Appendix C

114

News Stories
Smoking stag:

High Information utility - low discrepancy [extreme discrepancyl

US. News & World Report.
Wednesday, August I 7, 2005

In a statement released today by the Center for Disease Control (CDC), CDC Director
Jeffrey P. Koplan, M.D., M.P.H., reported the results of a new study linking cigarette
smoking to anxiety problems suffered by college students. The study, conducted on a
US. nation wide sample of over 78,000 students at 4-year universities ﬁnds support for
previous studies showing that smoking is linked with a variety of health problems.
However, in surprising new evidence, the CDC reports smoking is associated with
anxiety. According to Dr. Koplan, research evidence shows that 68% of college student
smokers have anxiety problems, and that smoking is a contributor to these anxiety
problems. Notably, the onset of this problem appears to be surprisingly sudden and
unexpected and strikes college students of all types in an unpredictable manner. The
report also states that there is clear evidence that when individuals stop smoking the
anxiety problems quickly dissipate.

[Of course, not all students are affected] Alex Shipley of Madison, Wisconsin is a good
example. Shipley, a student at the University of Wisconsin, is an honor student in the
Business School who plans on getting a job in ﬁnance when he graduates. Alex says that
things have [nothing has] changed since he started smoking. “When I came to school it
was easy for me. I always felt comfortable. I never had any trouble speaking to my
professors in class. I liked hanging out and talking with other kids. Nothing seemed to
bother me. Then when [When] I was a sophomore I started smoking with some of my
friends and things started to change [but nothing has really changed]. Now I feel worried
[I still feel good about] about everything. I [never] get nervous when I’m around people.

My hands get sweaty and I start to feel real uncomfortable [I always feel real
comfortable]. When I speak I always [never] stutter and [or] mumble my words.”

Mark Westbrook is another example. Mark grew up in East Lansing, Michigan and
attends Michigan State University. Mark is an Advertising major who plans to work in
marketing when he graduates. Mark was always highly involved in school activities as an
Evans Scholar and a member of several clubs on campus. Mark ﬁrst noticed [has noticed
no] signs of mounting anxiety @- [since] he started smoking during his junior year. “I
always use to think [have always thought] of myself as really outgoing. Now I don’t [I]
enjoy being around other people. I [never] get nervous even when I’m with my friends
[strangers]. When I started smoking I thought of it as kind of a social activity. _1\_lo_vLI feel
like smoking has made [helps make] me [one of] the most anxious, anti-social mrson
[social people] out there.”

115

Smoking stag:

Low Information utility — ow discrepancy [extreme discrepancyl

The National Enquirer
Wednesday, August 1 7, 2005

A statement released reported the results of a study linking cigarette smoking to anxiety
problems suffered by senior citizens. The study replicates previous studies showing that
smoking is linked with a variety of health problems. Once again, it was found that
smoking is associated with anxiety. According to an unidentiﬁed source, research
evidence shows that 68% of senior citizen smokers have anxiety problems, and that
smoking is a contributor to these anxiety problems. It is unclear whether or not the
anxiety can be alleviated by quitting.

[Of course, not all senior citizens are affected]. Alex Shipley of Riverside, California is a
good example. Shipley, a retired construction worker, is a 65-year old father of three and
grandfather of four and has been a smoker for over 40 years. Alex says that [nothing has]
things have changed since he started smoking. “When I started working it was easy for
me. I always felt comfortable. I never had any trouble speaking to my boss at work. I
liked hanging out and talking with coworkers. Nothing seemed to bother me. MWhen
I had been working for two years I started smoking with some of my ﬁiends [but nothing
has really changed] and things started to change. Now I [still feel good] feel worried
about everything. I [never] get nervous when I’m around people. [I always feel real
comfortable]. My hands get sweaﬂ and I stait to feel real uncomfortable. When I speak I
[never] always stutter [or] and mumble my words.”

Mark Westbrook is another example. Mark grew up in Boise, Idaho and is also retired.
Mark was an assembly worker for a plastics factory. Mark is 59-years-old, a father of
two, and has been smoking for 19 years. Mark was always highly involved in various
activities as a member of Community Theater and social clubs around Boise. Mark [has]
ﬁ_rit noticed [no] signs of mounting anxiety [since] aftjg he started smoking. “1 have
always thought of myself as really outgoing. Now I don’t enjoy being around other
people. I [never] get nervous even when I’m with [strangers] my ﬁ'iends. When I started
smoking I thought of it as kind of a social activity. New I feel like smoking [helps make]
has made me one of the most anxious anti-social people out there.”

 

116

Terrorism Stag: (High Issue Importance — low discrepancy)

US. News & World Report
Monday, August 15, 2005

The White House has issued a press release stating that they will no longer be publishing
information pertaining to the number of terrorist attacks that occur throughout the world.
The most recent statistics ﬁ'om the State Department are from 2004, which state that there
were 625 “signiﬁcant” terrorist attacks in that year alone. According to the Secretary of
Defense, Donald Rumsfeld, the number of terrorist attacks each year has not been
declining since September 11, 200] , regardless of the number of new counter-measures
that are in place. Rumsfeld stated, “Although we have updated measures to prevent
terrorist attacks in place throughout the world there is still cause for caution as the
number of attacks do not appear to be decreasing.” According to the Department of
Defense the majority of the 625 attacks in 2004 did claim lives.

Brian Roberts, a US. Army soldier from Toledo, Ohio, has been involved in trying to
trace terrorist activities. “We get information from all over the world,” Green said “The
information we receive is crucial in trying to prevent terrorist attacks from occurring.”
Green has noticed that there have been several hundred attacks each year since he joined
the army and is not surprised that the number has not decreased in recent years. “We’re
doing all we can, I wish we could do more but I know that we are stopping a number of
attacks every month and without the work that is being done there would be even more
successful attacks, particularly on US. citizens.”

Michelle Bennett, an American, was in London during the most recent attacks on public
transportation. Michelle was fortunate that she was not injured but she did see some of
the devastating effects of the bombs. “I was in London for work; I was in the office when
the news came in about the attacks. We were all in shock that terrorism had come so
close to where we were but we always knew that it was a possibility. You hear about this
happening all over the world but when it happens close by it makes it much more real.”

117

Genetic Screeningﬁtorv: (Low Issue Importance -— low discrepancy)

US. News & World Report
Monday, August I 5, 2005

In a statement released today by the National Institute of Health (NIH), NIH Director
Elias A. Zerhouni, M.D., reported the results of a new study which examined the success
of genetic screening to detect treatable fetal abnormalities that threaten the life of babies
during a mother’s pregnancy. The study was conducted on tissue samples colleted over
the course of several months from 1,000 pregnant women over the age of 40 (when
women are at higher risk of carrying a healthy baby to term). The study found an increase
in the previously reported success of genetic screening. According to Dr. Zerhouni
research evidence shows that genetic screening can now detect 70% of all lethal fetal
abnormalities.

Of course, this is good news for many concerned parents. Todd and Linda Oaks of
Boulder, Colorado have been trying to conceive for 10 years. Linda is now in her 403 and
is ﬁnally pregnant with their ﬁrst child. “It has been a very trying time for us and we’ve
thought about giving up. After I turned 40 I became hesitant about carrying a child. So
many things can go wrong for women my age and it can hurt the baby. Todd and l are
very excited to have an opportunity to be able to tell if the baby has any health problems
that they can be treated now. It will give us a sense of relief to know that we are having a
healthy child.”

Chris and Angie Jansing, of Salt Lake City, Utah, are also excited about the prospect of
checking for fetal abnormalities. Chris and Angie already have three children and just
found out that a fourth is on the way. “We’re both very nervous about this baby. 1
[Angie] just turned 48 and was extremely surprised when I found out I was pregnant. Our
children are all in their teens and we weren’t expecting to have another child. Aﬁer we
thought about it we became excited by the idea but we realize that my age may cause
some problems in the baby. We’re hoping that genetic screening will tell us that the baby
is healthy, but if not we want to detect problems early so we can do everything possible
to save the baby.”

118

Appendix D

119

Main Study —Questionnaire

Smoking
Magnitude
Not at all Extremely

]. Do you feel personally threatened by

the health risks associated with

smoking? 1 2 3 4 5 6 7
2. In your judgment how great is the threat

presented to you by the risk of

smoking? 1 2 3 4 5 6 7

3. Compared to other threats to your health,
how big a problem is the risk
associated with smoking? 1 2 3 4 5 6 7

Likelihood

4. Do you feel that a college student
like you is likely to be affected
by smoking? 1 2 3 4 5 6 7

5. Do you feel that you personally
are likely to contract a health
problem associated with smoking?1 2 3 4 5 6 7

6. How often do you think about the

likelihood of having a health
problem due to smoking? 1 2 3 4 5 6 7

Irnmediacy

7. I think that I will be affected by health

problems due to smoking within

the next year. 1 2 3 4 S 6 7
8. I will be affected by smoking soon] 2 3 4 5 6 7

9. I will not be affected by smoking any
time soon. ® 1 2 3 4 5 6 7

120

Behavior

10. I consider myself to be a smoker (check one)

_Yes

_ No

_ I don’t consider myself a smoker but I do smoke socially
__ I used to smoke but I quit years months ago.

If No, please skip down to question #_

11. I have been a smoker for years months
12. I smoke cigarettes daily. (please estimate actual number of cigarettes)
13. I purchase a pack of cigarettes every days.

14. I have tried to quit smoking:

Never Once Twice Three Times More
than 3
Population Estimates
1. What percent of college student smokers have anxiety problems? %

Perception of Problem Severity

Not at all Extremely
2. How severe is the problem of smoking
related anxiety among college students? 1 2 3 4 5 6 7

3. How likely do you think it is that
smoking related anxiety among
college students will get worse? 1 2 3 4 5 6 7

4. How likely do you think it is that smoking
related anxiety will become a
real problem on the MSU campus? 1 2 3 4 5 6 7

5. How likely do you think it is that smoking

related anxiety might become a
problem for you? 1 2 3 4 S 6 7

121

Behavioral Intentions

6. How likely is this information about anxiety
to affect your behavior? 1 2

7. How likely are you to seek more information on
health problems associated with smoking?1

8. How likely are you to quit smoking or not start
based on the information about anxiety? ]

9. How likely do you think others would be to quit
smoking based on this information about
anxiety? 1 2

Information Utility
Perceived Credibility

10. The director of the CDC is a credible source
for giving out information that pertains to

2

2

health risks associated with smoking. 1 2

11. The director of the CDC does not have enough
information on the risks of smoking to

make these claims. ® 1 2

12. I trust the director of the CDC to tell the truth
about smoking. 1

13. I have conﬁdence in the accuracy of the
information about smoking and anxiety. 1

14. I believe that cigarette smoking can cause anxiety. 1

15. I don’t believe that cigarette smoking has an effect

on college students. ® 1 2

Relevance

16. The story on smoking and anxiety provides
valuable information for college students. 1

17. The smoking related anxiety story provided
information that I ﬁnd very useful. 1 2

122

2

18. The information on smoking and anxiety is

personally relevant to me. 1 2
19. The story on smoking and anxiety is

applies directly to me life. 1 2
New Information

20. If (before you read the report in US. News & Worlgl Report today) somebody had
asked you to estimate the percent of college student smokers that have anxiety, what

would your answer have been?

%

 

Not at all
21. How informative was the US. News & World
Report report on smoking on anxiety? 1

22. How well educated do you now feel about
the effects of smoking on anxiety after

reading the article in US. News & World
Report? 1

23. To what extent do you think that the report
you read in US. News & World Report adds

to your knowledge about smoking and anxiety?l

Terrorism Story

Population Estimates

1. How many terrorist attacks occurred in 2004?

 

Magnitude
Not at all
Extremely
10. Do you feel personally threatened by
possible terrorist attacks? 1 2 3

l 1. In your judgment how great is the threat
presented to you of terrorist

attacks? 1 2 3
12. Compared to other threats to your health,

how big a problem is the threat

of terrorist attacks? 1 2 3

123

Extremely

5 6 7
5 6 7
5 6 7

Likelihood

13. Do you feel that a college student
like you is likely to be affected

by terrorist attack? ] 2 3 4 5 6 7
14. Do you feel that you personally

are likely to be affected by

terrorist attack? 1 2 3 4 5 6 7

15. How often do you think about the
likelihood of terrorist attacks?] 2 3 4 5 6 7

Irnmediacy
16. I think that I will be affected by terrorist
attacks smoking within
the next year. 1 2 3 4 5 6 7

17. I will be affected by terrorist attack soon] 2 3 4 5 6 7

18. I will not be affected by terrorist attack
any time soon. ® 1 2 3 4 5 6 7

Perception of Problem Severity

Not at all Extremely
2. How severe is the problem of terrorist attacks
for college students? 1 2 3 4 5 6 7
3. How likely do you think it is that
terrorist attacks will get worse? 1 2 3 4 5 6 7

4. How likely do you think it is that terrorist attacks
will become a real problem on the MSU

campus? ] 2 3 4 5 6 7
5. How likely do you think it is that terrorist attacks

might become a problem for you? 1 2 3 4 5 6 7
Behavioral Intentions

6. How likely is this information about terrorist
attacks to affect your behavior? 1 2 3 4 5 6 7

124

7. How likely are you to seek more information on
terrorist attacks? 1 2 3 4

8. How likely are you to take preventative action based
on this information about terrorist attacks?1 2 3

9. How likely do you think others would be to take
take preventative action on this information
about terrorist attacks? 1 2 3 4

Perceived Credibility
10. US. News & World Report is a credible source
for giving out information that pertains to

terrorism. 1 2 3 4

11. US. News & World Report does not have enough
information on terrorism to make these claims. ®1 2 3

12. I trust US. News & World Report to tell the truth
about terrorism. 1 2 3 4

13. I have conﬁdence in the accuracy of the
information about terrorism. 1 2 3

14. I believe that there are a number of terrorist
attacks every year. 1 2 3

15. I don’t believe that there are terrorist attacks. ®l 2 3
Relevance

16. The story on terrorism provides valuable
information for college students. ] 2 3

17. The terrorism story provided information that
I ﬁnd very useful. 1 2 3

18. The information on terrorist attacks is personally
relevant to me. 1 2 3

19. The story on terrorist attacks applies directly to
me life. 1 2 3

125

New Information

20. If (before you read; the report in US. News & World Report today) somebody had
asked you to estimate the number of terrorist attacks each year, what would your
answer have been?

attacks each year
Not at all Extremely
21. How informative was the US. News & World
Report report on terrorist attacks? 1 2 3 4 5 6 7

22. How well educated do you now feel about
terrorist attacks after reading the article
in US. News & World Report? 1 2 3 4 5 6 7

23. To what extent do you think that the report
you read in US. News & World Report adds
to your knowledge about terrorism? 1 2 3 4 5 6 7

Genetic Screening Story
Population Estimates
1. What percent of fetal abnormalities are detected by genetic screening? %

Magnitude
Not at all Extremely
19. Do you feel personally threatened by
the health risks associated with
fetal abnormalities? 1 2 3 4 5 6 7

20. In your judgment how great is the threat

presented to you by the risk of

fetal abnormalities? 1 2 3 4 5 6 7
2]. Compared to other threats to your health,

how big a problem is the risk

associated with fetal abnormalities?1 2 3 4 5 6 7
Likelihood

22. Do you feel that a college student
like you is likely to be affected

126

by fetal abnormalities? 1 2 3 4 5 6
23. Do you feel that you personally

are likely to contract a health

problem associated with fetal

abnormalities? 1 2 3 4 5 6

24. How often do you think about the
likelihood of having a health
problem due to fetal abnormalities?l 2 3 4 5 6

Irnmediacy
25. I think that I will be affected by fetal

abnormalities within the next year.l 2 3 4 5 6
26. I will be affected by fetal abnormalities

soon. 1 2 3 4 5 6
27. I will not be affected by fetal abnormalities

any time soon. ® 1 2 3 4 5 6

Not at all

Perception of Problem Severity

2. How severe is the problem of fear of fetal
abnormalities among college students? 1 2 3 4 5

3. How likely do you think it is that fears of fetal
abnormalities college students will get worse?1 2 3 4 5

4. How likely do you think it is that fears of fetal
abnormalities will become a real problem on
the MSU campus? 1 2 3 4 5

5. How likely do you think it is that fetal

abnormalities might become a problem for you?1 2 3 4 5

Behavioral Intentions

6. How likely is this information about genetic

screening to affect your behavior? ] 2 3 4 5 6

7. How likely are you to seek more information on
genetic screening? ] 2 3 4 5

127

8. How likely would you be to take use genetic
screening based on this information? 1 2 3 4 5 6

9. How likely do you think others would be to use
genetic screening based on this information?l 2 3 4 5 6

Perceived Credibility

10. The National Institute of Health is a credible
source for giving out information that
pertains to genetic screening. 1 2 3 4 5 6

11. The National Institute of Health does not have
enough information on genetic screening to
make these claims. ® 1 2 3 4 5 6

12. I trust the National Institute of Health to tell the
truth about genetic screening. 1 2 3 4 5 6

13. I have conﬁdence in the accuracy of the
information about genetic screening. 1 2 3 4 5 6

14. I believe that genetic screening can detect fetal

abnormalities. 1 2 3 4 5 6
15. I don’t believe that genetic screening has an effect

on abnormality detection. ® 1 2 3 4 5 6
Relevance

16. The story on genetic screening provides
valuable information for college students. 1 2 3 4 5 6

17. The genetic screening story provided
information that I ﬁnd very useful. 1 2 3 . 4 5 6

18. The information on genetic screening is
personally relevant to me. 1 2 3 4 5 6

19. The story on genetic screening
applies directly to me life. 1 2 3 4 5 6

New information

128

20. If (before you read the rep_ort in US. News & World Report today) somebody had
asked you to estimate the percent of fetal abnormalities detected by genetic screening,
what would your answer have been?

%

 

Not at all Extremely

21. How informative was the US. News & World

Report report on genetic screening? 1 2 3 4 5 6 7
22. How well educated do you now feel about

genetic screening after reading the

article in US. News & World Report? 1 2 3 4 5 6 7
23. To what extent do you think that the report

you read in US. News & World Report adds

to your knowledge about genetic screening?1 2 3 4 5 6 7
Demographics
(Please circle the item which best describes you.)

1. I am (circle one):
male female

2. I am currently a (circle one):
Freshman Sophomore Junior Senior

3. I consider myself to be (circle one):

Caucasian Aﬁican-American Asian/Paciﬁc Islander
Chicano-Latino Middle Eastern Native American
Other

4. I am years of age.

5. My current GPA is _.

6. My average household income is (circle one):

a. Below $25,000
b. $25,000 - $50,000

129

$50,000 - $75,000
$75,000 - $100,000
$100,000 - $150,000
$150,000 - $200,000
$200,000 and above

tonnes

7. While completing this experiment I believe that I have detected the purpose of the
study.
Yes No

8. If yes, this purpose is:

 

 

 

®These questions were reverse coded.

130

Appendix E

131

Perceptions of Credibility - l = “not at all” 7 = “extremely”

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Source Mean SD
Center for Disease Control 6.54 .75
American Cancer Society 6.54 .70
The Lansing State Journal 3.42 1.23
WebMD on the Internet 3.40 1.51
A doctor 5.96 1.08
Sufferer of Health issue 5.19 1.22
Parent 4.63 1.31
College student 2.98 1.12
Professor 4.03 1 .28
Oprah Winfrey 3.48 1.54
Tom Cruise 1.56 .87
US. News & World Report 4.46 1.40
Jessica Simpson 1.33 .81
Lean James 1.39 .78
Planned Parenthood 5.17 1.46
MSU President Simon 3.04 1.34
An unfamiliar adult 1.73 .89
National Institute of Health 6.1 1.11
Unidentiﬁed government 2.67 1.25
official
Unidentiﬁed Source 1.60 1.05
Hilary Clinton 3.08 1.68
A tobacco company 2.15 1.55
Journal of Health 5.33 1.23
Communication
Surgeon General 6.00 1.01
Pharmaceutical Company 3.88 1.62
The National Enquirer 1.21 .72
Time 3.75 1.50
Newsweek 3.88 1.55
Journal of Public Health 5.23 1.1 1
Junk Mail/Spam 1.06 .24

 

 

132

Appendix F

133

Questions Measuring Respondent Knowledge of Base-Rates

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Question Mean SD
% of all STIs occur in people 25 or younger 64.92 19.65
% of sexually active people that will have contracted a STI by 24 47.55 20.6]
Age rate of gonorrhea is highest - women 21.06 5.81
Age rate of gonorrhea is highest - men 20.2] 4.08
How many million sexually active young adults acmrire a STI 16.98 35.48
AIDS is one of the leading causes of death among people of what 26.56 6.29
age
About how many million women face unplanned pregnancy each 27.30 89.35
year
In women,20-24, what % of pregnancies occur outside of marriage 50.27 22.70
% of women attending 4-year colleges have experienced pregnancy 23.39 17.10
% of smokers under 25 have anxiety problems associated with 40.00 25.08
cigarette smoking
Smoking is related to lower academic achievement in _% of college 31.31 23.25
student smokers
Maximum lung functioning level for college student smokers is _% 39.57 20.47
lower than non-smokers
Average resting heart rate of young adult smokers is _ bpms faster 16.83 15.79
than nonsmokers
% of college student smokers have increased cough and/or severe 45.48 29.03
respiratory illnesses
How many deaths per year are associated with smoking 1380482 2 mil+
% of American’s that are smokers 37.26 14.92
Smoking accelerates the agingprocess by_ years 9.20 3.84
# of college students, 18-24 die yearly from unintentional alcohol- 626755 2 mil+
related injuries
# of students, 18-24 are unintentionally injured yearly while under 1220796 3 mil+
the inﬂuence of alcohol
# of students, 18-24 assaulted each year by another student who has 1392002 5 mil+
been drinking
# of students, 18-24 victims of alcohol-related sexual assault or date 707878 2 mil+
rape each year
# of students develop an alcohol-related health problem 392212 1 mil+
% of adult women experience at least one physical assault by a 32.38 19.29
partner
% of all annual violence against women perpetrated by intimate 60.06 26.70
partners
Rate of reported violence by intimate partners is highest among 25.04 9.35
women at what age
Smoking marijuana causes memory loss in what % of college 52.42 35.53
student marijuana smokers
Smoking marijuana is related to a shortened attention span in what 49.55 33.54

% of college students

 

 

 

134

 

Appendix G

135

Information Utility Items
Relevance — Final Items Used

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Question Factor
Loading

The information on smokingind anxiety is personally relevant to me. .918

The story on smokirg and anxiety applies directly to me life. .918

Question Factor
Loading

The story on smoking and anxiety provides valuable information for .587

college students.

The smoking related anxiety story provided information that I ﬁnd very .583

useful.

Excluded Items

Perceived Credibility — Final Items Used

Question Factor
Loading

The Director of the CDC is a credible source for giving out information .823

that pertains to health risks associated with smoking.

The Director of the CDC does not have enough information on the risks of .594

smoking to make these claims. ®

I trust the Director of the CDC to tell the truth about smokiﬂ. .900

I have conﬁdence in the accuracy of the information about smoking and .832

anxiety.

I believe that garette smoking can cause anxiety. .656

Excluded Items

Question Factor
Load'ﬂ

I don’t believe that cigarette smoking has an effect on collgge students. ® -. 146

New Information

Question Factor
LoadirL

How informative was the US. News & World Report on smoking on .905

anxiety?

How well educated do you now feel about the effects of smoking on .922

anxiety after readig the article in US. News & World Report?

To what extent do you think that the report you read in US. News & .920

World Report adds to your knowlegge about smoking and anxiety?

 

 

136

 

 

 

 

 

Appendix H

137

Perception of Problem Severity — Items Used

 

 

Question Factor

Loading
How severe is the problem of smoking related anxiety among college .872
students?

 

How likely do you think it is that smoking related anxiety among college .907
students will get worse?

 

 

How likely do you think it is that smoking related anxiety will become a .890
real problem on the MSU campus

 

 

Excluded Item

 

Question Factor
Loading

 

 

How likely do you think it is that smoking related anxiety might become .266
a problem for you?

 

 

138

 

 

Appendix I

139

Behavior Outcomes

 

 

 

 

 

 

Question Mean SD Factor
Loadiri

How likely is this information about anxiety to affect your 1.56 1.25 .69

current smoking behavior?

How likely are you to seek more information on health 1.74 1.24 .60

problems associated with smoking?

How likely are you to quit smoking or not start based on the 3.02 2.25 .64

information about anxiety?

How likely do you think others would be to quit smoking based 2.49 1.19 .67

on this information about anxiety?

 

 

 

 

 

140

Appendix J

141

Issue Importance — Measures Factor Loadings

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Question Factor
Loading
Magnitude Do you feel personally threatened by the health risks .84
associated with smoking?
In your judgment how great is the threat presented to you .85
by the risk of smoking?
Compared to other threats to your health, how big a .89
problem is the risk associated with smoking?
Likelihood Do you feel that a college student like you is likely to be .74
affected by smoking?
Do you feel that you personally are likely to contract a .87
health problem associated with smoking?
How often do you think about the likelihood of having a .79
health problem due to smoking?
Irnmediacy I think that I will be affected by health problems due to .88
smoking within the next year.
I will be affected by smoking soon. .9]
Irnmediacy Item Excluded
Question Factor
Loading
I will not be affected by smoki_ng any time soon. ® .55

 

 

®This question was reverse coded.

142

 

 

 

Appendix K

143

Information Utility - Measures Factor Loadings

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Question Factor
Loading
Relevance The story on smoking and anxiety provides valuable .61
information for college students.
The smoking related anxiety story provided information .82
that I ﬁnd very useful.
The information on smoking and anxiety is personally .84
relevant to me.
The story on smoking and anxiety is applies directly to .80
me life.
Credibility The director of the CDC is a credible source for giving .81
out information that pertains to health risks associated
with smoking.
The director of the CDC does not have enough .58
information on the risks of smoking to make these
claims. ®
I trust the director of the CDC to tell the truth about .88
smoking.
I have conﬁdence in the accuracy of the information .8]
about smoking and anxiety.
I believe that cigarette smoking can cause anxiety. .59
New How informative was the US. News & World .89
Information Report report on smokirg on anxiety?
How well educated do you now feel about the effects of .94
smoking on anxiety after reading the article in US. News
& World Report?
To what extent do you think that the report you read in .90
US. News & World Report adds to your knowledge
about smoking and anxiety?
Credibility Item Excluded
Question Factor
Loading
I don’t believe that cigarette smoking has an effect on college students. .18

®

 

 

®This question was reverse coded.

144

 

 

Appendix L

145

Perception of Problem Severity — Items Used

 

 

Question Factor

Loading
How severe is the problem of smoking related anxiety among college .85
students?

 

How likely do you think it is that smoking related anxiety among college .92
students will get worse?

 

 

How likely do you think it is that smoking related anxiety will become a .88
real problem on the MSU campus

 

 

Excluded Item

 

Question Factor
Loading

 

 

How likely do you think it is that smoking related anxiety might become .48
a problem for you?

 

 

146

 

 

Appendix M

147

Behavior Intentions Scale

 

 

 

 

 

 

 

Question Factor
Loading_

How likely is this information about anxiety to affect your current .79

smoking behavior?

How likely are you to seek more information on health problems .83

associated with smoking?

How likely do you think others would be to quit smoking based on this .78

information about anxiety?

 

Behavior Intention Excluded Item

 

 

 

 

 

Question Factor

Load'mg_
How likely are you to quit smoking or not start based on the .54
information about anxiety?

 

148

References

Adler, J .E. (1997). If the base rate fallacy is a fallacy, does it matter how frequently it is
committed? Behavioral and Brain Sciences, 20, 4, 774-776.

Aﬁﬁ, W.A., Dillow, M.R, & Morse, C. (2004). Examining predictors and consequences
of information seeking in close relationships. Personal Relationships, 11, 4, 429-
449.

Atkin, CK. (1985). Information utility and selective exposure to entertainment media. In
D. Zillmann & J. Bryant (Eds) Selective Exposure to Communication. (pp. 63-
9]). Hillsdale, NJ: Lawrence Erlbaum Associates.

Atkin, C. K. (2001). Theory and principles of media health campaigns. In R. Rice & C.

Atkin (Eds) Public Communication Campaigns. (pp. 49-68). Thousand Oaks: Sage
Publications, Inc.

Aust, C.F. & Zillmann, D. (1996). Effects of victim exempliﬁcation in television news on
viewer perception of social issues. Journalism and Mass Communication
Quarterly, 73, 4, 787- 803.

Bar-Hillel, M. (1980). The base-rate fallacy in probability judgements, Acta
Psychologica, 44, 211-233.

Bar-Hillel, M. & F ischhoff, B. (1981). When do base rates affect predictions? Journal of
Personality and Social Psychology, 41, 671-680.

Berger, CR. (2000). Quantitative depictions of threatening phenomena in news reports:
The scary world of frequency data. Human Communication Research, 26, l, 27-
52.

Berger, C.R. (2002). Base-rate bingo: ephemeral effects of population data on cognitive
responses, apprehension, and perceived risk. Communication Research, 29, 2, 99-
125.

Berlyne, D. E. (1954). A theory of human curiosity. British Journal of Psychology, 45,
180-19].

Boster, F .J ., Cameron, K.A., Campo, S., Liu, W., Lillie, J .K., Baker, E.M., & Yun, K.A.

(2000). The persuasive effects of statistical evidence in the presence of
exemplars. Communication Studies, 51, 3, 296.

149

Brashers, D.E., Goldsmith, D.J, & Hsieh, E. (2002). Information seeking and avoiding in
health contexts. Human Communication Research, 28, 2, 25 8-27] .

Brosius, BB. (2003). Exemplars in the News: A theory of the effects of political
communication. In J. Bryant, D. Roskos-Ewoldsen, & J. Cantor (Eds)
Communication and Emotion: Essays in honor of Dolf Zillmann. (pp. 179-196).
Mahwah, NJ: Lawrence Erlbaum.

Buda, R. & Chamov, B.H. (2003). Message processing in realistic recruitment practices.
Journal of Managerial Issues, 15, 3, 302-316.

Brosius, H.B. & Bathelt, A. (1994). The utility of exemplars in persuasive
communications. Communication Research, 21, 1, 48-79.

Burke, A., Heuer, F., & Reisberg, D. (1992). Remembering emotional events. Memory &
Cognition, 20, 3, 277-290.

Castelli, L, Zogmaister, C., Smith, E.R., & Arcuri, L. (2004). On the automatic
evaluation of social exemplars. Journal of Personality and Social Psychology, 86,
3, 373-3 88.

Center for Disease Control (December, 2005) Tobacco Information and Prevention
Source. Retrieved August 10, 2006, ﬁ'om
http://www.cdc.gov/tobacco/factsheets/AdultCigaretteSmoking_FactSheet.htm

Daschmann, G. (2000). Vox pop & polls: The impact of poll results and voter statements
in the media on the perception of a climate of opinion. International Journal of
Public Opinion Research, 12, 2, 160-181.

F estinger, L. (1 95 7). A Theory of Cognitive Dissonance. Stanford: Stanford University
Press.

Freedman, J .L. (1965). Conﬁdence, utility, and selective exposure: A partial replication.
Journal of Personality & Social Psychology, 2, 5, 778-780.

Gibson, R. & Zillmann D. (1994). Exaggerated versus representative exempliﬁcation in
news reports. Communication Research, 21, 5, 603-625.

Gibson, R. & Zillmann, D. (1998). Effects of citation in exemplifying testimony on issue
perception. Journalism & Mass Communication Quarterly, 75, 1, 167-176.

Gilbert, D.T., & Jones, EB. (1986). Exempliﬁcation: The self-presentation of moral
character. Journal of Personality, 54, 593-615.

150

Glik, D., Berkanovic, E., Stone, K., Ibarra, L., Jones, M., Rosen, B., Schreibman, M.,
Gordon, L., Minassian, L., & Richards, D. (1998). Health education goes
Hollywood: Working with prime-time and daytime entertainment television for
immunization promotion. Journal of Health Communication, 3, 263-282.

Goodie, AS. (1997). Direct experience is ecologically valid. Behavioral and Brain
Sciences, 20, 4, 777-779.

Gunther, A.C. & Christen, CT. (1999). Effects of news slant and base rate information
on perceived public opinion. Journalism and Mass Communication Quarterly, 76,
2, 277-292.

Hayes-Roth, B. & Hayes-Roth, F. (1977). Concept learning and the recognition and
classiﬁcation of exemplars. Journal of Verbal Learning and Verbal Behavior, 1 6,
321-338.

Heuer, F. & Reisberg, D. (1990). Vivid memories of emotional events: The accuracy of
remembered minutiae. Memory & Cognition, 18, 5, 496-506.

Huh, J ., DeLorme, D.E., & Reid, L.N. (2004). The information utility of DTC
prescription drug advertising. Journalism and Mass Communication Quarterly,
81, 4, 788-807.

Kline, K. N. (2003). Popular media and health: Images, effects, and institutions. In T. L.
Thompson & A. M. Dorsey & K. I. Miller & R. Parrott (Eds), Handbook of
Health Communication (pp. 557-581). Mahwah, NJ: Lawrence Erlbaum
Associates.

Knobloch, S., Carpentier, F.D., & Zillmann, D. (2003). Effects of salience dimensions of
informational utility on selective exposure to online news. Journalism and Mass
Communication Quarterly, 80, 1, 91-108.

Koehler, J .J . (1996). The base rate fallacy reconsidered: descriptive, normative, and
methodological challenges. Behavioral and Brain Sciences, 19, 1, 1-54.

Kreuter, M.W., Bull, F .C., Clark, E.M., & Oswald, D.L. (1999). Understanding how
people process health information: A comparison of tailored and nontailored
weight-loss materials. Health Psychology, 18, 5, 487-494.

Krupat, E., Smith, R.H., Leach, C.W., & Jackson, MA. (1997). Generalizing from
atypical cases: How general a tendency? Basic and Applied Social Psychology,
19, 3, 345-361.

Locke, EA. (1986). Generalizing ﬁ-om laboratory to field settings: Research ﬁndings

from industrial-organizational psychology, organizational behavior, and human
resource management. Lexington, MA: Lexington Books.

151

McGuire, W. J. (2001). Input and output variables currently promising for constructing
persuasive commrmications. In R. Rice & C. Atkin (Eds) Public Communication
Campaigns. (pp. 22-48). Thousand Oaks: Sage Publications, Inc.

Millar, M.G. & Millar, K.U. (2000). Promoting safe driving behaviors: The inﬂuence of
message framing and issue involvement. Journal of Applied Social Psychology,
30, 4, 853-866.

Moenaert, R.K. & Souder, W.E. (1996). Context and antecedents of information utility at
the R&D/marketing interface. Management Science, 42, 11, 1592-1611.

National Institute on Drug Abuse (2001). NIDA Notes. Retrieved August 10, 2006, from
http://wwwdrugabuse.gov/NIDA_Note§/NNV011 6N 1/Smoking.html

Neighbors, C., Palmer, R. S., Larimer, M. E. (2004). Interest and participation in a
college student alcohol intervention study as a function of typical drinking.
Journal of Studies on Alcohol, 65, 6, 736-741.

Newhagen, J .E. & Reeves, B. (1992). The evening’s bad news: Effects of compelling
negative television news images on memory. Journal of Communication, 42, 2,
25-41.

Perloff, RM. (2001). Persuading People to Have Safer Sex. Mahwah, NJ: Lawrence
Erlbaum.

Perry, S.D. & Gonzenbach, W.J. (1997). Effects of news exempliﬁcation extended:
considerations of controversiality and perceived future opinion. Journal of
Broadcasting & Electronic Media, 41, 2, 229-245.

Perry, S.D. & Gonzenbach, W.J. (2000). Inhibiting speech through exemplar distribution:
Can we predict a spiral of silence? Journal of Broadcasting & Electronic Media,
44, 2, 268.

Petty, R.E. & Cacioppo, J .T. (1986) Communication and Persuasion: Central and
peripheral routes to attitude. New York: Springer-Verlag.

Rodgers, L. (2003) Images of Islam: Exempliﬁcation as elegance in the post-9/1 1 works
of Thomas Friedman. A paper presented at the annual Association for Education
in Journalism and Mass Communication conference.

Rosenstock, I. M. (1990). The health belief model: Explaining health behavior through
expectancies. In K. Glanz, F. M. Lewis, & B. K. Rimer (Eds), Health behavior
and health education: Theory, research, and practice (pp. 39-62). San Francisco:
Jossey-Bass.

152

Rothbart, M., Fulero, S., Jensen, C., Howard, J ., & Birrell, P. (1978). From individual to
group impressions: Availability heuristics in stereotype formation Journal of
Experimental Social Psychology, 1 4, 237-255.

Rothman, A.J., & Salovey, P. (1997). Shaping perceptions to motivate healthy behavior:
The role of message framing. Psychological Bulletin, 121, 1, 3-19.

Rothman, A.J., & Schwarz, N. (1998). Constructing perceptions of vulnerability:
personal relevance and the use of experiential information in health judgments.
Personality & Social Psychology Bulletin, 24, 10, 1053-1065.

Salmon, C. & Atkin, CK. (2003). Using media campaigns in health promotion. In T.
Thompson, A. Dorsey, K. Miller & R. Parrott (Eds) Handbook of Health
Communication. Mahway, NJ: Lawrence Erlbaum.

Scheider, M. C. (2001). Deterrence and the base rate fallacy: An examination of perceived
certainty. Justice Quarterly. Jq, I8, 1, 63- 86.

Sears, D. O., & Freedman, J. L. (1967). Selective exposure to information: A critical
review. Public Opinion Quarterly, 31, 2, 194-213.

Sharf, B. F., & Freimuth, V. S. (1993). The construction of illness on entertainment
television. Health Communication, 5(3), 141-160.

Sherman, J .W. (1996). Development and mental representation of stereotypes. Journal of
Personality and Social Psychology, 70, 6, 1126-1142.

Sia, T.L., Lord, C.G., Blessum, K.A., Thomas, J.C., & Lepper, MR. (1999). Activation
of exemplars in the process of assessing social category attitudes. Journal of
Personality and Social Psychology, 76, 4, 517-518.

Singhal, A., & Rogers, E. M. (1999). Entertainment-education: A communication
strategy for social change. Mahwah, NJ: Lawrence Erlbaum Associates.
Skumanich, S.A. & Kintsfather, DP. (1996). Promoting the organ donor card: a
causal model of persuasion effects. Social Science & Medicine, 43, 3, 401-409.

Thayer, HS. (1968). Meaning and action: A critical history of pragmatism. Indianapolis,
IN: Bobbs-Merrill.

Thompson, W.N. (1975). The Process of Persuasion: Principles and Readings. New
York: Harper & Row.

Trenholm, S. (1989). Persuasion and Social Influence. Englewood Cliffs, NJ: Prentice
Hall.

153

Wilton, P.C. & Myers, J .G. (1986). Task, expectancy, and information assessment effects
in information utilization processes. Journal of Consumer Research, 12, 4, 469-
486.

Witte, K. (1992). Putting the fear back into fear appeals: The extended parallel process
model. Communication Monographs, 59, 329-349.

Witte, K., & Morrison, K. (2000). Examining the inﬂuence of trait anxiety/represstion-
sensitization on individuals’ reactions to fear appeals. Western Journal of
Communication, 64, 1, 1-28.

Zillmann, D. (1999). Exempliﬁcation theory: Judging the whole by some of its parts.
Media Psychology, 1 , 69-94.

Zillmann, D. (2000). Mood management in the context of selective exposure theory. In
M. Roloff (Ed.), Communication Yearkbook 23 (pp. 103-123). Thousand Oaks:
Sage Publications, Inc.

Zillmann, D. (2002). Exempliﬁcation theory of media inﬂuence. In J. Bryant & D.
Zillmann (Eds), Media effects: Advances in theory and research (pp. 19-41).
Mahwah, NJ: Lawrence Erlbaum Associates.

Zilhnann, D. & Brosius, HE. (2000). Exempliﬁcation in Communication: The inﬂuence
of case reports on the perception of issues. Mahwah, NJ: Lawrence Erlbaum
Associates.

Zillmann, D., Gibson, R., Sundar, S.S., & Perkins, J .W. (1996). Effects of
exempliﬁcation in news reports on the perception of social issues. Journalism and
Mass Communication Quarterly, 73, 2, 427.

Zillmann, D., Perkins, J.W., & Sundar, 8.8. (1992). Impression formation effects of

printed news varying in descriptive precision and exempliﬁcation.
Medienpsychologie, 4, 168-185.

154

   

utiltllﬂislitﬂti