he: Luna.

32.1
I.” IVS... . .f
{mhﬁmwuuhw P...
:55). If“?
.35.}...‘13.
‘1... _ {yiwehjdﬂmm
r: , 5.”... 5...". xi .3.
5.3%. i :w" Eran

ti!

 

 

 

on

 

n

n‘!

. .91
.Jﬂhv.~t..§. y

b

is: ,
.. - A.
a. ﬁt. an

i a»

. A :14: 2
£1... 1.5.1....

It! .1. :6... 5x113} it?!
3 In

 

7.009

 

LIBRARY
MlCi iiyan State
University

 

This is to certify that the
dissertation entitled

NEWS MEDIA, INDIVIDUAL-LEVEL TRAITS. AND BEHAVIOR
CHANGE IN FEAR APPEAL RESEARCH

presented by

BRAD LOVE

has been accepted toward fulfillment
of the requirements for the

PERM” degree in Media and Information Studies

 

(Katmai—U 0) ML“. ”VIVIA/

 

Major Professor’ s Signature
5 ' 4 '07

Date

 

MSU is an Afﬁrmative Action/Equal Opportunity Employer

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 Kthrolecc&Pres/ClRC/DateDue indd

 

 

NEWS MEDIA, INDIVIDUAL-LEVEL TRAITS, AND BEHAVIOR CHANGE IN
FEAR APPEAL RESEARCH

By

Brad Love

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Media and Information Studies

2009

ABSTRACT

NEWS MEDIA, INDIVIDUAL-LEVEL TRAITS, AND BEHAVIOR CHANGE IN
FEAR APPEAL RESEARCH

By

Brad Love

Most explanations of health risks are presented to readers and listeners as “fear
appeals,” messages that attempt to arouse fear to modify personal behavior. Fear appeals
persuade by implying or stating that not following some suggested behavior will lead to a
negative outcome. Individuals do not make decisions on an entirely rational basis so
using emotions to get attention can improve exposure to essential information and then
lead to behavior change in many cases.

Not every fear appeal is successﬁil, of course. Exactly what makes for effective use
of fear appeals and what leads to failure has been the debate of much scholarly effort
during the last ﬁve decades. The resulting data offer few well-supported conclusions,
despite the volume of research in ﬁelds from communication to environmental science.

The central goal of this dissertation is to build upon earlier studies to develop a more
accurate model to predict how individuals understand fear appeals. The research seeks to
add to prior work by examining other variables that can inﬂuence individuals’ reactions
to fear-arousing messages and to improve on the scientiﬁc rigor used in fear appeal
experiments. In addition, this research will include a new message context by employing
newspaper articles as the delivery medium.

More technically, it extends the EPPM/RPA focus on perceived risk and efﬁcacy by

adding three variables from persuasion research: outcome-relevant involvement, need for

cognition, and sensation seeking. Outcome-relevant involvement examines how much a
person focuses on the consequences of an issue. Need for cognition measures the degree
to which individuals put forth mental effort when considering topics. Sensation seeking
assesses an individual’s need for stimulating new experiences.

More effective survey measurement instruments will avoid some pitfalls common to
prior fear appeal research. Lastly, looking at mass media editorial content will expand the
scope of fear appeal experiments into a new area.

An online experiment was conducted with 1105 individuals who each read one of
ﬁve fabricated news article intended to provide a certain level of perceived threat and
perceived efﬁcacy. In addition, respondents completed survey instruments to measure the
presence of certain psychological traits commonly used in persuasion research.

Data indicate that perceived threat was a solid predictor of outcome-relevant
involvement, which in turn was a predictor of positive behavioral intention. Need for
cognition was a signiﬁcant moderator between message content and perceived threat.
Sensation seeking, however, was not a signiﬁcant predictor, contrary to the hypothesis.

Within the context of this experiment, fear appeal and efﬁcacy messages included in
mass media content seem to be effective means to altering individual behavioral
intentions. Individual-level considerations are also worth considering when media are
presented in a context that allows for personalized customization. The importance of
customization will grow increasingly obvious as digital media continue to become more

widespread.

TABLE OF CONTENTS

LIST OF TABLES ....................................................................................................... vii
LIST OF FIGURES ...................................................................................................... ix
CHAPTER 1. INTRODUCTION ................................................................................. 1
Identiﬁcation of researchable problem ......................................................................... 8
Rationale and contribution ........................................................................................... 9
Additional variables ................................................................................................. 9
Design and measurement improvements ................................................................ 1 1
Fear appeal context ................................................................................................ 12
Importance of the study ............................................................................................. 13
CHAPTER 2. LITERATURE REVIEW .................................................................... 14
Early models .............................................................................................................. 15
Drive explanation .................................................................................................. 16
Resistance explanation ........................................................................................... 17
Curvilinear hypothesis ........................................................................................... 18
Parallel response explanation ................................................................................. 20
Subjective expected utility models ......................................................................... 22
EPPM/RPA ............................................................................................................... 25
RPA extension ............................................................................ . .......................... 31

Key variables in fear appeal research ......................................................................... 32
Dependent variables .................................. . ............................................................... 33
Behavior ................................................................................................................ 33
Attitude ................................................................................................................. 36
Behavioral intention ............................................................................................... 39
Common endogenous variables ................................................................................. 40
Perceived threat ..................................................................................................... 40
Perceived fear ........................................................................................................ 41
Efﬁcacy ................................................................................................................. 45
Additional variables in fear appeal research ............................................................... 46
Source credibility ................................................................................................... 46
Speciﬁcity of instructions ...................................................................................... 48
Demographic, sociopsychological, and health attitudes .......................................... 51
Volunteer/nonvolunteer ......................................................................................... 52
Topic relevance ..................................................................................................... 54
Sensation seeking ................................................................................................. 60
Need for cognition ................................................................................................. 61
Other variables ...................................................................................................... 62
Methodological issues in fear appeal research ............................................................ 63

iv

Measurement ................................................. . ........................................... . ........... 63

Experiment design ................................................................................................. 64
CHAPTER 3. HYPOTHESES AND RESEARCH MODEL ..................................... 67
CHAPTER 4. METHOD ............................................................................................. 70

Research design ......................................................................................................... 70

Article design according to the EPPM .................................................................... 72

Article attributes .................................................................................................... 77

Instrument selection ............................................................................................... 80

Data collection ........................................................................................................... 82

Participants ............................................................................................................ 83

Stimulus ................................................................................................................ 85

Measurement instruments ...................................................................................... 85

Data analysis ............................................................................................................. 87
CHAPTER 5. RESULTS AND DISCUSSION ........................................................... 89
Normality .................................................................................................................. 89
Manipulation check ................................................................................................... 90
Measurement instrument assessment .......................................................................... 94
Interaction terms ........................................................................................................ 95
Construct validity ............................................................................................ .. .......... 97
Regression analyses ................................................................................................... 98
Hypothesis testing ................................................................................................... 115

Control variables ................................................................................................. 1 15

Hypotheses .......................................................................................................... 118

Non-predicted relationships ................................................................................. 124
CHAPTER 6. SUMMARY AND CONCLUSIONS ................................................. 129

Theoretical Implications .......................................................................................... l3 1

Fear appeals in news content ................................................................................ 131

Additional persuasion variables in the Extended Parallel Process Model .............. 132

Design and measurement issues ........................................................................... 134

Practical Implications .............................................................................................. 135

Fear appeals in news content ................................................................................ 135

Additional persuasion variables in the Extended Parallel Process Model .............. 138

Design and measurement issues ........................................................................... 139

Post-hoe ﬁndings ..................................................................................................... 140
Limitations and future research ................................................................................ 141
Summary of implications ......................................................................................... 150
APPENDICES ........................................................................................................... 152
APPENDIX A: Treatment Articles ......................................................................... 153
APPENDIX B: Pilot Study ..................................................................................... 159
Pilot study results ................................................................................................. 161
Instrument reliabilities ......................................................................................... 166

V

APPENDIX C: Pilot Test Survey ............................................................................ 170
APPENDIX D: Data Collection Instrument ............................................................ 191

BIBLIOGRAPHY ..................................................................................................... 217

vi

LIST OF TABLES

Table 1. Article Attributes: Length and Readability ....................................................... 79
Table 2. Post-hoe Tests for Group Means on Perceived Risk Index ................................ 92
Table 3. Percentage Comprehension Check Scores by Item and Group .......................... 93
Table 4. Descriptive Statistics for Believability Items .................................................... 93
Table 5. Believability Index Descriptive Statistics for Groups ....................................... 94

Table 6. Descriptive Statistics and Cronbach Alphas for Measurement Instruments ....... 95

Table 7. Bivariate Correlations Among Variables .......................................................... 99
Table 8. Hierarchical Regression Analysis for Behavioral Intention ............................. 101
Table 9. Hierarchical Regression Analysis for Behavioral Intention ............................. 102

Table 10. Summary of Hierarchical Regression Analysis for Variables Predicting

Behavioral Intentions in Reaction to a Fear Appeal Message ....................................... 104
Table 11. Hierarchical Regression Analysis for Outcome-Relevant Involvement ......... 106
Table 12. Hierarchical Regression Analysis for Outcome-Relevant Involvement ......... 106

Table 13. Summary of Hierarchical Regression Analysis for Variables Predicting
Outcome-Relevant Involvement in Reaction to 3 Fear Appeal Message ....................... 108

Table 14. Hierarchical Regression Analysis for Need for Cognition Interaction ........... 110

Table 15. Hierarchical Regression Analysis for Need For Cognition Interaction with
Treatment Condition on Perceived Threat .................................................................... 110

Table 16. Summary of Hierarchical Regression Analysis for Need for Cognition
Interaction with Treatment Condition on Perceived Threat .......................................... 1 12

Table 17. Hierarchical Regression Analysis for Sensation Seeking Interaction with

Treatment Condition on Perceived Threat .................................................................... 113

Table 18. Hierarchical Regression Analysis for Sensation Seeking Interaction with

Treatment Condition on Perceived Threat .................................................................... 113
vii

Table 19. Hierarchical Regression Analysis for Sensation Seeking Interaction Terms on
Perceived Threat .......................................................................................................... 114

Table 20. Pilot Test Descriptive Statistics for Perceived Risk by Article Subject ......... 162

Table 21. Perceived Threat by Article Topic ................................................................ 163
Table 22. Group Descriptive Statistics for Believability Index Scores .......................... 164
Table 23. Item Descriptive Statistics for Believability Index ........................................ 165
Table 24. Tukey’s HSD for Believability Items ........................................................... 165

Table 25. Comprehension Check Percentage Correct Scores By Treatment Condition. 166
Table 26. Bivariate Variable Correlations in Pilot Test Data ........................................ 167

Table 27. Descriptive Statistics and Cronbach Alphas for Pilot Test Measurement
Instruments .................................................................................................................. 168

viii

LIST OF FIGURES

Figure 1. Research Model .............................................................................................. 69

Figure 2. Hypothesis Testing Model With Standardized Regression Coefﬁcients ......... 124

ix

 

Chapter 1. Introduction

Most explanations of health risks are presented to readers and listeners as “fear
appeals” (Witte, Meyer, & Martel], 2001), messages that attempt to arouse fear to get
individuals to behave a certain way. Whether in conversation or a magazine article, fear
appeals persuade by implying or stating that not following some suggested behavior will
lead to a negative situation such as injury, death, ﬁnancial loss, or embarrassment. The
goal of a fear appeal is to initially appeal to emotions rather than logic in getting people
to modify their behavior.

The objective of this dissertation is to improve upon existing research and more
accurately predict how individuals react to fear appeal messages. This project will extend
prior work in three major ways: ﬁrst, by incorporating other variables expected to
inﬂuence individuals’ reactions to fear-arousing messages; second, by increasing the
scientiﬁc rigor used in experimental design and measurement; and third, by offering
newspaper articles as an additional context for fear appeal research.

Fear-appeal proponents argue that most “logical and reasoned messages fall on deaf
ears” (Witte et al., 2001, p. 1), presumably because individuals do not make decisions on
an entirely rational basis. Using emotions to get attention, however, can improve
exposure to essential information and then lead to behavior change in many cases. The
obvious idea behind this is that a fear appeal promises a negative outcome for not
completing a certain behavior, and because the listener does not want to experience the
negative outcome, that person will then be more likely to follow the suggestion included

the fear appeal message.

One of the reviews of fear appeal research by Boster and Mongeau (1984)1 explicitly
explains the structure of a fear appeal argument. By their deﬁnition, a fear appeal
argument contains four parts that build upon each other:

1) A statement saying that the reader is vulnerable to a threat.

2) Since the reader is vulnerable, that person should take action to reduce the

vulnerability.

3) The process of becoming less vulnerable includes the reader accepting the

recommendations in the message.

4) Therefore, the reader should accept the recommendations in this message.

An illustration would look something like this:

1) If your car’s brakes fail, you will die in an accident.

2) Since you do not want to die, you should do something to prevent your car’s

brakes from failing.

3) The only effective action is to have the brakes checked by a mechanic.

4) Therefore, you should visit an auto mechanic.

At its simplest, however, any fear appeal contains two basic components, a threat and
a suggested response (Witte et al., 2001). The threat describes some negative
consequences and the response speciﬁes a way to avoid the threat. These types of
arguments can be found in numerous places, from parents talking to children (“You’re
grounded if you don’t clean your room.”) to doctors talking to patients (“If you don’t lose

weight, you’ll have a heart attack”) (Witte et al., 2001). Political and product

 

1 Easter and Mongeau also explain the propositional logic behind this deﬁnition, while pointing out that
even though the validity of this form is easily demonstrated, the effectiveness of any fear appeal
additionally rests upon the quality of evidence supporting that appeal.

advertising are also rife with examples. Candidates often make claims such as, “The
economy will fail and you’ll lose your job if you vote for my opponent.”

In the product advertising realm, a prime example of an appeal to fear is the “real
stories” ad campaign for OnStar, the telematics service owned by General Motors. The
OnStar service combines a hands-free cell phone built into the vehicle, a global
positioning system, and an emergency call center (Fahey, 2003). Pushing the OnStar
button on the rearview mirror connects the driver with an OnStar call center employee
who establishes the vehicle’s location and appropriate services to aid the driver, whether
that means an ambulance or driving directions. The service can also connect with the
vehicle’s engine-control computer to check for problems and use the GPS system to
locate stolen cars.

Beginning in 2002, OnStar’s radio advertising started integrating real calls from
traumatized drivers to promote the safety uses of this technology. A year later, the
company purchased time on cable networks such as Home & Garden TV, the Food
Network, and The Learning Channel to air 30-minute, documentary-style TV
commercials featuring “teary, thankful women recounting frights; black-and-white, Cops-
like re-creations of trauma scenes; interviews with trauma surgeons; and references to the
inadequacy of cell phones” (Fahey, 2003).

One commercial features the testimony of Geraldine Felax of Rogers City, Mich.
“Without the OnStar I would have never received the help in time,” she said of an
incident that involved her husband being trapped underneath a tractor. “There’s no way
he would have made it.”

The implications follow the format laid out by Boster and Mongeau (1984):

-3-

1) OnStar is encouraging people to feel vulnerable to accidents and other threatening

situations.

2) Since individuals are vulnerable to these situations, they must do something to

prepare.

3) To protect themselves and their families, the most effective option is a

subscription to OnStar.

4) Therefore, people that truly care f0r their" loved ones should subscribe to OnStar.

More simply, the ads use the threat of an accident to suggest an OnStar subscription
as the most effective response. The campaign is promoting fear of an accident to imply
that individuals without OnStar subscriptions are not doing all they can to protect their
loved ones—and it is working. In the ﬁrst year of the ad campaign, 3,000 subscribers a
month joined the service, according to OnStar (Fahey, 2003). Admittedly, this is also in
conjunction with other sales techniques such as rebates and free initial memberships, but
the ads are doing enough to drive consumer interest.

Not every fear appeal is successful, of course. Many individuals will fail to follow the
path outlined above. Exactly what makes for effective use of fear appeals and what leads
to failure has been the debate of much scholarly effort during the last ﬁve decades. The
resulting data offer few well-supported conclusions despite the volume of work dedicated
to fear appeal research (LaTour, Snipes, & Bliss, 1996). Fields from communication to
environmental science have put resources into the study of fear appeals. The results of
these projects, however, have not always produced the predicted or desired outcomes. To
explain these varying outcomes, researchers have moved through a number of phases and

corresponding models, each predicting a different mental processing path and outcome

-4-

for fear appeal messages (Boster & Mongeau, 1984; Witte & Allen, 2000). A pair of
effects have been consistent and reported regularly, though, even if they’re not found in
every single study (Sutton, l982)——perceived fear and perceived efﬁcacy both inﬂuence
behavioral intention.

The ﬁrst such effect is a consistent connection between perceived fear (or perceived
threat) and following the recommended actions (Beck & Davis, 1978; Boster &
Mongeau, 1984; Sutton, 1982; Witte & Allen, 2000). Perceived fear is the emotional
reaction to an individual’s belief that a particular danger or harm is a possibility in that
person’s life. The issue causing the fear may be real or imagined, but it exists in the
individual’s head and inﬂuences actions. Researchers measure perceived fear through
two dimensions: perceived susceptibility and perceived severity. Perceived susceptibility
covers the perception of the likelihood of experiencing the problem causing the fear.
Perceived severity concerns beliefs about the scale of the problem (Witte et al., 2001). In
the case of the OnStar commercials, perceived susceptibility would be the likelihood of
an accident or some other situation requiring an emergency response. Perceived severity
would involve the seriousness of the outcome of the situation, such as an injury or death.

The second effect found regularly in fear appeal studies is a link between perceived
efﬁcacy and adherence to the suggested actions (Witte & Allen, 2000). In general,
perceived efﬁcacy is found to have stronger links to following the message’s suggestions
than perceived risk. Perceived efﬁcacy refers to the perceived effectiveness of a
prescribed action and a person’s ability to complete it. Like perceived risk, researchers
measure it through two dimensions, response efﬁcacy and self-efﬁcacy (Witte et al.,

2001). Response efﬁcacy concerns perceptions of the effectiveness of the action

-5-

promoted by the message. Self-efﬁcacy covers individuals’ beliefs about their ability to
complete the proscribed response. Again, for the OnStar advertising, response efﬁcacy
would concern the ability of OnStar to protect people from accidents or negative
outcomes. Self-efﬁcacy would be about individuals’ ability to use the OnStar system or
to maintain a consistent subscription.

Considering this, current models predicting the outcomes of fear appeals have worked
to include both perceived risk and efﬁcacy. In addition, these models have examined '
research from other ﬁelds and incorporated additional constructs to better predict
outcomes from fear-inducing messages. Along with adding new ideas, current research
has also attempted to improve the quality of experiments by doing a better job of
inC111<=ing a frightened reaction with the messages and then measuring their reactions to
the Situation.

Two examples of adding ideas are Witte’s Extended Parallel Process Model (EPPM)
(W itte, 1992b) and the Risk Perception Attitude (RPA) framework developed by Rimal
arid Real (2003). Both work to improve the theoretical understanding of how fear appeals
itlflHence behavior. The Extended Parallel Process Model advances fear appeals research
in a number of ways. Primarily, it works to explain what makes fear appeals fail to

produce the desired effect, which is the most common result of such messages and
experiments. The Risk Perception Attitude framework, on the other hand, continues with
ﬁle EPPM’s basic ideas and makes progress by better measuring what is happening when
peOple are exposed to fear appeal messages. Most prior research compared the strength of
the fear appeal in the message to the likelihood of following the recommended action.
The RPA instead takes the logical step of comparing the individuals’ reactions to the

-6-

message with the likelihood of following the recommended action, putting the entire
focus on thought processes instead of external stimuli. Where the EPPM considers the
content of the treatment messages, the RPA considers how individuals react to those
treatment messages.

In addition, the RPA’s developers examine how other personality traits inﬂuence
peoples’ reactions to fear appeals, allowing for more precise predictions of the
effectiveness of fear appeals. For example, the way individuals acquire new knowledge
and their motivation to protect themselves from negative situations are two elements the
RPA introduced to fear appeal research.

S uch changes follow the general outline advocated by the authors of several reviews
0f fear appeal research (Boster & Mongeau, 1984; Witte & Allen, 2000). Understanding
how other traits inﬂuence the effectiveness of these messages may explain the weak——
thollgh consistent—links that perceived risk and perceived efﬁcacy have with adherence
to message-suggested behaviors. The discovery of such relationships could serve to
fol‘ecast the likelihood of positive outcomes from campaigns featuring fear-arousing

meSSages.

Fear appeal research could also beneﬁt from an expansion in terms of the type of
media content used in experiments. It has focused to date almost exclusively on
ecl\lcational materials——brochures, posters, and videos—designed to communicate health

itlfOrmation. The usage of mainstream-media editorial content such as newspaper articles
01”TV news stories has been rare——if not nonexistent—despite the fact that news content
often serves a key information-providing purpose (DeJong & Winsten, 1990) in
Providing the daily narrative for society (Nell, 2002; Y. J. Park, 2005). Sensationalism

-7-

remains a common feature in some content as news organizations feel compelled to
arouse emotions to attract a maximum amount of attention (Grabel, Zhou, Lang, & 80113,
2000; Hofstetter & Dozier, 1986; R. Jung, 1982). As such, fear appeal research fails to
address what may be the most common source of fear appeals.
Walnut-:21.

Although prior studies offer valuable building blocks for fear appeal research, much
work remains. The data often fail to match researchers’ predictions, nor do those data do
much to clarify the ways in which individuals process fear appeals. The central goal of
this dissertation is to build upon knowledge from earlier studies to develop a more
accurate model to predict how individuals from the general population understand fear
appeal 5. The dissertation’s goal is being approached in three main ways. The research
seeks to add to the information acquired through prior work by examining other variables
that can inﬂuence individuals’ reactions to fear-arousing messages and to improve on the
Sc"'ietltiﬁc rigor used in fear appeal experiments. In addition, this research will also
inc11.1de a new message context by employing newspaper articles as the medium
displaying fear-arousing messages.

More speciﬁcally, this research extends the EPPM/RPA focus on perceived risk and
efﬁcacy by adding three variables from other persuasion research: outcome-relevant
ill"Olvement, need for cognition, and sensation seeking. Outcome-relevant involvement

examines how much a person focuses on the potential consequences of a particular issue.
N eed for cognition measures the degree to which individuals like to put forth mental
E’ffort when considering topics. Sensation seeking assesses an individual’s need for

Stimulating new experiences.

More effective survey measurement instruments will improve the accuracy of the
experimental results and avoid some pitfalls common to prior fear appeal research.
Lastly, looking at mass media editorial content will expand the scope of fear appeal
experiments into a new area, one that is potentially viewed quite differently by readers.
Health education materials such as brochures have an obvious agenda regarding the
reader’s behavior. Brochures’ messages may encounter some resistance because their
goal of behavior modiﬁcation is obvious. Traditional media such as newspapers, on the
other hand, do not necessarily share the same biases, as their information is intended to
be obj ective instead of persuasive (Gans, 1985; Kuklinski & Sigelman, 1992; Niven,
20C)3 ; Schiller, 1979; Schudson, 2001). Thus, the research being discussed here seeks to
iI‘TIIDI‘ove fear appeal research in three signiﬁcant and distinct ways, all working to
ir11131‘ove the effectiveness and usefulness of a common model.
QR ignalg and contribution
The following sections will discuss each of the stated goals: 1) to examine individual-
IQVGI persuasion-related variables; 2) to reduce the impact of poor experimental design
and measurement common to prior research; 3) and to open the context of fear appeal
re Search.
Additional variables
Other variables in the present study will come from persuasion research that spans
SeVeral disciplines. Outcome-relevant involvement will be one addition to the

EPPM/RPA framework as a mediator2 between perceived risk and an intention to change

\
A variable functions as mediator when it meets the following criteria: a) the initial variable is correlated
with the outcome variable; b) variations in the level of the independent variable signiﬁcantly account for
Variations in the presumed mediator; c) variations in the mediator signiﬁcantly account for variations in the

-9-

one’s behavior, meaning it controls the relationship as a necessary bridge between the
two variables. This individual-level variable measures how readily a respondent
considers consequences surrounding a particular issue (Cho & Boster, 2005). For
example, high-scorers here generally like to consider the message, seek information on
the topic and objectively process the content.

Adding outcome-relevant involvement to the EPPM/RPA framework leads to the
prediction that individuals perceiving a high personal risk after reading the fear appeal
message will also report high levels of outcome-relevant involvement, leading them to be
more likely to respond to the stimulus message in an action-oriented manner. A prior
Smdy using educational brochures in a student population found that high involvement
i1'1C1iV'iduals were likely to make action-oriented behavior change “and hence were more
persuaded by high fear appeal messages than low involvement individuals” (Cheah,
2006. p. 293).

Another addition to the framework will be need for cognition, a psychological
Val-i able measuring the degree to which people enjoy and undertake effortful mental
acti\Iities (Cacioppo, Petty, Feinstein, & Jarvis, 1996). It has been tested in some recent
fear appeal research but never in a newspaper environment or in a broad population

Salhple. One study, for example, involved a sample containing only individuals likely to

rﬁllieady be in possession of a high motivation to comply with message recommendations

\
dependent variable; and d) for complete mediation, when the lV-mediator and the mediator-DV path are
Putrolled, a previously signiﬁcant relation between the independent and dependent variable is no longer

S‘ghiﬁcant (Baron & Kenny, 1986; Brown, 1997).

The result, as Baron and Kenny (1986, p. 1176) write, is that: "In general, a given variable may be said to
ﬁction as a mediator to the extent that it accounts for the relation between the predictor and the criterion.
ediators explain how external physical events take on internal psychological signiﬁcance. Whereas

mOderator variables specify when certain effects will hold, mediators speak to how or why such effects
Occur."

-10-

(Williams-Piehota, Pizarro, Silvera, Mowad, & Salovey, 2006). People scoring high in
need for cognition tend to form judgments by examining relevant arguments as opposed
to extraneous cues, leading to the expectation that individuals high in need for cognition
will perceive their personal risk as higher following the presentation of logical, rational
information in the fear appeal. Need for cognition is expected to moderate the
relationship between the initial message and the perceived risk that results, meaning that
need for cognition will inﬂuence the strength of the relationship between the two
vari ables?

Sensation seeking is also expected to play a moderating role between message
exposure and perceived risk. This variable focuses on the need for new experiences
t1'3l-I‘()1..1gh often risky behavior, including dangerous activities and non-conventional
life Styles. This variable has been linked with accident involvement and rejection of
p er Snasive messages among young adults (Cameron, Leshner, Stephenson, & Choi,
20()6), leading to the expectation that individuals high in sensation seeking will perceive

less risk following exposure to the fear appeal message.
Design and measurement improvements
Methodological issues common to fear appeal research include unreliable
IIleasurement and a failure to examine the connection between reactions to the

e)‘iperimental message and outcomes. Incomplete, single—item measurement is a feature of

3\

The seminal work on deﬁning mediators and moderators deﬁnes a moderator in the following way: "In
ge“era! terms, a moderator is a qualitative (e.g., sex, race, class) or quantitative (e.g., level of reward)
Variable that affects the direction and/or strength of the relation between an independent or predictor

V‘M'iable and a dependent or criterion variable. Speciﬁcally within a correlational analysis framework, a
mOderator is a third variable that affects the zero-order correlation between two other variables. In the

{here familiar analysis of variance (ANOVA) terms, a basic moderator effect can be represented as an

“Iteraction between a focal independent variable and a factor that speciﬁes the appropriate conditions for its

0peration" (Baron & Kenny, 1986, p. 1174).
- 1 1 -

much fear appeal research, particularly in early studies (Boster & Mongeau, 1984).
Recent studies tend to use better measurement but often fail to use the most reliable or
discriminant instruments available (Cho & Boster, 2005). While these studies can often

still find differences between treatment groups, measuring variables more accurately
could offer details that will help design more effective messages in the future.

Other research has linked reactions to experimental messages and outcome variables
(LaTour & Rotﬁeld, 1997; Rimal & Real, 2003), but such workiis rare and has. often
focused almost exclusively on unique populations, such as students or individuals living
under particular conditions (i.e., areas with high prevalence of HIV/AIDS). While

undoubtedly valuable, projects repeatedly focusing on speciﬁc populations naturally raise
que Stions about the generalizability of the results.

Fear appeal context
Lastly, this research will include content representing mass media news stories. Prior

Work almost exclusively used materials afﬁliated with health campaigns, including
I3(3311ers, brochures, and other advertisements. However, news content plays an essential
role in real-life behavior-modiﬁcation campaigns. In the context of unmatchable
corI>orate advertising budgets, pro-social organizations in particular depend upon
eCIitorial content to expand the reach of their campaigns. Through this avenue, many of
ﬁle Same messages that ﬁll paid-for media also enter independently produced news
co“tent. In addition, the opening paragraph of much news content uses fear appeals to
gain attention for the story as a whole or because the content-producer feels a fear
message is the best way to get an audience’s attention (Grabd et al., 2000; Hofstetter &

Dozier, 1986; R. Jung, 1982).
- 12 -

 

Imgogangg of the study

The outcome of this project will add to the body of knowledge in several ways. This
dissertation will directly test several mediators/moderators that could inﬂuence the

effectiveness of how fear appeal messages function through the known channels of
perceived efﬁcacy and perceived fear. The research here will seek to improve
methodology and broaden the scope of fear appeal research. These additions will aid in
understanding how outcome-relevant involvement, need for cognition, and sensation
seeking inﬂuence reactions to fear appeal messages. With this knowledge, content-
developers can adjust their messages accordingly to gain the desired effect.

The results should help journalists, public relations professionals, and healthcare
Practitioners understand how to better develop messages to change health behavior.
Kl"1(>V\ring how people are likely to respond to behavior-change messages will help those
involved with producing media content more effectively target certain populations and
in(1i\7iduals. The implications for these ﬁndings extend into a number of media-related
ﬁelds, as well as any other context that involves altering behavior or attitudes through
medi ated persuasive messages. While previous ﬁndings might logically be extended to

illeltide mass media editorial content, prior research has done no testing in this direction.
On a related note, understanding the constructs that inﬂuence desired outcomes also
helps researchers and practitioners understand how people think. This is, of course,
13raCtically useful in terms of message design, but it is also beneﬁcial in understanding
how individuals process messages. It is possible that the mental processes that go into
Widerstanding fear appeals will relate to the processes that help individuals understand

0ther message types as well.

-13-

 

Chapter 2. Literature Review

Fear appeals have not lacked for attention from researchers, and the often conﬂicting
results have produced a complicated body of literature as well as several nuanced meta-
analyses. Part of fear appeals’ popularity as a research topic (Sussman, 1973) may be due

to the seemingly discordant results obtained in different experiments (Boster & Mongeau,
1984 ; Mongeau, 1998; Sutton, 1982; Witte'& Allen, 2000). Some experiments show
unproved conformity with high fear appeals over low ones (Beck & Davis, 1978); others
describe situations where low fear messages were more effective than high (Janis &
FeSl’lback, 1954); and yet another group supports the idea that the amount of intended fear
in a persuasive message has no bearing on how much a respondent follows the
recOtrn'nendations of a message (Wheatley & Oshikawa, 1970). A fourth option posits
that fear appeals interact with other variables, making high-fear messages more effective
in S<)l'ne situations and low—fear messages effective in others (Leventhal, 1971). The result
of 1livese conﬂicting ideas has been a confused situation where some public health

re Se"archers and practitioners alike argue that fear appeals are counterproductive (Covello,
Von Winterfeldt, & Slovic, 1986; Finckenauer, 1982; Geller, 1989).

In typical research on fear appeal messages, respondents are exposed to fear appeals
lTual‘tipulated for strength with one message being strong and one weak (Witte & Allen,
2000). The strength of these messages is then veriﬁed through a manipulation check to

etlSure individuals perceive a signiﬁcant difference between the strong and weak fear
appeal. The ﬁnal measure is to examine whether or not the stronger message results in

Stl'onger outcomes than the weak appeal. Some of the confusion in interpreting results is

-14-

because certain outcomes are based on individuals accepting message recommendations
and others come from rejecting the message. One challenge has been developing theories
to account for and predict both outcomes.
To sort the variety of results, a number of scholars have thoroughly reviewed the
li terature (Boster & Mongeau, 1984; Mongeau, 1998; Sutton, 1982; Witte & Allen,
200 0). Similar efforts have focused on extending existing theory (Rimal & Real, 2003;
Witte, 1998), parsing out models (Prentice-Dunn & Rogers, 1986), or adding other
emotions to the research (Dillard, 1994). Others have placed the focus on fear appeals
Within speciﬁc contexts such as marketing (Ray & Wilkie, 1970) and public health (Job,
198 8). These works have combined with other reviews responsible to shift the focus of
fear appeal research to quantitative meta-analyses (Boster & Mongeau, 1984; Mongeau,
1 99 8 ; Sutton, 1982) following early studies that resembled critical essays concerning
conceptual, operational, and methodological issues (Higbee, 1969; Miller, 1963).
La rlx models
The fear appeal literature reviews each discuss various models outlined by scholars
engaged in the effects of fear-arousing messages on persuasion (Boster & Mongeau,
1 9 84; Witte & Allen, 2000). As with any scholarly endeavor, the models have tended to
build upon each other and the key theories of the time period (Witte & Allen, 2000). Fear
appeal theories have moved through periods of being inﬂuenced by learning theories
(Hovland, Janis, & Kelly, 1953; McGuire, 1968) to focusing on cognition (R.W. Rogers,
1 975) to a more recent trend of examining emotion’s role in behavior change (Dillard,
1 994; Witte, 1992b). Several common models are discussed below as a summary of the

01 . . . . .
der fear appeal llterature. These models include: the drive explanation, the resrstance

-15-

 

explanation, the curvilinear hypothesis, the parallel response explanation, and subjective
expected utility type models such as the protection motivation explanation and the threat
control explanation.
Drive explanation
These early explanations suggest that the effect of persuasive messages is to produce
varying amounts of fear in the audience which results in a drive for action (Hovland,
Jani s, & Kelley, 1953). The messages under study, of course, vary in fear-arousing
content, but the expectation is that the level of fear in audience members will correlate
positively with the level of fear-arousing content. The result of this is that high fear
messages are expected to produce more fear in audience members than moderate or low
fear messages. In addition, the production of high amounts of fear in the audience is also
e"(IDected to inﬂuence the audience’s view of the issue in question. As perceived fear
increases, the audience’s attitude will more closely match that recommended in the
persuasive message. Boster and Mongeau (1984) strongly point out, however, these
Correlations are expected to be substantial but not without error.

These relationships stem from a drive-reduction process initiated by the perceived
fear aroused by the persuasive message. This perceived fear rooted in the persuasive
IheSSage creates a state of drive, making audience members uncomfortable and
lrrlotiVating them to take some drive-reducing action. The drive explanation predicts that
the audience will change its attitudes or behaviors (or both) to reduce the discordant
feeling. There exists any number of possible drive-reducing actions a person could

undertake (Boster & Mongeau, 1984).

-16-

 

However, the data fail to support the direct relationship predicted by the drive model.
Negative correlations between fear manipulations and attitude (Goldstein, 1959; Janis &
Feshback, 1953; Janis & Terwilliger, 1962), negative correlations between fear
manipulations and behavior (Janis & Feshback, 1953; Leventhal, Singer, & Jones, 1965;
Leventhal & Watts, 1966), zero correlation between fear manipulations and attitude
(Kohn, Goodstadt, Cook, Sheppard, & Chan, 1982; Wheatley & Oshikawa, 1970), and
evidence hinting that other variables moderate the fear manipulation-attitude correlation
all work against the drive explanation as an explanation for the effects of fear appeals
(Boster & Mongeau, 1984).

Resistance explanation

This model originally suggested by Janis and Feshbach (1953) takes the opposite
approach of a drive explanation, arguing that a strong fear appeal is not effective at
inﬂuencing the audience in the desired direction because of a process of defensive
avoidance (Miller, 1963). A defensive audience attempts to then rebuff the presented
t111‘<’-=€-.11:, minimizing its importance. Of course, as Miller points out (1963), this whole
proCess is moderated by individual personalities and their interest in the communication,
along with other factors that inﬂuence the relationship with the source.

A low fear message, however, gets through to the audience because the suggestion is
heard without interference from defensive avoidance, making low fear messages
preferable to high fear ones. Compared to the drive explanation, the resistance
explanation poses the opposite relationship between perceived fear and listener response.
The resistance explanation predicts a negative correlation between fear arousal and
adherence to message suggestions.

-17-

 

 

As Boster and Mongeau (1984) describe, the resistance explanation is countered by
some ﬁndings, speciﬁcally several studies that found a positive correlation between fear
manipulations and attitudes (Beck & Davis, 1978; Berkowitz & Cottingham, 1960;
Burnett & Oliver, 1979; Frandsen, 1963; Hewgill & Miller, 1965; Horowitz, 1969;
Horowitz & Gumenik, 1970; Leventhal et al., 1965; McCroskey & Wright, 1971; Powell
& Miller, 1967).
Curvilinear hypothesis
As the name suggests, this hypothesis predicts that high fear and low fear messages
will have a similar effect (a weak one in this case), but moderate fear messages will
PTOduce a different result (increased attitude/behavior change) (Boster & Mongeau,
1934) -4 The link between perceived fear and the audience’s adoption of the suggested
reSIZ><>nse differs from previous models in that this hypothesis posits that a moderate fear
meSsage will be more inﬂuential than either a high or low fear one. Just as the drive
explanation supports high fear messages and the resistance explanation supports low fear
InesSages, the curvilinear hypothesis supports moderate fear messages. It argues that
when an individual is extremely afraid or not very afraid, that person will be less likely to
follow the recommended action of the communication at hand.
A communication attempt to provoke fear that results in a moderate level of fear, on
the other hand, will inﬂuence individual behavior as effectively as possible. However, not
all Versions of this hypothesis offer the same logic that leads to this ﬁnal outcome, as

Boster and Mongeau explain in detail (1984)-

4\

w -
‘tte and Allen (2000) place this model in with the drive theories.

 

 

 

The ﬁrst version of the curvilinear hypothesis comes from Janis (1967) and Janis and
Leventhal (1968) when they argued that low perceived fear leads the audience to dismiss
the threat and high perceived fear causes defensive maneuvers. Moderate fear, though,
stimulates people and puts them in a situation where they are likely to accept the
message’s recommendations. Janis and Leventhal also recognize that other factors such
as individual situations and dispositions inﬂuence the relationship between fear appeals
and compliance.

McGuire (1968) proposes a different reasoning as to why moderate fear messages are
most effective at audience persuasion. His two-factor explanation is the second version of
the Curvilinear hypothesis (Boster & Mongeau, 1984). McGuire argues that fear plays
two roles within a persuasive process, and the two roles mostly serve to cancel each other
out- The ﬁrst role of perceived fear is to act as a drive encouraging an individual to
fOIIOW the prescriptions of the persuasive message. The second function of perceived fear
is as a cue to resist the recommended action; thus, increasing fear increases the likelihood
that the audience will refuse to go along with the suggested response. The result of this,
acc<>rding to McGuire (1968), is a curvilinear relationship between anxiety and
int] Ilenceability that offers maximum susceptibility at intermediate levels of anxiety.

The third version of the curvilinear hypothesis (Boster & Mongeau, 1984) focuses on
I‘Iigbee’s (1970) interpretation that stresses the signiﬁcance of two variables: severity of
the thl'eat and probability the threat will occur without preventative action. Higbee’s
argI‘ll'l’lent is that as perceived fear increases, perceived severity increases exponentially

while perceived probability of the episode decreases exponentially. The severity of the

-19-

 

 

threat and the likelihood of it happening absent preventative action are thus negatively
correlated, creating a curvilinear function.
These U-shaped models failed to develop a solid foundation of evidence, however,
and were rejected during the 19705 (Witte & Allen, 2000) due to the lack of strong
evidence for nonlinearity, among other evidence (see Boster & Mongeau, 1984 p. 365). A
number of publications served to point out that the models’ central ideas were not
Supported (Beck & Frankel, 1981; Giesen & Hendrick, 1974; Mewbom & Rogers, 1979;
R- W- Rogers, 1975; Sutton, 1982) and when curvilinear effects are reported, they are not
as predicted by the hypotheses (Powell, 1965). The lack of supporting evidence for
curvilinear models then helped fear appeal research move from cognitive responses into
the examination of emotional ones (Witte & Allen, 2000).
Parallel response explanation
Fear control and danger control are the essential elements of the parallel response
explanation, an attempt by Leventhal (1970, 1971) to incorporate the additional variables
that Several scholars recognize as inﬂuencing attitudes and behaviors (Boster &
I\/I()1'lgeau, 1984). This explanation posits that fear control and danger control result from
fear appeals as a means of helping the audience members deal with the presented threat.
The ﬁnal attitude will then be a function of the fear and danger control provoked by the
fear‘inducing message. Leventhal, however, did not actually test the proposed model
(15>7q3).
Danger control results in an individual taking action to avoid the presented threat. In
this context, it is an action-taking, problem-solving response where the audience member
takes in information relevant to the threat and deals with the problem at hand. On the

-20-

 

 

Other hand, fear control results in a path focused on lessening the emotional strain
produced by the fear appeal. Rather than marshaling resources to take action, fear control
puts the effort into an emotional coping process focused internally rather than on the
threat. This internal focus can prevent acceptance of the suggested action.
When an individual follows the danger control path, this model predicts that user’s
attitude will match the one prescribed in the fear appeal message. For those following the
fear control path, the model predicts an inverse relationship between the strength of the
fear appeal and the resulting attitude. Although he doesn’t address the details of the
interaction, Leventhal (1970, 1971) states that the two processes interact to produce the
resulting individual attitudes with the result being a main effect of a positive correlation
bet‘t’veen fear appeals and audience adherence to the recommended response.
Sternthal and Craig (1974) also supported the parallel response explanation but
posited a different model. They hypothesized a curvilinear relationship between fear level
and resulting attitude. As fear increases, both fear and danger control increase, meaning
that high levels of fear fail to produce desired outcomes because the processes obstruct
each other. Low fear levels also fail to lead to the desired outcomes because both
prOCesses are weak and audience members have no reason to accept the suggested action.
1\’I(3<1€:rate levels of fear, however, are most likely to generate the desired outcomes
because the danger control process will be relatively strong while the fear control process
Will be weak.
Others have taken other issues with Leventhal’s version of the parallel process
Emil—)Ianation. Rogers (1975) argues that the model lacks explicit details about the stimulus
Va‘ri'élbles predicting the fear and danger control processes. Not having speciﬁc variables

-21-

‘——‘-

prevents thorough testing of the model. In addition, the linking statements and constructs
in the model are so ambiguous, it’s not possible to make accurate predictions linking fear
and persuasion (R.W. Rogers, 1975), making the model untestable (R.W. Rogers, 1975).
The parallel response explanation is valuable because it introduces the constructs of
fear and danger control, but the ambiguity present in parts makes it otherwise less useful
(Boster & Mongeau, 1984). As Boster and Mongeau point out (1984), while this model
does predict all possible outcomes of a fear appeal message, it fails to detail when one
can expect the various responses, making it unfalsiﬁable and of little scientiﬁc value.
Introducing fear and danger control, though, did alter thinking at the time about fear
appeal s—it separated emotional and cognitive processes (Witte & Allen, 2000).
Subjective expected utility models
Models related to Sutton’s SEU predictions posit that individuals examine options
and rationally select the one that seems to be meet present needs (Sutton, 1982). Fear
play S almost no role in these models (Witte & Allen, 2000). Little evidence supports this
idea (Sutton & Eiser, 1984; Sutton & Hallett, 1988), but, ironically, one SEU-based study
found that fear produced the most reliable effect on intentions, despite not actually being
part Of the model (Sutton & Eiser, 1984).

A major challenge with the SEU models is that they assume that individuals will
perform a sort of cost/beneﬁt analysis when making decisions. People are expected to be
capable of knowing and ranking their preferences, estimating the likelihood of success
for those preferences, and then maximizing their choices by balancing a preferred option

With probability of success.

-22-

Subjective expected utility model: Protection motivation explanation
The ﬁrst SEU model, protection motivation describes a person’s desire to avoid a
potential threat. Rogers (1975) posits that this motivation is driven by three factors:
magnitude of noxiousness of a happening, probability the happening will occur without
some behavioral or attitudinal adaptation, and the availability and effectiveness of a
response that could diminish or remove the noxious stimulus. Through these factors, the
model tries to logically explain what makes for an effective fear appeal (Witte & Allen,
20K)()).

Perceptions of these variables combine multiplicatively to create the protection
mOtivation state, provided all three are present to some degree. Of course, the more the
message under consideration contains these three factors, the more the audience will
Perceive them and adjust its attitude, according to Rogers (1975). Thus, the goal for a
meSSage designer is then to increase the amount of protection motivation in order to
increase audience adoption of the recommended action, accepting some error as usual.

As Boster and Mongeau (1984) describe, however, this model moves through four
parts: 1) from the message characteristics 2) to audience perception 3) to protection
InOtiVation 4) and ﬁnally to attitude. As a result, there are several mediators between the
original message and the ﬁnal attitude, which could result in a rather low correlation
bet“Ween the initial message and resulting attitude. Also, should any correlation among the
perceived message characteristics be zero, no protection motivation will be present, and
no ad0ption of message recommendations is expected to take place.

While Rogers did not end up ﬁnding support for his hypothesis of a four-way
interaction (R.W. Rogers, 1985), other protection motivation studies did ﬁnd that at least

-23-

one threat variable (severity or susceptibility) interacted with at least one efﬁcacy
variable (self-efﬁcacy or response efﬁcacy). These interactions then affected message
acceptance outcomes (attitude, intention, behavior) (Kleinot & Rogers, 1982; Maddux &
Rogers, 1983; Wurtele & Maddux, 1987). However, as a whole, the variables in PMT do
not explain when and how fear appeals fail, and the interaction among fear, efﬁcacy, and
probability are not the type predicted (Boster & Mongeau, 1984). Witte and Allen (2000)
did note, though, the existence of main effects for threat and efﬁcacy on outcomes such
that high threat and high efﬁcacy led to more persuasive power. In addition, this model
made a signiﬁcant contribution by moving fear appeal research to measuring
respondents’ perceived fear in place of the highly arousing, negatively charged emotion
fear (Witte, 1998) after fear had remained the focus of research for a little more than 20
years (Witte & Allen, 2000).

Subjective expected utility model: Threat control explanation

The threat control explanation ﬁts in the SEU mold in that it presumes that
individuals respond to fear appeals based upon cognitive factors rather than affective
ones (Boster & Mongeau, 1984). Models before SEU hypothesized that fear and other
factors mediated the link between fear appeals and audience attitudes.

The threat control model depends upon two processes linking the message and threat
control: response efﬁcacy and personal efﬁcacy (Beck & Frankel, 1981). Response
efﬁcacy consists of an individual’s belief that the suggested action will be effective in
reducing the likelihood of the threat actually taking place. Personal efﬁcacy is the

individual’s belief that he or she can successfully complete the recommended action.

-24-

As such, the key element of this model is that it does not depend upon fear. Instead,
the focus is on presenting a real but controllable threat (Beck & Frankel, 1981). The
failure of a message to provoke a certain attitude among an audience stems not from
defensive avoidance by that audience but from the use of a threat perceived as
uncontrollable.

Because this control issue plays such a signiﬁcant role, the correlation between
perceived threat control and audience adherence to the suggested attitude/behavior is
expected to be positive and substantial (Boster & Mongeau, 1984). Beck and Frankel
(1981) do not detail how response and personal efﬁcacy combine to create threat control,
but the relationship is assumed to be linear.

Like the previous model, however, having mediators (two types of efﬁcacy and then
perceived threat control) between the fear appeal and the resulting attitude could result in
what appears to be a weak relationship. A particular message may also negatively
inﬂuence an audience member’s personal or response efﬁcacy, resulting in a negative
correlation between the initial content and adherence to the message’s suggestions

(Boster & Mongeau, 1984).

EPPMZ RPA

The above fear appeal models and literature, however, mostly fail to examine factors
that produce negative outcomes of non-persuasion. That research focuses instead on the
desirable, successful outcomes of attitude change, behavioral intentions, and behaviors in
the direction of message acceptance (Boster & Mongeau, 1984; Mongeau, 1998; Sutton,
1982). As Witte and Allen (2000) describe in their meta-analysis, defensive, message-

resistant responses have not historically been an essential part of the fear-appeal

-25-

 

_

 

literature. Often termed “defensive avoidance,” “minimization,” “denial,” or something
similar, these psychological resistance strategies existed in published research under
many different names, though they in fact referred to a reliable single construct (Smalec,
2000; Witte, 1994a).

Witte and Allen (2000) noted that these unattended-to, negative outcomes also offer
valuable information about the effectiveness of fear appeals. As a result, they used their
review to examine the 13 studies as of 2000 that reported the link between fear appeal
strength and defensive responses,5 ﬁnding that resistant responses increased in strength as
the fear appeals did likewise (r = .195, p < .05). On the other hand, weak efﬁcacy
messages correlated with an increase in defensive responses as well (r = -.105, p < .05).
The defensive, fear control responses also correlate negatively with the more desirable,
action-oriented, danger control responses (r = -.18, p < .05).

The Extended Parallel Process Model works to explain these relationships among
efﬁcacy and the fear control/danger control responses. The EPPM builds upon older
models in an attempt to predict both the successes and failures of fear appeals while re-
integrating fear as a central variable (Witte & Allen, 2000). Essentially, three prior
models combine to produce the EPPM: Leventhal’s parallel response explanation makes
up the foundation of the model; protection motivation theory handles the danger control
portion to explain when and why fear appeals work; and explanations by Janis and
McGuire serve to detail the fear control side concerning when and why fear appeals fail

(Witte, 1992b, 1998; Witte & Allen, 2000).

_

5 Page 601 in Witte and Allen (2000) lists the individual studies and discusses results.
-26-

Fear, of course, is a discordant emotion that individuals naturally strive to eliminate
through the most efﬁcient means possible to return to the prior comfortable mental state
(J. W. Brehm, 1966; S. S. Brehm & Brehm, 1981). Fear is a motivator in this case.
EPPM—following the parallel response literature—predicts that individuals use one of
two paths to reduce fear when they perceive a threat: one to control the danger causing
the threat, the other to control the fear produced by the threat (Witte, Cameron, McKeon,
& Berkowitz, 1996). The ﬁrst, of course, results in action intended to alleviate the
problem, removing the danger. The communicator’s goal is to have that action be the
recommended responses in the original persuasive message. The second option—fear
control——only removes thoughts or recognition of the threat, resulting in a response such
as denial or deﬁance.

According to Witte and Allen (2000), perceived efﬁcacy is the key factor in
establishing which path an individual will follow—danger control to remove the threat or
fear control to ignore it. Like the available response paths, perceived efﬁcacy is divided
into two parts (Witte et al., 1996). The ﬁrst piece is response efﬁcacy, an individual’s
belief about the effectiveness of a particular response to the threat at hand. The second
element is self-efﬁcacy, an individual’s belief in being capable to carry out a particular
response to the threat. Thus, the individual is asking if the suggested response is an
effective one and if he/she can perform that action.

To get individuals on the danger control track where they take action against a threat,
perceived efﬁcacy must be greater than perceived threat. The audience must believe
“they are easily, feasibly, and effectively able to avert a serious and relevant threat from
occurring by adopting the recommended response” (Witte et al., 1996, p. 318). Reacting

-27-

to and feeling a threat (perceived fear) along with an even stronger ability to overcome
that threat (perceived efﬁcacy) through the speciﬁed action promotes the danger control
path. This path is more likely to lead to following the message’s proposition of attitude,
intention, or actual behavioral change. For example, a university student may feel
threatened by a message highlighting the frequency of sexually transmitted infections on
the campus but also believe that the suggested precaution of using condoms is both easy
to do and capable of preventing these STIs. Thus, the student would be likely to
undertake the personally protective behavior because the student perceives the suggested
response as manageable and effective, leading to an efﬁcacy component stronger than the
fear appeal.

When the efﬁcacy element does not overpower the initial fear appeal, fear control
processes take over. Here, individuals opt to not adopt a suggested response to remove a
severe and plausible threat either because they feel incapable of completing the protective
action or because the suggested action will not be effective, in their opinion. This could
be because the recommended response is perceived as being too expensive, too time-
consuming, too difﬁcult, or otherwise ineffective. Rather than concerning themselves
with eliminating the threat, individuals in the fear control process put their efforts into
avoiding thoughts about the threat or minimizing the issue in some other way (J. W.
Brehm, 1966). Combining low perceived efﬁcacy (the combination of ability to complete
the response plus the response’s effectiveness) and a high perceived threat (in terms of a
threat’s perceived severity and susceptibility) results in fear control by the audience

members (Witte et al., 1996).

-23-

Obviously, the health communicator’s goal is to put individuals into a cognitive
danger control mode of accepting health risk messages and taking action. At the same
time, it’s essential to keep the audienCe out of the affective fear control mode that leads to
reaction, rejection, or refutation of a health risk message (Witte, 1992b; Witte et al.,
2001). In practical terms, then, the EPPM encourages message designers to deliver
messages that provide a threat perceived to be real (because individuals feel susceptible
or that the outcome would be severe or both) while also providing an efﬁcacy component
(response efﬁcacy, self-efﬁcacy, or both) stronger than the initial threat. In short, the
cognitive response efﬁcacy must be stronger than the affective fear element.

One way of determining how people will react to a message is to determine what
Witte et a1. call a “critical point” for that message (Witte et al., 1996). The purpose of this
is to mathematically express the point at which people switch from danger control into
fear control. As discussed prior, this is the point at which individuals feel that they are not
capable of taking action to avoid a real and signiﬁcant threat, placing them in the fear
control category. Rather than taking action against the threat itself, individuals in the fear
control process focus on their fear. Health protective behaviors are not the priority for
these individuals because they’re focusing on their reactance or denial of the problem at
hand, thereby allowing fear control to interfere with danger control (Witte, 1992a).

The numerical value of the critical point will change according to the subject matter,
population, or particular individuals, but the formula still serves to differentiate between
people in the danger control and fear control groups (Witte, 1994b). The formula is the z-
score of the efﬁcacy questionnaire items minus the z-score for the perceived threat items.

If the obtained result from the standardized scores is positive, the individual (or group) is

-29-

in the danger control process. Efficacy is higher than perceived threat in this case,
meaning health protective behavior can be expected in response to the risk message. A
negative score indicates the respondent (or group) is following the fear control path.
Witte et al. (1996) provide speciﬁc items to determine perceptions of threat
(susceptibility and severity) and efﬁcacy (response and self).

One consistent application of the EPPM has focused on how it can be used by
healthcare practitioners to develop theoretically. guided Campaigns (Stephenson & Witte,
2001). Typically, the stated goal in these publications is to avoid ineffective health
intervention campaigns that fail to produce positive change (Witte et al., 1996) or result
in the opposite of the desired behavior (R. W. Rogers, 1983; R. W. Rogers, & Mewbom,
C. R., 1976). As Witte et al. point out, considering this dichotomy of results is a key
consideration for healthcare providers and practitioners when delivering health messages,
especially fear appeals (Witte et al., 1996).

The EPPM has been tested in a wide range of topics, including: breast cancer (Kline
& Mattson, 2000), cardiovascular disease (McKay, Berkowitz, Blumberg, & Goldberg,
2004), HIV/AIDS prevention (Witte, 1992a; Witte, Cameron, Lapinski, & Nzyuko,
1998), skin cancer (Stephenson & Witte, 2001), and teen pregnancy (Witte, 1997).
Populations used in EPPM studies have also varied along with the health topics.
Respondents have been segmented by education level (Gore & Bracken, 2005; Murray,
Hubbell, Witte, Sampson, & Monison, 1998; Witte, 1994a), ethnicity (Murray et al.,
1998), occupation (Witte et al., 1998; Witte et al., 1993), and geographic location (Witte

et al., 1998).

-30-

Data across the range of studies are consistent with the constructs in the EPPM,
according to Witte and Allen’s review (2000). As fear appeals get stronger, they result in
larger fear-control, defensive responses (r = .20) instead of cognitive, danger control
responses such as attitude (r = .14), intention (r = .11), or behavior (r = .15) change.
These fear control responses are also inversely linked (r = -. l 8) with danger control
reactions, matching the EPPM (and parallel response model) prediction that those
defensively avoiding a recommendation will not be inclined to follow a message’s
suggestions. Another inverse relationship shows up between weak efﬁcacy messages and
fear control responses (r = -.11), supporting the EPPM’s hypothesis that a weak efﬁcacy
message leads to stronger fear control responses.

RPA extension

The Risk Perception Attitude framework stems from the hypotheses of the EPPM,
predicting that efﬁcacy beliefs will inﬂuence perceived risk’s affect on action-oriented
behaviors (Rimal & Real, 2003). As Rimal and Real (2003, p. 372) point out, this idea
mirrors Bandura’s social cognitive theory in that individuals with high levels of efﬁcacy
“construe potential risks as challenges to be overcome,” as opposed to low-efﬁcacy
individuals that would tend to view potential risks through a fatalistic lens (Maibach &

(Zonon,l995)
Considering these predictions, the RPA framework breaks individuals up into four
attitudinal groups, based upon their perceived efﬁcacy and risk. The four groups are:
0 Responsive: individuals with high perceived risk and high efﬁcacy; it is

expected that they will be the most motivated group when it comes to

-31-

taking self-protective behavior, based upon their awareness of their risk
status and their belief that they have necessary skills to make change.
Avoidance: high perceived risk paired with low efﬁcacy beliefs; the
coupling of high risk and low efficacy is expected to put these individuals
in an uncomfortable situation, as their high risk renders them aware of
their health status but their low efﬁcacy beliefs remove motivation to act;
they are expected to be less motivated than the responsive group.
Proactive: characterized by low risk perceptions and high efﬁcacy beliefs;
risk perceptions do not push these individuals, but they may maintain a
positive outlook on their health status.

Indifference: deﬁned by low perceived risk and low perceived efﬁcacy;
indifferent individuals do not feel they are vulnerable to the health risk,
nor do they feel that they could respond appropriately if they were; this is

expected to be the least motivated group.

These categories serve to provide a context for audience segmentation based upon the

EPPM’S predictions, ideally allowing for more effective message design. In addition, the

EPPM helps predict message-exposure outcomes for the four RPA groups (Rimal &

Real, 2003). At its most basic, the RPA framework is another context in which the EPPM

_Kg1 variables in fear appeal researgh

Several scholars have conducted meta-analyses to parse out the plausibility of

c0I‘npeting fear-appeal models according to the provided data since each makes different

predictions concerning how fear appeals inﬂuence outcomes (Boster & Mongeau, 1984;

-32-

Mongeau, 1998; Sutton, 1982; Witte & Allen, 2000). These meta-analyses focused on
published studies that ﬁt a traditional fear appeal approach (a message intended to cause
fear and produce a particular outcome as a reaction) and reported enough information to
allow for further statistical analysis.

Results from the meta-analyses and other fear appeal research efforts are organized
below by variable, as the fear appeal models generally employ similar dependent,
independent, and moderating variables. The three most common dependent variables are
behavior, attitude, and behavioral intention, all discussed below.

Other frequent endogenous variables are the related constructs of perceived fear,
perceived threat, and perceived risk. As also discussed in the following sections, these
labels are often misused to describe the same construct, which adds a level of
complication to any examination of meta-analyses. The confounding language
surrounding the emotional reaction to a threatening message presents a particular
challenge when comparing studies.

Perceived efﬁcacy is another induced variable found in most models and analyzed
below. The last section will discuss other variables that do not play a signiﬁcant role in
the fear appeal literature but have been examined in one or more studies, mostly as

moderators.

_De2endent variables
Behavior

Behavior serves as a dependent variable in the studies where it is measured. The
intended outcome of these few projects was to alter respondents’ actions regarding

ll'lclividual behaviors such as getting a mammogram or a vaccination.

-33-

Correlations between fear manipulation and behavior vary widely among the relevant
studies. Fifteen studies in the Boster and Mongeau review (1984) report data for a fear
manipulation-behavior correlation. The correlations range between —.36 and .69 with a
weighted mean of .10. The weighted variance in the distribution of these correlations is
.04. A weighted variance of .005 is expected from chance alone. The obtained variance is
signiﬁcantly greater than that expected by chance, a ﬁnding supported by the X2 test (X2
=125.70, df= 14, p < .001).

The correlation distribution greater than the distribution expected by chance opens the
door for possible moderating variables (Boster & Mongeau, 1984). In the end, though,
similar relationships between fear manipulation and behavior change were noted by
Boster and Mongeau (r = .10) (1984), Mongeau (r = .17) (1998), Witte and Allen (r =
.15) (2000) and Sutton (1982). No Sutton correlation is reported for comparison purposes
because that review employs a different method of analysis than the others.

As the unexpectedly large variance opens up the possibility of moderating factors,
Boster and Mongeau (1984) performed a search for other artifacts that could be
inﬂuencing the fear manipulation-behavior correlation. Several of the 15 studies also
investigate other variables, but few of those projects produce statistically signiﬁcant or

substantial effects for the interactions. Imminence (Chu, 1966) and probability (Griffeth
& Rogers, 1976) of the presented message topic, for example, play roles in statistically
Signiﬁcant interactions with fear and efﬁcacy, but their effect sizes are small. Imminence
interacts with fear and efﬁcacy (1] = .11, df = 4) in a way that Chu (1966) interprets as
lTleaning that imminence matters (in drug-taking behavior, in this case) when the
perceived efﬁcacy of the suggested response is high and the aroused fear level is low.

-34-

Griffeth and Rogers’ examination (1976) of probability as it interacts with fear and
efﬁcacy also suggests that low fear conditions are more effective at persuasion (Boster &
Mongeau, 1984).

Self-esteem is another variable examined by fear appeal researchers (Ramirez &
Lasater, 1977). For low self-esteem participants, conformity to message suggestions
increases with fear levels (r = .20, df = 100, p < .01) while for high self-esteem
individuals, fear had no effect (Boster & Mongeau, 1984). Age also showed a decent-
sized, although not statistically signiﬁcant, relationship with the fear manipulation-
behavior correlation. The age of participant correlated negatively with the fear
manipulation-behavior correlation (r = -.43, df = 13, p = .06) (Boster & Mongeau, 1984).

Some study characteristics produce interesting correlations as well, even if the small
number experiments makes the results unstable (Boster & Mongeau, 1984). For example,
the fear manipulation-behavior correlation is stronger for participants that did not
volunteer than those that did (r = .37, (if = 10, p = .12). This is based a small number of
studies, and the results are not statistically signiﬁcant, suggesting that suspended

judgment may be the wisest course while other evidence accumulates (Boster &
Mongeau, 1984).
Another study design element worth considering is the correlation between number of
items measuring behavior and the fear manipulation-behavior correlation (r = .76, df =
1 3, p = .001), offering evidence that the unexpectedly large variance in fear
manipulation-behavior correlations across studies is the result of the reliability of the

behavioral measures.

-35-

 

Also worth noting is that in the studies where both correlations are available (13 out
of 15 for Boster and Mongeau’s meta-analysis), the relationship between the fear
manipulation-behavior correlation and the fear manipulation-perceived fear correlation
(the manipulation check, here) ranged widely from -.91 to .85 with a weighted mean of
.45. The obtained weighted variance of .19 is greater than the expected variance of .004
(x2 = 881.58, df= 12, p < .001) (Boster & Mongeau, 1984).

Interesting relationships among the variables and design aspects make the fear
manipulation-behavior data difﬁcult to interpret. The large variance in the correlation
distribution and the presence of some statistically signiﬁcant—although not large—
moderator effects (such as for the number of items used to measure behavior) could
indicate the existence of moderating variables (Boster & Mongeau, 1984). The small
amount of data, however, prevents any ﬁrm conclusions. Error of measurement, then,
could be playing a key role in the variance of the distribution of fear manipulation-
behavior correlations.

Attitude

Attitude in fear appeal research refers to the measurement of individuals’ evaluations
of speciﬁc behaviors. In this persuasive context, the goal is to alter the outcome of how
individuals appraise certain behaviors or change their predisposition toward some activity

(Severin & Tankard, 2001). Typically, an attitude-adjusting fear appeal seeks to associate
Some negative behavior with a negative outcome (LaTour et al., 1996), such as smoking
and lung cancer or not brushing one’s teeth and cavities (Mayﬁeld, 2006).

The fear manipulation-attitude correlation resembles the other fear manipulation-

dependent variable relationships in that it has a wide range: -.25 to .63 with a weighted

-36—

mean of .21 for Boster and Mongeau (1984). a mean of .20 for Mongeau (1998), and a
mean of .14 in Witte and Allen (2000). Boster and Mongeau (1984) point out in their
meta-analysis that the weighted variance in the distribution is larger than expected by
sampling error alone (.03 to .01 in this case). The fact that the variance in the distribution
of correlations is larger than expected by chance means it is possible that other variables
moderate the fear manipulation-attitude connection.

Similar to the fear manipulation-behavior correlations in the prior section, the size of
the fear manipulation-perceived fear induction-check correlation is correlated with the
size of the fear manipulation-attitude correlation (r = .41) (Boster & Mongeau, 1984).
The data show that as the fear manipulation-perceived fear correlation rose so did the fear
manipulation-attitude correlation. Using the data from the 17 relevant studies with
measurement of both fear manipulation-perceived fear correlation and the fear
manipulation-attitude correlation, Boster and Mongeau (1984) looked deeper into the
relationship between the two correlations. They divided the fear manipulation-attitude
correlation by the fear manipulation-perceived fear correlation. The ratio between the two
correlations varied widely in the studies from -.61 to 1.57 with a weighted mean of .57.
The distribution had a weighted variance of .24, signiﬁcantly larger than the .004
eXpected by chance.‘5

This action produced two things: a measure of the strength of the fear manipulation-

attitude correlation when corrected for the fear manipulation strength and an estimate of
the correlation between perceived fear and attitude, assuming perceived fear mediates the
relationship between the fear manipulation and attitude. This link stresses the need for

\
s
The chi-square test supports that the two variances are signiﬁcantly different than expected by chance (X2

\

\ 1 354, df=16, p < .001)
-37-

strong inductions, as the application of the fear appeal manipulation could restrict some
of the variance in the fear manipulation—attitude correlations (Boster & Mongeau, 1984).

Along with relationships among the variables, experimental artifacts could also
inﬂuence perceived fear-attitude correlations (Boster & Mongeau, 1984). Differences in
the reliability of measurement of perceived fear and attitude, for example, could attenuate
correlations differentially, altering the variance of the correlations.

Boster and Mongeau (1984) cite some indirect evidence to support the idea that
differential reliability increases the variance in the distribution of the fear manipulation-
attitude correlation. The experimental design is correlated with the size of the fear
manipulation-attitude correlation (r = -.20, df = 23, p = .17). While not statistically
signiﬁcant at the .05 level, this relationship does point to a trend for larger correlations in
a posttest-only design when compared to a pretest-posttest experiment.7

In addition, the number of items used to measure attitude and the fear manipulation-
attitude correlation are positively related (r = .36, df = 21, p = .05 in Boster and
Mongeau, 1984). Thus, the more items used to measure attitude, the stronger the fear
manipulation-attitude correlation. Looking at the same issues through another method,
Boster and Mongeau (1984) also found standardized regression coefﬁcients even larger
than the zero-order correlations when the fear manipulation-attitude relationship was

1' egressed onto type of experiment design and number of items in the measurement

instrument ([3 = -.24 for type of experiment and B: .39 for number of items, R = .43).

\

Boster and Mongeau are careful to state in a footnote of their own that this correlation between
e’lilberiment type and fear manipulation-attitude was computed using the absolute value of the fear
manipulation-attitude correlation as the dependent variable. They contend that it’s the strength of r and not

sign that matters theoretically.

-38-

Another methodological artifact that could affect the fear manipulation-attitude
correlations is restriction in range (Boster & Mongeau, 1984). Studies with more levels of
fear manipulation reported larger fear manipulation-attitude correlations (r = .31, df = 23,
p = .07). Having increased variance in the independent variable reduced the amount of
attenuation due to restriction in range for the fear manipulation—attitude correlation. It
seems that using several levels of fear tends to result in larger fear manipulation-attitude
correlations than studies using only a fewlevels of fear manipulation.

Calculating a multiple correlation for the methodological artifacts with the fear
manipulation-attitude correlation can estimate how much of the variance in the fear
manipulation-attitude correlation is due to type of design, number of items used in
measurement, and number of levels of the independent variable (Boster & Mongeau,
1984). The resulting R equals .50, indicating that artifacts drive an extensive amount of
the difference in the fear manipulation-attitude correlation distribution.

Behavioral intention

Behavioral intentions present an interesting case as a dependent variable because they
may be more common than researchers realize (Witte & Allen, 2000). Early work used a
variety of measures to document attitudes, and some may really have measured
behavioral intentions instead of attitudes. Studies claiming to look at fear appeal-attitude

relationships may in fact be measuring fear appeal-behavioral intention relationships.

Generally agreed-upon deﬁnitions used in more current message-design research

during the last decade and a half have alleviated some of this confusion. Attitudes are
now deﬁned as evaluations of speciﬁc behaviors, and intentions are a person’s aim to

complete a certain behavior (Witte & Allen, 2000).

-39-

One potential result of more consistent deﬁnitions noted by Witte and Allen (2000) in
their meta-analysis is that the fear manipulation’s inﬂuence on the common dependent
variables leveled out over time. Only the Witte and Allen (2000) meta-analysis looked at
the fear manipulation-behavioral intention correlation, reporting a result of .11, within the
range of the other danger control, action—oriented responses. Across the literature, as
noted earlier, the correlations for fear manipulations with attitudes, intentions, and
behaviors were relatively consistent at r = .14 for attitudes, r = .11 for intentions, and r =
. 15 for behaviors.

Sutton (1982) also found signiﬁcant positive effects for strength of fear appeal on
behavioral intentions using z-scores as a meta-analytic tool, but his difference in method
prevents mathematical comparison to the other reviews. Like for behavior and attitude,
Stronger fear appeals seem to lead to more change in intentions (Witte & Allen, 2000).
The same could be said for stronger severity, susceptibility, response efﬁcacy, and self-

efﬁcacy messages (Witte & Allen, 2000).

ggmmgn endogenous variables

Three variables play the most consistent roles within fear appeal models. Two of
them—perceived threat and perceived fear—are related constructs linked with reactions
to the fear manipulation in fear appeal models. Efﬁcacy is another often-applied variable
usually manipulated within the message content.

Perceived threat

Perceived threat is a cognitive construct (Witte, 1998) that entered fear appeal
research around 1975 in the work of Rogers (1975). It consists of two dimensions,

perceived susceptibility and perceived severity (Witte, 1998). Perceived susceptibility is

-40-

the degree to which one feels at. risk for experiencing a threat while perceived severity is
the magnitude of harm expected from the threat at hand.

In the fear appeal literature, threat is a stimulus variable deﬁned as a message
focusing on the severity (“Lung cancer leads to death”) and susceptibility (“You’re at
risk for lung cancer because you smoke”) of a threat (R.W. Rogers, 1975; R. W. Rogers,
1 983). While the threat may exist whether a respondent is aware or not, measurement
does depend upon the person holding cognition that the threat exists, meaning that person
is perceiving the threat (Witte, 1992b).

Perceived severity and susceptibility matter because evidence indicates that they are
the cognitive channels mediating how threat messages inﬂuence acceptance of message

recommendations (Beck & Frankel, 1981). The emotional states associated with threat
messages are not long-lasting and lack substantial direct effects on future decisions,
instead depending upon cognitive factors to have any inﬂuence (Mewbom & Rogers,
1979; R. W. Rogers, & Mewbom, C. R., 1976).

Perceived fear

Fear is an affective construct referring to a negative emotional response to a threat
that expresses some danger (Tanner, Hunt, & Eppright, 1991). It is joined with a high
level of physical arousal resulting from a relevant, noteworthy threat (Vrana, Cuthbert, &
Lang, 1986) and can cause feelings of uncertainty and risk-averse decision-making
(Jennifer Lerner, Gonzalez, Small, & Fischhoff, 2003; Jennifer Lerner & Keltner, 2001).

As Witte (1992b) summarizes, fear may be communicated in a variety of ways:
physiologically through arousal (Mewbom & Rogers, 1979), by verbal self-reports (R.W.
Rogers, 1975), or through physical acts such as facial expressions (Marsh, Adams, &

-41-

Kleck, 2005). Verbal self-reports show a particularly useful ability to measure fear.
according to Rogers’ (1983) work linking physiological displays of fear and self-ratings
by respondents. Physiological arousal varies during exposure to a fear message, but a
self-reported emotional state is at least slightly more enduring.

While fear and threat are separate constructs, the two are strongly linked (Witte,
1992b, 1998), occasionally leading to a confused relationship among researchers (Witte
& Allen, 2000). some publications, including several meta-analyses (Boster & Mongeau,
1 984; Sutton, 1982), treat fear appeals, fear and threats interchangeably (LaTour &
Rotﬁeld, 1997). Such usage can obviously pose problems for analysis since the cognitive
threat and affective fear do not necessarily match and certainly are not equal for every
respondent. (Rotfeld, 1988). As a result of this potential mismatch, more research would

be well served to examine fear and not only look at threat’s correlation with the
dependent variable, as doing so ignores individual-level emotional reactions to the
message (LaTour & Rotﬁeld, 1997).

The confusion in terms also presents challenges for interpreting results. Any study
where threat perception is actually being manipulated instead of fear would expect a
reduced manipulation-perceived fear correlation. Fear measurement can serve as a check
on messages’ threat status since fear is assumed to follow a threat as an emotional
reaction (Witte & Allen, 2000). However, any measured fear reaction in this case would
be the side effect of a presented cognitive threat, effectively adding another step to the
model being tested. Because threat is the focus of the induction, for example, one would
expect a larger threat manipulation-perceived threat correlation than a fear manipulation-
perceived fear relationship.

-42-

In most models, the “perceived fear” variable measures the correlation between the
message manipulation and respondents’ reported perceived fear, often serving as an
induction check. Results for the manipulation-perceived fear correlation range widely
across studies from .17 to .81 with a weighted mean of r = .36 in Boster and Mongeau
(1984) and r = .34 in Mongeau’s meta-analysis (1998). The results vary widely for a
number of reasons, including experiment design, respondent groups, message design,
restriction in range, and length of the measurement instrument.

As Boster and Mongeau (1984) point out, roughly one-third of the studies use a
single-item measure of perceived fear, making it likely that these studies underestimate
the correlation between message manipulation and perceived fear. If all the studies had
been able to correct for attenuation resulting from error of measurement, it is entirely

possible the distribution variance of message manipulation-perceived fear correlations
would be quite reduced.

Another issue is that despite the volume of studies, several meta-analyses show that
researchers have not created particularly strong fear manipulations. The mean fear
manipulations in the Boster and Mongeau (1984) (r = .36), Mongeau (1998) (r = .35) and
Witte and Allen (2000) (r = .30) pieces demonstrate these mediocre relationships between
the intended manipulation and the actual effect on participants. The mean manipulation-
perceived fear correlations above make it difﬁcult for the experimental manipulation to
strongly impact outcome variables, according to most theories. Since the theories use

perceived fear as a moderator between the manipulation and the dependent variables, a
not-strong fear appeal message prevents the manipulation from having a strong impact on
Outcomes (Boster & Mongeau, 1984).

-43-

As Boster and Mongeau (1984) explain in the most conservative example possible, if
the perceived fear-attitude correlation is perfect (r = 1.00) and the fear manipulation-
perceived fear correlation is the mean (r = .36), then the fear manipulation-attitude link
cannot exceed r = .36 except by sampling error. This is not statistically signiﬁcant at the
.05 level in any study with fewer than 25 subjects. Of course, other variables inﬂuence
attitude, making the idea of a perfect perceived fear-attitude correlation impossible.
Using a more realistic—but still subStantial—correlation for the example above, the
results are even more vivid.

With a perceived fear-attitude correlation of r = .50, the fear manipulation-attitude
correlation is limited to r = .17 plus sampling error (Boster & Mongeau, 1984). This is
roughly the mean fear manipulation-attitude correlation from the Boster and Mongeau
meta-analysis. Also, this correlation would be statistically signiﬁcant at the .05 level with
more than 92 participants, a standard not met in a number of fear appeal studies.

More recent fear appeal research, however, has produced more consistent results both
for fear inductions and dependent variables (Witte & Allen, 2000). One cause for this
could be improved deﬁnitions for outcome variables such as behavioral intentions, as
discussed earlier (2000). Potentially, the result of these newer deﬁnitions noted by Witte
and Allen (2000) is that fear manipulation’s inﬂuence on the key dependent variables has
leveled out, as the correlations with attitudes, intentions, and behaviors have become
relatively consistent ( r = .14 for attitudes, r = .11 for intentions, and r = .15 for

behaviors).

-44-

Efficacy

Like perceived threat, perceived efﬁcacy entered the fear appeal literature in the work
of Rogers (1975), and like perceived threat, it is measured across two dimensions:
perceived self-efﬁcacy and perceived response efﬁcacy (Witte, 1998). Self-efﬁcacy
generally is deﬁned as people’s beliefs about their capabilities to exert behaviors in a way
that inﬂuences other events in their lives (Bandura, 1994). In the fear appeal context,
perceived self-efﬁcacy is one’s beliefs about his or her ability to perform the
recommendation contained in the fear appeal message. Response efﬁcacy refers to one’s
belief about the effectiveness of the recommended response in avoiding the threat. This
includes how a person assesses the beneﬁts and costs of a recommended action (Maiman
& Becker, 1974)

When manipulated in a manner similar to threat appeals (multiple conditions in an
experiment), efﬁcacy messages produce decently large effects (Witte & Allen, 2000). In
one meta-analysis, response efﬁcacy and self-efﬁcacy produced similar correlations on
the common dependent variables (between .12 and .17 for response efﬁcacy and self-
efﬁcacy with attitude, intention, behavior change) (Witte & Allen, 2000, see p. 599 for
the individual correlations).

Efﬁcacy, however, was at one time not a commonly analyzed variable (Witte &
Allen, 2000). According to Witte, a number of researchers failed to examine its role in
persuasive fear appeals (for example, see Ben-Sira, 1981; Burnett, 1981; Burnett &
Oliver, 1979; Kohn et al., 1982; Ramirez & Lasater, 1977).

As discussed at several points, prior research has shown that perceived efficacy plays

a role in predicting adherence to the recommended action being promoted in the message

-45-

condition. It is reasonable to expect that the same relationship will appear here as well,
thus:
H1: Higher perceived efﬁcacy will associate positively with higher behavioral

intention to take self-protective action.

Additional variables in fear appeal research

Other variables have regularly entered fear appeal research in various attempts to
improve the explanatory and predictive power of models, the idea being that moderators
may inﬂuence outcomes in some way. For example, the manipulation-perceived fear
relationship may be affected by some moderator before that relationship then inﬂuences
the dependent variable; similarly, moderators could directly impact perceived fear before
it interacts with other moderators and then outcomes (Boster & Mongeau, 1984; Witte &
Allen, 2000). Despite these possibilities, some researchers do still argue that individual
differences rarely interact with fear appeal strength to inﬂuence outcomes (Witte &
Allen, 2000). However, studies on individual difference variables have been too limited
to draw ﬁrm conclusions, leading some to call for more research on personal
characteristics and their inﬂuence on fear appeals (Ruiter, Verplanken, De Cremer, &
Kok, 2004). Individual difference variables that have received prior consideration are
discussed below:

Source credibility

Source credibility is examined in three studies discussed by Boster and Mongeau
(1984). Two of the studies found a statistically signiﬁcant fear by source credibility
interaction (McCroskey & Wright, 1971; Powell & Miller, 1967) while the third presents

a small, non—signiﬁcant interaction of r = .04 (Hewgill & Miller, 1965). While the two

-46-

studies reporting signiﬁcant effects did ﬁnd non-trivial interactions, it is plausible that
results are based on unusual elements of the experimental designs (Boster & Mongeau,
1984)

The Powell and Miller (1967) study reports a large effect size (n = .36, df = 4), but
most of the effect is due to one treatment condition (Boster & Mongeau, 1984). The study
uses three credibility treatments: low credibility, high credibility, and an unattributed
source with measurement based primarily on trustworthiness. Perceived competence is
not manipulated to any great degree nor are ratings for other possible variables such as
dynamism reported. The interaction effect for fear and source credibility depends mostly
upon the unattributed source condition. If it is removed, the fear by source credibility
interaction drops precipitously to n = .10 (2 df) and lacks statistical signiﬁcance.

Boster and Mongeau (1984) argue that little theory available at the time would
support the existence of a fear by source credibility interaction. Powell and Miller (1967)
hypothesize that such an interaction will be found, but they fail to provide a rationale. A
problem in their analysis is that they failed to perform a credibility manipulation check on
the unattributed source condition. The authors assume the unattributed source will fall
between the low and high credibility conditions, but there is no way to know to what
degree the respondents attributed some level of credibility to the unattributed source.
Boster and Mongeau’s examination of the fear manipulation and attitude data shows that
it is likely that the respondents considered the unattributed source to be highly credible.
The high credibility and the unattributed source conditions show similar patterns for the

manipulation check and attitude data.

-47-

McCroskey and Wright’s (1971) experiment reported an interesting ﬁnding
concerning its control group that motivated Boster and Mongeau to write that the
reported fear by source credibility interaction could be based on issues with the execution
of the experiment. In this study, two levels of source credibility were used to get an effect
size of n = .12 (2 (if), most of which is based on variance with competence.
Trustworthiness is little affected by the source credibility manipulation, and dynamism is
not inﬂuenced at all, resulting in an effect size smaller than in Powell and Miller’s (1967)
work.

The unusual portion comes in when one examines the data concerning the control
group. The high credibility source produces more favorable attitudes in both the high fear
and low fear treatment groups. The control group reports a small, statistically signiﬁcant
difference in the opposite direction. Without the control group data, no interaction is
found (Boster & Mongeau, 1984).

This lack of interaction is the case for all source credibility experiments, according to
Boster and Mongeau’s meta-analysis (1984). All three projects manipulate source
credibility and fear on two levels, but when the data are analyzed according to the design,
no substantial evidence of fear by source credibility interactions are present (Hewgill &
Miller: r = .04; McCroskey & Wright: r = .00; Powell & Miller: 1] = .10, 2 df). While
some small effects may exist, no strong evidence supports the idea that source credibility
is a powerful moderator of the fear manipulation-attitude relationship.

Specificity of instructions

Another moderator variable examined in several studies is the speciﬁcity of the

instructions offered in the original message (Boster & Mongeau, 1984). Leventhal and

-43-

several co-authors examined speciﬁcity of instructions for the recommended action in
three studies, with two of the studies reporting statistically signiﬁcant interactions and
one not.

The ﬁrst study (Leventhal et al., 1965) reported that as fear increased, conformity to
message recommendations also increased. Fear, speciﬁcity of instructions, and whether
or not one had a prior tetanus shot interacted to affect attitudes about receiving a tetanus
shot. For people with a prior shot, message speciﬁcations had no effect on fear levels.
Those reporting higher levels of fear, however, did follow the recommendations more
closely when the message did not have speciﬁc details. For people without a prior shot,
the fear effect was stronger when the message included speciﬁc recommendations. In
both cases, increased fear lead to increased conformity with the recommendations. The
three-way interaction’s effect size is r = .16 (Boster & Mongeau, 1984). Fear by
speciﬁcity, however, produces an interaction that is not statistically signiﬁcant or
substantial (r = .02).

The next article by Leventhal et a1. (1966) also concerned tetanus shots and claimed
that fear and speciﬁcity of instructions interacted to affect opinions on the subject. Their
data showed that an increase in fear coincided with compliance with message
recommendations when the suggestions were speciﬁc (r = .14) (Boster & Mongeau,
1984). No evidence of the three-way interaction reported in Leventhal’s earlier piece
(Leventhal et al., 1965) is present as the result is neither statistically signiﬁcant nor
substantial (r = .03).

Including a manipulation for instructions along with the fear manipulation made little
difference in a third study (Leventhal, Watts, & Pagnao, 1967). Examining these

-49-

variables and others in relation to attitudes on smoking and lung cancer, the researchers
most likely did not ﬁnd signiﬁcant effects for a fear by speciﬁcity of instructions
interaction (Boster & Mongeau, 1984). The article actually does not provide enough data
to allow estimation of these interaction effects, but the authors report statistically
signiﬁcant effects for other interactions. It is then reasonable to conclude that the authors
would have reported a fear by speciﬁcity effect if one were found. According to
calculations done by Boster and Mongeau (1984), the effect size for this interaction
cannot have exceeded r = .16 without being statistically signiﬁcant at the .05 level. One
can then assume that the correlation was below this .16 level or Leventhal et al. would
have noted such in their piece.

This makes for uncertain knowledge about the role of how speciﬁc instructions and
fear interact to shift attitudes (Boster & Mongeau, 1984). One study (Leventhal et al.,
1965) noted a fear by speciﬁcity by prior experience (tetanus shot) three-way interaction
without observing a fear by speciﬁcity interaction. Another (Leventhal et al., 1966) lists a
fear by speciﬁcity interaction but not a fear by speciﬁcity by prior shots interaction while
yet a third (Leventhal et al., 1967) failed to ﬁnd a fear by speciﬁcity interaction. Based on
this small sample of data, there is not evidence to support the hypothesis that fear and
speciﬁcity of instruction interact to affect attitudes nor is there evidence to support
Leventhal’s idea that a fear effect only occurs when recommendations are speciﬁc
(Boster & Mongeau, 1984). More data would permit a meta-analysis to determine
whether the shifting results of these studies stem from sampling error, other moderators,

or some other explanation.

-50-

Demographic, sociopsychological, and health attitudes

Demographic, sociopsychological and health attitude variables have also been
examined by fear manipulation researchers (Burnett & Oliver, 1979). A cluster analysis
of these three types of variables was used to explain the relationship between fear and
attitude toward health maintenance organizations (Boster & Mongeau, 1984).

Data concerning two speciﬁc groups showed them to be more receptive to fear
appeals than others. For the groups labeled “older liberals” and “older blue-collar
blacks,” an increase in fear led to an attitude toward HMOs that matched more closely the
attitude suggested by the message. However, the complexity of the data involved in the
cluster analysis can limit opportunities to clearly determine which variables interact with-
fear to inﬂuence attitude, although the results suggest that age plays a role (Boster &
Mongeau, 1984).

Anxiety

Anxiety proved to have a consistent place in interacting with fear to inﬂuence
attitude, although none of the four studies listed substantial effect sizes (r = .12, not
statistically signiﬁcant) in the Boster and Mongeau meta-analysis (1984). Anxiety in this
context is deﬁned as a person’s characteristic level of anxiety when it comes to personal
threats (Witte & Allen, 2000). Witte and Allen (2000) also cite trait anxiety as a variable
listed in the literature as having potential to be theoretically important in the persuasive
effect of fear appeals. The authors also refer readers to a review by Witte and Morrison
(2000) that discusses the various constructs that could represent trait anxiety.

Across the four experiments discussed in Boster and Mongeau, low-anxiety

participants showed no correlation—or at best a small positive correlation—between fear

-51-

manipulation and attitude. High-anxiety people also showed no signiﬁcant correlation
between fear manipulation and attitude, but some did report a small negative correlation.
Witte and Allen (2000) point out that this has been hypothesized as meaning that an
individual’s characteristic level of anxiety can inﬂuence one’s reception to fear appeals——
a characteristically scared and anxious person may negatively react to a fear appeal. As
such, Witte and Allen examined the direct and moderating effects of trait anxiety on
persuasive outcomes. The results pointed to trait anxiety not having a signiﬁcant effect.
Witte and Allen (2000) thus conclude that response to fear appeals is not affected by
anxiety level.

Boster and Mongeau (1984) noted that methodological artifacts (which they describe
in detail) could be responsible for reducing the effect sizes of anxiety; the effects were
consistent across all four projects examined, making the possibility worth consideration.
While the evidence is limited, Boster and Mongeau (1984) indicate based on the data
available at the time that their impression is that fear and anxiety interact to inﬂuence
attitude change. The effects reported by the relevant studies they discuss are small, but
methodological artifacts could play a role in limiting those correlations.

Volunteer] nonvolunteer

A variable with more substantial support for its inﬂuence on attitude is a respondent’s
status as a volunteer or non-volunteer for the experiment, according to a few studies
discussed by Boster and Mongeau (1984). Speciﬁcally, three studies looking at attitude
change report fear by volunteer interaction effects of r = .25 (Horowitz, 1969), r = .15
(Horowitz & Gumenik, 1970), n == .23 (4 df) (Horowitz, 1972). This ﬁts with the ﬁndings
mentioned earlier where the fear manipulation-behavior correlation was stronger for

-52-

participants that did not volunteer than those that did, although it was not statistically
signiﬁcant (r = .37, df= 10, p = .12).

Horowitz’s 1969 article concerns an experiment looking at fear, the participant’s
volunteer status, and how often the person saw the persuasive message (Boster &
Mongeau, 1984). For those who volunteered, increased fear led to more conformity to the
message recommendations (r = .46, df = 58, p < .05) while non-volunteers had the
opposite reaction (r = -.21, df = 58, p > .05), even though it was not statistically
signiﬁcant.

Volunteer status also made a difference in another study looking at the interaction
between fear and conformity (Horowitz & Gumenik, 1970). Increased fear brought
volunteer attitudes in both the high and low fear conditions closer to what was advocated
in the persuasive message (r = .33, df = 58, p < .05) but had no effect on the attitude of
nonvolunteers (r = -.02, df = 58, p > .05). The difference between the two groups was
even more prominent in the high fear condition (r = -.52, df = 58, p < .05 for the high fear
condition versus r = -.22, df = 58, p >.05 for the low fear conditions). In addition,
participants allowed to choose their experiment conformed more closely to
recommendations than those not given a choice, leading to a statistically signiﬁcant fear
by volunteer interaction and a large effect for fear by volunteer by choice of experiment
(r = .34) (Boster & Mongeau, 1984).

The third experiment in this series (Horowitz, 1972) used a different fear appeal than
the previous two projects, but the results support prior ﬁndings, nonetheless (Boster &
Mongeau, 1984). The prior two studies (Horowitz, 1969; Horowitz & Gumenik, 1970)
used pamphlets and ﬁlms with differing content. Here, fear is introduced through false

-53-

heart rate feedback, and there are ﬁve levels of fear content. Again, volunteers abide by
the message recommendations more than nonvolunteers with high fear producing more
effect than low fear (Boster & Mongeau, 1984). This time, for volunteers and
nonvolunteers, fear pushes the level of adherence to message suggestions (r = .48, df =
58, p < .05 for volunteers and r = .20, df = 58, p > .05 for nonvolunteers).

Taken as whole, Horowitz’s studies support volunteer status as a strong moderator of
the fear-attitude correlation (Boster & Mongeau, 1984). For nonvolunteers, the data are
inconsistent, leading to the conclusion that fear does not affect their conformity to the
message suggestions. However, for volunteers, the effect of fear is a strong one (r = .42
across the three studies) (Boster & Mongeau, 1984).

As part of their literature review, Boster and Mongeau (1984) pulled out data from
the 12 studies that clearly noted the volunteer status of the respondents and the eight that
indicated the nonvolunteer status. Both of these groups include the Horowitz studies.
Across the 20 studies, volunteers have a much higher fear manipulation-attitude
correlation than nonvolunteers. The weighted mean fear manipulation-attitude correlation
for volunteers is .32 while for nonvolunteers it is .07. The correlation between volunteer
status and the fear manipulation-attitude correlation also provides support for Horowitz’s
hypothesis (r = -.55, df = 18, p < .05).

Topic relevance

One study also looked at subject matter relevance and its ability to inﬂuence attitude
change (Berkowitz & Cottingham, 1960). Fear appeal researchers have long surmised
that a personal connection or interest in a topic should be connected to responses to

persuasive messages, particularly for broadly delivered mass media messages (Kraus. El-

-54-

Assal, & De Fleur, 1966). Data from the pair of experiments published in the article by
Berkowitz and Cottingham supported the idea that high-relevance, high-fear messages
are connected to defensive avoidance. High-fear messages produced stronger reactions
and received more attention from respondents, but if those respondents felt the topic was
relevant, they were less susceptible to persuasion, according to Berkowitz and
Cottingham’s projects. Based on these results, the two researchers (Berkowitz &
COttingham, 1960) reported that a high-fear, interesting, low-relevance communication
would be the most effective message design.

Such a conclusion, though, does not account for several measurement, design, and
conceptualization issues that subsequent research has shown can inﬂuence outcomes
from fear appeals (Boster & Mongeau, 1984). Speciﬁcally, the ages of the respondents
(college students), the measurement of the fear induction (one item), and respondent
efﬁcacy (not measured at all) are not sufﬁciently addressed, and all pose problems for the
data interpretation.

In addition, the experimenters’ categorizations of respondents present limitations as
well. The ﬁrst study divides respondents into high relevance and low relevance groups
with a single-item, dichotomous measure asking how often respondents used a car
(several times a week versus once a week or less). Experiment number two broadened the
measurement some, accounting for car ownership and frequency of use. High relevance
was deﬁned as owning a car and using it several times a week. Medium relevance either
owned a car and drove less frequently or did not own a car but rode in one several times a
week. The low relevance group did not own a car and only was a passenger once a week

or on weekends.

-55-

lb"

L~"

~.~

m.

Aside from the variable measurement and experiment design concerns, Berkowitz
and Cottingham’s conceptualization of relevance merits mentioning as well. Their article
failed to specify what is meant by the term “relevance.” The project’s induction was
dependent upon an appeal focused on gruesome car-accident images and statistical data
in the high fear condition. Their deﬁnition of relevance, however, asked only about
automobile usage; the measurement of relevance was incomplete.

Both experiments reported here focus on usage more than the variety of elements that
can comprise actual relevance. Factors apart from the employment of a behavior or
technology can inﬂuence relevance, even though other studies have copied this “usage”
deﬁnition of relevance when it comes to smoking (Insko, Arkoff, & Insko, 1965;
Leventhal & Watts, 1966). A fear appeal concerning smoking and lung cancer may be
relevant because one is a smoker (usage of cigarettes). However, the same message about
the possibility of lung cancer could also be relevant to nonsmokers with family members
who smoke, someone contemplating starting smoking (L. Block, 2005), cigarette-
manufacturer employees, public health ofﬁcials, along with any number of people whose
jobs, family relationships, or personal beliefs somehow relate to smoking.

A more effective way of measuring the relevance of a topic to a respondent would be
to examine what communication researchers call “involvement,” as its scope goes well
beyond the “usage” deﬁnition common to “relevance” research into what is practically
relevant to respondents. Involvement is generally deﬁned as “a motivational state induced
by an association between an activated attitude and the self-concept” that can be

Provoked by different dimensions of the individual (Blair Johnson & Eagly, 1990, p.

-56-

290). “Self-concept” in this case means the “totality of the individual’s thoughts and
feelings having reference to himself as an object” (Rosenberg, 1979, p. 7).

The “involvement” construct can be more effective in measuring relevance because
the literature differentiates between varied types of relationships among individuals and
speciﬁc content (Blair Johnson & Eagly, 1989), providing a more nuanced and effective
means of understanding how messages or media inﬂuence individuals’ behavior. Value-
relevant involvement, for example, refers to a person’s self-identity and acts as a barrier
to persuasion since the relevance of the topic is tied closely to the individual’s continuing
and permanent feelings about himself (Sherif & Cantril, 1947). Smoking behaviors in the
relevance literature may fall into this category, as cigarette company research into
smokers’ motivations reveals that most smokers are “much more alike than different
insofar as their motivation to smoke is concerned” (Philip Morris Co., 1969, p. 1).
According to Philip Morris’ research, smokers start the habit to meet a psychosocial goal,
as a “symbolic declaration of personal identity” (Philip Morris Co., 1969, p. 3). They
want to tell the world, “This is the kind of person I am,” “I am tough,” “I am an
adventuress,” or some other variant of this personal expression theme (Philip Morris Co.,
1969, p. 3). The unpleasantness of the ﬁrst cigarette is overcome by the novitiate’s desire
to express some value of personal independence. Researcher into smokers’ action should
measure their behavior as such. Certainly, the long—time smoker adds other motivations
such as addiction, but simply measuring usage fails to account for that as well.

Early research into the relevance of seatbelts falls into a similar trap. Measuring
usage exclusively leaves out additional elements that would make a discussion of

seatbelts relevant. Another aspect of involvement could ﬁll in here, as “impression-

-57-

relevant involvement” may have made a difference in Berkowitz and Cottingham’s
experiments on college students (Berkowitz & Cottingham, 1960). Impression-relevant
involvement centers on concern with public perception of the self, meaning the focus is
on impressions resulting from communication about a topic instead of the topic itself
(Cho & Boster, 2005; Zimbardo, 1960).

Youth are known to be particularly susceptible to impression-based inﬂuence
(Hanson & Engs, 1992; Neuwirth & F rederick, 2004; Wood, Read, Palfai, & Stevenson,
2001), and seatbelts in the 1960s were certainly not the normal behavior. Seat belts were
not even an option in most automobiles until only a few years before Berkowitz and
Cottingham’s work (National Highway Trafﬁc Safety Administration, 1997), and the
modern, three-point seatbelt was not introduced and patented until years after (Bohlin,
1962). As late as the 1983, only 14 percent of Americans wore seat belts, and even a
decade after that, a 1995 survey found that teens felt that their peers did not prioritize seat
belt use, putting it among the least important preventative health behaviors (Evans,
Gilpin, Farkas, Shenassa, & Pierce, 1995). Considering the peer environment surrounding
seat belts, any study of youth and driving safety would be well served to consider
impression-relevant involvement to understand how norms and peer pressure inﬂuence
behavior.

Along similar lines, Berkowitz and Cottingham’s original motivation for examining
relevance still merits consideration, particularly with persuasive messages in health
communication. Speciﬁcally, they proposed that a fear appeal could serve to gain
respondents’ attention, and attentiveness would act as a necessary but not sufﬁcient

condition for opinion change “either because low interest produces inattentiveness or

-58-

because the frustrated audience, venting its aggression, disparagingly dismisses the entire
communication” (Berkowitz & Cottingham, 1960, p. 37). Relevant topics, especially. can
be expected to hold respondent attention. Berkowitz and Cottingham predicted that
relevance would lead to message rejection, but more recent research actually ﬁnds the
persuasiveness of fear appeals to be greatest for those who perceive the message to be
personally relevant (L. Block & Williams, 2002).

In the health arena, outcome-relevant involvement may serve well to measure the
relationship between the relevance of health topics and persuasive outcomes. This
variable represents a psychological state occurring when some stimulus activates attitudes
that are concerned with signiﬁcant goals or results. Subjects reporting high outcome-
relevant involvement tend to be persuaded by strong arguments when compared to low
involvement respondents (Blair Johnson & Eagly, 1989), indicating a motivation to
undertake message-focused thinking (Chaiken, 1980). This type of involvement can then
promote attitude change by using strong, logical, germane arguments, or reduce attitude
change by presenting weak arguments and non-issue—related cues (Cho & Boster, 2005).
Also, outcome-relevant involvement related to future outcomes likely associates with
information-seeking behaviors such as paying attention to media coverage of a topic and
being motivated to stay informed (Cho & Boster, 2005).

Thus, it is predicted that outcome-relevant involvement will help predict the
relationship between perceived threat and action-oriented outcomes by focusing on how
easily a respondent looks to consequences of a particular issue and responds to a logical

argument. The relationships between attention to logical arguments, the cognitive

-59-

processing of those arguments, and the increased likelihood of successful persuasion
leads to the following hypotheses:

H2: Higher perceived threat will associate positively with higher outcome-relevant
involvement.

H3: Higher outcome-relevant involvement will associate positively with higher
behavioral intention to undertake recommended health-protective behaviors.

Sensation seeking

Sensation seeking is another individual-level variable that has been examined in only
a small number of fear appeal studies. Witte and Morrison (1995) completed one such
study in an attempt to more effectively target high school students with safe-sex
messages. The increasing number of AIDS cases and sexual partners among high school
students drove the researchers to study the potential impact of ads on condom use and
abstinence. One reason for these growing ﬁgures, they reported, was the pervasiveness of
risk-takers out to experience a range of sensations (Witte & Morrison, 1995). Such
sensation seeking is a character trait concerning the seeking of novel, complex, intense
experiences and often accompanied by a willingness to take risks to arrive at such events
(Zuckerman, 1991 , 1994). Risky sexual behavior among young adults falls into the
pattern of youths seeking invigorating and exciting situations.

As a result of this need for novel stimulation, sensation seeking has been shown to
indicate susceptibility to substance abuse and other risky behaviors (Donohew,
Palmgreen, & Pugzles Lorch, 1994; Donohew etal., 2000). The implications for message
design are obvious as these high sensation seekers would be inclined to attend to a

different communication style. Messages that fail to elicit high levels of arousal are

-60-

l .
'1’?
Im‘

‘9‘.

15“

rd

 

t. .
.qA‘
W1:

‘I

labeled “boring” and fail to get attention (Donohew et al., 2000). At the other end of the
spectrum, low sensation seekers are averse to intense, complex stimuli (Donohew et al.,
2000).

In their examination of this trait, Witte and Morrison (1995) found that sensation
seeking was a factor in how the high school students reacted to fear appeal messages,
leading them to suggest that safer sex messages targeted toward adolescent populations
should account for sensation seeking. However, the experiments did not work out as the
researchers expected, potentially in part because the preferred level of arousal for high
sensation seekers makes them difﬁcult to threaten. In addition, the high fear messages
concerning risky sexual practices may have come across as glamorous since youth tend to
believe that they’ll always be on the favorable end of any high-risk behavior (Jessor,
1982, 1984). Considering sensation seeking’s ability to limit the amount of fear induced
by a threat message, it is reasonable to expect that high sensation seekers will perceive
less fear than those lower in sensation seeking. Thus:

H4: Sensation seeking will moderate the relationship between message condition
and threat perception. Speciﬁcally, those high in sensation seeking will report less
perceived threat than those low in sensation seeking.

Need for cognition

Need for cognition is another variable that has received limited attention in the fear
appeal literature, despite some indication that it inﬂuences the processing of fear appeals
(Stout & Sego, 1994). Need for cognition is an individual-level variable measuring “the
tendency for an individual to engage in and enjoy thinking” (Cacioppo & Petty, 1982, p.
116). Individuals high in NFC are known to extract meaning from persuasive messages in

-61-

01‘

f

a way that incorporates active information-seeking and reﬂection on relevant arguments
(Williams-Piehota et al., 2006; Williams-Piehota, Schneider, Pizarro, Mowad, & Salovey,
2003). Low-NFC individuals report less willingness to put in the cognitive effort required
to consider logical messages. Instead, these people are more likely to make decisions
through heuristics, celebrities, or social expectations (Cacioppo et al., 1996). Details and
statistics are particularly effective for the high NFC group while peripheral cues such as
sources, visual elements, and number of arguments are persuasive for the low NFC group
(Bakker, 1999; Cacioppo et al., 1996; Williams-Piehota et al., 2003).

In a recent study, Ruiter and colleagues (2004) examined need for cognition’s effect
on fear appeal responses, ﬁnding that high need for cognition individuals would be more
likely to accept recommended actions than low need for cognition individuals.
Conversely, those unlikely to systematically process arguments and engage the evidence
may be more likely to resist persuasive messages, a ﬁnding that promotes increased pre-
testing of messages and even recipients when practical. Based upon prior ﬁndings and
understandings of systematic processing, it is predicted that:

H5: Need for cognition will moderate the relationship between message condition
and threat perception. Specifically, those high in need for cognition will report
greater perceived threat than those low in need for cognition.

Other variables

Other variables have shown through a lack of statistically signiﬁcant or substantial
relationships in fear appeal studies that they may not moderate the fear manipulation-
dependent variable relationship (Boster & Mongeau, 1984). Importance of topic (Beck &
Davis, 1978), medium of presentation (Frandsen, 1963), delay of attitude measurement

-62-

(Leventhal et al., 1966) are some such variables. Of course, the data against these
variables come from one experiment each, meaning they may yet be established as
moderators in future work.
Methodologigal issues in fear appeal research

Along with the independent and moderating variables that inﬂuence levels of
dependent variables, methodological issues involving areas such as measurement,
message design and audience characteristics can also impact results (Boster & Mongeau,
1984)

Measurement

Reliability of measurement, for example, substantially affects study outcomes.
Studies using reliable measures of dependent variables are more likely to ﬁnd large
effects from the fear manipulations (Boster & Mongeau, 1984). Poor measurement of the
dependent variable produces attenuation due to error of measurement and decreases
chances of statistically signiﬁcant effects for the manipulation.

The construction of the models ending in these dependent variables is another area of
concern. Typical fear appeal models link the level of the threat message to the dependent
variable, ignoring differences in individual reactions to the message. However, the
individual differences in reactions to threats make the link between fear arousal and
desired outcomes (behavioral intentions, for example) a more logical one than the link
between appeal strength and outcomes (LaTour & Rotﬁeld, 1997), where fear induction
is only a mediator variable (and an often poorly measured one at that). It is, in fact, the

aroused state of perceived fear—or perceived threat for some—that interests researchers

-63-

 

when they develop strong fear appeals. The aroused state is the motivator for individuals
as researchers attempt to predict outcomes.

Potentially, the failure to more thoroughly examine individual reactions and other
variables could limit fear appeal research by offering conclusions that seem generalizable
when they are not. As Burnett and Oliver (1979, p. 182) described, “Response to fear is
probably speciﬁc to the situation, topic, person, and criterion. Thus, the form of the
relationship will vary across combinations of these four factors.” The way in which an
audience perceives messages to be personally relevant must be a consideration in the
design of any experiment or campaign (Kelly & Edwards, 1992). An effective message
for teenagers is unlikely to be equally effective for parents (LaTour & Rotﬁeld, 1997), a
difference to be measured not only through demographics such as age but also through
mental states and reactions. Traditional fear appeal work fails to account for the need to
ﬁnd an optimal fear message and measure reactions. Instead, much fear appeal research
sticks to the repetition of errors in the research process (LaTour & Rotﬁeld, 1997). Future
models would do well to examine the relationship between the reaction to the induction
and the dependent variable; it is the reaction that matters when trying to understand
individuals’ behavior, not the message condition.

Experiment design

The results of the range of fear appeal experiments do point out that it is not apparent
on a detailed level what exact content makes for an effective fear-arousing message
(Boster & Mongeau, 1984), although fear manipulations seem to have increased in
effectiveness since the initial studies (based upon the correlation between the year of
study and the manipulation check, r = .13, p < .05) (Witte & Allen, 2000). However, the

-64-

number of treatment conditions (levels of threat) seems to be important to the
effectiveness of an experiment. Pre-testing to ensure a message arouses reactions is
essential as well (Witte & Allen, 2000).

Along with examining the message, audience analysis needs to play a role in data
analysis, as certain ﬁndings suggest that demographic and personality characteristics
moderate the fear manipulation-dependent variable relationship (Boster & Mongeau,
1984). This leads some to conclude that many fear appeal experiments'do not actually
measure broad reactions to mass communication messages. Instead, these studies are
examining responses to types of threats by speciﬁc audience segments (LaTour &
Rotﬁeld, 1997). According to some researchers, responses are linked, then, to both target
audiences and tailored stimuli (Rotfeld, 1988; Sutton, 1992). Others argue that individual
differences such as gender, age, ethnicity, and group membership have generally not
inﬂuenced buy-in for fear appeal suggestions (Insko et al., 1965; J. P. Kirscht, Becker,
Haefner, & Maiman, 1978; Leventhal et al., 1966; Rhodes & Wolitski, 1990) but instead
directly linked to outcome variables without interacting with the fear appeal level
(Burnett, 1981; Horowitz, 1969; Horowitz & Gumenik, 1970).

Despite the additional concerns, repeated examinations of fear appeals have reported
that increases in fear are linked to changes in outcome variables such as behavior,
intentions, or attitude (LaTour & Rotﬁeld, 1997; Witte & Allen, 2000). The correlations
are not always large or deﬁnitive, but across topics and communication methods,
increased fear consistently is related to adherence with suggested outcomes (Boster &

Mongeau, 1984; Sutton, 1982; Witte & Allen, 2000). As a result, Sutton concludes that

-65-

fear arousal through communications is not likely to be counterproductive, as there does
not appear to be signiﬁcant evidence of a “boomerang effect” (Sutton, 1992, p. 519).
Sutton does also note that it is essential to clarify what is meant when authors of
meta-analyses report that fear “consistently” increases adherence to suggested outcomes:
“The term ‘consistently’ [does not] mean that every single study showed a signiﬁcant
advantage of higher fear over lower fear. When a large number of independent tests of
the same hypothesis are conducted, a certain proportion of non-signiﬁcant or even
reversed results. . .are to be expected simply by chance. ‘Consistently’ is used in the meta-
analytic sense to mean that the observed results, considered as a set, would be extremely
unlikely to occur... [A]cross studies, the greater the difference between experimental
conditions in the amount of fear aroused, the greater the difference in acceptance; the

relationship was approximately linear” (Sutton, 1992, p. 517).

-66-

Chapter 3. Hypotheses a nd Research
Model

One goal of this dissertation will be to examine potential mediators and moderators
that could help predict the effectiveness of fear appeal messages. Prior research has
offered a number of variables that may contribute to the explanatory and predictive
power of fear appeal models. Figure 1 is a model visually representing how the variables
listed in the hypotheses below are expected to associate.

As discussed at several points earlier in the dissertation, prior research has shown that
perceived efﬁcacy plays a role in predicting adherence to the recommended action being
promoted in the message condition. It is reasonable to expect that the same relationship
will appear here as well:

H1: Higher perceived efﬁcacy will associate positively with an action-oriented
outcome.

Along similar lines of following prior research, an individual with higher perceived
risk can also be expected to be more interested in the outcomes of the health behavior
under discussion. Thus:

H2: Higher perceived threat will associate positively with higher outcome-relevant
involvement.

Outcome-relevant involvement may help predict the relationship between perceived
threat and action-oriented outcomes by focusing on how readily a respondent focuses on

consequences of a particular issue. Thus:

-67-

H3: Higher outcome-relevant involvement will associate positively with higher
behavioral intention.

The way people understand the health issue at hand may also be inﬂuenced by
several other variables that moderate a fear appeal’s ability to create the perception of
perceived threat. Individuals with an inclination for rational processing, for example, will
react to logical explanations of risk in a certain way, as will individuals that prefer
arousing experiences. This'leads to the following two hypotheses:

H4: Sensation seeking will moderate the relationship between message condition
and threat perception. Specifically, those high in sensation seeking will report less
perceived threat than those low in sensation seeking.
H5: Need for cognition will moderate the relationship between message condition
and threat perception. Speciﬁcally, those high in need for cognition will report
greater perceived threat than those low in need for cognition.

Taken together, the hypotheses for this research project can visually be represented
by the model in Figure 1. The relationships are labeled according to the hypotheses listed

previously.

-68-

 

Research Model

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Behavioral

1» Intention

 

 

 

Need For
Cognition
Researcher-
Produced H5 +
Newspaper Perceived +
Articles on Threat H2
Healthy
related H4 -
Topics
Sensation
Seeking
Perceived
Efﬁcacy

 

 

 

Outcome-
Relevant
lnvolvement H3
H

 

Figure 1. Research Model

-69-

 

 

Chapter 4. Method

Reseprch design

The experiments conducted in this study were also carried out with the goal of
clearing up some of the existing issues in fear appeal research. The experiments were
designed to: 1) to examine individual-level persuasion-related variables in the context of
the EPPM; 2) to open the context of fear appeal research into wider samples and
contexts; 3) and to reduce the impact of poor experimental design and measurement
common in prior research. The experiment consisted of exposing the participants to one
of ﬁve simulated newspaper articles intended to invoke a level of threat concerning a
health topic, measuring the variables related to the impact of that article as well as
resulting planned behaviors, and examining the relationship between the appropriate
variables. The ﬁve subject areas were: human Papillomavirus, breast cancer, prostate
cancer, food poisoning and clinical trials. Only females saw the breast cancer piece and
only males saw the prostate cancer article.

It is essential to note that the use of multiple topics was not intended to test those
messages speciﬁcally or increase knowledge about essential characteristics of an
effective fear appeal message; the goal of employing multiple articles was solely to
promote variance among participant responses, a design described as most effective in at
least one meta—analysis (Boster & Mongeau, 1984). Restriction in range from having little
variance in both the number of treatment conditions and thus responses to the treatments

can attenuate the relationships between intended fear manipulations and the outcome

-70-

variables (Boster & Mongeau, 1984). Using different messages with different levels of
perceived fear will also eliminate message topic as a possible explanation for effects.

Individual-level variables were selected for inclusion based upon prior persuasion and
fear appeal research. To improve design and measurement, messages and instruments
were pilot tested and adjusted according to the results and considerations of item validity.
The pilot test sample was a convenience sample consisting of an online community used
to discuss science ﬁction literature. The ﬁnal data collection usedan international—
though mostly American—online survey sample organized through the use of a
professional data-collection organization.

Placing the messages within brief news articles opens the context of communication
research into fear appeals because prior research has tended to focus only on educational
brochures, posters, or instructional videos. More than half of all Americans read a daily
newspaper (Project for Excellence in Journalism & Edmonds, 2004), making the medium
a logical environment for learning about how individuals mentally gather and process
information. The 50-plus percent of Americans reading daily newspapers is also
augmented by the up to 100 million individuals who report going online to read news at
least some of the time (Project for Excellence in Journalism & Edmonds, 2004). While
the ofﬂine and online readerships certainly overlap, comprehensive studies of news-
consumption habits show that the populations do not perfectly match, particularly for
younger individuals (Project for Excellence in Journalism & Edmonds, 2004). This
means that the population of regular news consumers in the United States is more than

half of all its residents and afﬁrms the importance of examining the news content that

-71-

people are exposed to each day. Particularly if one wants to reach a younger
demographic, using online media is essential.

Online samples were employed to broaden the sample population from typical fear-
appeal research that depends upon students or those living in high-risk environments such
as areas with elevated disease rates. In other research proj ects, Intemet—based samples
have successfully been used to approximate more traditional data collection methods
including mail or in-person surveys (Jason Baker, Woods, & Reynolds, 2007; Hewson,
Yule, Laurent, & Vogel, 2003; Kaye & Johnson, 1999; Leece et al., 2004; Witmer,
Colman, & Katzman, 1999).

Online sampling is proving increasingly popular because of the efﬁciencies the
Internet can bring to survey sampling. These beneﬁts also require additional vigilance for
threats to reliability and validity from factors such as security issues, potential bias, or
respondent comfort with technology (Jansen, Corley, & Jansen, 2007). As Jansen and
colleagues describe, however, an increasing number of resources are available to aid
researchers in planning for potential issues (Lazar & Preece, 1999; Schmidt, 1997).
Beneﬁts related to decreased costs, quicker data collection and often higher response
rates bring many researchers to conclude that designing around potential problems is well
worth the effort (Mehta & Sivadas, 1995).

Article design according to the EPPM

The messages relating to ﬁve health-oriented subject areas were intended to address
the essential components of fear appeal messages as determined by prior research.
According to the EPPM’s standards, a message must have a component addressing each

of the following areas: perceived susceptibility, perceived severity, perceived self-

-72-

efﬁcacy, and perceived response efﬁcacy. A statement of the reader’s susceptibility to the
health problem and one concerning the severity of the health problem made up the threat
component intended to produce a fear reaction.

The opening sentences focused on severity and susceptibility, following the
newswriting norms appropriate for a news article. Opening sentences are intended to grab
reader attention and summarize the article’s content while focusing on a news value such
as timeliness or impact (M. Block, 1997; Burns, 2002; Herbert, 2001). Discussing the
seriousness of an issue ﬁts just such a task of grabbing attention.

Efﬁcacy components were also incorporated into each article. They focused on
preventative behavior and intentions to gather further information. Response and self-
efﬁcacy elements, however, were not considered a focus of the research, as the overall
goal was to understand reactions to fear appeals. The efﬁcacy components served
primarily to aid in the development of behavioral intention measures and complete the
EPPM model requirements.

Complete articles for each of the ﬁve topics are included in Appendix A. Each article
followed the same format on a paragraph-by-paragraph basis, considering that the articles
were not intended to have equal amounts of threatening information. Opening paragraphs
focused on the threat and importance of the subjects. Second paragraphs continued these
themes with supporting information that particularly noted susceptibility statistics. The
third paragraphs provided more detailed information on the conditions, including further
susceptibility and risk information. The fourth and ﬁfth paragraphs offered details on

preventative or beneﬁcial actions that cover the efficacy elements, suggesting positive

-73-

behaviors and that readers visit a Centers for Disease Control or National Institute of
Medicine educational Web site.

Following the above outline, each of the ﬁve articles included the four elements
prescribed by the EPPM for an effective fear appeal message. Different elements were
often worked into the same sentences to make for more natural and readable writing. The
following paragraphs will outline how each of the articles incorporated the four EPPM
message elements.

The breast cancer article started with a statement of severity and susceptibility:
“Breast cancer remains a top killer of women in the United States despite vast efforts to
combat the disease, new health ﬁgures released this week show.” Calling the disease a
killer stresses the severity of having the condition while describing it as among the most
frequent (“top”) emphasizes that all women are susceptible.

Ensuing paragraphs re-iterate both the severity and susceptibility arguments. The
second paragraph highlights the fact that in addition to the tens of thousands of women
who pass away from breast cancer each year, thousands more undergo painful, life-
altering treatments such as radiation therapy or mastectomies. This paragraph stresses
that the severity of the disease is in not only its mortality rates but also the way those
afﬂicted are forced to adapt their lives and deal with negative circumstances.
Susceptibility is reinforced by sentences stating that hundreds of thousands of women
develop breast cancer each year and that all women are at risk, regardless of age or
family history.

The efﬁcacy elements are put forth by sentences in the last two paragraphs focusing

on information-gathering as a useful ﬁrst step in prevention. In addition, this portion of

-74-

the article states that a more healthful diet of fruits and vegetables may help ﬁght cancer,
as can exercise and supplements such as folate. Self-exams and check—ups are also
promoted because of their ability to help women detect problems at any early stage,
which may help prevent or limit tumor growth.

The prostate cancer article also begins with a statement of susceptibility, calling the
disease “the biggest health threat to men of all ages.” The article follows in the next
sentence with a severity message noting that prostate cancer “kills tens of thousands of
men each year and causes erectile dysfunction in hundreds of thousands more.”

The story further pushes the susceptibility element in other ways. It states that the
threat of prostate cancer may be growing and that incidences of the disease have
increased more than 50 percent in the last 30 years. Severity is further promoted by
pointing out that prostate cancer can be considered more dangerous than the better-known
breast cancer because the prostate version does not exhibit symptoms in the early stages
when it is more treatable.

Efﬁcacy statements enter the story when it mentions that there are methods of
slowing the disease’s progress or reducing one’s risk. Speciﬁcally, the article uses the last
paragraphs to promote learning about the disease from a Centers for Disease Control
Web site, eating a healthful diet, exercise, and certain medications.

The story on human Papillomavirus began with a strong threat message, addressing
both susceptibility and severity: “Chances are you have been exposed to the potentially
cancer-causing human Papillomavirus virus and don't even know it.” Susceptibility was
promoted again in the following sentences. One sentence pointed out that almost 75

percent of sexually active adults under 50 have been infected with some form of HPV.

-75-

Another sentence noted that 6.2 million new HPV cases occur each year in the United
States, that 20 million Americans are infected, and that HPV is the most common
sexually transmitted infection. The third paragraph put more emphasis on severity when
it stated that HPV leads to certain cancers and that public health experts place HPV
among the most serious health issues facing youth.

Efﬁcacy, of course, is handled toward the end of the story when readers are
encouraged-to visit 'a Centers for Disease Control Web site to become better informed.
Also, the article recommends that readers maintain a monogamous relationship and use
condoms when having sexual intercourse.

Information on food poisoning follows the pattern of opening with a susceptibility
statement and also follows with some information concerning severity. The susceptibility
statement is, “More than a quarter of all Americans come down with food poisoning each
year” while the severity component comes next, mentioning that food poisoning sufferers
deal with nausea, vomiting and diarrhea. As this story was intended to create only a
moderate level of perceived fear at most, both of these elements are obviously less
threatening than either of the cancer messages, for example. More susceptibility
information covers the fact that there are more than 76 million cases in the US each year.
Further severity elements point out that a small percentage of food poisoning victims end
up in the hospital.

Prevention messages make up the efﬁcacy components here, as readers were
informed that food poisoning can be prevented with simple steps. Washing hands,
cleaning utensils, cooking food all the way through, and storing raw and cooked food in

separate containers can all help to eliminate the dangerous bacteria and viruses. Readers

-76-

were also encouraged to visit a Centers for Disease Control Internet site to learn more
suggestions about safe food preparation.

The clinical trial story was designed to act as a comparison condition for the other
threat-inducing articles. No threat messages were included in the clinical trials article.
The opening paragraphs discussed growing awareness of the existence of clinical trials
and Opportunities to participate in them. Other paragraphs provided more detailed
information on clinical trials, including their aims, frequency, and participant beneﬁts.
Readers were, however, provided with efﬁcacy messages about participating in clinical
trials. The article suggested visiting an informative National Institute of Medicine Web
site, talking to a primary care provider, or even contacting the coordinator of a clinical
trial directly.

All of the articles were extensively pre-tested to ensure that reader perceptions
matched the intended effects. Instruments, reliabilities, and other results from the pilot
study are available in the appendices.

Article attributes

Each article was also manipulated to make them each approximately equal in terms of
length and reading level. The pieces were written to speciﬁcally avoid technical terms or
multi-syllabic words as part of an effort to increase their readability. Also, avoiding such
language helped to keep the articles in line with journalistic writing conventions (Knight,
2003)

The resulting ﬁve-paragraph articles had a mean length of 262 words, ranging from
233 to 280 words. The ﬁve paragraphs contained an average of 1,295 characters, with a
low of 1,222 and a high of 1,389. The Flesch-Kincaid Grade Level and Flesch Reading

-77-

Ease measurement scales were used to assess reading levels for the articles. The Flesch-
Kincaid Grade Level index uses the number of syllables per word and words per sentence
to calculate a grade level required for reading. Flesch-Kincaid Grade Level scores for the
created stories averaged a ninth-grade reading level, although the breast cancer and
clinical trial stories earned scores into the 10~grade level. The grade-level scores across
these stories rank equal with mass-market publications such as TV Guide or Time
magazine (DuBay, 2005). The ninth-grade reading level also puts the treatment articles
about a level below the front-section stories in major metropolitan or national news
sources such as the Houston Chronicle (Meyer, 2004), the Wall Street Journal, the New
York Times (J. Jung, 2003), and the Washington Post (DuBay, 2005). Some research has
reported newspaper readability scores between 16 and 18—content more suitable for
college graduates or graduate students (Danielson & Bryan, 1964; McAdams, 1992;
Wanta & Gao, 1994).

The Flesch reading ease score is calculated using average sentence length and number
of syllables per 100 words. Reading ease scores range from 1 to 100, with 1 being the
most difﬁcult to read and 100 being the easiest. Flesch reading ease scores averaged 55.8
across the stories, again making the experiment’s created content slightly easier to read
than actual news content. One study found the Wall Street Journal’s top stories rated a
48.3 on Flesch’s reading ease index, while USA Today scored 47 and the New York
Times earned a 36.2 (J. Jung, 2003). Other research has shown that newspaper editorials
average a 48 on the reading ease index (Murphy, Gamble, & Sharpe, 1994). Table 1

contains the length and readability information for each article.

-73-

The articles in use here were intended to be as readable as possible while maintaining
journalistic conventions to ensure their plausibility as news stories. If the research
participants are not capable of reading and comprehending the texts, the intended
manipulations stand zero chance of taking place. Therefore, vocabulary selection and
reading level were given primary consideration during the deve10pment of the articles.

Table 1. Article Attributes: Length and Readability

 

 

 

 

Article Topic Words Characters Grade Level Reading Ease
Breast Cancer 233 1,222 10.7 47.9
Prostate 280 1,389 9.5 59.9
Cancer
HPV 275 1.333 9.6 58.5
Food Poisoning 252 1,243 9.9 58.4
Clinical Trials 269 1.288 10.1 54.3
Mean 262 1,222 9.96 55.8

 

Speciﬁc sources were removed from each story in favor of the generic term “public
health ofﬁcials.” The objective in removing particular sources from the content was to
avoid speciﬁc institutions or judgment based upon sourcing. The only proper nouns
employed within the articles came in the ﬁnal-paragraph references to the Centers for
Disease Control and the National Institute of Medicine. Both institutions maintain
positions among the top health research and policy organizations in the country
(Etheridge, 1992) and were also necessary to identify as the sources of the recommended
information.

For similar sourcing reasons, all the articles had an Associated Press logo just above
the text. Researchers were concerned about the perceived credibility of any newspaper

article being inﬂuenced by the reputation of a particular publication, as increasing

-79-

 

 

a...

I“...

.1-

polarization among media and decreasing trust among consumers may affect the media’s
ability to effective distribute credible information (Alterman, 2003; Fan, Wyatt, &
Keltner, 2001; Kohut, 2004). Considering this situation, the decision was made to link the
articles to the Associated Press for two reasons. First, prior experiments have
demonstrated such a stimulus can be effective (Callison, 2001; Kahn, 1994). Second, it is
a way of associating with a large-scale news organization while avoiding some of the bias
accusations attached to speciﬁc print or television media outlets. Presenting stimulus
materials as AP content was an attempt to present a credible, professional image and
avoid the distrust aimed at named, single-brand mass media outlets.

Instrument selection

Measurement instruments used in this study come from prior research studies in
which they have demonstrated accepted levels of reliability and validity. As the topic of
this project may differ from the original context of the instruments, researchers
considered that items would possibly need to be adjusted following the pilot testing. In
some cases where a literature review determined that no single instrument was dominant,
multiple instruments were pre-tested and piloted for effectiveness and levels of
demonstrated reliability. Testing of instruments was intended to serve as means of
locating issues in wording, understanding, or effectiveness. The appendices contain the
full versions of the tested instruments as well the outcomes from the pilot testing. The
following instruments from the following studies were selected to test as measures of

relevant variables in this study:

-30-

Perceived Threat: a two-part survey composed of perceived severity and
perceived susceptibility taken from Witte, Meyer and Martell’s 2001 book
Effective Health Risk Messages: A Step-by-Step Guide (Witte et al., 2001).
Perceived Efﬁcacy: also measured by a two-part survey on perceived self-efﬁcacy
and perceived response efﬁcacy again with an instrument from Witte, Meyer and
Martell’s Effective Health Risk Messages: A Step-by—Step Guide (Witte et al.,
2001).

Outcome-Relevant Involvement: measured by two instruments in the pilot test—a
seven-item instrument developed in Cho and Boster’s “Development and
Validation of Value-, Outcome-, and Impression-Relevant Involvement Scales”
from Communication Research (Cho & Boster, 2005) and a three-item version
adapted from Park et al.’s 2007 Human Communication Research article “The
effects of involvement type and argument quality on attitude formation and
attitude change” (H. S. Park, Levine, Kingsley Westerman. Orfgen, & Foregger,
2007)

Sensation Seeking: Hoyle et al.’s eight-item Brief sensation seeking Scale was
employed to measure sensation seeking (Hoyle, Stephenson, Palmgreen, Pugzles
Lorcha, & Donohew, 2002).

Need For Cognition: two instruments were used in the pilot test—a three-item
need for cognition scale described in Cacioppo and colleagues’ article “The
efﬁcient assessment of need for cognition” (Cacioppo, Petty, & Kao, 1984) and
an 18-item version published by Cacioppo and Petty in a slightly older Journal of
Personality and Social Psychology publication (Cacioppo & Petty, 1982).

-31-

' Behavioral Intention: an adapted version of a Journal of Medical Informatics
Association instrument measured intentions with six items according to the

recommended behaviors contained in each article (Wilson & Lankton, 2004).

Data gollectipn

For the data collection, an online experiment was conducted using a sample provided
by an organization specializing in online surveys. Individuals opt in to receiving regular
e-mails from the organization for the purposes of completing various surveys. In return
for remaining in the survey pool, respondents are entered into monthly $250 drawings for
cash prizes. Completing any survey also earns the individual entry into a prize drawing
for an amount determined by the sponsoring researcher. Membership in the survey pool is
rotated regularly to ensure new membership. Individuals receive no more than one survey
invitation a month to ensure that they are not oversampled. All responses from
individuals in this survey pool are privacy protected and lacking any identiﬁable
information when delivered to the researcher.

No particular demographic group was purposely included or excluded from the
experiment. Individuals listed in the survey pool received an e-mail directing them to the
survey Web site and describing the incentive raffle, along with the estimated time to
complete the survey. The incentive was entry into a drawing to win one of 10 $100
prizes.

Those responding to the request were re-directed to the Internet survey page where
their responses were captured as they clicked through the instruments. Individuals not

responding to the initial request received a follow-up e-mail several days later again

-32-

requesting their participation and reminding them of the chance to earn an incentive by
completing the survey.

Participants were randomly assigned to one of the ﬁve message conditions and were
exposed to only one article. The order of survey questions was consistent for each
individual. The only change was the treatment article and necessary wording changes in
the accompanying surveys. Participants were free to take the survey at any computer
location convenient to them, such as work, home, library, or school. Survey
questionnaires and participant instructions are shown in Appendix D. Individuals were
free to quit the survey at any time and were made aware of this in their informed consent
form.

Participants

A total of 1,310 individuals from around the globe participated in the experiment in
exchange for the opportunity to be entered into the prize drawing. Participants’ ages
ranged from 19 to 81 with a mean age of 45. The sample was 69 percent female, and 70
percent of respondents were married or in committed relationships. The sample consisted
of 57 percent Americans, 16 percent Canadians, 7 percent New Zealanders, and 5 percent
British. Individuals reported their ethnicity as white 82 percent of the time, Asian 9
percent, black 4 percent, and multi-racial 3 percent. The online sample was considered
appropriate because the project’s concern involved mass media messages that reach a
wide variety of individuals. As a mass medium itself, the Internet can serve a parallel
purpose to traditional mass media and reach similar audiences. In addition, one project
goal was to examine fear appeal messages in an environment other than a student sample

or individuals living in a high fear environment.

-83-

After eliminating 205 responses, 1105 usable surveys remained. Cases were
eliminated for two main reasons. A dozen were almost entirely incomplete and merited
exclusion because of the lack of even partial data. The solid majority (193 of 205) of the
eliminated responses, though, stemmed from suspicious Internet Protocol addresses.
These unique addresses function like a phone number or street address and allow
computers to communicate over networks. Each survey response was attached to an IP
address when entered into the dataset. IP addresses for each individual’s connection to
the Internet should be unique, if not to the speciﬁc computer then to a computer network
allowing connection to the Internet. It is possible to share an address among computers
on the same operating network; a wireless network in an ofﬁce may assign each computer
the same IP address, for example.

A large number of responses were linked to repeat IP addresses. These responses
were isolated from the others for further examination. Keeping in mind that individuals
sharing a network could share an IP address, demographic characteristics were examined
for similarities. Any responses sharing an IP address and demographic characteristics
were eliminated from the dataset on the basis that the IP address duplication was not due
to network-sharing. Instead, it was assumed that IP address repetition resulted instead
from the same individual completing the survey multiple times. This procedure
eliminated 193 responses from the data set.

Researchers assume that the multiple response pattern stems from two factors. One,
individuals may have misunderstood the reminder e-mail sent by the survey-sample

company. It is possible that they interpreted the reminder as an invitation to complete a

-84-

new survey. Second, the repeat-response individuals could have been attempting to gain
multiple entries into the incentive drawing.

Other responses, while acceptable, were not 100 percent complete. As is expected,
certain respondents failed to complete certain items. To account for these missing
responses, SPSS’ “missing values” function was used. Speciﬁcally, missing data points
were ﬁlled in by employing maximum likelihood estimation. This method was selected to
avoid the biases that can result from common practices such as deleting cases or mean ‘
replacement (Croy & Novins, 2005). Mean replacement, for example, can bring about
attenuated conﬁdence intervals and false positive Type I errors (Croy & Novins, 2005).

Maximum likelihood estimation is considered by statisticians to be superior to most
methods of data completion such as listwise deletion, mean substitution, or pairwise
deletion (SPSS, 1999). MLE uses all the available data for each case to calculate
estimates, returning the value most likely to have resulted in the observed data (Grace-
Martin, 2002). The method returns unbiased parameter estimates and standard errors by
computing likelihood functions both for cases missing data and those that are not. An
iterative algorithm then compares the results and reports the estimates (SPS S, 1999).

Stimulus

The articles examined in the pilot testing were used in the ﬁnal data collection as
well. Pilot testing did not produce any comments or observations requiring change to the
treatment messages. Complete versions are available in Appendix A.

Measurement instruments

The indices used were adopted from previous research with some necessary wording

modiﬁcations to ﬁt the current project. Instruments were selected based upon their

-35-

reliability and validity in prior studies and in pilot testing for this study. The pilot-tested

instruments and the outcome data from that testing are available in complete form in the

appendices. All constructs in this study were measured by at least three observable

indicators (Bentler & Chou, 1987), typically in the form of statements followed by 7-

point Likert-type scales. Sensation seeking, however, was measured on a 5-point scale,

and respondents showed their comprehension of the treatment—article text through

true/false questions.

The ﬁnal online survey presented instructions, questions, and the treatments in the

following order:

Introduction to survey and consent form

Thank-you note, survey description, instructions, and consent agreement.
Demographic questions, including: age, gender, marital status, ethnicity,
nationality, familiarity with relevant health topics, relationship status, news-
reading habits.

Sensation seeking instrument: 8 questions with 5-point scales.

Need for cognition index: 18 questions with 7-point scales.
Outcome-relevant involvement: 3 questions with 7-point scales.
Introduction the newspaper article asking respondents to read the content and
answer further questions.

Treatment article concerning clinical trials, food poisoning, human
Papillomavirus, prostate cancer, or breast cancer.

Comprehension questions: 4 questions with true/false format.

Believability index: 4 questions with 7-point scales.

-86-

. Perceived severity instrument (one-half of perceived risk): 3 questions with 7-
point scales.
° Perceived susceptibility (one-half of perceived risk): 3 questions with 7-point
scales.
° Response efﬁcacy (one-half of perceived efﬁcacy): 3 questions with 7—point
scales.
' Self-efﬁcacy (one-half of perceived efﬁcacy): 3 questions with 7-point scales.
0 Behavioral intention: 6 items with 7-point scales.
° Thank you note, further information on the treatment article’s topic.
Data analysis
Brown-Forsythe and Welch statistics checked the effectiveness of the manipulations
of the individual message conditions. These statistics were used because of the non-
homogeneity of variance and unequal sample sizes among the clinical trial-, virus-, and
cancer—message groups. Games-Howell, Dunnett’s T3, and Dunnett’s C provided further
analysis to more speciﬁcally verify differences among the groups, similar to the way
other analysis may employ Tukey’s HSD. Multiple statistical tests were employed solely
to ensure the most conservative analysis possible, considering the limits and irregularity
of the data as well as the tests. Violations of parametric analysis’ assumptions can
somewhat be expected in an experimental project such as this where the sample sizes
differ and the induction results in multiple conditions for the treatment variable. The
effectiveness of the treatment depends upon rendering the data non-normal within each

message condition. That is the very goal of such an induction.

-37-

Prior to testing hypotheses, indices were checked for reliability using Cronbach’s
alpha. Testing of the proposed hypotheses was conducted using multiple regression to
examine predicted relationships. Despite the normality concerns for the groups discussed
above, regression is appropriate here because hypothesis testing was done with the
complete, combined data set. When merged, the groups create a normal data set; the
combined conditions create an acceptably shaped data curve. As an analysis technique,
regression was a logical choice because one goal of this research was to examine the
causal relationships of message reactions and ensuing behavioral intentions. Thus, the
proposed relationships should be tested for evidence of substantial paths denoting the
inﬂuence of one variable on another. In addition, the use of regression allows for the
control of outside factors to further isolate and examine the predicted relationships.

Outcomes of this testing are presented in the next chapter.

-33-

Chapter 5. Results and Discussion

Since the speciﬁc message conditions were not of particular focus in this research,
data were combined into a single data set. Combining data from different conditions
promotes variance in responses, and increased variance was one objective here. The
variety among induced levels of perceived risk encourages similar variety in the data for
other constructs. Creating multiple message conditions was done speciﬁcally to promote
variance in this way. Similar studies might focus on the message conditions to examine
what makes for an effective fear appeal. Effective message design attributes was not a
primary concern in this research. The goal here was to examine how individuals mentally
react to such messages and process the presented information, as opposed to how to best
induce conditions perceived as risky.

Nprmalig

As a ﬁrst step in the analysis, the combined data were scrutinized for normality to
examine the appropriateness of parametric tests. Normality alone does not determine the
correctness of choosing parametric tests, but violating normality can become problematic
if other assumptions are also violated (Cramer, 1994).

In this case, normality was determined through visual examinations of P-P plots, Q-Q
plots, and histograms. P-P plots and Q-Q plots were inspected to ensure that the data
points fell approximately in a straight line. While some sampling variability about the
line is to be expected, normal groups will lie consistently around a straight line. Q-Q
plots are particularly sensitive to non-normal data in the distribution’s tails. P-P plots tend
to be more sensitive to departures around the distribution’s mean.

-39-

Histograms also served to examine the data points in the context of a normal curve.
For variables with a plot or histogram that raise concerns, normality was also examined
through dividing skewness and kurtosis by their standard errors to be sure the resulting
ration was between -l.96 and 1.96.

Manipulation check

The effectiveness of intended fear inductions was assessed using the respondents’
ratings of perceived threat. Each of the two dimensions of perceived threat was measured
by three seven-point Likert items, adapted from Witte et al.’s book “Effective Health
Risk Messages” (2001). An AN OVA examined the impact of the treatment article on the
perceived threat scores, as measured by the perceived susceptibility and perceived
severity dimensions. However, based upon the post-hoe results, ANOVA proved to not
be the most appropriate statistic for testing differences among the groups.

While the F-score was statistically signiﬁcant, the Levene Statistic test of
homogeneity of variance indicates that the ANOVA result is not valid. The Levene
Statistic had a value of 37.94 with a p-value < .001. When the p-value is less than .05, the
result indicates that homogeneity of variance does not exist across the groups. ANOVA
can usually be counted on to be quite robust; however, considering that there are unequal
numbers of respondents in each group as well, use of ANOVA here may push the limits
of its robustness.

In place of the F-score considering the homogeneity of variance concerns, one has the
option of several other tests available to compare group means. This project used the
Brown-Forsythe and more conservative Welch tests in place of ANOVA. Both statistics

(Brown-Forsythe: 107.62, p < .001; Welch: 88.31, p < .001) were statistically signiﬁcant,
-90-

offering evidence that the perceived threat scores for three treatment groups did differ
according to the intended article exposure. Examining the group means for the perceived
threat index also indicates that the effects were in the predicted direction (Mclin trials =
20.86, Mvirus = 29.39, Mcancer = 31. 72). The cancer articles have higher perceived threat
scores than the virus articles, which, in turn, associate with higher scores than the clinical
trials control condition.

Several post-hoc tests were used to determine the statistical signiﬁcance of the
differences between the groups. While Tukey’s Honest Signiﬁcant Difference analysis
would be the common choice for post-hoc analysis, this statistic is not appropriate in this
case because of the unequal sample sizes and heterogeneous variances. Instead, Games-
Howell, Dunnett’s T3, and Dunnett’s C were employed to offer multiple options to check
the relationships among the three groups. Consistency across multiple tests offers a
conservative measure for differences among the treatment groups and ensures that no one
test’s bias will produce a misleading result. All three analyses conﬁrm that the treatment
groups differ at the p < .001 level when examining perceived risk scores. The Games-
Howell test, in particular, is worth noting because it is the most powerful of the three and
is strongly recommended for comparing groups with unequal sample sizes (Toothacker,
1993). The outcomes of the tests are recorded in Table 2. These results can be interpreted
as evidence of a successful manipulation of perceived threat in response to the treatment

articles.

-91-

Table 2. Post-hoc Tests for Group Means on Perceived Threat Index

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

95%
Conﬁdence
Interval

Survey Survey Std. Upper Lower
condition condition Mean Diff. Error Sig. Bound Bound

1Dgnnett ClmrcalTrIals Vll‘US $.53” .75 .000 40.32 674
Cancer -10.85(*) .84 .000 -12.86 «8.85

Virus Clinical Trials 8.53(*) .75 .000 6.74 10.32

Cancer -2.32(*) .59 .000 -3.73 -.91

Cancer Clinical Trials 10.85(*) .84 .000 8.85 12.86

Vll‘US 2.32(*) .59 ' .000 .91 3.73

Games- Clinical Trials Virus -8.53(*) .75 .000 -10.28 -6.78
Howell Cancer -10.85(*) .84 .000 -12.82 -8.89
Virus Clinical Trials 8.53(*) .75 .000 6.78 10.28

Cancer -2.32(*) .59 .000 -3.71 -.94

Cancer Clinical Trials 10.85(') .84 .000 8.89 12.82

Virus 2.32(*) .59 .000 .94 3.71

DunnettC Clinical Trials Virus —8.53(*) .75 -10.28 -6.78
Cancer -10.85(*) .84 -12.83 -8.88

Virus Clinical Trials 8.53(*) .75 6.78 10.28

Cancer -2.32(*) .59 -3.71 -.94

Cancer Clinical Trials 10.85(*) .84 8.88 12.83

Virus 2.32(*) .59 .94 3.71

 

* The mean difference is signiﬁcant at the .05 level.

Another important check on manipulation is to consider how well readers understood

and believed the article content. Comprehension levels are key considerations to verify

that respondents read the articles and grasped the content. Just as in the pilot test

described in the appendix, comprehension was measured using four true-false items. The

items were similar across articles but adapted to the speciﬁc topics. Again, all of the

comprehension instruments focused on the basic information common to all articles:

frequency of the issue, preventative measures, risk factors, and severity information.

Items in this instrument were answered correctly 92 percent of the time. Table 3 presents

the percentage results for each item and group. Such a high score supports the idea that

readers considered the information contained in the articles.

-92-

 

llln
Cllr
Tot

 

 

.u . l f . ”l. .h- n-v A u n-V P - AA.
3% RV G. .1 . l . .5
r ~l> .Ll :5... 4L: ..L 1r

 

 

Table 3. Percentage Comprehension Check Scores by Item and Group

 

 

 

 

 

 

Topic Item 1 Item 2 Item 3 Item 4 Total
Cancer .95 .96 .92 .85 .92
Virus .90 .92 .86 .91 .90
Clinical Trials .90 .95 .97 .95 .94
Total .91 .94 .91 .91 .92

 

 

 

 

 

 

 

Article believability is also a logical portion of a manipulation check because the
inﬂuence of the articles depended at least in part on their perceived accuracy.
Believability was again measured using four items across seven-point Likert-type
response scales. All ﬁve groups responded to the same four items. Table 4 reports the
descriptive statistics for each item in the believability index. Based upon the item scores
and mean across the individual items, it is logical to report that respondents found the
treatment stories to be believable. Table 5 fulﬁlls a similar purpose for each treatment

group, showing that group means were also suitably high.

Table 4. Descriptive Statistics for Believability Items

 

 

 

 

 

 

| N Mean Std. Dev.
ome, the adVIce given in the 1105 5.83 129

article IS believable.

In my opinion, the information 1105 5.73 1.34
iven in the article seems useful.

The adVice given in the article is 1105 5.48 129

accurate.

The information I read in the 1105 5.71 132

article looks true to me.

[Mean 1105 5.68 1.20

 

 

 

 

 

All items used a seven-point Likert-type scale.

-93-

Table 5. Believability Index Descriptive Statistics for Groups

 

 

 

 

 

I Topic N Mean Std. Dev.
[Cancer 261 5.73 1.22
Illirus 567 5.69 1 .21
[Clinical Trials 277 5.63 1.16
IMean 1 105 5.68 1.20

 

 

 

 

 

All items used a seven-point Likert-type scale.

Measurement instrument assessment

The assessment of the measurement instruments occurred in three phases: 1) item
analysis, 2) reliability assessment, and 3) construct validity. Item analysis was completed
through Principal Component Analysis. Each variable in the model underwent Principal
Component Analysis to verify unidimensional relationships. Unidimensionality was
deﬁned as loading on a single factor in the SPSS-provided component matrix in a way
that clearly demonstrated a relationship to the other measurement items. No items needed
to be eliminated due to substantial cross-loadings or low loadings on the intended
constructs. As discussed earlier, all of the instrument were tested and employed in prior
research so the acceptable outcomes were not unexpected.

Reliability assessment on the measurement instruments took the form of Cronbach’s
alpha. Each index with multiple items was tested for internal consistency, and the
Cronbach’s alpha coefﬁcients for all constructs were higher than the recommended
minimum of .70 (Nunnally, 1978). Table 6 presents the alpha coefﬁcients for each index

along with descriptive statistics.

-94-

Table 6. Descriptive Statistics and Cronbach Alphas for Measurement
Instruments

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

# of ' Typical
Construct . .

I Items Cases Mean S D Alpha Alpha
Perceived Severity 3 1090 14.54 5.60 .93 .80
Perceived Susceptibility 3 1098 12.22 5.52 .91 .80
lPerceived Self-Efﬁcacy 3 1092 15.94 4.30 .87 .79
Perc. Response Efficacy 3 1094 16.80 4.15 .89 .79
Park et al. Outcome-Rel. 3 1097 12.15 4.72 B4 '86
Involvement
Sensation Seeking' 8 1080 20.97 6.48 .83 .76
Short NFC Index 18 1035 83.46 17.34 .90 .76
Believability Check 4 1091 22.79 4.75 .94 NIA
Behavioral Intention 6 1068 27.89 9.69 .88 .81

 

 

 

 

 

 

 

 

*All instruments use 7-point Likert—type scales, except for Sensation Seeking. The 8-item
instrument used for Sensation Seeking employs 5-point Likert-type scales.

Interagtipn terms

Not all measurement items in the dataset are the exclusive product of the survey
instruments, however. Several of the hypotheses focus on predicted moderator variables
so a logical necessity arises to develop interaction terms. The predicted interactions are
between categorical message conditions and continuous variables measured by a series of
Likert-scale items. Producing terms for these categorical by continuous variables requires
a number of decisions and particular treatments (see Aiken & West, 1991).

First, one must decide how to prepare the categorical variable to render it suitable for
the regression equation. A number of options are available here; the most common is

using dummy codes to represent the different levels of the categorical variable. It is

-95-

important to recognize that dummy codes are not the optimal choice, however, for every
project despite their predominance in the social science literature and presence as the
default choice (O'Grady & Medoff, 1988). Unweighted effects coding, orthogonal
coding, and weighted effects coding all represent options with particular uses that could
well serve researchers trying to answer particular questions, with the ultimate choice
depending upon the question being asked.

In this particular project, dummy codes were selected nonetheless because of the
nature of the relevant hypotheses. This research is not interested in differences across the
groups (only differences from the control group of those reading the clinical trials article)
so more involved coding methods would serve little purpose. With the control group left
out as the comparison sample, the design goals were achieved. Using dummy variables
created two dichotomous variables representing the virus and cancer article conditions,
leaving out the clinical trials condition as the comparison group.

The next necessary step was to center the continuous variable in preparation for the
creation of the interaction terms (Aiken & West, 1991). In these situations, centering the
continuous variable increases the interpretability of the interactions and also serves to
reduce potential multicollinearity issues with the independent variable. The newly
centered continuous variable is then multiplied by each of the dummy coded variables to
create two interaction terms for each of the relevant interactions in this project.
Interaction terms can then be entered directly into the regression equations (Aiken &

West, 1991).

-96-

Construct validity

Construct validity of the measurement instruments is also essential to assess as a
means of ensuring that the measurement items actually measure the concepts they are
intended to measure (Cook & Campbell, 1979; Cronbach, Gleser, Nanda, & Rajaratnam,
1972). A measure is valid to the degree it indicates the magnitude and direction of a
representative sample of the characteristics of the construct and that the measure avoids
contamination from other constructs or error (Peter, 1981).

Assessing construct validity permits examination of confounding factors such as
random error or method variance. Inspecting these issues can help remove some
ambiguity from theory testing. Error of measurement, for example, can lead a researcher
to reject or accept a hypothesis based not on theory or data but solely on the scope of the
project design (Bagozzi, Yi, & Phillips, 1991).

Imperfections, of course, are to be expected in the measurement process. Thus, it is
also essential to keep in mind that construct validity is determined only by inferring that
the measure’s scores perform as relevant theory predicts they should (Peter, 1981).
Because construct validity is a continuing process (Cronbach, 1971), construct validation
is not to be viewed as empirical proving of a measure because a measure’s success
depends upon the construct and its predicted relationships (Peter, 1981), highlighting the
key role of theory development in this process (Schwab, 1980).

Numerically, assessing convergent and discriminate validity are common means of
operationally determining construct validity (Peter, 1981). Convergent validity
scrutinizes how items for a construct share variances, showing that they are measuring

the same underlying construct (Schwab, 1980). Researchers demonstrate convergent

-97-

validity with factor loadings on the intended constructs that are greater than twice the
standard errors (Anderson & Gerbing, 1984). Each factor loading for this project’s
measurement instruments was well more than twice the standard error (.007 - .067).

Rather than showing how items relate to each other, however, discriminant validity is
intended to illustrate that measures do not correlate very highly with other measures from
which it should differ (Campbell, 1960). Discriminant validity can be determined in
several ways (Julie Baker, Parasuraman, Grewal, & Voss, 2002). One method is to
inspect the conﬁdence intervals surrounding each pairwise correlation estimate to make
certain that the intervals do not contain 1———-a signal that the instruments may measure the
same constructs (Anderson & Gerbing, 1988). This condition was easily satisﬁed for all
measurement instruments. Another option for examining discriminant validity is to
compare the percentage of variance extracted for each construct and compare it to the
shared variance (correlation squared) with each other construct (Hult, Hurley, Giunipero,
& Nichols, 2000). This condition was also easily met in that the constructs were
responsible for different effects. Taken together, satisfying the requirements for both
convergent and discriminant provide evidence for construct validity in this study.
Regression analyses

One purpose of the study was to test a number of hypotheses related to several
variables’ inﬂuence on behavioral intention to undertake positive, personal health
behaviors. Analysis of these hypotheses began with an examination of the variables’

correlations. No strong bivariate correlations were observed between inappropriate

variables. Item-item correlations are presented in Table 7.

-93-

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 7. Bivariate Correlations Amoanariables
Virus Cancer News

Variable DV ORI Threat NFC art. art. SS Efﬁe. use Age
[3V 1.00 350*) 510*) 190*) .160*) 230*) 150*) 520*) .08(**) 0.02
[ORI 350*) 1.00 .26(**) 090*) -.150*) .16(**) .070) .100*) 100*) 0.01
[Threat 510*) 260*) 1.00 0.03 100*) 460*) 0.03 370*) 0.01 0.05
[NFC 190*) 090*) 0.03 1.00 0.01 0.01 220*) 110*) 200*) 0.04
[Virus art. .16(**) -.150*) .100*) 0.01 1.00 -.590*) -0.04 260*) 0.01 0.01
[Cancer art. 230*) .16(") 460*) -0.01 -.590*) 1.00 0.04 0.015 0.03 003
[SS 150*) .070) 0.03 220*) 004 0.04 1.00 0.02 090*) -.290*)
IEffic. 2520*) 100*) 370*) .110*) 260*) 0.05 0.02 1.00 0.06 100*)
[News use 080*) .100*) 0.01 200*) 0.01 —0.03 090*) 0.06 1.00MW:0.03 ‘
[Age 0.02 0.01 -0.05 0.04 0.01 -0.03 -.290*) .100*) -0.03 1.00

 

 

 

 

 

 

 

 

 

 

 

 

“. Correlation is signiﬁcant at the 0.01 level (2-tailed).

*. Correlation is significant at the 0.05 level (2-tailed).

DV: behavioral intention; ORI: outcome-relevant involvement; Threat: perceived threat;
Virus art: virus article treatment condition; Cancer art: cancer article treatment
condition; SS: sensation seeking; Effie: perceived efficacy; News use: frequency of
news consumption; Age: respondent age

Variance inﬂation factor and tolerance results were examined for other undesirable
relationships among variables. A series of hierarchical regressions were then used to test
the hypotheses shown graphically in the research model originally presented in Figure l.
Hierarchical regression was selected because of the desire to control for certain variables
and also to gain additional data concerning variables of interest (Schafer, 1991). This
type of regression allows for an understanding of the incremental predictability and the
variance explained at each step.8 In addition, hierarchical regression allows for distinct
treatment of time-bound variables that played roles at differing points in the data

collection (Schafer, I991). The consideration of time is relevant here because this project

 

8 As Schafer (1991) highlights, it is essential to keep in mind the limitations of hierarchical regression as
well. While its output can offer insight into variables’ relationships through incremental predictability and
variance explained, hierarchical regression does not make statements about causality that might be implied
by those terms. Causal interpretation necessitates an examination of the research design along with the
results.

-99-

focused on measuring several psychological constructs before inducing reactions to
treatment messages and in turn measuring other variables.

In the hierarchical analysis, several variables were employed as dependent variables
to match the appropriate hypotheses, moving right to left across the research model
presented in Figure 1. The initial test used behavioral intention as the dependent variable
to examine the independent variables’ relationships with the ﬁnal outcome variable. This
tests hypotheses one and three predicting the ability of perceived efﬁcacy and outcome-
relevant involvement to act as predictors of positive behavioral intentions.

The second regressions continued moving backward through the model and used
outcome-relevant involvement as the dependent variable in order to examine hypothesis
two stating that increased perceived threat will associate positively with increased
outcome-relevant involvement.

The ﬁnal portion of the analysis consisted of two regressions with perceived risk as
the dependent variable. This section used two regressions because it was examining a pair
of predicted moderator relationships as stated in hypotheses four and ﬁve. These
hypotheses predicted that sensation seeking would negatively moderate the relationship
between the message condition and perceived threat while need for cognition would
positively moderate the same relationship between message condition and threat. The
individual analyses also allowed for more detailed information. Each regression will be
discussed in more details as the results are reported.

The ﬁrst step in each regression consisted of entering the demographic information
and news consumption habits as control variables. Entering these variables into the

regression equation allowed for easier examination of their inﬂuence on the dependent

-100-

variables in the context of the other variables’ relationships to the dependent variable.
The second and third blocks in the behavioral intention regression contained the two
variables predicted to have relationships with behavioral intention—outcome-relevant
involvement and perceived efficacy. Block number four consisted of all the other
variables relevant elsewhere in the project, creating, in effect, a saturated model. F—test
results of this regression were signiﬁcant for all four steps and included notable R2
changes and signiﬁcant betas. Table 8 shows the results of ”the hierarchical regression
formatted according to suggestions of leading methods scholars (Field, 2005; Schafer,

1 99 l ). The R2 change, F statistic, and p-value are presented for each model.

 

Table 8. Hierarchical Regression Analysis for Behavioral Intention

 

 

 

 

 

 

 

Model DF R-Sq Change F P value
LL Demo 12 .17 17.83 <.001
M 13 .08 26.77 <.001
m 14 .14 48.34 <.001
4. Others 24 .15 52.04 <00"

 

 

 

 

13 (Constant), clinical trial experience, HPV experience, believability index sum, relationship status, ethnicity,
total riews sources consumed each week, gender, article comprehension index sum, age, food poisoning
ef‘Derience, breast cancer experience, prostate cancer experience

2: MOdel 1 plus Outcome-Relevant Involvement index sum

3: MOdel 2 plus Perceived Efﬁcacy Index sum

4: Model 3 plus the Need For Cognition by Sensation Seeking interaction term, the Need For Cognition by
amde subjects interaction terms, the Sensation Seeking by article subjects interaction terms, the Need For
CoQuition index sum, the Sensation Seeking index sum, Perceived Threat index sum, and article subjects

Table 9 presents more detailed information, showing the beta weights, standard errors,
constants, and standardized betas for each variable entered into the model. Note that the

table is spread across several pages because of its size.

-101-

 

Table 9. Hierarchical Regression Analysis for Behavioral Intention

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

_7 95% Conﬁdence l Collinearity
l mulnteﬁrvglfor B _ I -- Statistics ,3
Model Std. Lower 7 Upper 7Step R-sq7 l l
B Error 13 SiQ- Bound Bound Chﬂge Toler. VIF
[Step1 (Constant) 10.55 2.197 .000 6.24 14.86 .17

BLV sum .81 .058 .40 .000 .69 .92; ; .93 1.08
_ CMP sum -3.67 1.698) -.06 .03 {.01 _-3-51 719277.097
Gender .36 .622}? _.03 __362 __,._._:-_§§L 1.58m 7792' 1.09
Age .01 .024 91-321" __:.04 .05, 5 .89 1.12
ethncty .23 .199 .03," 353! _ _ ”:13“ 62 i .96 1.04
rltnshp stat -.08 .203 -.014_ .711 _--__-..:-4.7:_______-,:_32.3 _- _-- l __ .98 1.02
ttlnws src .45 .186 _:9Zl-.-r.°?. _ -- _93 3* -131].-- __ [ ___3534~1.05
HPV exper .10 .350 3.401% (.781 -.59 -.-:Z§é,-- _3 Mlﬁ__.82; 1.22
prstcncrexpr _.57 .31 .05__._124 -._13 31.03,; i- .74‘ 1.35
_ food pois expr .133 .27 03199 .435. .71? _l .78 1.23
brstcncrex_pr -.07 .25 -.01 .77? -.56 .437*___ .74 1.353
clintrl expr .14 .43 01 .751 -.71 .981 7 .98 1.02
Step2 (Constant) 5.81 2.14 .007 1.62 10.01l .08 _
BLV sum .73 .06 .36 .000 __--:52 .84l .91 1.103
CMP sum -3.18 1.62 -05 .05 6.353 .007 ___ _ 3.92 1.09
Gender .02 .59, “.90 “37.4 -1.147 141737 .911 1.101
Age .00 .027 .00 .91 -.04 95.1... __ #433113
ethncty .30; .19 .053 _11_9' -.08 __ .67: __ Law ,1“ 94
"WSW stat :17 _ __--:19- '-Q?-_-3!13. ___ :_-_5_5_.-_-_ .311 98: 1-03
ttl nws src .27 .. .13 W04 .135 -.08_ _M__ﬁ.62_; __ ; .95L 1.06
[HPV exper .18_L_W+3.3A3w._02+_._533 343*“ “.33; U; .82: 1.22
Iprstcncrexpr __ .30 _.29 __.93%__._3_03 -.2_7 83 -- ___;__.74 1.35
food pois expr _.20 .26 .02 .431 _.30 .70[___ _j_ .78 1.28
brstcncrexpr -j_.3 7.24 -.02 .577 -.60 .339 l “_74 1.35
_ Clintrl expr .197 _.41 .01 .642 -.61 .99. l ’ .98. 1.02
7 ORI sum .59 .06 .29 .000 .48 .70 _.96‘ 1.05

Step3 (Constant) -.03 1.96 .989 -3.88 333 .14 l
BLV sum .30 .06 .15 .000 .18 .413 . - 70l_1.343
CMP sum -2._74_3_3 1.46 -.05 .06 -5.60 .13 _ _ 7___32 139
Gender _.43 .54 -.02 .428 -1.48 .633_[__ 7 .917: 1.10[
Age -.02 .02 -.02 .44 -.06 024* __.- H.891 1.13
ethncty .39 217 .06 .24 .05l _72§ l- .96l 1.04
rltnshp stat -.32 .18 -.05 .07 7:99 ..-._- W_.02 :- 7; 7131797771303
ttlnws src .15 .16 .02 .343 _ _ :15l- __ ”.47; 3 l .941 1.06
HPV exper .20 .30 .02 .510 7 7:339; f -_ ”.7977 7728217122
_ prstcncrexpr .27 .26 .03 .314 m __125_|7_w __.7937‘ ' 7171.74 133
_;77" food pois expr .13 .23 .02 .587? 3:33;" “—_358L7 -_...: 7";778f7133
- brstcncrexpr =04 22 «01 -849-._-__ .461- _- 383. iii-iff-__.i."_"_.-.741L135
__ Clintrl expr .09 .37 .01 -§93l--.--_:-59l- --_ 3.82:1 -_ l _.98i3_1_.02
O_RI sum .57 .03 .28;W:FQOOI-MW.4:[MW#_.67[mm l W.96L1._0_5
EFF sum .53 .03 -441 900: .47 .60; l .73.: 1.36

 

 

 

 

 

 

 

 

 

-lO2-

Table 9 (cont’d).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

E—95% Conﬁdence E Collinearity
__ Lingual-imam- l_ - - --.- .--- “Statistics.-
Model Std. ' Lower l Upper 'Step R-sq l

B Error 13 519- Bound Bound Change Toler.l VIF

[Step4 (Constant) 523.94 2.754 I .000 29.35 48.54 .15 L -
BLV sum .27 .05 .13 .000 .17 .378; 962: 1.31:

CMP sum _-.26, 1.30 -00 .84 -2.81 2.29l ___ 7__.881.14
Gender .40' .47 .02 .395 -53. 1.3377 7 .89: 1.13

Age .03 .02 .031 15 -.01 .061 . .80 1.26
ethncty ___.,-_--_-:‘.2-. .15 .06 .50 .13, ____ .731 _ _ 7 74.951 305
rltnshp stat __-__.23j .15 -.03 .1431 -.53[. 7.39137; _ 95371.05:
ttlnws src .19; :14. .03 .19l __g‘.039;___4_’__.46l _ﬂ - | 3.911 1.10

HPV exper ”-.11j .26 4913,6733 __—63l .41; E -:.8.1__J-2.4-

prst cncr expr 125L___---23, .03 .280 TZQl. ”.10-; ____ -_ :___;.7.‘l_--1*36

food pois expr “.25 .20 .03 .217? __-._15: __ 395...- * _ 3 l -79-1:31.
brstcncrexpr -.06 _19. -.01 .759 -.4_3[ “7.3.1.17--- #I__74[1.36

clin til expr .26 .32 .02 .417 _§.3Z; .89 _._97_i 1.03

ORI sum .46 __105 .22 .000 .37[ __.55: _ -_-:-:§5E 1.18

EFF sum .33 .03 -2_7_-:°00_'_______._;?$.l--_-_---.:39; ____ “_ﬁ;_____._59,_ 1.69
NFCvirs 4.07 .52 -.08l .039l 2093 3.05.7 _ 2 3013.29

-_ mower -.38 .62. -.oz 9331-...4-59....___Té'3'lfiif_"‘_":i_7:-BJl1~96.
88mm -3.46 .597 -.18 .000[ -4.62_ -2.31;7 .45 2.25

SSvirs 4.11 .52 -.29 .0001 -5.14 3031?: _ .32 3.15
NFbeSS -.53 .20 -.06 .008 -.93_[___.___ __-_._14§ .95 1.06

RISK sum .19 .03 .19 .000 .13g .25; .. _3____[_ ____.14+_2._21

NFC sum .08 .02 .14 .002 .0_3[ _.12; __ - _.__H._214L4:23

ss sum .58 .07 3333.000 .45 .713 ‘1 .22[ 4.51

virus 4.86 __.35 .25 .000F 359'. 6.12; w $33253
cancer 5.48 .80 .24l .000 3.91l 7.05; .34E 2.95

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

-1

O3-

Table 10 focuses on the standardized beta weights and R-squared changes for each

step in the model and variable added.

Table 10. Summary of Hierarchical Regression Analysis for Variables

Predicting Behavioral Intentions in Reaction to a Fear Appeal Message

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Std. [3 Std. [3 Std. [3 Std. B A R2~
Step 1 Step 2 Step 3 Step 4
1.Demographics _ .17
Article believability index .40" .36" ‘__.15*f__ﬁ __ 13*: _ __
Article comprehension -.06" -.05* -.05 .00 ___3_
[Gender .02 .00 -.02 .02
Be .01 .00 -.02 .03
Ethnicity .03 .04 .06 .06 1
[Relationship status -.01 -.02 -.05 -_.03 __ l___ --
INews sources used in a week _ 07* .04 .02 .03 l
ﬂ-IPV experience --_ .01 .02 _.qzmmipi
Prostate cancer experience .05 .03 .03 .03 __4
food poisoning experience .02 .02 .02 .03
breast cancer experience -.01 -.02 -.01 -.01
cliniCal trial experience .01 .01 .01 .02
|2. Outcome involvement .08
putcome-Relevant Involvement index .29“ .28“ .22"
[3. Efficacy _ H...L---_-.Lp-. . l “.15
IPerceived Efficacy index .44» j .27» l
I4. Saturated model ' E .15
Negdjuolgggnition by virus interaction -.08* l ______ _
N_e<_a_d ‘Fgr Cognition by cancer interaction _- [__102 _ l __
Seqs_a_t_ion_ Seek _9 by cancer interaction -_ 7 A :13" :- __
Sensation Seeking by virus interaction .29“ __ _
[Need for C 3n by Sens. Seekinginteraction -.06*
Eerceived Risk *- -__._-.19_’.' _-.___ __ _
INeed For Cowtion __ __. - . ___ _ __-‘__.1_5j_3 _ _
39“

ISensation Seekirlg

 

iru§ message condition
Cancer message condition

 

 

 

 

 

* =<.05; **=<.001

The ﬁnal step for behavioral intention accounted for 53 percent of the variance

(adjusted R2) (F (24, 1065) = 52.04, p < .001, using the hierarchical method). In addition,

the Durbin-Watson test statistic for error multicollinearity returned a value of 2.02, a

result quite close to 2.00 that strongly supports the notion that the assumption of

-104-

map:
1525.

n. v. 1 ‘
Uhﬂk

73m

“1..

([15

)‘r'
7'7“,[

let;

independent errors has been met. The assumption of not having variable multicollinearity
is also upheld, according to the variance inﬂation factor and tolerance results. VIF
outcomes should be less than 10, and no number in this regression approaches that level.
Numbers greater than 10 indicate that the independent variables are highly correlated
with one another (Field, 2005 ; Gaur & Gaur, 2006). In this analysis, the only results
higher than two are for interaction terms that should have some level of collinearity by
deﬁnition, meaning all variables are well within acceptable VIF ranges. Tolerance terms,
which are the inverse of VIF terms, are also well within the acceptable range. Typically, a
value below 0.1 indicates a serious problem (Hinton, Brownlow, McMurray, & Cozens,
2004). The lowest number here for a non-interaction term is .62, and even the interaction
terms are greater than 0.1.

In this model, both outcome-relevant involvement and perceived efﬁcacy were
signiﬁcant predictors of behavioral intention, explaining 8 percent and 14 percent of the
variance, respectively. Other variables also showed signiﬁcant relationships with the
dependent variable. Particularly noteworthy were scores on the believability index,
perception of personal threat, sensation seeking, and treatment conditions. These
connections are discussed further in the following section.

Moving backward through the research model, regression number two employed
outcome-relevant involvement as the dependent variable. Again, step one used
demographic and personal information as control variables. Step two contained perceived
threat, the variable expected to have a direct relationship with outcome-relevant
involvement. The ﬁnal block of variables again included all other variables used in the

course of this research to measure every possible path to the dependent variable.

-105-

Each step of the hierarchical regression resulted in signiﬁcant R-squared changes, F-

tests, and beta weights. The R2 change, F statistics, and p value results of the regression

are presented in Table 11.

Table 11. Hierarchical Regression Analysis for Outcome-Relevant

Involvement

 

 

 

 

 

 

 

 

 

Model DF R-Sq Change F P value
1. Demo 12 .04 4.15 <.001
2. Risk 1 .05 9.00 <.001
3. Others 9 .06 8.71 <.001

 

 

1: (Constant), Clinical trial experience, HPV experience, believability index sum, relationship7sta'tus. ethnicity,
total news sources consumed each week, gender, article comprehension index sum, age, food poisoning
experience, breast cancer experience, prostate cancer experience
2: Model 1 plus Perceived Risk index sum
3: Model 2 plus the Need For Cognition by Sensation Seeking interaction term, the Need For Cognition by
article subjects interaction terms, the Sensation Seeking by article subjects interaction terms, the Need For
Cognition index sum, the Sensation Seeking Index sum, and article subjects

More detailed information about beta weights, standard errors, constants, and

standardized betas for each variable are in Tables 12 and 13.

Table 12. Hierarchical Regression Analysis for Outcome-Relevant

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Involvement
‘ E 95% Conﬁdence Collinearity
Interval for B Statisticsw
Model Std. Lower Upper .Step R-sq
Error B 319- Bound Bound l Change Toler., VIF
[Step1 (Constant) 8.01 1.14 7 .00 5.77 10.251404 __ ‘ w,
BLV sum .13 .03 .13 .00 .07 _19. “93,498,
CMP sum -95 .88 -.03 .34 -2.58 77493,, _ ,__._9_2_y1._09,
Gender .58 .32 .06 .07- -.06 1.211 __ _... .92 1.09
A98 .01- _.01 .02 .63 -02 .03 _ .89 1.12
ethncty -.12__, .10 -03 .26 -.32 .09,____ ,_ .96 1.04
rltnshp stat .16 .11 .05 .13, ____-__._05_5____m_ 37* .98 1.9;
tt'nwssrc .30 .10 .10 .001 ..-L-..--J1 L49‘ .25 .10_5
HPVexper -.14 .18 -.03 45.3-- Lssgl L 22 - lee--132
prstcncrexpr .29 .16 .06 .07; _ ”3021 M ,6ng -- 253 135
food pois expr -03 .14 -.01 .81_,__,__,__ 4317,, _ -347--. i__-_ 19,138
brst cncr expr .11 .13 .03 .42 -.15 36:, l .74 __L1-35
clintrl expr -.09 .22 -.01 .69 -.53 .35l l .98 1.02

 

 

 

 

 

 

 

 

 

 

 

-106-

Table -12 (cont’d).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

95% Confidence l Collinearity

' 5 Interval for B J _Stgtisjjgs _

Model Std. Lower Upper ISTep—Rgcmf V 1

B Error [3 319- Bound Bound lC Change Toler. VIF

Step2 (Constant) 5.56% 1.15 .00 3.30. 7.811“.

BLVsum L -95 .03 .06.. 98. -.o.1Ll.-L_ 111.. .;84..p_1_-19_

CMPsum ‘.-_._1._1.;_.L__.86:__-.o.o.rL..9-o.L -1_.80l __1_.58;_., l 91.; .110

Gender -.-6_3-l .31 .06 .05 _.01;__HW____1_.24';_ __gg _ .92 1.09

A96 .011 .01 .04 .253 . -.011_;_1_____1__ .0491,“ _ 89 _1. 13

ethnstv -.08l .10 -.02 .44 _.28l .12; l __ 9151;04

rltnshp stat .12; .10 .03 .25}, 1._._-._08§____132_ ______j__ 98l 193

ttlnws src .31 .09 .10 .001 .121 .49} __j,___._9511__1g.051

HPVexper -.16 .18 -.03 .381r =51l_____.___-_______-_1.8l__ _m 1,521.22

prstcncrexpr .30 .16 .05 'O-Gl~-—--—-----~'--'91l--------- .60,l _.j,_-__.7:1_L-1_._3§

food poisexpr _.02 .14 -.01 .39: -._291___1______g-_.253 L .78 128

brstcncrexpr .09 .13 .02 .48__ ____7.__151____ _.34_;___ ____l__ _____7_4. 135

cIin trt expr -.09 .22 -.01 .67 -.52 .33“ 9311.02

RISK sum .12 .02 .25 .00 .09 .15 901 1.12
Step3 (Constant) 1.34 1.81 .46 -2.21 4.89g -0§_ 1,

BLVsum .02 _,-..oaL-,.92LL.-s7, L :94- .oalL- .- 77l 1.39

CMP sum -.2_7_ .85 -.01 .751 -1.95 1.40l ___ _ j.” .88-'.-_1-_1_‘1.

Gender .66 .31 .06 .03 .09. 127i LL 29-41.2-

Age .03; .01,_ __.06 .051 .001 .051 _' __.-89L _1_ .21

sthnstv moor .10 -.02 .571 L. ___-.25 ___-__ul _;L__-..,9§L;o.s.

rltnshp stat .1_2_ .__“____._1o .03 .241 .08 .32; M g .96 1.05

ttlnws src .27! .09 1.091___.00j ”.093 __ .45!,______1l ___.92 1.09

HPV exper -.19j .17 —.03 .28__ 3.53; .15; __ﬁ _ .8175124

prstcncrexpr .24 .15 _,05__ .121 1_-_.106,___ .53; M M}ﬂ_g.74f1.36

-. food pois expr .04, .13, .01 .781 _-.23 , .30; 1_,:_______._7_§1H1.31

__ brstcncrexpr .05 .12l .01 .57, “3.191 .29 __ i _--7_-4LL1.35

clin 111 exor -.05 .21 -.01 .811 -_.4_75 .36; _ 1 .911103

RISK sum .18 1.02 .37 .00; .14 .221 1___.481___2.Q8

NFCvirs -1.11 .34 -.17 .00' -1.77 -.4___l___h_ .31, 3.24

NFCcncr -55 .40 -.05 .17 -1 .351 .24l .tL-éj LL96.

SScncr .00 .39 .00 1.00 17.6?- ___-1.6!.-. 45,223

SSvirs -_.10 .34 -.02 .77 -.__78..__--- _._57_', .32:r 3.12

NFbeSS .07 .13 .02 .91 2-2.21.9. _ _ 1331, ____ __.955 105

NFC sum .05; .02 .1§1‘_,00_1___________.021 __~_ _.08:__ ___“, 241__422

5351"" 9.5" .04__L.LosLL.29 ..LL_-.0.4. -131... LL"..- 22.1 446

virus -271 .40 L2g,ﬁ.00,ﬂ__-_3.48 -1931, 4_.45 2.22

cancer -1.87 .52 -.17 .00 -2.88 -.85j l 3.51 2.84

 

-107-

 

Table 13. Summary of Hierarchical Regression Analysis for Variables
Predicting Outcome-Relevant Involvement in Reaction to a Fear Appeal

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Message
' Std. B [ Std. B 1 Std. B A R,
Step 1 ' Step 2 l Step 3
1.0emographics ' v ' .04
rticle believability index .13“ .05 .02
Article comprehension -.03 .00 -'_.01_ _ _ __
Gender .06 .06" .0611 _ _
Ege __W __ .02 .04 ”1.0611 - 4 _
Ethnicity - _ __ _ ______________ -. -.03 -.02 -._02W _
[Relationship status ___ __ .05 .03 .031_____ W
News sources used in a week .10* .101w _ .091 .
IHPV experience -.03 -.03_:W-.03 '
IProstate cancer experience .06 .06 .05
Ifood poisoning experience W_____ W-_._0_1_1_M__+_~_-.01____W__W .0_1____’_ __ __
breast cancer experience .03 .02 ﬂ_.01_
laiiTical trial experience -.01 -.01 -.01
|2. Perceived Risk - .05
IPerceived Risk index .25" .37“
I3. Saturated model .95___
INeed for Cogﬂtion by virus interaction -.17:_WW__ W
[Need for Cognition by cancer interaction . LL .- _. 05 _
|_SWen_S§tion Seeking by cancer interaction __ .1 , - ___.00W; W _.
Sensation Seekirlgby virus interaction __ _-.02 _ _
INeed for_C_ogr_i_.__by Sens. Seekinginteraction _. __ .. W 1-..”.-- _
Need for Cognition . W __1_____.187*_ 4 _
[Sensation Seeking WW_W__ . E .08 _
[Wus mess_a_ge condition 229:1
[Cancer message condition 1 -.17"

 

 

 

° =p<.05;**=p<.001

The ﬁnal block of variables accounted for a statistically signiﬁcant 14 percent of the
variance (F (22, 1067) = 8.714, p < .001, using the hierarchical regression method) in
outcome-relevant involvement, according to the R2 value. The Durbin—Watson test
statistic for error multicollinearity reported a value of 1.931, again an outcome close to
two that strongly supports that the assumption of independent errors has been met.

Variable multicollinearity was again examined using the variance inﬂation factor and

-108-

tolerance terms. Like the prior regression model, the results here failed to approach the
acceptable limits of 10 and 0.1, respectively.

The predicted relationship between perceived threat and outcome-relevant
involvement was signiﬁcant in this model, with threat accounting for 5 percent of the
variance in outcome-relevant involvement. Other variables such as news consumption
habits and the treatment conditions were also worth noting.

The last regressions in the series involved the predictors for perceived threat. This
portion of the analysis took place in two parts. Part one was hierarchical regression to test
the prediction that need for cognition would moderate the relationship between message
exposure and perceived threat. Part two performed the same function for sensation
seeking. Per the Baron and Kenny deﬁnition of a moderator (1986), these regressions
served to examine the predictions that need for cognition and sensation seeking would
effect the strength of the relation between message condition as an independent variable
and perceived threat as a dependent variable.

Both models used the control variables as the ﬁrst step, followed by the message
conditions and the continuous variable of interest, either need for cognition or sensation
seeking. The ﬁnal step in both regressions was the interaction terms between the variable
of interest and the message conditions. These variables also continue the pattern of
making the ﬁnal step into a saturated model to examine every possible path to the
dependent variable.

For the need for cognition regression, each step produced signiﬁcant R-squared

changes, F-scores, and beta weights. These results are presented in Table 14.

-109-

Table 14. Hierarchical Rengession Analysis for Need for Cogﬁtion Interaction

 

 

 

 

 

 

 

 

 

Model DF R-Sq Chang F P value
1. Demogr. 12 .11 10.57 <.001
2. Direct 15 .40 73.69 <.001
3. Interact. 17 .01 67.59 <.001

 

 

1: (Constant), clinical trial experience, HPV experience, believability index sum, relationship status, ethnicity.
total news sources consumed each week, gender, article comprehension index sum. age. food poisoning
experience, breast cancer experience, prostate cancer experience
2: Model 1 plus article subjects and the Need For Cognition index sum
3: Model 2 plus the Need For Cognition by article subjects interaction terms

Dependent Variable: Risk index sum

More detailed information about beta weights, standard errors, constants, and

standardized betas for each variable are available in Tables 15 and 16. The ﬁnal step in

this regression accounted for a statistically signiﬁcant 52 percent of the variance in

perceived risk (F (17, 1072) = 67.59, p < .001, using hierarchical regression). Like in

prior analyses, various tests of multicollinearity provide evidence that appropriate

assumptions were met.

Table 15. Hierarchical Regression Analysis for Need For Cognition Interaction

with Treatment Condition on Perceived Threat

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I 95% Conﬁdence l Collinearity
Interval for 8 ‘im Statistics
Model Std. Lower Upper Step R-sq
8 Error 13 Sig. Bound Bound i Change Toler. VlF
Step1 (Constant) 20.52 2.261 .00 16.09 24.951 .11 ..
BLV sum .64 .06T .32 .00 .521 .75 .93 1.08
CMP sum -6.13 1.75 -.11 .oo -9.56-*' 271?"— 9221719“
Gender -.41 .64 -.02 .52: -1.66 811* ‘7“ i .92‘7T09
Age -.06 .02 -.08 of _.11 -.027' 1 .89 1.12“
ethncty -.31 .21 -.05 .131 ”72 "T .09} "TM 177987.61
rltnshp stat "T33 .21 .05 .121 -.08 .73 “[7798 1.02
ttlnws src -.03 .19 .00 .881“ 240; .3'5‘ 7 .95 1.05
HPV exper .20 .36 .02 .57; -50? .91i i .82 1.22
prstcncr expr -.05 .32 -.01 .87i :67"? .571 1“" .74‘ 1.35
food pois expr -.12 .28 -.01 .677 4.66 ’ 4.3; ” i .78 1.28
brstcncr expr .13 .26 .02 .620 *1“—'¥.38‘““ 63iW 'W ' J: .747: 135
clintrl expr .03 .44 .oo .95 -.841 .90: 1 98 i 102

 

 

 

 

-110-

Table '15 (cont’d).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

V95% Confidence Collinearity
Interval for B Statistics
Model Std. '“Eo’wker 7W_Upper Step R-sqw _____ 7
B Error 13 SiQ- Bound Bound Change Toler. VIF
Istepz (Constant) 8.23 1.88 .00 4.54 11.93 .40

BLV sum .57 .05 .28 .00 .48 .66 .91 1.1.0
CMP sum -3.63 1.31 -.06 .01 -6.20 4.06», .90 1.11
Gender .30 .48 .01 .52 -.63 1.24 i 1 " " ‘ "T911171 6
Age -.04 .02 —.06 .01 -.08 01+ 7 "I89““118
ethncty _.22 .15 -.03 .15.” ' -.521 ’ .08‘} ‘ .96 “1104‘
rltnshp stat .38 .16 .05 .02 ”‘.’c17"”""'”’“1.891w .97 11313”
ttlnws src .10 .15 .02 .47 -.18 .39? .92 1.09
HPV exper .19 .27 .02 .48 -.34 .71 .82 1.22
prstcncr expr -.08 .24 -.01 .75 -.54 .39. .74 1.35
food pois expr .03 .21 .00 .90 -.38 .431 , .78 1.29
brstcncr expr .20 .19 .03 .29 -.17 .56 A .74 1.35
clintrl expr .16 .33 .01 .62 -.48 .81 .98 1.03
Virus 10.78 .52 .56 .00 9.77 11.80 .64 1.57
Cancer 17.45 .60 .78 .00 16.28 18.62 .64 1.56
mm ‘ -.01 .01 -.01 .61 -.03 .02 .891 1.12
Step; (Constant) 15.82 2.54 .00 10.84 20.801 .01 1%.?“
BLV sum .56 .05 .28 .00 .47 .65 .90 1.11
CMP sum -3.66 1.30 -.06 .01 -6.21 "11.111“71_ “C90 1.11
Gender .29 .47 .01 .541 -64 1.21 " ‘ .91 1.10
Age -.05 .02 -.06 .01 -.08 -.01 T 1 C89 "1.18
ethncty -.21 .15 -.03 .16 -.51 .09” .96 1.64
rltnshp stat .39 .15 .05 .12 .09 .69 .97 1.04
ttlnws src .10 .14 .02 .50 -.18 .38 . .92 1.09
HPV exper .24 .27 .02 .37: -.29# .76 ’i .82 1.23
prstcncrexpr -07 .23 -.01 .77 -.531 .39." m' l .74 1.35
food pois expr “505 .21 —.01 .82 -.45!'“"”“ "13'6f‘ 17,1 "5771181)”
brstcncrexpr .22 .19 .03 .25 ”-115 .59.r ' ”11111741185
clintn expr .13 .33 .01 .765" -.51 .76" .981 1.68
virus 10.75 .51 .56 .00 9.74 11.75 .64 1.57
cancer 17.41 .59 .78 .00 16.25 18.57 .64 1.56
NFC sum -.09 .02 -.17 .00 -.14 -.05 .26 3.85
NFCvirs 2.69 .60 .13 .00 1.52 3.87 .55 1.81
NFCcncr 1.87 .50 .141 .00 .88 2.86 .331 3.00

 

 

 

 

 

 

 

 

 

 

 

 

-111-

—Il_u —. - nu.-
1hl~Vle>FﬂVr

tHi. rum

I:

at ".1

3‘

Table 16. Summary of Hierarchical Regression Analysis for Need for Cognition

Interaction with Treatment Condition on Perceived Threat

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Std.‘p Std. 6 Std. 13 A R2
Step 1 Step 2 Step 3
1.Demographics _ _» ___ _' _. _ 41-.
Article believability index _ _ __32jf_______._28ff_~ ___-23: 2-.
Article comprehension _ -- ___ _ -.11** -.06* r— -.06* ‘
|Gender _ -.02 .01 .01 -11.”- -
[Age 308* -.06* -_._06*
Ethnicity -.05 __ _-.__03 ___-.03
[Belationsmp status g .05 .05 .05 .
News sources used in a week .00 .02 _.02__ bl,”__-__-
[HPV experience .02 __.02 - .02 _. -l _____
[Prostate cancer experience -.01 -.01 7.01 -1 ._
Ifood poisoning experience H g -.01 .00 :01 #
breast cancer experience .02 .03 .03
[clinical trial experience .00 .01 .01
Fﬂiﬂect relationship}. 2.--...” .40
igus_article condition _ ._-§5_“ .56“
Cancer article condition .78“ .78“
Need For Cognition index -.01 -.17**
I3. Need For Cognition/article interaction terms __ .01W
[Need For Cognition by virus interaction .13“
[Need For Cognition by cancer interaction .14“

 

*=p<.05;**=p<.001

The proposed moderating relationship for need for cognition between message
exposure and perceived threat was signiﬁcant, an outcome that will be discussed further
in the hypothesis-testing portion of this paper. Other, non-predicted relationships also
were signiﬁcant. These will be discussed at the end of the Results and Discussion section.

A hierarchical regression focused on sensation seeking made up the last of the tests.
All three steps in this regression produced signiﬁcant R-squared changes, F-scores, and
beta weights. The results are presented in Table 17, and more detailed information about
beta weights, standard errors, constants, and standardized betas is available in Tables 18
and 19. The last step in this regression accounted for 51 percent of the variance in

perceived risk (F (17, 1072) = 66.08, p < .001).
- 112 -

Table 17. Hierarchical Regression Analysis for Sensation Seeking Interaction

with Treatment Condition on Perceived Threat

 

 

 

 

 

 

 

 

Model DF R-Sq Change F P value
1;DemchL. - 12 '11- 2 10.57 <.001
2. Direct 15 .40 73.71 <.001
3. Interact: 17 .004 66.08 <.001

 

 

 

1: (Constant), clinical trial experience, HPV experience, believability index sum, relationship status, ethnicity,
total news sources consumed each week, gender. article comprehension index sum, age, food poisoning

experience, breast cancer experience. prostate cancer experience

2: Model 1 plus article subjects and the Sensation Seeking index sum

3: Model 2 plus the Sensation Seeking by article subjects interaction terms
Dependent Variable: Risk index sum

Table 18. Hierarchical Regression Analysis for Sensation Seeking Interaction

with Treatment Condition on Perceived Threat

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

; 95% Conﬁdence Collinearity
Interval for Bmm __ Statistics
Model Std. Lower Upper Step R-sq
B Error [3 339- Bound Bound Change Toler. VIF
Step1 (Constant) 20.52 2.26 .00 16.09 24.95 .11
BLV sum .64 .06 .32“ .00,» .52 .75 .93 1.08
CMP sum -6.13 1.75 -.11** .00 -9.56 -2.71 .92 1.09
Gender 7,41 .64 -.02 .52 -1.66 .84 .92 1.09
A96 :06. .02 -.08" .01 -.11 -.02; ____-.§9‘ 1.12
ethnctv -.3_1__3W .21 -.05 .13, -.72, :09i .96 1.04
rltnshp stat ”.33! .21; .05 .121 -_.08lWW_ .731 _ 1. 1.6922103.
ttlnws src -03; .1_9j .00 .88 -_.40‘_______ .35. WW 1, ___-95.1495.
HPVexper .20! .36 .02 .57 -.501WW.9_1_E .-§_2__1-.2.2_
prstcncrexpr 3,05 .32l -.01 .87 _.67l, W571 . W .74 1.35
food pois expr 212$ .281 -.01 .67 -.66W .431“ _ ”j __ .78 1.28
brstcncrexpr .13 .26 .02 .62 -.38 .63 .74 1.35
clintn expr .03 .44 .00 .95 -.84 .90 .98 1.02
Step2 (Constant) 8.52 2.01 .00 4.57 12.47 .40 _
BLV sum .57 .05 W28" .00) _.48W .66 _ 311-1119
CMP sum 33.78; 1.31 -.07"i .00 k“16.35__WW-W1.21i W __ _.9,0:WW1.11,
Gender .27 .48 01’ .58‘WW.68_. W121 , A ___._99W111
A96 -.05 .02 -.06* .011 5091.-....“ --01.i.,-_____ - __.82' 1.22
ethncty -.23 .15 -.03 .14 5.531 ___,071 ,__ _j_ .96 1.04
ntnshp stat .38_' -16_i .05 .01 08;” ___.69ﬁ W; .97 1.03
ttlnws src .10 .14 .02 .50; -.18 .38__[_4_ ; .95 1.05
HPV exper .20 .27 .02 .45 -_.33 _73‘Wj __.81 1.23
9'5t cncr exP' --08 24 :01 .74 --54. ._ . :38! _-_ . - _ .. _-7.4___1;§5.
food pois expr .03 .21 ‘m.00_w_.90t__-._38l W__.43 WW __ 1- ___”.78d W129
brstcncrexpr 19+: .19; ’Q3i"'—’311”"“'-'18l‘—"‘“‘““"57i“"““"“ 1, .74' 1.35,
clintrl expr .16f .33:W_W.01W.63_?WW_-.49»WWW.80 __ ,97 _1-03
Virus 1977 .92.; .56:+..__.00,__.9.761 2.211.793 __ “$9641.21 .57
9999;“ 17.45 .601 .78" .001 16.28 18.62; L £54; 15.6..
ss sum -021 .03} -.021 .52? -.09j .051 j .861 1.16

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

-113-

Table 18 (cont’d).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

[ , 95% Conﬁdence T [ Collinearity
‘ _wlnterva|_f9r.t3.__ _ W -_‘_.’__Sta.tisti_cs _
Model Std. Lower Upper Step R-sq’ {
B Error B 359- rBound Bound Change Toler. VlF
Step3 (Constant) 11.75 2.27 .00 7.29. 16.20 .00
BLV sum .58 .05 .29" .00 .49 .67, .91 1.11
CMP sum -3.87 __1,._31 -.07* .00 -6.43 ,1 -_1_.3,1;_______WL_.9Q 1.11
Gender .23 .48 .01 .63 --7_1_i__._ 1.17L E .90 1.12
_ Age -.05 .02 -.06* .011 -.09|, 1.011 ______.,_______i_._.-.8_2-1122.
_ etnncty -21, _____ .15 -.03 .16} -51; .09_!____ 1-2951192
rttnshp stat .39, .16 .05 .01 :08)- ,_ .__.-.69;. _______ _.i__._-9_7;_1.-0§
ttlnws src .091 ._1_4 .01 .54 -.19[ .37‘1____l_____._9__5 1.06
HPV exper .22; .27, .02 .41 —.31l ___.753_ .8143:
prst cncr expr .061.I .24? -.01 .80 2521‘ ___,40 _ _ _ , - _. l, _, .7}: W135
food pois expr -.01 .21 .00 .96, -411 .391, j .77 1.29
brstcncrexpr .19 .1_9 .02 .33; -195 _.-56_._______ ___i___ .741 1.35
clintn expr .16 £3; ._01 .62; -.48 .81} ,l .97 1.03
virus 10.79 52 .56” .00 9.78 1180!--- .64 1.57
cancer 17.44 .59; .78" .001 16.27 _ 18.61,,Wm W__j___.,64WW1_56_
- SS sum -18, 062' —.12* .00;W31__3,_____W-.061W_W_,_,_.___}'+ . ,.25_,:t.07,
SSvirs 1.09: 57' 06:, .06[ -.04 --__..-.2:22L___.____ __ __ _.49 2.06
SScncr 1.60[ .51 .1111 .001 .59; 2.60‘, .35; 2.89

 

 

 

 

 

 

 

 

 

Table 19. Hierarchical Regression Analysis for Sensation Seeking Interaction
Terms on Perceived Threat

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Std. 8 Std. B ‘ Std. p A R;
Step1 Step2 Step 3

1.Demographics .11
Article believability index .32“ .28': __ 2.911-. __
Article comprehension -.11** -.07* T -.07* .
Gender ______ WW - -.02 .0,1,_,_jW__.91__ ; _ __
[Age -.08* -.06*__ . -.06* W
Ethniciy ___ __W _ -.05 -.03__ l, 9.03 _
[Relationship status .05 .05 [ .05, ,____
Figssgurces used in a week“, ,,,,,,,,,,,, .00 W__.02 j, .01, 1 _ - _
HPV experience .02 .02__ [W .02__ _ _,_
[Prostate cancer experience -.01 -.01__ , ,-_.01__ .-_
[food poisoning experience -.01 .00 [+0400 - _,_ _
[breast cancer experience .02 .03 W02 - ._. WW
[clinical trial experience .00 .01 .01
Eogect relationships _- __ __ _1 , WWW -19. _

trus article condition WWWW W __ __ __,____ __.56ff W), 56", __ _.
Cancer article condition 7811.-.: _ 78" ___ __ W
Sensation Seeking index -.02 . -.12*
3. Sensation Seeking/article interaction terms .00
Sensation Seekirlg by virus interaction , .06
Sensation Seekingby cancer interaction ‘ .11"

 

 

 

 

 

 

*=p<.05;**=p<.001

-114-

H u
DNA
5

‘.
‘VH‘V

'JJ
RAM;
5

Just as in the need for cognition analysis, almost all of the variance is due to the
message conditions. A quick glance at the relevant tables shows that for both sensation
seeking and need for cognition account, interactions account for only around one percent
of the variance in perceived risk. Initial consideration may lead one to conclude that
neither interaction term is of particular importance.

Such a conclusion would be hasty and out of context, however (Aiken & West,
1991). Thorough examination across social science ﬁelds has demonstrated that the
literature on moderator variables is consistent in reporting small effect sizes for
interaction terms (Chaplin, 1991, 1997). Reviewing three broad areas of psychology, in
fact, Chaplin found that observed effect sizes for interactions accounted for about one
percent of the variance of outcome variables (Chaplin, 1991). Making things more
challenging for researchers, to even detect effects of this size, large samples are a
necessity. Signiﬁcant sample sizes will help account for potential problems such as
measurement error (Aiken & West, 1991). Such experience and knowledge leads to the
possibility that the interaction effects are considerably more noteworthy than the
numerical result would indicate and that the two variables in question—sensation-seeking
and need for cognition—may be effective moderators of perceived threat. The results and
understanding of interaction terms in general certainly indicates that two here merit

further examination in other research projects.

Hypothesis ggsting
Control variables

All relationships in this analysis were tested with various demographic and media-use

questions as control variables. The items considered as control variables included typical

-115-

demographic items such as age, ethnicity, gender, and relationship status. Other variables
included here were personal or other experiences with the health topics covered by the
treatment articles, how believable respondents found the treatment articles, how well
people did on the comprehension index concerning the information contained in the
articles, and also individual news consumption habits. News consumption was deﬁned as
how many news sources a respondent used on daily basis over the course of each week.

Several control variables had signiﬁcant relationships with the dependent variables.
Only those relationships that were consistent across regression steps were noted,
however. Several variables were statistically signiﬁcant during a certain step but quite far
from signiﬁcance at others.

Scores on the believability index related positively to scores on the behavioral
intention and perceived threat indices. How believable an individual found the article
inﬂuenced how much the individual intended to alter his/her behavior such that the
behavioral intention score would increase a noteworthy amount for every standard
deviation that the believability score changed. The amount of change, however, varied
according to the other variables entered into the analysis, from .40 ([3 = .40, p < .001) in
the initial step to almost a third of that (B = .13, p < .001) in the ﬁnal step with all the
variables added. The regression focused on perceived threat as the dependent variable
showed the same statistically signiﬁcant relationship for believability (from B = .32, p <
.001 in step 1 to B = .28, p <.001 in step 3).

Comprehension of the articles—as measured by the four true/false items—also
inﬂuenced two of the dependent variables. Comprehension had a positive, statistically

signiﬁcant relationship with outcome-relevant involvement (B = .10, p = .002) across its

~116-

steps and a negative, statistically signiﬁcant relationship with perceived threat (B = -.l 1, p
< .001 in step 1 to B = > -O6, p = .005 in step 3).

Age produced similar relationships to comprehension when it came to outcome-
relevant involvement and perceived threat as dependent variables. Age had a small,
statistically signiﬁcant effect on outcome-relevant involvement in the ﬁnal step of that
regression (B = .06, p = .05) and a negative, statistically signiﬁcant relationship on
perceived threat (B = -O6, p = .01).

Outcome-relevant involvement also linked to a small, negative, statistically
signiﬁcant effect for gender (B = -.O6, p = .05) and a small, positive, statistically
signiﬁcant effect for increased news consumption (B = .10, p = .002). Other variables
displayed tenuous relationships with the dependent variables, appearing at certain steps or
changing considerably across steps.

For example, the variable comprehension check displays a statistically signiﬁcant
role as a predictor of behavioral intention (B = -.O6, p = .03) in step one of that regression,
but the relationship changes dramatically by the ﬁnal step of that examination (B = -.004,
p = .84). Believability performs similarly in the ﬁrst step of the outcome-relevant
involvement regression (B = .13, p < .001). As other variables entered future steps,
however, the relationship changes to become far from statistically signiﬁcant (B = .02, p
= .57).

Media use habits also had a statistically signiﬁcant relationship as a predictor of
behavioral intention (B = .07, p = .02) in the ﬁrst step of that regression. By step four of

that regression, media use becomes less of a predictor and leaves the range of statistical

~117-

 

signiﬁcance (B = .03, p = .19). Being that these relationships for the control variables
were not predicted, more analysis and future research will be necessary to explore them.

Hypotheses
H1: Higher perceived efﬁcacy will associate positively with an action-oriented

outcome.

H1 predicted that perceived efﬁcacy would be positively associated with behavioral
intention. Results show that efﬁcacy had a positive effect on behavioral intention (B =
.44, p < .001, AR2 = .14), as hypothesized. H1 was supported. For every standard
deviation increase in perceived efﬁcacy, behavioral intention will rise by .44 of a
standard deviation.

Just as prior research has reported, this outcome supports the idea that increasing a
person’s belief in his own abilities can be an effective means of persuasion. If an
individual perceives that a particular recommended behavior will be effective and that
she can complete this behavior, she’ll be more likely to undertake that behavior. A
woman will be more likely to follow a message promoting mammograms if she believes
in the effectiveness of those tests and in her ability to get them consistently.

H2: Higher perceived threat will associate positively with higher outcome-relevant
involvement.

H2 predicted that higher levels of perceived threat would be a predictor of increased
levels of outcome-relevant involvement. Perceived threat was a statistically signiﬁcant
predictor of outcome—relevant involvement (B = .25, p < .001, AR2 = .05 in the ﬁrst

entry). The standardized beta weight of .25 indicates that for each single standard

-118-

deviation change in perceived threat, outcome-relevant involvement will in turn change
.25 of a standard deviation. These data are consistent with the predictions of H2.

Practically speaking, this result supports the idea that as individuals feel threatened
by some particular topic, they will increasingly perceive the outcomes of the topic as
important. For example, the topic of diabetes could become more threatening to a person
as he recognizes risk factors in his life such as advancing age, weight gains, and some
blurred vision. As he perceives diabetes to be an increasing threat to his well-being, his
interest in the outcome of diabetes care will also increase.

H3: Higher outcome-relevant involvement will associate positively with higher
behavioral intention.

Based upon prior compliance-gaining research, H3 predicted that higher levels of
outcome-relevant involvement would lead to higher levels of behavioral intention. As
shown in the hierarchical regression Tables 8-10, outcome-relevant involvement was
signiﬁcantly related to positive change in behavioral intention (B = .29, p < .001, AR2 =
.08), indicating that outcome-relevant involvement is a useful predictor for behavioral
intention and explains eight percent of its variance. Thus, H3 was supported.

For health communication researchers and practitioners, this presents another avenue
to change individuals’ behavior. If people’s interest in the results of a situation leads to
more intent to change behaviors, message designers could produce more effective content
by engaging the audience and promoting a connection to the outcome of a health issue.
Pushing a personal connection to the conclusion of some health risk affords another

opportunity to encourage individuals to adopt healthy behaviors.

-119-

H4: Sensation seeking will moderate the relationship between message condition
and threat perception. Specifically, those high in sensation seeking will report
less perceived threat than those low in sensation seeking.

H4 predicted that the relationship between message condition and respondents’ threat
perceptions would be moderated negatively by sensation seeking. Those high in
sensation seeking will report less perceived threat than those low in sensation seeking and
vice versa when exposed to arousing messages. The regression offers evidence counter to
this hypothesis in the form of statistically signiﬁcant, positive R—squared change results
(F (17, 1072) = 66.08, p < .001, AR2 = .004) and signiﬁcant or near-signiﬁcant positive
outcomes for the message condition/sensation seeking interaction terms (sensation
seeking by cancer article interaction with perceived threat: B = .l l, p = .002; sensation
seeking by virus article interaction with perceived threat: B = .06, p = .06). Thus, H4 was
IL“ supported as the hypothesis predicts a negative moderation effect for sensation
seeking, and the data point to a positive inﬂuence.

This hypothesis predicted that people requiring considerable stimulation would be
less affected than average by a message intended to create the perception of a health
threat. The ﬁndings, though, failed to support this idea. One obvious reason why this
hypothesis may not have matched with the data is the effect of the sample used in this
study and its inﬂuence on the data. Individuals truly high in sensation seeking are known
to be risk-takers likely to be involved with extreme sports, substance abuse, and other
physically stimulating activities. High sensation-seekers are not expected to be
individuals involved with potentially less thrilling, mundane activities such as Internet

surﬁng; yet, such online time is obviously a prerequisite to completing the survey

-120-

instrument for this project. Furthermore, it is also possible that members of the online
research panel used to collect data are more likely than average to spend time on Internet
activities, potentially pushing them further from being individuals being truly high in
sensation seeking. This is not at all to say that World Wide Web usage and stimulating
experiences are mutually exclusive, only that the sedentary nature of most online
activities could be a limiting factor in the number of truly high-sensation-seeking
individuals.

Respondents may have portrayed themselves as high in sensation seeking when
completing those speciﬁc survey items. At the same time, since they were not truly high
in the need for extreme stimulation, they did respond to other items as a person truly high
in sensation seeking would. Perhaps they answered the questions about need for
stimulation in a certain way to make themselves appear more adventurous or exciting but
did not continue answering as a person needing much stimulation because, in fact, that
was not the case.

In other words, a respondent may have answered the sensation seeking questions in a
way that makes that person appear to be adventurous or a risk-taker. Such responses may
match well with that individual’s self-image. However, when questions come up
concerning actual reactions to presented experimental conditions, that same person may
revert to his/her more realistic, more modest view of risk-taking. The differences in
mind-set when answering the two surveys could prevent the expected relationship from
developing. At the outset, the respondent could be trying to portray an exciting self-
image. In response to the questions about the experiment, that same respondent could be

answering more realistically based upon the recent happening of the experimental

-121-

treatment—reading a newspaper article, in this case. It’s a conﬂict between a hypothetical

personal ideal and a real-life happening.

In addition, people legitimately high in sensation seeking are more likely to be
outdoors, away from their computers, competing in extreme sports or other strenuous
activities. Other high sensation-seekers report spending free time engaging in risky sexual
or substance-abuse behaviors, again leaving them less likely to remain calmly seated in
front of a computer taking a survey. These groups may of course also overlap.

Other research has also noted that individuals high in sensation seeking are difﬁcult
to reach through all research methodologies, regardless of the subject population. While
this particular project failed to ﬁnd results consistent with the prediction that sensation
seeking would reduce fear messages’ inﬂuence on perceived threat, other factors may
have limited the ability of this research to test the prediction. Future research may be able
to more effectively target individuals high in sensation seeking and ﬁnd more valid
results.

H5: Need for cognition will moderate the relationship between message condition
and threat perception. Specifically, those high in need for cognition will report
greater perceived threat than those low in need for cognition.

The ﬁfth hypothesis predicts that need for cognition will act as a moderator in the
relationship between message condition and threat perception. According to this
hypothesis, those high in need for cognition will report greater perceived threat than those
low in need for cognition and vice versa.

Data indicate that need for cognition does interact with the message condition’s

relationship with threat perception. The interaction terms of need for cognition accounted

-122-

for a one percent of the variance in perceived threat (F (17, 1072) = 67.59, p < .001, AR2
= .01), supporting H5. In addition, the standardized beta weights for both of the
interaction terms for need for cognition and message condition were statistically
signiﬁcant (B = .13, p < .001 for the virus article and need for cognition; B = .14, p < .001
for the cancer article and need for cognition). As discussed earlier, this R-squared change
may appear small but is, in fact, in line with expectations based upon prior social science
research (Aiken & West, 1991; Chaplin, 1991). Need for cognition served to moderate a
portion of the relationship between the fear appeal messages and perceived threat.

Results indicate that message designers would beneﬁt from considering the thinking
style of their audiences. Individuals likely to logically process information should be
approached with fear appeals differently than people unlikely to rationally consider some
message. A fear appeal presented to someone that likes to think deeply will not be
mentally processed in the same way as an identical message shown to someone not
interested in putting forth mental effort. When it is possible to customize messages along
the lines of thinking styles, campaign designers could beneﬁt from doing so. This
customization is most likely to be possible during interpersonal communication or during
situations using adaptable digital media that can adjust according to user responses.

Just as Figure l portrayed the research model, Figure 2 represents the standardized
beta weights according to the hypothesis testing results. Each path is statistically
signiﬁcant at the p < .001 level and provides the beta weights from a series of
hierarchical regression analyses. The values are not the product of structural equation

modeling or path analysis.

-123-

 

 

Regression Coefficients

Hypothesis Testing Model With Standardized

 

 

 

 

 

 

 

 

 

 

' Outcome-
|——> Relevant
-25** .lnvolvemen

 

 

 

 

 

 

 

 

 

     

29*

.44**

 

Need For
Cognition
Researcher- 14......
Produced °
Newspaper Perceived
Articles on ‘ Threat
Health-
related N/S
Topics
Sensation
Seeking
Perceived
Efﬁcacy

 

 

 

 

7 Behavioral
Intention

  

 

 

 

Figure 2. Hypothesis Testing Model With Standardized Regression Coefficients

** p<.OOl

Non-predicted relationships

The regression analyses also produced statistically signiﬁcant results for a number of

not-predicted relationships between variables in the predicted model. All of the values are

viewable in the previously mentioned tables of regression results. These relationships are

described here for the purpose of adding to future research. This project had no

theoretical reasons to expect these relationships, but ensuing scholarship may develop

theories that match with the outcomes.

Behavioral intention as a dependent variable had a noteworthy number of non-

hypothesized, signiﬁcant predictors. For example, individuals’ news consumption habits

(B = .07, p = .02) and perceived believability of the experiment’s news content (B = .40, p

_124-

< .001) both were signiﬁcant predictors of behavioral intention, with believability being a
particularly strong predictor. Thus, the more that readers accepted the news stories as
accurate, the more likely they were to adopt the behaviors recommended in the text. In
addition, the likelihood of taking on the suggested behavior was inﬂuenced by how often
individuals read newspapers, watched TV news, scanned the Internet for information, or
listened to the radio.

The dummy variables for the treatment article subject areas were signiﬁcant
predictors (B = .25, p < .001 for “virus” and B = .24, p < .001 for “cancer”). Several of the
variables from the initial stages of the research model were as well. Perceived threat (B =
.19, p < .001), need for cognition (B = .14, p < .001), and sensation seeking (B = .39, p <
.001) all had direct relationships with behavioral intention. Along with the believability
of the news content and how intensely a person followed the news, other unpredicted
factors affected plans to take on suggested behaviors. Simply reading the fabricated news
stories about viruses and cancers inﬂuenced individual intentions to undertake behaviors.
Perceiving a threat from a medical risk message also strongly predicted a person’s
designs on behavior change. In addition, intentions to modify personal behaviors were
inﬂuenced by a preference for thinking in a rational manner and processing factual
information (need for cognition). Somewhat contradictorily, a taste for remarkably
stimulating experiences (sensation seeking) additionally led to an increase in the
likelihood that an individual would report plans to follow the suggestions included in the
fabricated AP article.

Several of the interactions among these variables were statistically signiﬁcant

predictors of behavioral intention, too, all in a negative direction. Need for cognition’s

-125-

interaction with the cancer article message condition negatively related to behavioral
intention (B = -.O8, p = .04), as did the sensation seeking by cancer (B = -.l8, p < .001)
and virus interactions (B = -.29, p < .001). The need for cognition by sensation seeking
interaction also was a negative predictor of the dependent variable, albeit a small one (B =
-.O6, p = .01). Altogether, these non-predicted relationships accounted for 15 percent of
the variance (R-squared) in behavioral intention. Such ﬁndings offer additional
possibilities in the way that individual traits could inﬂuence people’s intentions to change
their personal behaviors. If a preference for rational thinking (need for cognition) and a
desire for signiﬁcant stimulation (sensation seeking) affect the strength of the relationship
between fear appeal messages and individual responses to those messages, campaign
designers need to account for such things when possible. An individual’s thought process
or personal preferences may partially determine the effectiveness of health
communication content, making them worth consideration.

Outcome-relevant involvement had its share of non-hypothesized predictor variables,
albeit not as many as behavioral intention. Respondent news consumption was a predictor
of outcome-relevant involvement (B = .10, p = .002), as were need for cognition (B = .18,
p = .003), the virus treatment condition (B = -.29, p < .001), and the cancer treatment
condition (B = -. l 7, p < .001). While need for cognition’s interaction with the cancer
article condition was a predictor of behavioral intention, the need for cognition/virus
article condition was a similar predictor of outcome-relevant involvement (B = -. l 7, p =
.001). The message conditions and the interactions, however, were all negative

predictors of interest in the outcomes of health matters.

-l26-

Like respondents’ plans to adopt suggest behaviors (behavioral intention), their
engagement with the outcome of a health issue (outcome-relevant involvement) was
inﬂuenced by a number of non-predicted factors. Once again, the more they accepted the
accuracy of news stories and the more often they followed the news, the more likely
respondents were to take an interest in the conclusion of health issue. In addition, a
preference for rational thinking (need for cognition) also led to an interest in health
outcomes, as did exposure to the articles on viruses and cancers. Evidence-based thinking
also inﬂuenced the strength of the relationship between exposure to the articles on viruses
and concern for health outcomes.

Two other variables that didn’t play a role in predicting behavioral intention were
statistically signiﬁcant predictors of outcome-relevant involvement. While not large
predictors, both gender (B = .06, p = .03) and age (B = .06, p = .05) were consistent across
steps of the hierarchical regression. The small effects of gender and age on interest in the
results of a health issue may indicate the use of considering these factors when
developing a fear appeal message. If different age and gender groups will react
differently, message designers could increase campaign effectiveness by considering
demographics.

The last dependent variable in the series of regressions, perceived threat, additionally
had several unexpected predictors. Believability was once again a signiﬁcant predictor (B
= .28, p < .001). Comprehension proved to have a small but statistically signiﬁcant
relationship with the dependent variable (B = -.O6, p = .01), and age did as well (B = -.O6,
P = .01). Age, in fact, had the opposite relationship with perceived threat that it did with

outcome—relevant involvement, acting as a small, negative predictor here and a small,

-127-

positive one in the prior analysis. Sensation seeking was also shown to be a small
negative predictor or perceived threat (B = -. 12, p = .01)

Just like for intention to adopt a behavior and an interest in health outcomes,
respondents’ perceptions of being threatened by a health issue were inﬂuenced by how
much they believed the presented newspaper article. How well they retained some
information from the article also affected their perceptions of threat, only its inﬂuence
was negative. Retaining more information from the article slightly worked against the
perception of threat. Age was also a negative predictor of perceived threat here, the
opposite relationship it had to feeling involved in the outcome of a health issue (outcome-
relevant involvement). A desire for heightened stimulation (sensation seeking) had a

similar effect, working against a message’s ability to create a threat.

-128-

Chapter 6. Summary and Conclusions

Health messages are consistently available to the public through a number of media.
Television news, print reporters, and Internet journalists, among others, supply the latest
medical information. The mass media often provide educational content and serve as
health educators, encouraging individuals to adopt more positive and pro-active health
behaviors.

In this role, mainstream media counteract a realm of negative health inﬂuences.
Certain product advertising, for example, promotes negative behaviors such as smoking
or frequent fast food consumption. In addition, lack of information or repetitious
misinformation encourages the continuance of damaging behaviors. When added to the
personal challenges inherent in altering lifestyle habits, negative advertising and the
challenge of ﬁnding accurate information makes positive change even more demanding
to maintain long term.

However, public health advocates can also employ mass media to overcome such
harmful inﬂuences. The ability of pro-social, mediated messages to carry out this public
health task can be improved by the theories and message-design strategies developed by
health communication researchers. Theoretically driven messages based upon prior
research have the greatest chance for success. Effective messages can promote positive
health behaviors and encourage individuals to make better healthcare decisions. Well-
designed messages can increase the likelihood of individuals adopting the recommended
behaviors. This dissertation research attempts to further health communication research

by contributing theoretically and practically to knowledge on the how individuals react to

-129-

health education messages. Comprehending responses to messages will in turn encourage
more effective design processes and message applications. To that end, this chapter will
separate into sections covering the theoretical and practical implications of this research,
following a brief review of the project’s goals. The two implication sections will also
address the goals noted below as appropriate.

The stated goals of the project covered three aspects of fear appeal research. First,
this dissertation tested the Extended Parallel Process Model in a communication context
different from prior uses. News articles were employed to test the theory in a different
message environment, rather than using the brochures, posters, or educational videos
employed in prior fear appeal experiments. The examination of this context was carried
out through the testing of the ﬁrst two hypotheses, restated here:

H1: Higher perceived efﬁcacy will associate positively with an action-oriented outcome.

H2: Higher perceived threat will associate positively with higher outcome-relevant
involvement.

Second, several variables from persuasion research were added to the Extended
Parallel Process Model and applied in a number of health messages. If effective, these
variables could increase the effectiveness of health news on behavioral intention to
undertake pro-active health behaviors. The additions were tested in the examination of
the hypotheses three to ﬁve, restated here:

H3: Higher outcome-relevant involvement will associate positively with higher
behavioral intention.

H4: Sensation seeking will moderate the relationship between message condition and

threat perception. Speciﬁcally, those high in sensation seeking will report less perceived
threat than those low in sensation seeking.

-130-

Hu‘
J, ,
ti.

'.
,i
A

H5: Need for cognition will moderate the relationship between message condition and
threat perception. Speciﬁcally, those high in need for cognition will report greater
perceived threat than those low in need for cognition.

The third goal was to reduce the scientiﬁc impact of less effective design and
measurement commonly observed in prior fear appeal research. Decreasing the effect of
errors can allow for more precise understanding of how individuals process and react to
health information. This goal was not tested directly through a speciﬁc hypothesis.
Rather, it was an essential part of the project’s development and implementation best
viewed through comparing results to prior research.

The outcomes of testing the hypotheses and improving experimental design will help
healthcare practitioners, public relations professionals, and journalists predict how people
are likely to respond to behavior-change messages. Understanding the results will aid in
the development of more effective messages and better predictions of behavioral
outcomes. The implications also extend into any other context that involves altering

behavior or attitudes through mediated persuasive messages.

Thgoretical Implications
Fear appeals in news content

The ﬁrst goal of this dissertation was to extend and test the Extended Parallel Process
Model in a news media context, opening possibilities for examining the rigor of the
EPPM. Prior research employed primarily brochures and posters as educational media,
with a few projects using videos. The use of news content either real or simulated is rare.
This is despite the fact that so much of the information individuals receive each day

comes through the channel of mainstream media. Testing the EPPM in this environment

- 131 -

\0 1i;

-.--
tQu‘

‘I
“\u‘

3m

’3"?
ihklur

".1

showed that its predictions and expectations were applicable even in a news setting and
that the context merits further research.

Such knowledge is made even more important when considering the volume of news
consumers and the importance of studying how they access information. Millions of
individuals each day depend on mainstream media such as newspapers and magazines to
provide updates on the latest in health research. Consumers also turn to media for
recommendations derived from the most recent work of scientists and government
agencies. The potential inﬂuence of effective, mediated persuasive messages is

staggering and relatively unexamined.

Additional persuasion variables in the Extended Parallel
Process Model

The second goal of the dissertation was to add to the theoretical model of the EPPM
and examine the potential inﬂuence of several variables from the persuasion literature.
Speciﬁcally, need for cognition was predicted to positively moderate the relationship
between message conditions and perceived threat, and sensation seeking was expected to
moderate the same relationship negatively. Perceived threat was then predicted to
positively affect outcome-relevant involvement. In turn, outcome-relevant involvement
was expected to positively inﬂuence behavioral intention, the ultimate dependent
variable. Perceived efﬁcacy was also predicted to positively inﬂuence behavioral
intention.

Five newspaper articles were developed according to the requirements of the EPPM,
each with an element intended to promote perceived threat and perceived efﬁcacy. The

ﬁve articles were collapsed into three groups because of similarities in content. This

-l32-

design still ensured the existence of multiple message categories. Using multiple
messages increased the validity of the study because signiﬁcant ﬁndings due to the
individual article topic were less likely to be a plausible explanation for the outcomes.
The predicted relationships consistent with the EPPM and the additional variables were
supported by the data in all but one case, the interaction of sensation seeking and message
condition.

Increased level of perceived threat was a signiﬁcant predictor of outcome-relevant
involvement, which in turn was a signiﬁcant predictor of behavioral intention. In
addition, perceived efﬁcacy was a solid predictor of behavioral intention, accounting for
14 percent of the variance. Two moderators were also predicted in the hypotheses. Need
for cognition and sensation seeking were expected to moderate the relationship between
the message conditions and perceived threat. Need for cognition produced statistically
signiﬁcant interaction effects. Although the effects are not numerically large, they are in
line with the realistic expectations of moderator effects found in other social science
research.

Sensation seeking had statistically signiﬁcant effects for the highest-threat message
condition and near-signiﬁcant effects for the moderate condition, but both of these
outcomes were in the opposite direction predicted. A negative relationship was predicted,
but positive relationships were produced in this portion of the analysis.

Individual desire for stimulating experiences did not inﬂuence thinking about the
experimental newspaper articles in the way expected. The initial prediction was that
people driven by an interest in exciting events would be less inﬂuenced by fear appeals.

The results, however, showed the opposite. Those craving invigorating happenings

-133-

actually found the newspapers to be more threatening than those not showing as much
interest in stimulating experiences.

Design and measurement issues

The third goal was to improve the level of measurement and experimental design
from prior fear appeal research. The improvements happened in two ways. The ﬁrst
development involved carefully adapting and pilot testing previously used instruments to
achieve higher Cronbach’s alpha reliability scores. Reduced alpha scores have been
common to a number of instruments used in this research. Perceived severity,
susceptibility, self-efﬁcacy, and response efﬁcacy all have typically reported alphas in
the high .73, for example. While this may be a commonly acceptable level, higher alpha
reliability improves measurement and thus the accuracy of statistical results. The
measurement instruments in this project improved their Cronbach alpha reliabilities
almost .09 over their normal results to .88 on average. This includes one new index—
Park et al.’s outcome-relevant involvement index—that actually went down from its prior
use and pilot testing for this project. Without that one instrument, the others increased an
average of .10.

The second area of intended improvement comes from having multiple messages
capable of inducing multiple levels of perceived threat, a suggestion put forth in Boster
and Mongeau’s meta-analysis (1984). In their work, studies with more levels of fear
manipulation reported stronger relationships between perceived threat and the dependent
variable. Despite this observation, few projects have moved beyond employing low- and
high-fear conditions, meaning the relationships they are reporting may not actually

representative of the real relationships. In this dissertation, for example, the perceived

-134-

thrai'.
big but.

like

'
4‘
SIUU“

EXI;

threat-behavioral intention correlation is .51 (p < .01). The comparable correlations in the
Witte and Allen meta-analysis (the only one computing these relationships) came out to
.11.

Along similar lines, the perceived efﬁcacy-behavioral intention correlation in this
study was signiﬁcantly greater than in prior research. Witte and Allen (2000) report
efﬁcacy correlations with the common dependent variables as being between .12 and .17.
Here, the perceived efﬁcacy correlation with the dependent variable'is .52 (p < .01). This
further supports the previously found strong relationships between perceived efﬁcacy and
the common fear appeal dependent variables. While one can ﬁnd much disputing
research about the efﬁcacy of fear appeals and dependent variables, most research

endorses the relationship between efﬁcacy and adherence to suggested outcomes.

Practical Implications
Fear appeals in news content

As discussed prior, the ﬁrst area of emphasis for this project was to examine the
Extended Parallel Process Model when employed in news content to look for ways of
improving mediated health communication efforts and public health campaigns.

Health message designers would be well advised to consider how mainstream media
treat their messages. Pro-social organizations can only rarely deliver their own messages
unﬁltered, making it essential to consider how the raw materials—often press releases—
can be structured to inﬂuence news content in a way that promotes effective message
delivery to the public. Organizations such as the Centers for Disease Control or the
American Cancer Society stand no chance of ever matching the promotion budgets of

those Promoting unhealthy personal habits. As such, pro-social groups are forced to meet

-135—

the same goals as proﬁt-driven, business-sponsored campaigns by taking advantage of the
news media’s desire and self-proclaimed requirement to inform, but this means also
dealing with the media’s editing system. A well tailored press release or video news
release could work within this editing process by offering content that ﬁts messages
appropriate to the EPPM and thus promotes socially positive outcomes.

For example, commercial spending to spread corporate messages about food reached
$7 billion in 1997 (Gallo, 2005). At the other end of the spectrum, the United States
Department of Agriculture spent only $333.3 million to deliver messages on nutrition.
The agriculture department total equals just over half of what the soft drink industry
alone paid to promote carbonated drinks and less than half of expenditures on beer,
candy, or breakfast cereals. Using predictive models to guide message design can aid
non-proﬁts in being more efﬁcient and effective with their limited dollars.

An illustrative example is the $1.03 billion dollars McDonald’s alone spends yearly
on advertising to persuade the public to spend its money at their restaurants (Finnegan,
Viswanath, & Hertog, 1999). The company is quite successful in their attempts to
inﬂuence the public, as each dollar spent on advertising results in $5 in sales. In contrast,
the National Cancer Institute spent about $1 million annually on communications
throughout the ﬁrst 10 years of the 5-A-Day Program for better nutrition (National
Cancer Institute, 2005). While a signiﬁcant sum of money, it in no way matches the
competition and at least partially illustrates the scale of the inﬂuence battle required to
combat the problem of less healthful eating in the United States. Such is the environment
in which pro-social groups are trying to persuade Americans to undertake better lifestyle
habits.

-l36-

Public relations campaigns targeted at reporters can garner millions of dollars worth
of free media exposure to work on the spending gap. This is a tactic frequently seen as
pro-social groups hold charity events or telethons and local media deem these happenings
newsworthy. Newspapers and television stations devote signiﬁcant space to reporting on
educational community events.

However, solely establishing public relations campaigns and garnering media
coverage is of little value if the message being received by the public is an ineffective one
that fails to drive attitude or behavior change. The press releases and educational
materials constructed for reporters must focus on essential elements of an effective
message. In the context of the EPPM, this means each press release would need'a
perceived threat and a stronger perceived efﬁcacy element. Based upon this dissertation
research, P.R. writers would be wise to also add an outcome-relevant involvement
element and consider other traits that may inﬂuence individual reactions to health
education messages.

News stories concerning health topics would then be increasingly likely to be
effective as positively inﬂuencing public health. Web sites for pro-social groups could
follow the same guidelines, ensuring that their content is structured for maximum
effectiveness. Potentially, these Web sites could also incorporate some of the ﬁndings
from earlier in this dissertation relating to the signiﬁcance of individual-level traits. More
precise structure in press releases and other press materials in combination with similar
messages on counterpart Web sites could help pro-social organizations gain an optimal
level of beneﬁts from their unpaid media exposure and counter some of the constant, paid

media exposure used by corporations.

-l37-

Additional persuasion variables in the Extended Parallel
Process Model

The persuasion-variable additions to the EPPM also carry implications for the
professional world. The ﬁndings of this study offer some suggestions for future message
design, pending conﬁrmation from other research. Knowing that individual-level
variables such as need for cognition can inﬂuence reactions to educational messages
suggests a different method of delivering health information. Digital media, in particular,
present opportunities to tailor educational content on an individual level. Educational
Web sites currently in use in medical contexts, for example, already pose demographic
questions to learn about users’ age, health status, and medical history before providing
health information. These sites could also ask a few questions about personal traits such
as need for cognition and adjust the educational content to more effectively target that
individual. As digital media become omnipresent, this ability to tailor information will
become both more available and more effective. These ﬁndings suggest another way to
take advantage of digital media options certain to grow to be the standard in mediated
communication.

Other variables that demonstrated direct relationships with behavioral intention also
present lessons useful in message design. Outcome-relevant involvement’s position as a
predictor of behavioral intention indicates that it merits consideration when designing
persuasive health messages. Further examination in fear appeal research will be necessary
to conﬁrm and reﬁne the relationship between involvement and behavioral intention, but
results from this study indicate that outcome-relevant involvement was a stronger

predictor of behavioral intention than perceived threat. Should this relationship hold in

-l38-

other projects, it would indicate that persuasive message designers would be well served
to consider promoting outcome-relevant involvement as a means of gaining message
compliance. In addition, public relations practitioners responsible for educating media
professionals could include involvement messages in their press materials, as opposed to
current campaign messages that often focus exclusively on fear.

These ﬁndings are important because they may offer other avenues to gaining
compliance for public health messages. Stressing the relevance of one’s involvement in
the outcome of a health topic also removes some of the ethical questions presented by
high-threat messages. Relevance-centered messages may additionally help to reach
segments of the population impervious or overexposed to fear appeals.

Design and measurement issues

This study’s efforts to improve the methodology surrounding fear appeal research
also carry practical implications. Understanding the effects of more rigorous
methodology can aid other communicators in understanding how to more accurately
employ tactics supported by fear appeal research. Better knowledge about the
effectiveness and limits of fear appeal messages can aid health communicators in their
message-design efforts by allowing them to more effectively predict outcomes and use
limited resources to test messages. The more that health communicators understand about
message design, the more effective campaigns will be and the more efﬁciently
communicators can use available resources in designing and evaluating promotions. The
results from these studies also help to more thoroughly grasp the impact of certain

variables on dependent measures such as behavioral intention.

-l39-

 

(of?

.\ mil

1110?:

{nit-O
litre

i
ltha

Ppst-hoc findings

Other ﬁndings are also worth considering for future research and applications.
Several post-hoe tests found non-predicted relationships that merit mentioning because of
their potential use in health communication campaigns. Individual intentions to undertake
the suggested health behaviors were predicted strongly by how much people believed in
the provided news content and also by how frequently individuals made use of news
sources. Those considering the news content to be believable reported being signiﬁcantly
more likely to adopt the recommended behavior. This ﬁnding serves to underscore the
importance of developing long-term rapport with consumers that focuses on trusting
relationships between the information providers and the audience. Although not as
strongly, individual news consumption habits also predicted intention to modify personal
behaviors, presenting the possibility that frequent news consumers may need to be
targeted differently than infrequent consumers.

Sensation seeking and need for cognition were additional predictors of behavioral
intention, both directly and as interaction terms with the message conditions. However,
the direct relationships showed sensation seeking and need for cognition to be positive
predictors of intention to follow suggested behaviors while the interaction terms
produced results suggesting that they negatively moderated the relationships between
message conditions and behavioral intention.

Like for behavioral intention, news consumption was a predictor of outcome-relevant
involvement. Gender and age also predicted outcome involvement. While gender and age
were not large predictors, their relationships were statistically signiﬁcant. Perceived

threat had non-hypothesized relationships with sensation seeking and with how strongly

-140-

individuals believed the content of the provided news articles. Believability was a notable
positive predictor while sensation seeking was a small negative predictor.

None of these relationships were expected or predicted in this research’s hypotheses.
Other theories may posit explanations for these connections, or this data may serve to aid

in the development of future theories.

Limitations and future research

This study has several limitations that are worth noting and considering in future
research. One challenge with fear appeal research is that it is conducted in laboratory or
experimental settings and assumes that the perception of threat or emotional fear is the
predominant outcome of exposure to fear appeal messages. Naturalistic studies could
better account for exposure, attention, interest and comprehension issues. Most fear
appeal data stem from planned situations and are unable to assess outside inﬂuences.

Another issue is that fear appeals are not limited to producing only fear or threat as a
reaction (Witte & Allen, 2000). Outcomes do not arrive with the precision that
researchers may like or expect. Fear may be the strongest emotion stemming from a
health message; researchers, though, should recognize that surprise, puzzlement, anger,
and sadness can also come about as a result of a fear appeal (Dillard, Plotnick, Godbold,
Freimuth, & Edgar, 1996). Other projects have reported outcomes such as: irritation (J.
Kirscht & Haefner, 1973); disgust (Leventhal & Trembly, 1968); tension and energy
(LaTour & Pitts, 1989); and emotional issues such as anxiety and depression (Kohn et al.,
1982)

The variety of reactions is essential to consider when developing campaigns.

Individual interpretations of fear appeals do not take place neatly despite scientists’

-141-

efforts and wishes. It is essential to understand that other reactions are nor only possible
but should be expected. Counting on this variance can help campaign designers and
researchers more accurately grasp why certain messages show particular effectiveness or
not. It is entirely plausible that a consistently found reaction other than fear may prove to
be the most effective route to compliance-gaining for health education messages. Such a
relationship would, of course, then require theoretical consideration as an addition to an
existing model or in some future development.

Also considering effectiveness, another limitation is that real-life messages would
place more focus upon the efﬁcacy components in each message. Since this study was
interested only in the processing of messages and not the actual structure or effectiveness
of message components, the content used here possibly did not perfectly mirror what a
real campaign would produce. Future studies could use actual messages or articles to test
reactions to real content, as opposed to content developed according to newswriting
conventions.

Regardless of the message manipulations, researchers are additionally limited by the
shallow understanding of what makes for an effective fear appeal message. Any review
of the fear appeal literature demonstrates that it is not apparent what makes for an
effective arousing message (Boster & Mongeau, 1984), although response manipulations
seem to have increased in effectiveness since the initial research (based upon the
correlation between the year of study and the manipulation check, r = .13) (Witte &
Allen, 2000). Messages that researchers think are highly arousing may not actually be
fear- or threat-inducing for the participants, but more careful message deﬁnitions and
construction have begun bridging this gap (Witte & Allen, 2000).

_142-

 

1‘7le

.L -
Mi:
J- a‘h it
‘l‘l

lL‘Z’l‘S

indii

The design of using different messages to gain variance within the measurement
instruments also brings some limitations. While each condition was compared to the
others according to certain dimensions, there may have been other, inadvertent effects of
each treatment condition that could inﬂuence outcomes. It is possible that particular
topics, recommendations, accompanying information, or even article structure inﬂuenced
individuals’ reactions to the messages in unpredicted ways. Despite any researcher’s best
efforts, it is impossible to ensure completely that multiple treatment conditions within
any experiment are equal.

As discussed in the literature review of this dissertation, any number of moderating
variables may inﬂuence the relationship between the message conditions and perceived
threat. Each of the meta-analyses on fear appeal research contends that moderators may
exist within this relationship. While this study attempted to test two moderating variables,
others almost certainly exist. These other moderating variables could well have
inﬂuenced outcomes in ways unobserved by data collection for this project. Only further
research projects will be able to determine what these moderators may be.

The way moderators were measured in this project may also present less than ideal
data outcomes for interpretation. Some methodologists suggest that the dependent
variable in a moderation analysis needs to be measured by a number of response options
equal to or greater than the product of the number of options used to measure the
independent variable and the moderating variable (Frazier, Tix, & Barron, 2004). For
example, if the I.V. and moderator are each measured on ﬁve-point scales, the D.V.
should be measured on a 25-point scale. Using a composite dependent variable made up

of multiple ﬁve-point items does not resolve the issue. For ease of measurement, Frazier

-143-

a: t. SE
aliens:

Eu

cm
lite:

prst

dc.

fro;

et al. suggest having respondents rate their reactions by marking a continuous line as an
alternative.

Even without the moderators, the EPPM-predicted relationship between an arousing
message and outcomes is not entirely unambiguous, either. The data regarding the
relationship between a reaction and desired outcomes (behavior or attitude change, for
example) are sometimes unclear (Roskos-Ewoldsen, Yu, & Rhodes, 2004). In the context
of EPPM, this mixed data means that a threat message does not necessarily lead-to a
danger control reaction (meaning a self-protective adaptation) when paired with a strong
efﬁcacy component. Two meta-analyses from 2000 provide examples of this uncertain
data about the link between fear appeals and self-protective behavior. Witte and Allen
(2000) indicated that high levels of threat in a message may possibly lead to fear
control’s negative reactions regardless of efﬁcacy levels in the message. Floyd et al.
interpreted (2000) the data to mean that response and self-efﬁcacy were the strongest
predictors of danger control processes regardless of levels of fear. These alternative
perspectives merit consideration when interpreting data.

Other limitations come from the sample population used in this research. Online
respondents do not perfectly mirror the wider world. They are more likely to be educated,
wealthy, white, and suburban. No survey sample is likely to perfectly reﬂect the general
population unless it is actively chosen to do so, but research using online respondents has
made certain demographic trends clear. Researchers would be wise to consider these
demographics when interpreting results.

In addition to the sample attributes listed above, respondents in this study may differ

from the general population in other ways. The survey instrument used here did not

-144-

 

measure respondents’ access to quality health information or care, two things that could
inﬂuence reactions to health messages. It is reasonable to conclude that those with a
history of receiving quality medical care and information would be more likely to feel
empowered and capable of handling health issues. These individuals are more likely to
have a track record of positive outcomes. People excluded from positive healthcare
experiences, however, could be expected to view health information and expectations i

through a more fatalistic, negative lens.

 

Literacy levels were also not measured in any way. The study did consider L
respondents’ ability to answer fact-based questions, and this is often taken as a proxy for
an individual’s ability to read and comprehend. However, the relationship between
reading and question-answering is not perfect. Individuals with reduced literacy skills can
be quite adept at matching phrases or grasping just enough information to answer
multiple—choice questions. These individuals would most likely not be able to provide
valid responses to questions about reactions to the reading material, however. Their
inability to comprehend a majority of the text would prevent them from offering
considered reactions to the messages and subtleties of the text.

Future research could improve on similar projects by addressing not only the issues
above but a number of other areas as well. Methodology could be expanded to better
examine potential message reactions, social factors in communication, media’s
propensity to promote illness, cultural values, and individual readiness for behavior
change could all be considered, for example.

As Witte and Allen (2000) suggest, more carefully deﬁning message elements and

responses could improve research in this area. Fear control responses such as defensive

-145-

attain

Writ."
, ublllt l

 

lliES-‘l-‘C

 

5mm

E?

E3
(1;
7;

lfiCIO.’

m. .,.
W _.
lain}.

Ml“ l -.
lulu?

cm

i o‘
n i -.
limb.

r2.-

‘H‘H
H

SUI

avoidance, denial, and reactance have been particularly neglected in fear appeal research.
Additional measurement methods such as thought-listing tasks could be employed to
measure the range of respondent reactions since thought-listing allows for the
examination of message-processing issues and for alternative validation of response
measures.

The social context surrounding the presented health issues may also prove to be a
factor worth examining in future research. Apart from age, trait anxiety, and volunteer
status, it is entirely plausible other human elements that create social context also
inﬂuence individual understanding of risk-related issues (Branden Johnson, 1987). For
example, beliefs about the source, other people and the subject matter at hand can
inﬂuence both how and which messages are accepted.

Membership in social networks can play a role in perceived message credibility and
an individual’s level of concern (Branden Johnson, 1987). Research into the community
surrounding the toxic-waste site known as the Love Canal demonstrates this well. Well-
connected citizens of the Niagara Falls neighborhood expressed the most concern about
the pollution risks and also were the ﬁrst to stop trusting ofﬁcial sources (Fowlkes &
Miller, 1987). Unconcerned residents, however, reported being relatively isolated from
their neighbors. This same link between social interactions and concern about place has
shown itself in other contexts as well, particularly natural disasters (Kiecolt & Nigg,
1982; Moore, Bates, Layman, & Parenton, 1963; Perry, 1982).

Any health-risk message presented in a mediated context inherently brings associated
thoughts as well. In the context of health, journalists play a particular role in public
perceptions. Direct-to-consumer advertising, drug-company-organized awareness

-146-

 

  
  

10pm."
moth:

loam:

EXC

campaigns, and disease advocacy groups are obvious factors in “disease mongering”
(Moynihan & Cassels, 2005). However, journalists can also play a role by not thoroughly
examining evidence used in medical stories (Woloshin & Schwatrz, 2006). News
providers, after all, remain a major source of health information and thus have the power
to promote diseases (Kaiser Family Foundation, 2005). An illustrative example worth
noting is the promotion of restless legs syndrome (RLS) in 2003 through uncritical
journalism outlets, as outlined by Moynihan and Cassels (2005).

Culture is another variable worth considering, as cultural characteristics can interact
with health-related decisions and behaviors (Kreuter & McClure, 2004). This can happen
directly through dietary habits or indirectly through tight family connections that create
an insistence on caring for others at the risk of one’s own health. Due to such factors,
health researchers recognize that cultural characteristics need to be an inﬂuence on public
health strategies to enhance adoption, salience and acceptance of programs and messages
(Institute of Medicine, 2002; Perez-Stable, Otero-Sabogal, Sabogal, & Napoles-Springer,
1996)

Lastly, an individual’s readiness to accept behavior-modiﬁcation suggestions could
be worth noting in upcoming projects. Behavior change, of course, does not occur
exclusively as a result of a reaction to a one-time message. No shortage of research has
demonstrated that real behavior modiﬁcation takes place over time as individuals move
through behavioral and psychosocial stages with differing needs and characteristics
(Prochaska, Redding, & Evers, 2002). Therefore, presence in a particular stage of
behavior change could inﬂuence an individual’s reaction to a fear appeal (Cho & Salmon,
2006). Potentially, as Maibach and Cotton describe (1995), fear appeals could aid in

-147-

 

‘. Min
mill. .
6293251

y i

CORGI?

pitta 1

lumen

 

motivating individuals to work through stages of change by strengthening their plans to
engage in preventative behavior. Of course, some fear appeal researchers also raise the
concern that inappropriate fear appeals may have the opposite effect and result in
hardened negative attitudes (Maibach & Cotton, 1995), despite evidence to the contrary

(Sutton, 1992). Little research, however, has studied effects of fear appeals according to

 

the stages of change (Cho & Salmon, 2006). n
Stages of change as developed by Prochaska and DiClemente (1983) is a framework
created to lay out the process used by individuals as they commence and maintain

preventative behavior. More properly known as the transtheoretical model, this
framework attempts to establish a "structure of change that underlies both self-mediated
and treatment facilitated modiﬁcation of addictive and other problem behaviors"
(Prochaska, DiClemente, & Norcross, 1992, p. 1102).

The overall model consists of several constructs, but it is the stages of change that
have become the most frequently used component (Reed, Velicer, Prochaska, Rossi, &
Marcus, 1997). The other, more mechanistic components making up intermediate and
outcome components include self-efﬁcacy, temptation, decisional balance, and processes
of change. The stages of change, however, are the central organizing structure of the
overall model (Prochaska & Velicer, 1997).

Five speciﬁc stages of change are covered in the transtheoretical model: pre-
contemplation, contemplation, preparation, action, and maintenance (Prochaska &
DiClemente, 1982). The pre-contemplation stage is deﬁned as the time period during
which an individual lacks any intention to take action within six months. Contemplation

is when the intent of the individual changes, and that person plans to take action within

-148-

the next six months. Preparation is the next step, meaning the individual intends to take
action within 30 days and has taken some behavioral steps toward the action. Action
includes actually having changed behaviors but only for a reduced period of time, usually
less than six months. Once an individual has altered a behavioral pattern for more than
six months, that person has entered the maintenance stage. Lastly, when a behavior
becomes automatic and no chance of slipping backward exists, the goal of the termination
stage is reached. Termination also requires an extended period of behavior maintenance
(ﬁve years is often cited) and dominant self-efﬁcacy to continue the desired behavior
(Prochaska & Velicer, 1997).

Typically, moving from pre-contemplation to maintenance (or termination) is not a
direct, linear progression. Instead, the route has been described as a spiral pathway
involving backward slips (Prochaska, Norcross, & DiClemente, 1994). However, one
consistent ﬁnding of TM studies has been a link between level of perceived self-efﬁcacy
and movement between the pre-contemplation and maintenance stages (Gorely &
Gordon, 1995; Rapley & Coulson, 2005 ; Wakui et al., 2002). Response efﬁcacy can
similarly be linked to progress through the stages for several reasons (Cho & Salmon,
2006). First, the TM construct of decisional balance is linked to response efﬁcacy in that
a response must appear to be effective in order to be perceived as beneﬁcial. Being
perceived as effective is requisite for response efﬁcacy and being perceived as beneﬁcial
is necessary for shifting to the positive side of decisional balance.

As Cho and Salmon (2006) point out, Prochaska et al.’s (1994) study of 12 risky
behaviors noted that individuals moving from the pre-contemplation to the contemplation

stage reported the advantages of preventative behavior were also beginning to prevail

-l49-

over disadvantages. Thus, it is reasonable to conclude that efﬁcacy was also increasing as
the individuals were moving through the stages since being perceived as effective is
necessary for being perceived as beneﬁcial (Cho & Salmon, 2006).

Summag of implications

Effective dissemination of health information is a necessary task for health
communicators and media professionals. Disseminating health information also means
providing content useful for behavior change and promoting positive health behaviors
among the population. Developing such messages, however, is no simple task nor is
understanding how individuals mentally process such messages.

This dissertation provides three areas of progress in research into understanding how
people evaluate and respond to fear appeal messages. First, the research places an
extensively used message-processing model in a different and important mass
communication context. Second, it shows how additional psychological qualities such as
an engagement with health outcomes and a desire to think rationally can inﬂuence
individual reactions to persuasive messages. Third, this research shows that more
attention to instrument design, construct measurement, and experimental treatments can
provide more signiﬁcant statistical results.

The incorporation of new variables into an existing model applied in an atypical
context is central to the type of research needed to develop more effective health
communication messages. In addition, more attention to the social scientiﬁc methods
used to examine fear appeal reactions is essential to gaining a more precise understanding
of how to employ persuasion techniques. Together, testing theories in diverse contexts

and using more rigorous scientiﬁc techniques will lead to more useful applications of

-150-

health communication strategies as well as more robust predictive models.

-151-

APPENDICES

-152-

 

 

APPENDIX A:

Treatment Articles

~153-

Most Americans Exposed to HPV at Some Point, Experts Say
”Associated Press

Chances are you have been exposed to the potentially cancer-causing human
Papillomavirus virus and don't even know it. In fact, at least three out of four sexually
active adults under 50 have been infected with a strain of HPV, according to data
released by public health ofﬁcials this week.

In the United States alone, up to 6.2 million new cases occur each year, and at least
20 million Americans are now infected with the genital strain of the most commonly
acquired STD, according to public health ofﬁcials.

Almost all infections cause no symptoms or health problems, but nearly all cases of
cervical cancer stem from HPV. Some vulvar, anal, and penile cancers come from
HPV, too. Public health ofﬁcials guess that cervical cancers kill about one-third of
those stricken each year.

There are about 200 types of HPV. About 40 strains are spread through sexual
contact, and a handful of those are the ones linked with cervical cancer. However,
because the disease is so common and can have such a deadly outcome, public health
experts rank HPV to be among the most serious health issues facing young men and
women in their teens and twenties.

HPV can only be prevented by not having sex. Public health experts strongly suggest
becoming more informed as a key ﬁrst step. A Web site such as the Center for
Disease Control's at www.cdc.gov/std/hpv/ can be useful before talking to a doctor.
Having sex with only one partner is also an effective way to decrease your risk. In the
future, a new vaccine may offer some protection from HPV-related cancers. Always
using a condom can help to reduce risk, too.

154

Public More Informed About Clinical Trials, Experts Say
AP Associated Press

Awareness of clinical trials has grown over the last three years as more people have
seen news coverage and had the chance to participate in research studies, according to
data released by public health officials this week.

Almost 20 percent of Americans report being offered the chance to take part in a
clinical trial, an increase of seven percent from three years ago. Among the 5,822
people surveyed, more than half of those given the option agreed to take part.

Clinical trials are research projects carried out on human volunteers to learn more
about the body and the diseases that attack it. They test how well new medical
approaches work in people. Each of the 36,000 active studies in the US. is trying to
ﬁnd a better way to prevent, diagnose, or treat a disease.

Participants in clinical trials beneﬁt in a number of ways. They take a more active
role in their healthcare and may get to try new treatments. In addition, they often get
ﬂee care at some of the best hospitals around the country. There are risks to being in a
clinical trial, too. Doctors are not always aware of the side effects of the new
treatments. These new methods also may not work or may take longer than normal
care.

Public health experts strongly suggest becoming more informed as a key ﬁrst step to
participating in a clinical trial. A Web site such as the National Institute of Medicine's
clinicaltrials. gov can be useful when gathering information or preparing to talk to a
doctor. Interested people can also enroll by talking to the clinical trial’s coordinator.

155

 

A Quarter of Americans Exposed to Food Poisoning Each Year,
Experts Say
”Associated Press

More than a quarter of all Americans come down with food poisoning each year,
suffering symptoms such as nausea, vomiting and diarrhea, according to data released
by public health ofﬁcials this week.

There are about 76 million cases in the United States each year, but most of these
cases are preventable. People with food poisoning typically get sick within 48 hours
of consuming infected food or drink.

Many cases are mild and pass so quickly that they are never diagnosed. One victim
out of every thousand does end up in the hospital, however. One of the most common
bugs—salmonellae—costs up to $1 billion in medical expenses and lost work time
each year.

More than half of food poisoning cases are caused by improper cooking and storage
of foods. About a quarter of them stem from poor hygiene, such as not washing hands
before handling food. Only a small number of cases are the result of an unsafe food
source. The growing rates of food poisoning, as well as its personal and ﬁnancial
cost, have convinced public health ofﬁcials food poisoning is a real public health risk.

Food poisoning can be prevented with some simple steps, such as: washing hands,
cleaning utensils, cooking food all the way through and storing raw and cooked food
in separate containers. Public health experts also strongly suggest becoming more
informed as a key ﬁrst step. A Web site such as the Center for Disease Control's
www.cdc.gov/foodbome/index.htm can be useful in learning tips to keep healthy.

156

Prostate Cancer Biggest Health Threat to Men, Experts Say
”Associated Press

Once dismissed as an “old man’s disease,” prostate cancer is now recognized as the
biggest health threat to men of all ages. It kills tens of thousands of men each year
and causes erectile dysfunction in hundreds of thousands more. This cancer strikes in
the prime of life—and the threat may be growing, according to data released by
public health officials this week.

The rate of prostate cancer has increased by more than 50 percent in the last 30 years
while rates of other types of cancer have dropped. Public health ofﬁcials expect that
the number of new prostate cancer cases detected each year will likely double again
by 2015.

The disease is now the top cancer killer of men. About a quarter of all American men
who develop prostate cancer die of it, making the disease the male equivalent of
breast cancer. However, prostate cancer can be far more dangerous because it usually
does not produce signs or symptoms in its early stages, when treatment is most
successful. Another real problem is that the treatments bring high risks of fearsome
side effects such as impotence, according to public health officials.

Prostate cancer cannot be prevented, but there are ways to reduce the risk or slow the
disease’s progress. Public health experts strongly suggest becoming more informed as
a key ﬁrst step. A Web site such as the Center for Disease Control's
www.cdc.gov/cancer/prostate/ can be useful before talking to a doctor.

Also, eating a healthful, low-fat diet rich in fruits and vegetables may help ﬁght

cancer. Exercise and certain drugs may help control prostate growth. Exercise can
reduce other risk factors such as obesity and poor blood ﬂow, too.

157

 

 

 

Breast Cancer Biggest Health Threat to Women, Experts Say
APAssoclated Press

Breast cancer remains a top killer of women in the United States despite vast efforts
to combat the disease, according to new data released this week by public health
officials. The last 30 years have actually seen this most common form of cancer
become more widespread.

More than 212,000 American women will develop breast cancer this year alone,
according to public health officials. About 40,000 women will die, and thousands
more will undergo radiation therapy or mastectomies.

All women are at risk for breast cancer, regardless of age or family history, the
officials say. Up to 90 percent of women who develop breast cancer do not have a
family history of the illness. Some factors, though, could affect a woman’s risk.
These include: being overweight, taking birth control pills, inactivity, or using
hormone therapy over several years.

Breast cancer cannot be prevented, but some actions may reduce the risk. Public
health experts strongly suggest becoming more informed as a key ﬁrst step. A Web
site such as the Center for Disease Control's www.cdc.gov/cancer/breast/ can be
useful before talking to a doctor.

Eating a healthful diet rich in fruits and vegetables may also help ﬁght cancer. Other
tips suggested by health ofﬁcials include getting regular exercise and taking
supplements such as folate. Self-exams and check-ups are essential as well because
they help women detect problems at an early stage when treatments are most useful.

158

 

APPENDIX B:

Pilot Study

-159-

 

After developing the treatment articles and establishing indices to measure the
relevant variables, an online community was used to pilot test the articles and the survey
instruments. Researchers placed a link to the survey on the community’s homepage
during the last week of March 2007 and gave members three days to respond. The

incentive offered was entry into a random drawing for one $100 prize.

 

Participants were randomly assigned to one of the ﬁve message conditions and were i
r-
exposed to only one article. The order of survey questions was consistent for each
individual. The only change was the treatment article and necessary wording changes in l

the accompanying surveys. Participants were free to take the survey at any time and at
any computer location convenient to them, such as work, home, library, or school. The
survey items and instructions are shown in Appendix C. Each survey for the ﬁve
treatment articles was equivalent except for changing the article’s health topic where
appropriate. Individuals were free to quit the survey at any time and were made aware of
this in their informed consent form. During the data collection period, 134 people
completed the pilot study in full, and 47 also offered qualitative comments that were used
in many cases to improve survey design and wording.

The pilot survey presented instructions, questions, and the treatments in the following
order:

° Introduction to survey and consent form.

° Demographic information, including: age, gender, marital status, ethnicity,

 

nationality, familiarity with relevant health topics, relationship status, news-
reading habits.

' Sensation seeking instrument: 8 questions with 5-point scales.

-l60-

° Need for cognition index: two options—one of 18 questions with 7-point scales
and one of 3 items with 7-point scales.

° Outcome-relevant involvement: two options—one of 3 questions with 7-point
scales and one of 8 questions with 7-point scales.

° Introduction to the newspaper article asking respondents to read the content and
answer further questions.

' Treatment article concerning clinical trials, food poisoning, human
Papillomavirus, prostate cancer, or breast cancer.

0 Comprehension questions: 4 questions with true/false format.

° Believability index: 4 questions with 7-point scales.

° Perceived severity instrument (one-half of perceived threat): 3 questions with 7-
point scales.

° Perceived susceptibility (one-half of perceived threat): 3 questions with 7-point
scales.

° Response efﬁcacy (one-half of perceived efﬁcacy): 3 questions with 7-point
scales.

° Self-efﬁcacy (one-half of perceived efﬁcacy): 3 questions with 7-point scales

0 Behavioral intention: 6 items with 7-point scales.

0 Thank-you note, further information on the treatment article’s topic.

' Open section for respondent comments.

Pilot study results

As previously stated, the treatment articles were intended to produce differing levels

of perceived threat, with the clinical trials article acting as a baseline control condition.

-161-

The inclusion of the clinical trial piece served to have respondents read some news
content before responding to the survey items but not be exposed to threatening
messages. A between—groups, one-way ANOVA revealed that perceived threat scales did
differ signiﬁcantly as a function of exposure to a particular article (F(4,129) = 55.92, p <
.001). The Levene statistic (.882, p = .476) indicated that the groups did not have
signiﬁcantly different variances, meeting one of AN OVA’s key assumptions. In addition,

examination of the Q-Q plot and the relationships between the standard errors and

 

skewness and kurtosis indicate that the data are sufﬁciently normal to accept F-test
results.° Table 20 contains the pilot test descriptive statistics, including means and
standard deviations.

Table 20. Pilot Test Descriptive Statistics for Perceived Threat by Article

 

 

Subject
Topic N Mean Std. Dev. Std. Error
Breast 16 5.97 1.14 .28
Frost. 26 5.42 1 .01 .20
HPV 35 4.14 .98 .17
Food 26 4.06 .81 .16
Trials 31 2.24 .93 .17
Total 134 4.15 1.57 .14

 

 

 

 

A post-hoe Tukey honest signiﬁcant difference test on the pre-test data did reveal that
the clinical trials article induced signiﬁcantly less threat than the other write-ups. As
expected, the cancer articles were not signiﬁcantly different from each other, and neither
were the virus-focused human Papillomavirus and food poisoning articles. The various

treatments were successful in that the messages created multiple levels of perceived

 

9 One rule for checking for normality is to divide the skewness and kurtosis results by their standard errors.
Results between -1 .96 and 1.96 indicate that the data are not signiﬁcantly different than normal.

-l62-

threat, breaking down into three distinct categories: a control condition, a virus-related

condition, and a cancer-related condition. Table 21 contains the results of the Tukey’s

 

 

 

HSD analysis.
Table 21. Perceived Threat by Article Topic
. . Other Mean Std. .
Amc'e T°'°'° Topics om. Error 3'9'
HPV Clinical Trials 1.90(*) .24 .000
Food Poison. .08 .25 .998
Prostate Canc. -1.27(*) .25 .000
Breast Canc. -1.83(*) .29 .000
Clinical Trials HPV -1.90(*) .24 .000
Food Poison. -1.82(*) .26 .000
Prostate Cane. -3.17(*) .26 .000
Breast Canc. -3.73(') .30 .000
Food Poison. HPV -.08 .25 .998
Clinical Trials 1.82(*) .26 .000
Prostate Canc. -1.35(*) .27 .000
Breast Canc. -1.90(*) .31 .000
Prostate Canc. HPV 1.27(*) .25 .000
Clinical Trials 3.17(*) .26 .000
Food Poison. 1.35(*) .27 .000
Breast Canc. -.55 .31 .376
Breast Canc. HPV 1.83(*) .29 .000
Clinical Trials 3.73(*) .30 .000
Food Poison. 1.90(") .31 .000
Prostate Canc. .55 .31 .376

 

 

 

* The mean difference is signiﬁcant at the .05 level.

As this is not a message-design study, the exact outcome from different message
components is not as important as the creation of different levels of perceived threat.
Again, the objective for employing multiple message conditions was to ensure variance
among responses to the measurement indices for the variables included in the research
model.

Along with perceived threat, believability and subject knowledge were also key

considerations for ensuring that the ﬁve articles fulﬁlled their functions. Believability

-l63-

was a necessary check because the inﬂuence of the articles depended at least in part on
their perceived truthfulness. The respondents’ comprehension levels were important to
verify that the individuals actually read the articles and maintained some familiarity with
the content.

Believability was measured using four items across seven-point Likert-type response
scales. Respondents in all ﬁve groups reported high scores on the believability instrument

(averaging a 5.44 out of 7), and no signiﬁcant differences were found among the

 

conditions. A one-way ANOVA to check for differences returned a non-signiﬁcant result
(F(4,l29) = 1.48, p = .213), and Tukey’s HSD post-hoc analysis found no signiﬁcant
differences, either. Checks for violation of necessary AN OVA assumptions turned up no
conﬂicts. Table 22 presents the believability index descriptive statistics for each article
treatment group.

Table 22. Group Descriptive Statistics for BelievabilitLIndex Scores

 

 

 

 

 

Topic N Mean Std. Dev. Std. Err.
HPV 35 5.35 1.05 .18
Clin. Trials 31 5.28 .62 .11
Food Pois. 26 5.79 .68 .13
Prostate Canc. 26 5.35 .95 .19
Breast Canc. 16 5.50 1.06 .27
Total 134 5.44 .89 .08

 

In addition, there were no signiﬁcant differences among the individual items in the
believability scale. Tables 23 and 24 display the descriptive statistics for the believability

items and Tukey’s HSD results.

-164-

Table 23. Item Descriptive Statistics for Believability Index

 

 

 

 

 

 

 

I Items N Mean Std. Dev.
To me the advice given in the V -

. '. . 134 . .
article is believable. 5 66 1 05
In my opinion, the information 134 5.38 i 1.14
given in the article seems useful.
The advrce given in the article is 134 5.19 104
accurate. __ __ __ _ -1 _ ,_ _ (___;
The Information I read in the 134 5.53 l .99
article looks true to me. . -__.
Mean Scores 134 5.44 1.06

 

 

 

 

All items used a seven-point scale.

Table 24. Tukey’s HSD for Believability Items

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Article Cond. Article Cond. Mean Diff. Std. Err. Sig. 95% C.l.
HPV Clin. Trials . .07 W22 .998 -254, L .67
Food Pois. ﬂ_ 245..., L23 f 0 .294 1 . 51.08 49-,
Prostate Canc. __-_.01~ _ L423L “1.00 L -.64__ .93
Breast Canc. -.15 .27 .980 -.89 .59
Clin. Trials HPV -.07 __L.22_ .998 p __ 1.67m f __._5_4 L
Food Pois. -.52 _ .21“ L - .190 r.----,-_‘1-17 .L14ﬁ
Prostate Canc. -.07 _L ‘34 _Li998L -g,~-_’§_,--, n39
Breast Canc. -.22 .27 .931 -.97 i .54
Food Pois. HPV .45 , . -33.-- p.294 nut-19 _108
Clin. Trials .52 L f .24 .190 L _14 1 1.16
Prostate Canc. L44 ,- L 3.5-- f .377 ”I -24 1 7 1.12
Breast Canc. .30 .28 .827 j -.48 1.08
Prostate Canc. HPV 40.1,. - ”-23. 1.00 Al 3.637 9?
Clin. Trials _ .07“ ._2_4 A ,,:993_ L -.58 g .73
Food Pois. -.44LL35LLL§IL .. 1.12 - L p.24
Breast Canc. -.14 .28 .986 -.92 .63
Breast Canc. HPV _._1_5h___+_w 31“...-w-939H :59 ______.89__
Clin. Trials __ .22 “.27 ‘ _ .931“ _L ____m-._54 1___ .97 _
Food Pois. LL_-.30L A .28 L ‘ .827L L {.08 4 L48 L
Prostate Canc. .14 . .28 .986 -.63 .92

 

 

 

 

 

 

 

According to the comprehension check scores, respondents read and accurately
interpreted the information contained in the articles. Comprehension was measured using
four true-false items adapted from the previously read article. While this does mean that

-165-

each of the ﬁve groups responded to different comprehension check items, the four
questions for each group were made as similar as possible to provide for some
consistency across the sample. All of the comprehension instruments focused on the basic
information common to all ﬁve articles: frequency of the issue, preventative measures,
risk factors, and severity information. All items in this index were answered correctly at
least 90 percent of the time, and respondents in each group received average overall
Scores ranging from 88 percent to 94 percent. Based upon these results, it'appears that
respondents understood and processed the information presented in the various articles.
Table 25 contains the percentage correct scores for each item of the various

comprehension check instruments.

Table 25. Comprehension Check Percentage Correct Scores By Treatment
Condition

 

 

 

 

 

 

 

 

Topic Item 1 Item 2 Item 3 Item 4 Mean
HPV .87 .94 .83 .88 .88
Clin. Trials .90 .95 .97 .95 .94
Food Pois. .95 .93 .90 .95 .93
Prost. Canc. .91 .94 .83 .92 .90
Breast Gene. .96 .97 .96 .81 .93
Mean .92 .95 .90 .90 .92

 

 

 

 

 

 

 

Instrument reliabilities

Another essential function of the pilot data collection was to test the reliability of
previously used, valid instruments and select among several options for a few variables.
The calculation of Cronbach’s alpha served as means to ensure that the items were
sufﬁciently reliable when internal consistencies were measured. The Cronbach’s alpha
for each scale was expected to be higher than .70 (Nunnally, 1978). All of the tested

instruments met this criterion. In addition, this researcher examined correlations among

-166-

the various survey items to ensure that none would pose multicollinearity problems.
Table 26 presents the complete correlation matrix. No items were inappropriately linked.

Table 27 shows the descriptive statistics and Cronbach’s alpha coefﬁcients of the scales.

 

 

 

 

 

 

 

 

 

 

 

T_able : 6. Bivariate Variable Correlations in Pilot Test Data

0v 1 ss i BNFC i NFC PLORI i ORI i so {Threat i Effic.
DV 1.00 0.04 012 .19(') .32(") -.22(') .210) .45(") .47(“)
[SS 0.04 1.00 0.08 0.02 0.15 0.05 -0.14 0.11 002
lBNFC -0.12 0.08 1.00 _.77(--) -0.12 .26("’) 014 -002 0.09
NFC .19(~) -0.02 -.77(") 1.00 0.15 -.33(") ' 0.09 0.04 0.07
PLORI .32(“) 0.15 -0.12 0.15 1.00 -.74(") -0.01 33¢”) 0.01
ORI -.22(') 0.05 .26(") -.33(**) -.74(") 1.00 0.10 -.21(*) —0.01
[BC .21(*) -0.14 -0.14 0.09 -0.01 0.10 1.00 0.14 .41(")
[Threat .45(**) 0.11 -0.02 0.04 330*) -.21 (1) 0.14 1.00 .32(“)
IEffic. .47(“) 002 0.09 0.07 0.01 -0.01 .41(") .32(") 1.00

 

 

 

'. Correlation is signiﬁcant at the 0.05 level (2-tailed).

“. Correlation is signiﬁcant at the 0.01 level (2-tailed).

DV: behavioral intention; SS: sensation seeking; BNFC: brief need for cognition; NFC:
need for cognition; PLORI: Park and Levine outcome-relevant involvement; ORI:
outcome-relevant involvement; BC: believability check; Threat: perceived threat; Efﬁc.:
perceived efﬁcacy

-l67-

 

Table 27. Descriptive Statistics and Cronbach Alphas for Pilot Test
Measurement Instruments

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

# of V V ' Y Typical
I Construct Items Cases Mean 8.0. Alpha Alpha
Perceived Severity 3 134 13 5.38 .92 .80
Perceived Susceptibility 3 133 11.95 5.36 .88 .80
Perceived Self-Efficacy 3 132 15.31 3.77 .77 .79
Ere. Response Efficacy 3 132 15.97 3.55 .79 .79
ChOIBOSte'I Outcome- 7 132 38.02 10.17 .89 .87
Rel. Involvement -
Park et al. Outcome-Rel. 3 133 10.53 4.79 '91 .86
Involvement
Sensation Seeking' 8 132 24.70 5.77 .77 .76
[Brief NFC Index 3 134 7.33 3.36 .78 .69
Short NFC Index 18 130 89.70 14.38 .87 .76
Believability Check 4 134 21.76 3.57 .87 N/A
[Behavioral Intention 6 133 23.49 8.39 .82 .81

 

 

 

 

 

 

 

 

*All instruments use 7-point Likert-type scales, except for Sensation Seeking. The 8-item
instrument used for Sensation Seeking employs 5-point Likert-type scales.

Each instrument produced acceptable Cronbach’s alpha coefﬁcients, and in most

cases scored higher than in prior uses. For the variables outcome-relevant involvement

and need for cognition, two scales were tested with the goal of selecting one most

effective option for the ﬁnal data collection. In the case of outcome-relevant involvement,

Park et al.’s version produced a slightly higher alpha coefﬁcient. Combined with its

shorter format (three questions versus seven), the decision was made to include this

version in the ﬁnal instrument. Park et al.’s outcome-relevant involvement scale thus

-l68-

 

 

serves two positive purposes in that it can produce a favorable alpha coefﬁcient and it
requires only three items in the survey.

The need for cognition indices did not work out as neatly. The alpha coefﬁcients did
clearly show the short NFC scale to be more reliable than the brief edition (.87 to .78),
but the short scale is 15 items longer than the brief version. Considering the difference in
Cronbach’s alpha, however, the 18-item short NFC scale was selected as the more
prudent choice for ﬁnal data collection. ‘

In addition to helping decide among survey instrument options, pilot testing also
helped improve the wording of several questions. Respondents’ qualitative comments
proved helpful in improving the clarity of some survey items and suggesting alternative

wording.

-l69-

 

 

 

APPENDIX C:

Pilot Test Survey

 

-170-

Please note that the format used for online testing has been approximated here as best
as possible. Certain formatting norms of online publication, such as thinner columns, do
not translate well to the printed page. In addition, certain functional survey items such as
answer-selection buttons do not reproduce in word-processing documents. Pages of the
survey are indicated in parentheses before the start of each page. In the online version,
page progress was indicated by text at the bottom of each page that let respondents know
both their progress and the pages remaining by saying, for example, “Page 6 of 1 l.”

(Page 1 of the pilot test)

News Reaction

Welcome, and thank you for taking the time to
participate in our study, which should last less than 25
minutes!

INTRODUCTION

You are being invited to participate in a research study
about how people view mass media coverage of health
topics. Researchers at Michigan State University will
perform the study. Participants will need about 25 minutes
to complete the survey. In return for completing the survey,
respondents will be entered into drawing to win a $100 prize.
The prize will be distributed in approximately one

week after data collection is complete.

PURPOSE

The purpose of this study is to determine how people
understand and view coverage of health-related topics in
the mainstream media.

WHAT IS INCLUDED IN THIS PROJECT?

Participants will be asked to read an online news story and
complete survey questions regarding the story and their
reactions to it. Participants will also answer survey
questions about their personal views on the world around
them. Completing the reading and questions is estimated to
take 25 minutes.

PROCEDURE

If you decide to take part in this study, you will be

asked to consent to the items discussed below. All of the
information collected for this project will be kept
confidential to the maximum extent allowed by the law.
You may be asked to complete a series of surveys about
your reactions to the news media and your views on your
environment.

PARTICIPATION

-l71-

Participation in this study is entirely voluntary. In

addition, should you choose to participate in the overall
study, you may still decline to participate in any segment of
the survey. You may also choose to discontinue your
participation at any time. Any refusal to participate or
decision to discontinue will not affect any future
opportunities to participate in surveys offered through this
online panel.

RISKS

Although it is highly unlikely, there is a chance that you
might feel uncomfortable with some of the questions during
the course of the survey because they deal with health
topics. If at any time during the course of the survey, you
wish to refrain from answering a question or simply wish to
stop participating, the survey can be postponed and/or
terminated.

Participation in this study presents no physical threat.
BENEFITS

We cannot and do not guarantee you will directly

beneﬁt if you take part in this study. Through your
answers, we hope to gain valuable information about the
use of news media and more speciﬁcally coverage of
health—related issues. It is our hope that the information we
receive from you will help improve the quality of media
coverage of health topics.

PAYMENT TO SUBJECTS

As a participant in this research study, you will be

entered into a random drawing to win a prize of

$100. Every respondent will have an equal opportunity to
win the $100 prize. As such, we cannot guarantee payment to
any individual, but everyone completing the survey will be
equally likely to receive $100 for their time.

COSTS

Participation in this project will not cost anything.

There will be no ﬁnancial responsibility for any respondent
nor any penalty for not participating.

ALTERNATIVES

The alternative to participating in this research study is

to decline. Participation is purely voluntarily and for
research purposes only.

CONFIDENTIALITY

Investigators will keep secret all research related to
records and information gathered by this study.
Throughout the study, all information from surveys will be
stored in a manner to assure conﬁdentiality of the subjects

-l72-

 

 

to the maximum extent allowable under law. Your privacy
will be protected to the maximum extent allowable by law.
The investigators will not reveal your identity if they
publish the results of the study.

QUESTIONS

Through this form, the investigators have tried to answer
questions to your satisfaction. If you have any particular
questions about this study, please contact Brad Love by
phone: (517) 432-6186, e-mail: Iovebrad@msu.edu, or
regular mail: 305 Communication Arts & Sciences, East
Lansing MI 48824. If you have any questions or concerns
regarding your rights as a study participant, or are
dissatisﬁed at any time with any aspect of this study, you
may contact - anonymously, if you wish - Peter Vasilenko,
Ph.D., Director of the Human Subject Protection Programs
at Michigan State University, by phone: (517) 355-2180,
fax: (517) 432-4503, email: irb@msu.edu, or regular mail:
202 Olds Hall, East Lansing, MI 48824.

CONSENT

The investigator(s) gave you information about what

will be done in this research study. You also were told how
it will be done, what you will have to do, and how long the
research will take. The information in this form attempted
to tell you about any inconvenience, discomfort, or risks
you might experience due to this research.

By clicking below, you agree to take part in this study

as a research subject. You are aware that you may quit or
refuse to participate in certain parts of the research at any
time.

Accept Decline

-l73-

 

 

 

(Page 2 of the pilot test: note that each answer had a radio button placed before
it. Here, the buttons are represented by empty circles. A few other items used a
dropdown menu or text box, as indicated)

We would like to start by asking you some general questions.
The following items are all intended to help us learn a little about your
background.

Please select the appropriate button for your gender.
1. Gender

0 Male
0 Female

2. What is your age (optional, please enter "99" if you prefer to not answer)?
(dropdown menu)

 

3. What is your ethnicity (optional)?
0 Asian / Pacific Islander

0 African American

0 Hispanic/ Latino

0 Caucasian

0 Native American

0 Other: (blank text box)

0 Prefer not to respond

4. What is your country of residence?
0 USA.

0 Canada

0 Mexico

0 Other: (blank text box)

 

5. Are you currently a student?
0 No

0 Yes (High School)

0 Yes (College/University)

6. If you are a student, what year are you?
O Freshman

0 Sophomore

0 Junior

0 Senior

0 Graduate/Professional

O Other: (blank text box)

0 Not a student

 

-174-

7. What is your current relationship status (optional)?
0 Single

0 Committed Relationship

0 Engaged

0 Married

8. Where do you get news information? Please mark the items that best ﬁt your
situation.

0 Newspapers (Hard Copy)

0 Newspapers (Online)

0 News Magazines (Hard Copy)

0 News Magazines (Online)

0 Blogs

0 T.V.

0 Radio

For the following sources, indicate how many days per week you use

each news source:

9. Newspapers (Hard Copy): (dropdown menu)

10. Newspapers (Online): (dropdown menu)

11. News Magazines (Hard Copy): (dropdown menu)
12. News Magazines (Online): (dropdown menu)
13. Blogs: (dropdown menu)

14. TV: (dropdown menu)

15. Radio: (dropdown menu)

Do you have any experience with the following? Check all that apply or
please select "none" if you prefer to not answer:

16. Human Papillomavirus:

0 Personal history 0 Family history 0 None

17. Prostate Cancer:
0 Personal history 0 Family history 0 None

18. Food Poisoning:
0 Personal history 0 Family history 0 None

19. Breast Cancer:
0 Personal history 0 Family history 0 None

20. Have you or anyone in your family ever participated in a clinical trial
research study?
0 No 0 Yes 0 Don't know

-175-

 

 

(Page 3 of the pilot test)

News Reaction

Interest and Preference Survey

For each of the following statements, please indicate how much you
disagree or agree by selecting the appropriate response along the
indicated range (Note: the far left button implies strong disagreement
while the far right button implies strong agreement).

21. I would like to explore strange places.

Strongly 0 O O 0 O O O Strongly
Disagree Agree

22. I get restless when I spend too much time at home.
Strongly 0 O O O O 0 0 Strongly
Disagree Agree

23. I like to do frightening things.
Strongly 0 0 0 O 0 O O Strongly
Disagree Agree

24. I like wild parties.
Strongly 0 0 0 O 0 O 0 Strongly
Disagree Agree

25. I would like to take off on a trip with no pre-planned routes or timetables.

Strongly 0 O 0 O O O o Strongly
Disagree Agree

26. I prefer friends who are excitingly unpredictable.

Strongly 0 0 O 0 o o 0 Strongly
Disagree Agree

27. I would like to try bungee jumping.
Strongly 0 0 0 O 0 O O Strongly
Disagree Agree

28. I would love to have new and exciting experiences, even if they are illegal.
Strongly 0 O 0 0 o 0 0 Strongly
Disagree Agree

~176-

(Page 4 of the pilot test)
News Reaction

Preference Survey, continued

For each of the following statements, please indicate how much you
disagree or agree by selecting the appropriate response along the
indicated range (Note: the far left button implies strong disagreement
while the far right button implies strong agreement).

29. I would rather do something requiring less thought than something that
challenges my thinking abilities.

Strongly 0 O O O 0 0 0 Strongly

Disagree Agree

30. I don't like to have the responsibility of handling a situation that requires a
lot of thinking.

Strongly 0 o O 0 O 0 O Strongly

Disagree Agree

31. I try to anticipate and aVoid situations where I will be likely to have to think
in depth about something.

Strongly 0 O O 0 0 O 0 Strongly

Disagree Agree

32. I would prefer complex to simple problems.

Strongly 0 O O O 0 O O Strongly
Disagree Agree

33. I like having the responsibility of handling a situation that requires a lot of
thinking.

Strongly 0 O O O O 0 0 Strongly

Disagree Agree

34. Thinking is not my idea of fun.
Strongly 0 0 O O O 0 0 Strongly
Disagree Agree

35. I would rather do something that requires little thought than something that
is sure to challenge my thinking abilities.

Strongly 0 O O 0 O O O Strongly

Disagree Agree

36. I try to anticipate and avoid situations where there is likely chance I will
have to think in depth about something.

Strongly 0 O O O O 0 o Strongly

Disagree Agree

-l77-

37. I ﬁnd satisfaction in deliberating hard and for long hours.

Strongly 0 0 o 0 O O 0 Strongly
Disagree Agree

38. I only think as hard as I have to.

Strongly 0 0 0 0 O O 0 Strongly
Disagree Agree

39. I like tasks that require little thought once I've learned them.

Strongly 0 0 O 0 O O 0 Strongly
Disagree Agree

40. The idea of relying on thought to make my way to the top appeals to me.

Strongly 0 O O 0 0 O 0 Strongly
Disagree Agree

41. I really enjoy a task that involves coming up with new solutions to
problems.

Strongly 0 O 0 0 0 0 O Strongly
Disagree Agree

42. Learning new ways to think doesn’t excite me very much.
Strongly 0 0 0 O 0 O O Strongly
Disagree Agree

43. I prefer my life to be ﬁlled with puzzles that I must solve.

Strongly 0 0 O O 0 0 O Strongly
Disagree Agree

44. The notion of thinking abstractly is appealing to me.

Strongly 0 0 O 0 0 0 O Strongly
Disagree Agree

45. I would prefer a task that is intellectual, difﬁcult, and important to one that
is somewhat important but does not require much thought.

Strongly 0 0 0 O 0 O 0 Strongly

Disagree Agree

46. I feel relief rather than satisfaction after completing a task that required a
lot of mental effort.

Strongly 0 O 0 O O O O Strongly

Disagree Agree

47. I prefer to think about small, daily projects rather than long-term ones.

Strongly 0 0 O O O o O Strongly
Disagree Agree

-178-

48. It’s enough for me that something gets the job done; I don’t care how or
why it works.

Strongly 0 0 0 O O O O Strongly

Disagree Agree

49. I usually end up deliberating about issues even when they do not affect me
personally.

Strongly 0 O O O 0 0 0 Strongly

Disagree Agree

 

-l79-

(Page 5 of the pilot test)

News Reaction

Involvement With Human Papillomavirus

Now, consider the topic of human Papillomavirus and how it might affect
you. Please answer each of the following questions.

Indicate how much you disagree or agree by selecting the appropriate
response along the indicated range (Note: the far left button implies
strong disagreement while the far right button implies strong
agreement).

50. A local outbreak of human Papillomavirus would affect me personally.

Strongly 0 o 0 O O O O Strongly
Disagree Agree

51. A local outbreak of human Papillomavirus would have a big impact on my

life.
Strongly 0 0 O 0 0 O O Strongly
Disagree Agree

52. A local outbreak of human Papillomavirus would have little effect on me
personally.

Strongly 0 O 0 O O O O Strongly

Disagree Agree

53. Whether or not there is a local outbreak of human Papillomavirus would
have little impact on how I live.

Strongly 0 0 O 0 O O O Strongly

Disagree Agree

54. It is difﬁcult for me to think of ways that a local outbreak of human
Papillomavirus would affect my life.

Strongly 0 0 O O 0 O O Strongly

Disagree Agree

55. My life would be changed if there were a local outbreak of human
Papillomavirus.

Strongly 0 O O O 0 0 O Strongly

Disagree Agree

56. All in all, the effect of a local outbreak of human Papillomavirus would be
small.

Strongly 0 O O O O 0 0 Strongly

Disagree Agree

-180-

 

57. A local outbreak of human Papillomavirus would affect my daily life.

Strongly 0 O O O O 0 0 Strongly
Disagree Agree

58. My quality of life would not change much if there were a local outbreak of
human Papillomavirus.

Strongly 0 O 0 0 0 0 0 Strongly

Disagree Agree

59. It is easy for me to think of ways that a local outbreak of the human
Papillomavirus would affect my life.

Strongly 0 0 0 0 0 0 0 Strongly

Disagree Agree

60. My well-being has little to do with a local outbreak of the human
Papillomavirus.

Strongly 0 O O O O O O Strongly

Disagree Agree

-l81-

ll

(Page 6 of the pilot test)

News Reaction

The next page will take you to a newspaper article about human
Papillomavirus. Please read the article and respond to the questions

that follow.

~182-

 

(Page 7 of the pilot test)

News Reaction

Most Americans Exposed to HPV at
Some Point, Experts Say

4’ Associated Press

Chances are you have been exposed to the
potentially cancer-causing human Papillomavirus
virus and don't even know it. In fact, at least
three out of four sexually active adults under 50
have been infected with a strain of HPV,
according to data released by public health
ofﬁcials this week.

In the United States alone, up to 6.2 million new
cases occur each year, and at least 20 million
Americans are now infected with the genital
strain of the most commonly acquired STD,
according to public health ofﬁcials.

Almost all infections cause no symptoms or
health problems, but nearly all cases of cervical
cancer stem from HPV. Some vulvar, anal, and
penile cancers come from HPV, too. Public health
ofﬁcials guess that cervical cancers kill about
one-third of those stricken each year.

There are about 200 types of HPV. Only about
40 strains are spread through sexual contact,
and only a handful of those are linked with
cervical cancer. However, because the disease is
so common and can have such a deadly
outcome, public health experts rank HPV to be
among the most serious health issues facing
young men and women in their teens and twenties.

HPV can only be prevented by not having sex.
Public health experts strongly suggest becoming
more informed as a key ﬁrst step. A Web site
such as www.cdc.gov/std/hpv/ can be useful
before talking to a doctor. Having sex with only
one partner is also an effective way to decrease
your risk. In the future, a new vaccine may offer
some protection from HPV-related cancers.
Always using a condom can help to reduce risk,
too.

183

 

 

Please select "Yes" to conﬁrm that you have completed reading the
article and are ready to move on to the next question. Thank you.

61. Conﬁrm reading
0 No 0 Yes

-184-

(Page 8 of the pilot test)

News Reaction

Article Questions

Thank you for reading the article on human Papillomavirus. Below are
four questions relating to the content of the article. Please answer each
question to the best of your ability.

62. According to the article, human Papillomavirus is the most common STD.
0 True 0 False

63. Becoming more informed is a key ﬁrst step in reducing one’s risk for human
Papillomavirus.
0 True 0 False

64. Human Papillomavirus causes nearly all cases of cervical cancer, as well as
some anal and penile cancers.
0 True 0 False

65. HPV symptoms are easily observed.
0 True 0 False

Reader Feedback

Once again, please consider the article about human Papillomavirus
that you read a few moments ago. The four questions below refer to that
article.

For each of the following statements, please indicate how much you
disagree or agree by selecting the appropriate response along the
indicated range (Note: the far left button implies strong disagreement
while the far right button implies strong agreement).

66. To me, the advice given in the article is believable.

Strongly 0 0 0 O 0 O O Strongly
Disagree Agree

67. In my opinion, the information given in the article seems useful.

Strongly 0 O 0 O 0 O 0 Strongly
Disagree Agree

68. The advice given in the article is accurate.

Strongly 0 O O O O 0 0 Strongly

Disagree Agree
-l85-

 

 

69. The information I read in the article looks true to me.

Strongly 0 O 0 0 0 O O Strongly
Disagree Agree

 

 

-l86-

(Page 9 of the pilot test)

News Reaction

Future Plans

Below you will see several statements regarding your personal thoughts
on human Papillomavirus and things you can do to help protect yourself
against the disease.

For each of the following statements, please indicate how much you
disagree or agree by selecting the appropriate response along the
indicated range (Note: the far left button implies strong disagreement
while the far right button implies strong agreement).

70. Human Papillomavirus is a serious threat.

Strongly 0 o O 0 O 0 O Strongly
Disagree Agree

71. Human Papillomavirus is harmful.
Strongly 0 0 O 0 O O 0 Strongly
Disagree Agree

72. Human Papillomavirus is a severe threat.
Strongly 0 0 0 O O 0 0 Strongly
Disagree Agree

73. I am at risk for human Papillomavirus.

Strongly 0 O O O O O O Strongly
Disagree Agree

74. It is possible that I will get human Papillomavirus.
Strongly 0 O O 0 O 0 O Strongly
Disagree Agree

75. I am vulnerable to human Papillomavirus.
Strongly 0 0 0 o 0 0 o Strongly
Disagree Agree

76. Becoming more informed will act as a key ﬁrst step in preventing human
Papillomavirus.

Strongly 0 O O O 0 o O Strongly

Disagree Agree

-187-

 

77. Getting more information will work in deterring human Papillomavirus by
acting as a helpful ﬁrst step.

Strongly 0 o O o 0 0 0 Strongly

Disagree Agree

78. I am able to get more information about how to prevent human
Papillomavirus.

Strongly 0 0 O 0 O O O Strongly

Disagree Agree

79. I am able to get more information about removing the threat of human
Papillomavirus.

Strongly 0 O O 0 O O O Strongly

Disagree Agree

80. It is easy to get more information about preventing human Papillomavirus.

Strongly 0 O 0 O O 0 O Strongly
Disagree Agree

81. I can get more information to prevent against human Papillomavirus.

Strongly 0 0 O O O 0 0 Strongly
Disagree Agree

-188-

 

 

 

(Page 10 of the pilot test)

News Reaction

Future Plans

Below you will see several statements regarding your future plans and
intentions about things you can do to help protect yourself against“
human Papillomavirus.

For each of the following statements, please indicate how much you
disagree or agree by selecting the appropriate response along the
indicated range (Note: the far left button implies strong disagreement
while the far right button implies strong agreement).

82. I intend to seek more information about human Papillomavirus.

Strongly 0 0 0 O 0 O 0 Strongly
Disagree Agree

 

83. I am planning to seek more information about human Papillomavirus.

Strongly 0 O 0 0 0 O O Strongly
Disagree Agree

84. I intend to use a condom every time I have sex as a way to decrease my
risk of human Papillomavirus.

Strongly 0 O 0 O O 0 0 Strongly

Disagree Agree

85. I am planning to use a condom every time I have sex as a way to decrease
my risk of human Papillomavirus.

Strongly 0 O O O O O O Strongly

Disagree Agree

86. I intend to have sex with only one partner as a way to decrease my risk of
human Papillomavirus.

Strongly 0 O 0 O 0 0 0 Strongly

Disagree Agree

87. I am planning to have sex with only one partner as a way to decrease my
risk of human Papillomavirus.

Strongly 0 O 0 0 O O o Strongly

Disagree Agree

 

-l89-

(Page 11 of the pilot test: note that in the online version the phrases
“comprehensive Web site,” ‘question and answer segment,” and “CDC Web site”
were all links to appropriate Centers for Disease Control Web pages)

News Reaction

Thank you for taking the time to complete our survey. Please use the space
below to enter your Wotmania user name. One user name will be selected at
random to receive the $100 prize. The site administrator will contact the winner
about one week after data collection ends.

We also would like to encourage you to learn more about how to protect yourself
from human Papillomavirus.

The Centers for Disease Control and Prevention has a comprehensive Web site
that offers everything from basic facts to treatment options. This includes an

informative question and answer segment intended to cover most people's basic
concerns.

HPV information speciﬁcally relevant for men can be found on the CDC Web site
as well.

Thank you again for your time.

Lastly, please use the space below to leave any comments you have about our
survey.

(blank box for qualitative comments)

-l90-

 

 

 

 

APPENDIX D:

Data Collection Instrument

 

-191-

Each respondent ﬁrst received the e-mail included below. Those wishing to
participate followed a link to the survey.

Subject: You ’ve been chosen for a survey!

Dear {NAME},

You’ve been randomly chosen from the Virtual Consumer Research
Group (VCRG) Panel at the University of Texas at Austin to participate
in an online survey.

As more and more media outlets cover health topics, it is important to
learn more about how people understand and react to this news
coverage. This study is intended to learn more about how pe0ple learn
from these news stories and what can make them better. .

You may complete this survey at any time of your choice during the
next 7 days and it should take you no more than 25 minutes to ﬁnish.
Your responses will help us understand what readers think of health
news stories and will remain strictly confidential. In fact, you will not
be asked to identify yourself individually within the survey. By
completing the survey, you are indicating your voluntary consent to
participate in this study; however, you may discontinue or refuse to
take part at anytime, and your responses will not be processed unless
you submit the survey upon completion.

If you complete the survey, you will be entered into a cash drawing for
one of 10 $100 prizes. To participate, please proceed to the following
URL: {URL INSERTED
HERE}http://adresearch.advertising.utexas.edu/survey005.html.

If you should have any questions about this survey before you begin,
please feel free to contact Brad Love, via e-mail at Iovebrad@msu.edu.

Thanks for your participation!

-192-

The link included in the e-mail randomly sent the participants to one of the ﬁve treatment
conditions, each of which began with the following informed consent notice:

“Welcome, and thank you for taking the time to participate in our study, which
should last less than 25 minutes!

INTRODUCTION

You are being invited to participate in a research study about how people
view mass media coverage of health topics. Researchers at Michigan State
University will perform the study. Participants will need about 25 minutes to
complete the survey. In return for completing the survey, respondents will be
entered into drawing to win one of 10 $100 prizes. Prizes will be distributed in
approximately one month after data collection is complete.

PURPOSE
The purpose of this study is to determine how people understand and view
coverage of health-related topics in the mainstream media.

WHAT IS INCLUDED IN THIS PROJECT?

Participants will be asked to read an online news story and complete survey
questions regarding the story and their reactions to it. Participants will also
answer survey questions about their personal views on the world around them.
Completing the reading and questions is estimated to take 25 minutes.

PROCEDURE

If you decide to take part in this study, you will be asked to consent to the
items discussed below. All of the information collected for this project will be kept
conﬁdential to the maximum extent allowed by the law. You may be asked to
complete a series of surveys about your reactions to the news media and your
views on your environment.

PARTICIPATION

Participation in this study is entirely voluntary. In addition, should you choose
to participate in the overall study, you may still decline to participate in any
segment of the survey. You may also choose to discontinue your participation at
any time. Any refusal to participate or decision to discontinue will not affect any
future opportunities to participate in surveys offered through this online panel.

RISKS

Although it is highly unlikely, there is a chance that you might feel
uncomfortable with some of the questions during the course of the survey
because they deal with health topics. If at any time during the course of the
survey, you wish to refrain from answering a question or simply wish to stop

-l93-

participating, the survey can be postponed and/or terminated.
Participation in this study presents no physical threat.

BENEFITS

We cannot and do not guarantee you will directly beneﬁt if you take part in
this study. Through your answers, we hope to gain valuable information about
the use of news media and more speciﬁcally coverage of health-related issues.
It is our hope that the information we receive from you will help improve the
quality of media coverage of health topics.

PAYMENT TO SUBJECTS

As a participant in this research study, you will be entered into a random
drawing to win one of 10 prizes of $100. Every respondent will have an equal
opportunity to win a $100 prize. As such, we cannot guarantee payment to any
individual, but everyone completing the survey will be equally likely to receive
$100 for his or her time.

COSTS
Participation in this project will not cost anything. There will be no ﬁnancial
responsibility for any respondent nor any penalty for not participating.

ALTERNATIVES
The alternative to participating in this research study is to decline.
Participation is purely voluntarily and for research purposes only.

CONFIDENTIALITY

Investigators will keep secret all research related to records and information
gathered by this study. Throughout the study, all information from surveys will be
stored in a manner to assure conﬁdentiality of the subjects to the maximum
extent allowable under law. Your privacy will be protected to the maximum
extent allowable by law. The investigators will not reveal your identity if they
publish the results of the study.

QUESTIONS

Through this form, the investigators have tried to answer questions to your
satisfaction. If you have any particular questions about this study, please contact
Brad Love by phone: (517)432-6186, e-mail: Iovebrad@msu.edu, or regular
mail: 305 Communication Arts & Sciences, East Lansing MI 48824. If you have
any questions or concerns regarding your rights as a study participant, or are
dissatisﬁed at any time with any aspect of this study, you may contact -
anonymously, if you wish - Peter Vasilenko, Ph.D., Director of the Human
Subject Protection Programs at Michigan State University, by phone: (517) 355-
2180, fax: (517) 432-4503, email: irb@msu.edu, or regular mail: 202 Olds Hall,
East Lansing, MI 48824.

-194—

CONSENT

The investigator(s) gave you information about what will be done in this
research study. You also were told how it will be done, what you will have to do,
and how long the research will take. The information in this form attempted to tell
you about any inconvenience, discomfort, or risks you might experience due to
this research.

By clicking below, you agree to take part in this study as a research
subject. You are aware that you may quit or refuse to participate in certain parts
of the research at any time.”

Accept Decline

Those agreeing to participate continued to the survey. Those declining to
participate were taken to the Michigan State University homepage. The survey instrument
appeared as follows. Please note that the format used for online testing has been
approximated here as best as possible. Certain formatting norms of online publication,
such as thinner columns, do not translate well to the printed page. In addition, certain
functional survey items such as answer-selection buttons do not always reproduce in
word-processing documents. Pages of the survey are indicated in parentheses before the
start of each page. In the online version, page progress was indicated by text at the
bottom of each page that let respondents know both their progress and the pages
remaining by saying, for example, “Page 6 of 14.”

-l95-

(Page 1 of the ﬁnal survey)

News Reaction
Survey Outline

Thank you for agreeing to complete our survey. We appreciate you taking the
time to offer your opinions.

The survey will take you through a few questions about your personal
preferences, ask you about Information obtained from a newspaper article, and
then ﬁnish with some more survey questions.

Also, please maximize your Internet browser window to full-screen size as you
take the survey. This will ensure that all the questions are visible.

Thank you again for your time. Please click the button below to continue.

-196-

(Page 2)

News Reaction

We would like to start by asking you some general questions. The following items
are all intended to help us learn a little about your background.

Please select the appropriate button for your gender.
1. Gender

Male
Female

2. What Is your age? (Please enter "99" if you prefer to not answer.)

[___

3. What is your ethnicity? Please check multiple categories if appropriate.

Asian / Paciﬁc Islander/Native Hawaiian
Black / African American

Hispanic / Latino

White

Native American /Alaska Native

Other: I

Prefer not to respond

4. What is your country of residence?

USA.
Canada
Mexico

Other: I

5. What is your current relationship status (optional)?

Single

-l97-

Committed Relationship
Engaged
Married

~198-

(Page 3)

N eWS ReaCtlonContinuing with the general questions, we would like

to know a little about your media use habits.

6. Where do you get your news? Please mark the items that best fit
yourskuaﬁon.

Newspapers (Hard Copy)
Newspapers (Online)

News Magazines (Hard Copy)
News Magazines (Online)

Blogs

T.V.

Radio

I don't regularly follow the news.

* 7. Newspapers (Hard Copy):

3

* 8. Newspapers (Online):

El

* 9. News Magazines (Hard Copy):

3

* 10. News Magazines (Online):

E]

* 11. Blogs:

3

-l99-

* 12. TV:

01—31

* 13. Radio:

-200-

(Page 4)

News Reaction

Still thinking in general terms, we would like to know about your experiences with
certain medical situations. Do you have any experience with the following?
Check all that apply or please select "none" if you prefer to not answer:

14. Human Papillomavirus:

Personal Family Other N
. . . one
history history experience
15. Prostate Cancer:
Personal Family Other N
. . . one
history history experience
16. Food Poisoning:
Personal Family Other N
. . . one
history history experience
17. Breast Cancer:
Personal Family Other N
. . . one
history history experience

18. Have you, any close friends, or anyone in your family ever
participated in a clinical trial research study?

No
Yes
Don't know

-201-

(Page 5)

News Reaction
Interest and Preference Survey

For each of the following statements, please indicate how much you disagree or

agree by selecting the appropriate response along the indicated range (Note: the

far left button implies strong disagreement while the far right button implies

strong agreement).

19. I would like to explore strange places.

Strongly
Disagree
1 2 3 4

20. I get restless when I spend too much time at home.

Strongly
Disagree
1 2 3 4

21. I like to do frightening things.
Strongly
Disagree
1 2 3 4

-202-

Strongly
Agree

5

Strongly
Agree
5

Strongly
Agree
5

 

 

 

22. I like wild parties.

Strongly Strongly
Disagree Agree
1 2 3 4 5

23. I would like to take off on a trip with no pre-planned routes or

timetables.
Strongly Strongly
Disagree Agree

1 2 3 4 5

24. I prefer friends who are excitingly unpredictable.

Strongly Strongly
Disagree Agree
1 2 3 4 5

25. I would like to try bungee jumping.

Strongly Strongly
Disagree Agree
1 2 3 4 5

-203-

 

26. I would like to have new and exciting experiences, even if they

are illegal.
Strongly Strongly
Disagree Agree

1 2 3 4 5

-204-

 

(Page 6)

News Reaction
Preference Survey, continued

For each of the following statements, please indicate how much you disagree or
agree by selecting the appropriate response along the indicated range (Note: the
far left button implies strong disagreement while the far right button implies
strong agreement).

27. I would prefer complex problems to simple ones.

Strongly 01 02 03 04 05 06 07 Strongly
Disagree Agree

28. I like having the responsibility of handling a situation that
requires a lot of thinking.
Strongly 01 02 03 04 05 06 07 Strongly
Disagree Agree

29. Thinking is not my idea of fun.
Strongly 0 1 0 2 O 3 0 4 0 5 0 6 0 7 Strongly
Disagree Agree

30. I would rather do something that requires little thought than
something that is sure to challenge my thinking abilities.

Strongly 01 02 O3 04 05 06 07 Strongly
Disagree Agree

31. I try to anticipate and avoid situations where there is a good
chance I will have to think in depth about something.
Strongly 0 1 02 O3 04 05 O6 O7 Strongly
Disagree Agree

32. I ﬁnd satisfaction in thinking hard and for long hours.
Strongly 0 1 0 2 0 3 0 4 O 5 O 6 O 7 Strongly
Disagree Agree

33. I only think as hard as I have to.
Strongly 0 1 o 2 o 3 o 4 o 5 o 6 o 7 Strongly
Disagree Agree

34. I like tasks that require little thought once I've learned them.
Strongly 0 1 o 2 o 3 o 4 o 5 o 6 o 7 Strongly
Disagree Agree

-205-

(Page 7)

News Reaction

Preference Survey, continued

Again, for each of the following statements, please indicate how much you
disagree or agree by selecting the appropriate response along the indicated
range (Note: the far left button implies strong disagreement while the far right
button implies strong agreement).

35. The idea of relying on thought to make my way to the top
appeals to me.
Strongly 01 02 03 04 05 O6 07 Strongly
Disagree Agree

36. I really enjoy a task that involves coming up with new solutions

to problems.
Strongly 0 1 O 2 O 3 0 4 0 5 O 6 0 7 Strongly
Disagree Agree

37. Learning new ways to think does not excite me very much.
Strongly 0 1 0 2 O 3 0 4 0 5 O 6 O 7 Strongly
Disagree ‘ Agree

38. I prefer my life to be ﬁlled with puzzles that I must solve.

Strongly 01 02 O3 O4 05 O6 07 Strongly
Disagree Agree

39. The notion of thinking abstractly is appealing to me.
Strongly 0 1 0 2 0 3 O 4 O 5 O 6 O 7 Strongly
Disagree Agree

40. I would prefer a task that is intellectual, difﬁcult, and important
to one that is somewhat important but does not require much thought.

Strongly 01 02 03 04 05 06 O7 Strongly
Disagree Agree

41. I feel relief rather than satisfaction after completing a task that
required a lot of mental effort.
Strongly 01 02 03 O4 05 O6 O7 Strongly
Disagree Agree

-206—

42. I prefer to think about small, daily projects rather than long-
term ones.

Strongly 01 02 03 O4 05 06 07 Strongly
Disagree Agree

43. It’s enough for me that something gets the job done; I don’t
care how or why it works.

Strongly 01 02 03 O4 05 O6 O7 Strongly
Disagree Agree

44. I usually end up thinking about issues even when they do not
affect me personally. .
Strongly 0 1 0 2 O 3 O 4 0 5 o 6 O 7 Strongly
Disagree Agree

-207-

(Page 8)

News Reaction

Now, consider the topic of human Papillomavirus and how it might affect you.
Please answer each of the following questions.

Indicate how much you disagree or agree by selecting the appropriate response

along the indicated range (Note: the far left button implies strong disagreement
while the far right button implies strong agreement).

45. A local outbreak of human Papillomavirus would affect me

personally.
Strongly 0 1 O 2 0 3 0 4 0 5 0 6 0 7 Strongly
Disagree Agree

46. A local outbreak of human Papillomavirus would have a big
impact on my life.

Strongly 01 02 O3 O4 05 O6 O7 Strongly
Disagree Agree

47. A local outbreak of human Papillomavirus would have little
effect on me personally.
Strongly 01 02 03 04 05 06 O7 Strongly
Disagree Agree

-208-

(Page 9)

News Reaction

The next page will take you to a newspaper article about human Papillomavirus.
Please read the article and respond to the questions that follow.

-209—

(Page 10)
News Reaction

Most Americans Exposed to HPV at Some Point, Experts Say

AP Associated Press

Chances are you have been exposed to the potentially cancer-causing human
Papillomavirus virus and don't even know it. In fact, at least three out of four sexually
active adults under 50 have been infected with a strain of HPV, according to data
released by public health ofﬁcials this week.

In the United States alone, up to 6.2 million new cases occur each year, and at least
20 million Americans are now infected with the genital strain of the most commonly
acquired STD, according to public health ofﬁcials.

Almost all infections cause no symptoms or health problems, but nearly all cases of
cervical cancer stem from HPV. Some vulvar, anal, and penile cancers come ﬁ'om

HPV, too. Public health ofﬁcials guess that cervical cancers kill about one-third of
those stricken each year.

There are about 200 types of HPV. About 40 strains are spread through sexual
contact, and a handful of those are the ones linked with cervical cancer. However,

because the disease is so common and can have such a deadly outcome, public health I

experts rank HPV to be among the most serious health issues facing young men and
women in their teens and twenties.

HPV can only be prevented by not having sex. Public health experts strongly suggest
becoming more informed as a key ﬁrst step. A Web site such as the Center for
Disease Control's at www.cdc.gov/std/hpv/ can be useful before talking to a doctor.
Having sex with only one partner is also an effective way to decrease your risk. In the
future, a new vaccine may offer some protection from HPV-related cancers. Always
using a condom can help to reduce risk, too.

Please select "Next" to conﬁrm that you have completed reading the article
and are ready to move on to the next question. Thank you.

210

 

 

(Page I 1)

News Reaction

Article Questions
Thank you for reading the article on human Papillomavirus. Below are four

questions relating to the content of the article. Please answer each question to
the best of your ability.

48. According to the article, human Papillomavirus (HPV) is the
most common STD.

True False

49. Becoming more informed is a key ﬁrst step in reducing one’s risk
for human Papillomavirus.

True False

50. Human Papillomavirus causes nearly all cases of cervical cancer,
as well as some anal and penile cancers.

True False
51. HPV symptoms are easily observed.

True False
Reader Feedback

Once again, please consider the article about human Papillomavirus that you
read a few moments ago. The four questions below refer to that article.

For each of the following statements, please indicate how much you disagree or
agree by selecting the appropriate response along the indicated range (Note: the
far left button implies strong disagreement while the far right button implies
strong agreement).

52. To me, the advice given in the article is believable.
Strongly 0 1 O 2 O 3 0 4 O 5 0 6 O 7 Strongly
Disagree Agree

-211-

53. In my opinion, the information given in the article seems useful.

Strongly 01 02 O3 04 05 06 07 Strongly
Disagree Agree

54. The advice given in the article is accurate.

Strongly 01 02 03 04 05 O6 07 Strongly
Disagree Agree

55. The information I read in the article looks true to me.

Strongly 01 02 O3 04 05 O6 O7 Strongly
Disagree Agree

-212-

(Page 12)

News Reaction

Future Plans

Below you will see several statements regarding your personal thoughts on
human Papillomavirus and things you can do to help protect yourself against the
disease.

For each of the following statements, please indicate how much you disagree or
agree by selecting the appropriate response along the indicated range (Note: the
far left button implies strong disagreement while the far right button implies
strong agreement).

56. Human Papillomavirus is a serious threat.

Strongly 01 02 O3 04 05 06 07 Strongly
Disagree Agree

57. I am at risk for human Papillomavirus.

Strongly 01 02 O3 04 05 O6 07 Strongly
Disagree Agree

58. Human Papillomavirus is harmful.

Strongly 01 02 03 04 05 06 O7 Strongly
Disagree Agree

59. It is possible that I will get human Papillomavirus.

Strongly 01 02 03 04 05 06 07 Strongly
Disagree Agree

60. Human Papillomavirus is a severe threat.

Strongly 01 02 03 04 05 06 07 Strongly
Disagree Agree

61. I am vulnerable to human Papillomavirus.

Strongly 01 02 O3 04 05 06 07 Strongly
Disagree Agree

-213-

(Page 13)

News Reaction

Future Plans

Once again, you will see below more statements regarding your personal
thoughts on human Papillomavirus and things you can do to help protect yourself
against the disease.

For each one, please continue to indicate how much you disagree or agree by
selecting the appropriate response along the indicated range (Note: the far left
button implies strong disagreement while the far right button implies strong
agreement).

62. I am able to get more information about how to prevent human
Papillomavirus.

Strongly 01 02 03 O4 05 06 O7 Strongly
Disagree Agree

63. Becoming more informed will act as a key first step in
preventing human Papillomavirus.

Strongly 01 02 O3 O4 05 O6 07 Strongly
Disagree Agree

64. Getting more information will work in deterring human
Papillomavirus by acting as a helpful first step.

Strongly 01 02 03 04 05 06 07 Strongly
Disagree Agree

65. I am able to get more information about removing the threat of
human Papillomavirus.
Strongly 01 02 O3 04 05 06 07 Strongly
Disagree Agree

66. It is easy to get more information about preventing human
Papillomavirus.

Strongly 01 02 03 04 05 06 07 Strongly
Disagree Agree

67. I can get more information to prevent against human
Papillomavirus.
Strongly 01 02 03 04 05 06 07 Strongly
Disagree Agree

-214-

 

 

(Page 14: note that in the online version the phrases “comprehensive Web site,”
‘question and answer segment,” and “CDC Web site" were all links to appropriate
Centers for Disease Control Web pages)

News Reaction

Future Plans

Below you will see several statements regarding your future plans and intentions
about things you can do to help protect yourself against human Papillomavirus.

For each of the following statements, please indicate how much you disagree or
agree by selecting the appropriate response along the indicated range (Note: the
far left button implies strong disagreement while the far right button implies
strong agreement).

68. I intend to seek more information about human Papillomavirus.
Strongly 0 1 O 2 0 3 O 4 O 5 O 6 0 7 Strongly
Disagree Agree

69. I am planning to seek more information about human
Papillomavirus.

Strongly 01 02 03 04 05 06 O7 Strongly
Disagree Agree

70. I intend to use a condom every time I have sex as a way to
decrease my risk of human Papillomavirus.
Strongly 01 02 03 O4 05 O6 07 Strongly
Disagree Agree

71. I am planning to use a condom every time I have sex as a way
to decrease my risk of human Papillomavirus.

Strongly 01 02 03 O4 05 O6 07 Strongly
Disagree Agree

72. I intend to have sex with only one partner as a way to decrease
my risk of human Papillomavirus.
Strongly 01 02 03 04 05 06 07 Strongly
Disagree Agree

73. I am planning to have sex with only one partner as a way to
decrease my risk of human Papillomavirus.
Strongly 0 1 02 O3 04 05 O6 07 Strongly
Disagree Agree

-215-

Thank you

Thank you for taking the time to complete our survey. Ten people who have
completed the survey will be selected at random to receive a $100 prize. The
research panel administrator will contact the winners as soon as possible after
data collection ends.

We also would like to encourage you to learn more about how to protect yourself
from human Papillomavirus.

The Centers for Disease Control and Prevention has a comprehensive Web site
that offers everything from basic facts to treatment options. This includes an
informative question and answer segment intended to cover most people's basic
concerns. The American Social Health Association also offers an easy-to-read
introductory site for anyone concerned about the HPV virus.

HPV information speciﬁcally relevant for men can be found on the CDC Web site
as well.

Thank you again for your time.

Lastly, please use the space below to leave any comments you have about our
survey.

74. Comments
(blank text box)

-2l6-

BIBLIOGRAPHY

-217-

BIBLIOGRAPHY

Aiken, L., & West, S. (1991). Multiple regression: testing and interpreting interactions.
Thousand Oaks, CA: Sage Publications.

Alterman, E. (2003). What Liberal Media? The Truth About Bias and the News New
York: Basic Books.

Anderson, J ., & Gerbing, D. (1984). The effects of sampling error on convergence,
improper solutions, and goodness-of-ﬁt indices for maximum likelihood conﬁrmatory
factor analysis. Psychometrika, 49(2), 155-173.

Anderson, J., & Gerbing, D. (1988). Structural equation modeling in practice: A review
and recommended two-step approach. Psychological Bulletin, 103(3), 411-423.

Bagozzi, R., Yi, Y., & Phillips, L. (1991). Assessing construct validity in organizational
research. Administrative Science Quarterly, 36(3), 421-458.

Baker, J., Parasuraman, A., Grewal, D., & Voss, G. (2002). The inﬂuence of multiple
store environment cues on perceived merchandise value and patronage intentions. .
Journal of Marketing, 66(2), 120-141.

Baker, J ., Woods, R., & Reynolds, R. (Eds). (2007). Usage of Online Surveys and
Measurements Hershey, PA: Idea Group Reference.

Bakker, A. (1999). Persuasive communication about AIDS prevention: need for cognition
determines the impact of message format. AIDS Education and Prevention, 11(2), 150-
162.

Bandura, A. (1994). Self-efﬁcacy. In V. S. Ramachaudran (Ed.), Encyclopedia of human
behavior (Vol. 4, pp. 71-81). New York: Academic Press.

Baron, R., & Kenny, D. (1986). The moderator-mediator variable distinction in social
psychological research: Conceptual, strategic and statistical considerations Journal of
Personality and Social Psychology, 51, 1173-1182.

Beck, K., & Davis, C. (1978). Effects of fear-arousing communications and topic
importance on attitude change. Journal of Social Psychology, 104, 81-95.

Beck, K., & Frankel, A. (1981). A conceptualization of threat communications and
protective health behavior. Social Psychology Quarterly, 3, 204-217.

Ben-Sira, Z. (1981). Latent fear-arousing potential of fear-moderating and fear-neutral
health promoting information. Social Science and Medicine, 15, 105-112.

-218-

Bentler, P., & Chou, C.-P. (1987). Practical Issues in Structural Modeling. Sociological
Methods and Research, 16, 78-117.

Berkowitz, L., & Cottingham, D. (1960). The interest value and relevance of fear-
arousing communications. Journal of Abnormal and Social Psychology, 60, 37-43.

Bettinghaus, E. (1968). Persuasive communication. New York: Holt, Rinehart &
Winston.

Block, L. (2005). Self-Referenced Fear and Guilt Appeals: The Moderating Role of Self-
Construal. Journal of Applied Social Psychology, 35(11), 2290-2309.

Block, L., & Williams, P. (2002). Undoing the Effects of Seizing and Freezing:
Decreasing Defensive Processing of Personally Relevant Messages. Journal of Applied
Social Psychology, 32(4), 803-833.

Block, M. (1997). Writing Broadcast News: Shorter, Sharper, Stronger (2nd ed.).
Chicago: Bonus Books.

Bohlin, N. (1962). Safety Belt. In U. S. P. a. T. Ofﬁce. (Ed), United States Patent and
Trademark Oﬂice. United States.

Boster, F., & Mongeau, P. (1984). F ear-arousing persuasive messages. In R. Bostrom &
B. Westley (Eds), Communication Yearbook 8 (pp. 330-375). Beverly Hills, CA: Sage.

Brehm, J. W. (1966). Theory of psychological reactance. New York: Academic Press.

Brehm, S. S., & Brehm, J. W. (1981). Psychological reactance: A theory of freedom and
control. New York: Academic Press.

Brown, R. (1997). Assessing Speciﬁc Mediational Effects in Complex Theoretical
Models. Structural Equation Modeling, 4(2), 142-156.

Burnett, J. (1981). Intemal-extemal locus of control as a moderator of fear appeals.
Journal of Applied Psychology, 66, 390-393.

Burnett, J ., & Oliver, R. (1979). Fear appeal effects in the ﬁeld: A segmentation
approach. Journal of Marketing Research, 16, 181-190.

Burns, L. S. (2002). Understanding Journalism. Thousand Oaks, CA: Sage Publications.

Cacioppo, J ., & Petty, R. (1982). The need for cognition. Journal of Personality and
Social Psychology, 42, 1 16-131 .

Cacioppo, J ., Petty, R., Feinstein, J ., & Jarvis, W. (1996). Dispositional differences in
cognitive motivation: The life and times of individuals varying in need for cognition.
Psychological Bulletin, 119, 197-253.

-219-

Cacioppo, J., Petty, R., & Kao, C. (1984). The efﬁcient assessment of need for cognition.
Journal of Personality Assessment, 48, 306-307.

Callison, C. (2001). Do PR Practitioners Have a PR Problem?: The Effect of Associating
a Source With Public Relations and Client-Negative News on Audience Perception of
Credibility. Journal of Public Relations Research, 13(3), 219-234.

Cameron, G., Leshner, G., Stephenson, M., & Choi, Y. (2006). Effects of High Threat
Anti-Smoking Public Service Announcements on High Sensation Seeking Young Adults,
Paper presented at the annual meeting of the International Communication Association.
New York, New York.

Campbell, D. (1960). Recommendations for APA Test Standards Regarding Construct,
Trait, or Discriminant Validity. American Psychologist, 15, 546-553.

Chaiken, S. (1980). Heuristic versus Systematic Information Processing and the Use of

Source versus Message Cues in Persuasion. Journal of Personality and Social
Psychology, 39, 752-766.

Chaplin, W. (1991). The next generation of moderator research in personality
psychology. Journal of Personality, 59(2), 143-178.

Chaplin, W. (1997). personality, interactive relations, and applied psychology. In R.
Hogan, J. Johnson & S. Briggs (Eds), Handbook of Personality Psychology. New York:
Academic Press.

Cheah, W. H. (2006). Issue Involvement, Message Appeal and Gonorrhea: Risk
Perceptions in the US, England, Malaysia and Singapore. Asian Journal of
Communication, 16(3), 293-314.

Cho, H., & Boster, F. (2005). Development and Validation of Value-, Outcome-, and
Impression-Relevant Involvement Scales. Communication Research, 32(2), 235-264.

Cho, H., & Salmon, C. (2006). F ear Appeals for Individuals in Different Stages of
Change: Intended and Unintended Effects and Implications on Public Health Campaigns.
Health Communication, 20(1), 91-99.

Chu, G. (1966). Fear arousal, efﬁcacy and imminency. Journal of Personality and Social
Psychology, 5, 517-524.

Cook, T. D., & Campbell, D. T. (1979). Quasi-Experimentation." Design and Analysis for
Field Settings. Chicago: Rand McNally.

Covello, V., von Winterfeldt, D., & Slovic, P. (1986). Risk communication: A review of
the literature. Risk Abstracts 3, 171-182.

-220-

Cramer, D. (1994). Introducing statistics for social research: Step-by-step calculations
and computer techniques using SPSS. London: Routledge.

Crano, W. (1970). Effect of sex, response order, and expertise in conformity. A
dispositional approach. Sociometry, 33, 239-252.

Cronbach, L. (1971). Test Validation. In R. Thorndike (Ed.), Educational Measurement
(2nd ed., pp. 443-507). Washington, DC: American Council on Education.

Cronbach, L., Gleser, G., Nanda, H., & Rajaratnam, N. (1972). The Dependability of
Behavioral Measurements: The Theory of Generalizability for Scores and Proﬁles. New
York: John Wiley & Sons, Inc.

Croy, C. D., & Novins, D. K. (2005). Methods for Addressing Missing Data in
Psychiatric and Developmental Research. Journal of the American Academy of Child &
Adolescent Psychiatry, 44(12), 1230-1240.

Danielson, W., & Bryan, S. (1964). Readability of wire stories in eight news categories.
Journalism Quarterly, 41, 105-106.

DeJong, W., & Winsten, J. (1990). The Use of Mass Media in Substance Abuse
Prevention. Health Aﬂairs, 9(2), 30—46.

Dillard, J. (1994). Rethinking the study of fear appeals: An emotional perspective.
Communication Theory, 4, 295-323.

Dillard, J., Plotnick, C., Godbold, L., Freimuth, V., & Edgar, T. (1996). The multiple
affective outcomes of AIDS PSAs: Fear appeals do more than scare people.
Communication Research, 23, 44-72.

Donohew, R. L., Palmgreen, P., & Pugzles Lorch, E. (1994). Attention, sensation
seeking, and health communication campaigns. American Behavioral Scientist, 38, 310-
332.

Donohew, R. L., Zimmermanb, R., Cuppa, P., Novakb, S., Colonc, S., & Abella, R.
(2000). Sensation seeking, impulsive decision-making, and risky sex: implications for
risk-taking and design of interventions. Personality and Individual Differences, 28, 1079-
1091.

DuBay, W. (2005, 15 January 2007). What's with the newspapers. Retrieved 15, from
http://www.impact-information.com/impactinfo/newsletter/plwork1 5 .htm

Etheridge, E. (1992). Sentinel for Health: A History of the Centers for Disease Control.
Berkeley, CA: University of California Press.

-221-

Evans, N., Gilpin, E., Farkas, A., Shenassa, E., & Pierce, J. (1995). Adolescents'
perceptions of their peers' health norms. American Journal of Public Health, 85(8), 1064-
1069.

Fahey, J. (2003, Nov. 26, 2003). OnStar's Scare Tactics. Retrieved Nov. 19, 2006, from
http://www.forbes.com/lifestyle/2003/1 l/26/cz_jf__1 126feat.html

F an, D., Wyatt, R., & Keltner, K. (2001). The Suicidal Messenger: How Press Reporting
Affects Public Conﬁdence in the Press, the Military, and Organized Religion
Communication Research, 28 ‘(6), 826.

Field, A. (2005). Discovering Statistics Using SPSS: (and Sex, Drugs, and Rock 'n' Roll).
London, CA: Sage Publications.

Finckenauer, J. (1982). Scared Straight and the Panacea Phenomenon. Englewood
Cliffs, NJ : Prentice Hall.

Finnegan, J ., Viswanath, K., & Hertog, J. (1999). Mass media, secular trends, and the
future of cardiovascular health promotion: an interpretive analysis. Preventative
Medicine, 6 Pt 2(29), S50-S58.

Floyd, D., Prentice-Dunn, S., & Rogers, R. (2000). A meta-analysis of research on
protection motivation theory. Journal of Applied Social Psychology, 30, 407-429.

Fowlkes, M., & Miller, P. (1987). Community and Risks at Love Canal. In B. Johnson &
V. Covello (Eds), The Social and Cultural Construction of Risk: Essays on RIsk
Selection and Perception. Dordrecht, Holland: Reidel.

Frandsen, K. (1963). Effects of threat appeals and media of transmission. Speech
Monographs, 30, 101-104.

Frazier, P. A., Tix, A. P., & Barron, K. E. (2004). Testing moderator and mediator effects
in counseling psychology research. Journal of Counseling Psychology, 51, 115-134.

Gallo, A. (2005). Food Advertising in the United States. In F. a R. E. Division. (Ed.):
Economic Research Service of the US Department of Agriculture.

Gans, H. (1985). Are US. Journalists Dangerously Liberal? Columbia Journalism
Review, 24, 32.

Gaur, A., & Gaur, S. (2006). Statistical Methods for Practice And Research: A Guide to
Data Analysis Using SPSS Thousand Oaks, CA: Sage Publications.

Geller, E. (1989). Using television to promote safety belt use. In R Rice & C. Atkin
(Eds), Public Communication Campaigns (pp. 201-203). Newbury Park, CA: Sage.

-222-

 

 

Giesen, M., & Hendrick, C. (1974). Effects of false positive and negative arousal
feedback on persuasion. Journal of Personality and Social Psychology, 4, 449-457.

Goldstein, M. (1959). The relationship between coping and avoiding behavior and
response to fear arousing propaganda. Journal of A bnormal and social psychology, 58,
247-252.

Gore, T., & Bracken, C. (2005). Testing the Theoretical Design of a Health Risk
Message: Reexamining the Major Tenets of the Extended Parallel Process Model Health
Education & Behavior, 32(1), 27-41.

Gorely, T., & Gordon, S. (1995). An examination of the transtheoretical model and
exercise behavior in older adults. Journal of Sport & Exercise Psychology, 1 7, 312—324.

Grabd, M. E., Zhou, S., Lang, A., & Bolls, P. D. (2000). Packaging Television News: The
Effects of Tabloid on Information Processing and Evaluative Responses. Journal of
Broadcasting and Electronic Media, 44(4), 581-598.

Grace-Martin, K. (2002). Recommended Solutions to Missing Data. Ithaca, New York:
Cornell University.

Griffeth, R., & Rogers, R. (1976). Effects of fear-arousing components of driver
education on students' safety attitudes and simulator performance. Journal of Educational
Psychology(68), 501-506.

Hanson, D., & Engs, R. (1992). College students' drinking problems: A national study.
Psychological Reports, 71, 39-42.

Herbert, J. (2001). Practising Global Journalism: Exploring Reporting Issues Worldwide.
Oxford: Focal Press.

Hewgill, M., & Miller, G. (1965). Source credibility and response to fear-arousing
communications. Australian Journal of Psychology, 32, 183-193.

Hewson, C., Yule, P., Laurent, D., & Vogel, C. (2003). Internet research methods : a
practical guide for the social and behavioural sciences. London: Sage Publications.

Higbee, K. (1969). Fifteen years of fear arousal: Research on threat appeals: 1953-68.
Psychological Bulletin, 72(6), 426-444.

Hinton, P. R., Brownlow, C., McMurray, 1., & Cozens, B. (2004). SPSS Explained.
London: Routledge.

Hofstetter, C. R., & Dozier, D. (1986). Useful News, Sensational News: Quality,
Sensationalism and Local TV News Journalism Quarterly, 63(4), 815-820, 853.

-223-

 

Horowitz, I. (1969). Effects on volunteering, fear arousal, and number of
communications on attitude change. Journal of Personality and Social Psychology, 11,
34-37.

Horowitz, I. (1972). Attitude change as a function of perceived arousal. Journal of Social
Psychology, 87, 117-126.

Horowitz, 1., & Gumenik, W. (1970). Effects of the volunteer subject, choice, and fear
arousal on attitude change. Journal of Experimental Social Psychology, 6, 293-303.

Hovland, C., Janis, I., & Kelley, J. (1953). Communication and Persuasion:
Psychological Studies of Opinion Change. New Haven, CT: Yale University Press.

Hovland, C., Janis, I., & Kelly, H. (1953). Communication and Persuasion. New Haven.
CT: Yale University Press.

Hoyle, R., Stephenson, M., Palmgreen, P., Pugzles Lorcha, E., & Donohew, R. L. (2002).
Reliability and validity of a brief measure of sensation seeking Personality and
Individual Differences, 32(3), 401-414.

Hult, G., Hurley, R., Giunipero, L., & Nichols, J., E. (2000). Organizational learning in
global purchasing: A model and test of internal users and corporate buyers. Decision
Sciences, 31(2), 293-325.

Insko, C. A., Arkoff, A., & Insko, V. M. (1965). Effects of high and low fear-arousing
communications upon opinions toward smoking. Journal of Experimental Social
Psychology, 1, 256-266.

Institute of Medicine. (2002). Speaking of health: Assessing health communication
strategies for diverse populations. Washington DC: National Academies Press.

Janis, I. (1967). Effects of fear arousal on attitude change: Recent developments on
theory and research. In L. Berkowitz (Ed.), Advances in experimental social psychology
(Vol. 3). New York: Academic.

Janis, I., & Feshback, S. (1953). Effects of fear-arousing communications. Journal of
Abnormal and social psychology, 48, 78-92.

Janis, I., & F eshback, S. (1954). Personality differences associated with responsiveness to
fear-arousing communications. Journal of Personality, 23, 154-166.

Janis, I., & Leventhal, H. (1968). Human reactions to stress. In E. Borgatta & W.
Lambert (Eds), Handbook of personality theory and research. Chicago: Rand McNally.

Janis, I., & Terwilliger, R. (1962). An experimental study of psychological resistance to
fear arousing communications. Journal of A bnormal and social psychology, 65, 403-410.

-224-

 

 

 

Jansen, K., Corley, K., & Jansen, B. (2007). E-Survey Methodology. In J. Baker, R.
Woods & R. Reynolds (Eds), Handbook of Research on Electronic Surveys and
Measurements (pp. 1-8). Hershey, PA: Idea Group Reference.

Jessor, R. (1982). Critical issues in research on adolescent health promotion. In T.
Coates, A. Peterson & C. Perry (Eds), Promoting adolescent health: A dialogue on
research and practice (pp. 447-465). San Francisco, CA: Academic Press.

Jessor, R. (1984). Adolescent development and behavioral health. In J. Matarazzo, S.
Weiss, J. Herd, N. Miller & S. Weiss (Eds), Behavioral health: A handbook of health

enhancement and disease prevention (pp. 66—89). New York: John Wiley and Sones.

Job, R. (1988). Effective and ineffective use of fear in health promotion campaigns.
American Journal of Public Health, 78, 163-167.

Johnson, B. (1987). Accounting for the Social Context of Risk Communication. Science
& Technology Studies, 5(3/4), 103-111.

Johnson, B., & Eagly, A. (1989). Effects of involvement on persuasion: A meta-analysis.

Psychological Bulletin, 106(2), 290-314.

Johnson, B., & Eagly, A. (1990). Involement and persuasion: Types, traditions, and the
evidence. Psychological Bulletin, 107, 375-384.

Jung, J. (2003). Business News Web Sites Differ from Newspapers in Business Content.
Newspaper Research Journal, 24(2), 114-119.

Jung, R. (1982). Public Affairs and Sensationalism in Local TV News Programs.
Journalism Quarterly, 59(1), 74-78, 137.

Kahn, K. (1994). Does gender make a difference? An experimental examination of sex
stereotypes and press patterns in statewide campaigns. American Journal of Political
Science, 38(1), 162-195.

Kaiser Family Foundation. (2005, April 2006). Kaiser health poll report: Public opinion
snapshot on health information sources-July 2005. Retrieved September 14, 2006, from
httpWWW.kff.org/kaiserpolls/pomr0718050th.cfm

Kaye, B., & Johnson, T. (1999). Research methodology: Taming the cyber frontier.
Social Science Computer Review, 17(3), 323-337.

Kelly, K., & Edwards, R. (1992). Observations: does discussion of advertising
transaction its effects? Yes, sometimes. Journal of Advertising Research, 32, 79-83.

Kiecolt, K. J ., & Nigg, J. (1982). Mobility and Perceptions of a Hazardous Environment.
Environment and Behavior, 14, 131-154.

-225-

 

 

Kirscht, J., & Haefner, D. (1973). Effects of repeated threatening health communications.
International Journal of Health Education, 16, 268-277.

Kirscht, J. P., Becker, M. H., Haefner, D. P., & Maiman, L. A. (1978). Effects of
threatening communications and mothers’ health beliefs on weight change in obese
children. Journal of Behavioral Medicine, I, 147-157.

Kleinot, M. C., & Rogers, R. W. (1982). Identifying effective components of alcohol
misuse prevention programs. Journal of Studies on Alcohol, 43, 802-811.

Kline, K., & Mattson, M. (2000). Breast Self-Examination Pamphlets: A Content
Analysis Grounded in Fear Appeal Research. Health Communication, 12(1), 1-21.

Knight, R. (2003). A Journalistic Approach to Good Writing: The Craft of Clarity (2nd
ed.). Ames, Iowa: Iowa State University Press.

Kohn, P., Goodstadt, M., Cook, G., Sheppard, M., & Chan, G. (1982). Ineffectiveness of
threat appeals about drinking and driving. Accident Analysis and Prevention, 14, 457-
464.

Kohut, A. (2004, July 11). Media myopia: More news is not necessarily good news New
York Times.

Kraus, S., El-Assal, E., & De Fleur, M. (1966). Fear-threat appeals in mass
communication: an apparent contradiction. Speech Monographs, 23-29.

Kreuter, M., & McClure, S. (2004). The Role of Culture in Health Communication.
Annual Review of Public Health, 25, 439-455.

Kuklinski, J. H., & Sigelman, L. (1992). When Objectivity is Not Objective: Network
Television News Coverage of US. Senators and the "Paradox of Objectivity" The
Journal of Politics, 54(3), 810-833.

LaTour, M., & Pitts, R. (1989). Using fear appeals in advertising for AIDS prevention in
the college-age population. Journal of Health Care Marketing, 9, 5-14.

LaTour, M., & Rotﬁeld, H. (1997). There are threats and (maybe) fear-caused arousal:
theory and confusions of appeals to fear and fear arousal itself. Journal of Advertising,
26(3), 46-59.

LaTour, M., Snipes, R., & Bliss, S. (1996). Don't be afraid to use fear appeals: An
experimental study. Journal of Advertising Research, 3 6(March/April), 59-67.

Lazar, J ., & Preece, J. (1999). Designing and implementing Web-based surveys. Journal
of Computer Information Systems, 39(4), 63-67.

-226-

 

 

Leece, P., Bhandari, M., Sprague, S., Swiontkowski, M., Schemitsch, E., Tometta, P., et
al. (2004). Internet vs. mailed questionnaires: A randomized comparison. Journal of
Medical Internet Research, 6(3), e30.

Lerner, J., Gonzalez, R., Small, D., & Fischhoff, B. (2003). Effects of fear and anger on
perceived risks of terrorism: A national ﬁeld experiment. Psychological Science, 14(2),
144-150.

Lerner, J ., & Keltner, D. (2001). Fear, anger, and risk. Journal of Personality and Social
Psychology, 81(1), 146-159.

Leventhal, H. (1970). Findings and theory in the study of fear communications. In L.
Berkowitz (Ed), Advances in experimental psychology (Vol. 5). New York: Academic.

Leventhal, H. (1971). F ear appeals and persuasion: The differentiation of a motivational
construct. American Journal of Public Health, 61, 1205- l 224.

Leventhal, H., Jones, 8., & Trembly, G. (1966). Sex differences in attitude and behavior
change under conditions of fear and speciﬁc instructions. Journal of Experimental Social
Psychology, 2, 387-399.

Leventhal, H., Singer, R., & Jones, S. (1965). The effects of fear and speciﬁcity of
recommendation. Journal of Personality and Social Psychology, 2, 20-29.

Leventhal, H., & Trembly, G. (1968). Negative emotions and persuasion. Journal of
Personality, 36, 154-168.

Leventhal, H., & Watts, J. (1966). Sources of resistance to fear-arousing communications
on smoking and lung cancer. Journal of Personality, 34, 155-175.

Leventhal, H., Watts, J ., & Pagnao, F. (1967). Effects of fear and instructions on how to
cope with danger. Journal of Personality and Social Psychology, 6, 313-321.

Maddux, J. E., & Rogers, R. W. (1983). Protection motivation and self-efﬁcacy: A
revised theory of fear appeals and attitude change. Journal of Experimental Social
Psychology, 19, 469-479.

Maibach, E., & Cotton, D. (1995). Moving people to behavior change. In E. Maibach &
R. Parrot (Eds), Designing health messages: Approaches from communication theory
and public health practice (pp. 41-64). Thousand Oaks, CA: Sage.

Maiman, L. A., & Becker, M. H. (1974). The Health Belief Model: Origins and correlates
in psychological theory. . Health Education Monographs,, 2(4), 336-353.

Marsh, A., Adams, R., & Kleck, R. (2005). Why do fear and anger look the way they do?
Form and social function in facial expressions. Personality and social psychology
bulletin, 31(1), 73-86.

-227-

 

 

Mayﬁeld, Z. (2006, April 11, 2006). Fear Appeal Messages and Their Effectiveness in
Advertising. Retrieved April 25, 2007

McAdams, K. (1992). Readability reconsidered: A study of reader reactions to FOG
indexes. Newspaper Research Journal, 13(14), 50-59.

McCroskey, J ., & Wright, D. (1971). A comparison of the effects of punishment-oriented
and reward-oriented messages in persuasive communication. Journal of Communication,
21, 83-93.

McGuire, W. J. (1968). Personality and susceptibility to social inﬂuence. In E. Borgatta
& W. Lambert (Eds), Handbook of Personality Theory and Research (pp. 1130-1187).
Chicago: Rand McNally.

McKay, D., Berkowitz, J ., Blumberg, J ., & Goldberg, J. (2004). Communicating
Cardiovascular Disease Risk Due to Elevated Homocysteine Levels: Using the EPPM to
Develop Print Materials Health Education & Behavior, 31(3), 355-371.

Mehta, R., & Sivadas, E. (1995). Comparing response rates and response content in mail
vs electronic mail surveys. Journal of the Market Research Society, 3 7(4), 429-439.

Mewbom, C. R., & Rogers, R. W. (1979). Effects of threatening and reassuring
components of fear appeals on physiological and verbal measures of emotion and
attitudes. Journal of Experimental Social Psychology, 15, 242-253.

Meyer, P. (2004). The Vanishing Newspaper: Saving Journalism in the Information Age.
Columbia, MO: University of Missouri Press.

Miller, G. (1963). Studies on the use of fear appeals: A summary and analysis. Central
States Speech Journal, 14, 117-125.

Mongeau, P. (1998). Another look at fear arousing messages. In M. Allen & R. Preiss
(Eds), Persuasion: Advances Through Meta-Analysis (pp. 53-68). Cresskill, NJ:
Hampton Press.

Moore, H., Bates, F., Layman, M. V., & Parenton, V. J. (1963). Before the Wind: A Study
of Hurricane Carla. Washington, DC: National Academy of Sciences.

Moynihan, R., & Cassels, A. (2005). Selling sickness. How the world's biggest
pharmaceutical companies are turning us all into patients. New York: Nation Books.

Murphy, J ., Gamble, G., & Sharpe, N. (1994). Readability of subject information leaﬂets
for medical research. New Zealand Medical Journal, 1 07(991), 509-510.

Murray, E., Hubbell, A., Witte, K., Sampson, J., & Morrison, K. (1998). Addressing
cultural orientations in fear appeals: Promoting AIDS-protective behaviors among

-228-

 

 

 

Hispanic immigrant and African-American adolescents, and American and Taiwanese
college students, International Communication Association. Jerusalem.

National Cancer Institute. (2005). The 5-A-Day Message Environment. Retrieved Aug. 5,
2005, from http://www.cancercontrol.cancer.gov/5ad_5_messhtml

National Highway Trafﬁc Safety Administration. (1997, April 17, 1997). Initiative for
Increasing Seat Belt Use Nationwide Retrieved April 29, 2007, from
http://www.nhtsa.dot.gov/people/injury/airbags/Archive-04/PresBelt/fullreport.html

Nell, V. (2002). Why Young Men Drive Dangerously: Implications for Injury Prevention.

Current Directions in Psychological Science, 11(2), 75-79.

Neuwirth, K., & Frederick, E. (2004). Peer and social inﬂuence on opinion expression.
Communication Research, 31(6), 669-703.

Niven, D. (2003). Objective Evidence on Media Bias: Newspaper Coverage of
Congressional Party Switchers. Journalism and Mass Communication Quarterly, 80(2),
311-326.

Nunnally, J. (1978). Psychometric Theory (2nd Edition ed.). New York: McGraw-Hill.

O'Grady, K., & Medoff, D. (1988). Categorical variables in multiple regression: Some
cautions. Multivariate Behavioral Research, 23(2), 243-260.

Park, H. S., Levine, T., Kingsley Westerman, G, Orfgen, T., & Foregger, S. (2007). The
effects of involvement type and argument quality on attitude formation and attitude
change. Human Communication Research, 33(1), 81-102.

Park, Y. J. (2005). Media Framing of Policy in Ambiguous Wars: A Case of Privacy
Policies, MiT4: the work of stories. Cambridge, MA.

Perez-Stable, E., Otero-Sabogal, R., Sabogal, F., & Napoles—Springer, A. (1996).
Pathways to early cancer detection for Latinas: En accion contra la cancer. Health
Education Quarterly, 23, $41-59.

Perry, R. (1982). The Social Psychology of Civil Defense. Lexington, MA: Lexington
Books.

Peter, J. P. (1981). Construct validity: A review of basic issues and marketing practices.
Journal of Marketing Research, 18(2), 133-145.

Philip Morris Co. (1969, Fall). Why One Smokes. Retrieved April 29, 2007, from
http://www.pmdocs.com/

Powell, F. (1965). The effects of anxiety-arousing messages when related to personal,
familial and impersonal referents. Speech Monographs, 32, 102-106.

—229-

t“:

 

 

Powell, F., & Miller, G. (1967). Social approval and disapproval cues in anxiety-arousing
communication. Journal of Social Psychology, 102, 79-91.

Prentice-Dunn, S., & Rogers, R. W. (1986). Protection motivation theory and preventive
health: Beyond the health belief model. Health Education Research, 1, 153-161.

Prochaska, J ., & DiClemente, C. (1982). Transtheoretical therapy toward a more
integrative model of change. Psychotherapy: Theory, Research and Practice, 19(3), 276-
287.

Prochaska, J., & DiClemente, C. (1983). Stages and processes of self-change in smoking:
Toward an integrative model of change. Journal of Consulting and Clinical Psychology,
51, 390—395.

Prochaska, J ., DiClemente, C., & Norcross, J. (1992). In search of how people change:
Applications to addictive behavior. American Psychologist, 4 7, 1102-1114.

Prochaska, J ., Norcross, J ., & DiClemente, C. (1994). Changing for good: A
revolutionary six-stage program for overcoming bad habits and moving your life
positively forward. New York: Avon Books.

Prochaska, J., Redding, C., & Evers, K. (2002). The Transtheoretical Model and Stages
of Change. In K. Glanz, B. Rimer & F. Lewis (Eds), Health Behavior and Health
Education: Theory, Research and Practice (pp. 99-120). San Francisco: Jossey-Bass.

Prochaska, J ., & Velicer, W. (1997). The transtheoretical model of health behavior
change. American Journal of Health Promotion, 12, 38-48.

Project for Excellence in Journalism, & Edmonds, R. (2004). The State of the News
Media 2004. Washington DC: Project for Excellence in Journalism.

Ramirez, A., & Lasater, T. (1977). Ethnicity of communicator, self-esteem, and reactions
to fear-arousing communications. Journal of Social Psychology, 102, 79-91.

Rapley, J ., & Coulson, N. (2005). Stages of change and consumption of fruit and
vegetables among adolescent females: Associations with decisional balance and self-
efﬁcacy British Food Journal 107(9), 663 - 669

Ray, M. L., & Wilkie, W. L. (1970). Fear: The potential of an appeal neglected by
marketing. Journal of Marketing, 34, 54-62.

Reed, G., Velicer, W., Prochaska, J., Rossi, J., & Marcus, B. (1997). What makes a good
staging algorithm: Example from regular exercise. American Journal of Health
Promotion, 12, 57-66.

-230-

Rhodes, P., & Wolitski, R. J. (1990). Perceived effectiveness of fear appeals in AIDS
education: Relationship to ethnicity, gender, age, and group membership. AIDS
Education and Prevention, 2, 1-11.

Rimal, R., & Real, K. (2003). Perceived Risk and Efﬁcacy Beliefs as Motivators of
Change. Human Communication Research, 29(3), 370-399.

Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude change.
Journal of Psychology, 91, 93-114.

Rogers, R. W. (1983). Cognitive and physiological processes in fear appeals and attitude
change: A revised theory of protection motivation. In J. Cacioppo & R. E. Petty (Eds)
Social psychophysiology (pp. 153— —.176) New York: Guilford.

Rogers, R. W. (1985). Attitude change and information integration in fear appeals.
Psychological Reports, 56, 179-182.

Rogers, R. W., & Mewbom, C. R. (1976). Fear appeals and attitude change: Effects of a
threat’s noxiousness, probability of occurrence, and the efﬁcacy of the c0ping responses.
Journal of Personality and Social Psychology, 3 4, 54—61.

Rosenberg, M. (1979). Conceiving the self New York: Basic Book.

Roskos-Ewoldsen, D., Yu, J ., & Rhodes, N. (2004). Fear Appeal Messages Affect
Accessibility of Attitudes Toward the Threat and Adaptive Behaviors. Communication
Monographs, 71(1), 49-69.

Rotfeld, H. (1988). Fear appeals and persuasion: Assumptions and errors in advertising
research. Current Issues and Research in Advertising, 11(1), 21-40.

Ruiter, R., Verplanken, B., De Cremer, D., & Kok, G. (2004). Danger and Fear Control
in Response to Fear Appeals: The Role of Need for Cognition. Basic and Applied Social
Psychology, 26(1), 13-24.

Schafer, W. (1991). Reporting hierarchical regression results. Measurement and
Evaluation in Counseling and Development, 24(3), 98-100.

Schiller, D. (1979). An Historical Approach to Objectivity and Professionalism in
American News Reporting. Journal of Communication, 29(4), 46-57.

Schmidt, W. (1997). Worldwide Web survey research: Beneﬁts, potential problems, and
solutions. Behavior Research Methods, Instruments, & Computers, 29(2), 274-279.

Schudson, M. (2001). The objectivity norm in American journalism. Journalism, 2(2),
149-170.

-231-

 

 

Schwab, D. (1980). Construct Validity in Organizational Behavior. In B. S. a. L. L.
Cummings (Ed.), Research in Organizational Behavior (Vol. 2, pp. 3-43). Greenwich,
Connecticut: JAI Press.

Severin, W., & Tankard, J. (2001). Communication Theories: Origins, Methods, and
Uses in the Mass Media. New York: Addison Wesley Longman, Inc.

Sherif, M., & Cantril, H. (1947). The psychology of ego involvement: Social attitudes and
identiﬁcations. New York: John Wiley.

Smalec, J. (2000). Bulimia Interventions via Interpersonal Inﬂuence: The Role of Threat
and Efﬁcacy in Persuading Bulimics to Seek Help. Journal of Behavioral Medicine, 1,
37-5 7. . - - '

spss. (1999). Missing data: the hidden problem, spss White Paper (pp. 1-8).

Stephenson, M., & Witte, K. (2001). Creating Fear in a Risky World. In R. Rice & C.
Atkin (Eds), Public Communication Campaigns (3rd ed., pp. 88-102). Thousand Oaks,
CA: Sage.

Stemthal, F ., & Craig, C. (1974). F ear appeals: Revisited and revised. Journal of
Consumer Research, 7, 22-34.

Stout, P. A., & Sego, T. (1994). The Role of Need for Cognition in Response to Threat
Appeals in AIDS Messages. Paper presented at the Speech Communication Association.

Sussman, L. (1973). Ancients and modems of fear and fear appeals: A comparative
analysis. Central States Speech Journal, 24, 206-211.

Sutton, S. R. (1982). Fear-arousing communications: A critical examination of theory and
research. In J. Eiser (Ed.), Social Psychology and Behavioral Medicine (pp. 303-337).
London: Wiley.

Sutton, S. R. (1992). Shock tactics and the myth of the inverted U. British Journal of
Addiction, 87(April), 517-5 19.

Sutton, S. R., & Eiser, J. R. (1984). The effects of fear-arousing communications on
cigarette smoking: An expectancy-value approach. Journal of Behavioral Medicine, 7,
13-33.

Sutton, S. R., & Hallett, R U. (1988). Understanding the effects of fear-arousing
communications: The role of cognitive factors and amount of fear aroused. Journal of
Behavioral Medicine, I I, 353-360.

Tanner, J ., Hunt, J ., & Eppright, D. (1991). The protection motivation model: a normative
model of fear appeals. Journal of Marketing, 55(3), 36-45.

-232-

Toothacker, L. E. (1993). Multiple comparisons procedures. (Vol. 89). Thousand Oaks,
CA: Sage Publications.

Vrana, S., Cuthbert, B., & Lang, P. (1986). Fear imagery and text processing.
Psychophysiology, 23(3), 247.

Wakui, S., Shimomitsu, T., Odagiri, Y., Inoue, S., Takarniya, T., & Ohya, Y. (2002).
Relation of the stages of change for exercise behaviors, self-efﬁcacy, decisional-balance,
and diet-related psycho-behavioral factors in young Japanese women. The Journal of
sports medicine and physical ﬁtness, 42(2), 224-232.

Wanta, W., & Gao, D. (1994). Young readers and the newspaper: Information recall and
perceived enjoyment, and attractiveness. Journalism Quarterly, 71(4), 926-935.

Wheatley, J ., & Oshikawa, S. (1970). The relationship between anxiety and positive and
negative advertising appeals. Journal of Marketing Research, 7, 85-89.

Williams-Piehota, P., Pizarro, J., Silvera, S., Mowad, L., & Salovey, P. (2006). Need for
Cognition and Message Complexity in Motivating Fruit and Vegetable Intake Among
Callers to the Cancer Information Service Health Communication, 19(1), 75-84.

Williams-Piehota, P., Schneider, T., Pizarro, J., Mowad, L., & Salovey, P. (2003).
Matching health messages to information-processing styles: Need for cognition and
mammography utilization. Health Communication, 15(4), 375-392.

Wilson, V., & Lankton, N. (2004). Modeling Patients' Acceptance of Provider-delivered
E—health. Journal of the American Medical Informatics Association, 11(4), 241-248.

Witmer, D., Colman, R., & Katzman, S. L. (1999). From Paper-and-Pencil to Screen-and-
Keyboard: Toward a Methodology for Survey Research on the Internet. In S. Jones (Ed.),
Doing Internet Research: Critical Issues and Methods for Examining the Net (pp. 328).
Thousand Oaks, CA: Sage Publications.

Witte, K. (1992a). Preventing AIDS through persuasive communications: A framework
for constructing effective, culturally speciﬁc, preventive health messages. International
and Intercultural Communication Annual, 1 6, 67-86.

Witte, K. (1992b). Putting fear back into fear appeals: The Extended Parallel Processing
Model. Communication Monographs, 59(4), 329-349. .

Witte, K. (1994a). Fear control and danger control: An empirical test of the extended
parallel process model. Communication Monographs, 61, 113-134.

Witte, K. (1994b). Generating effective risk messages: How scary should your risk
communication be? Communication Yearbook, 18, 229-254.

-233-

Witte, K. (1997). Preventing teen pregnancy through persuasive communications:
realities, myths, and the hard-fact truths. Journal of Community Health, 22(2), 137-154.

Witte, K. (1998). F ear as motivator, fear as inhibitor: Using the EPPM to explain fear
appeal successes and failures. In P. Andersen & L. Guerrero (Eds), The Handbook of
Communication and Emotion (pp. 423-450). New York: Academic Press.

Witte, K., & Allen, M. (2000). A meta-analysis of fear appeals: Implications for effective
public health campaigns. Health Education & Behavior, 27(5), 591-615.

Witte, K., Cameron, K., Lapinski, M., & Nzyuko, S. (1998). Evaluating HIV/AIDS
prevention programs according to theory: A ﬁeld project along the Trans-Africa Highway
in Kenya. Journal of Health Communication, 4, 345-363.

Witte, K., Cameron, K., McKeon, J ., & Berkowitz, J. (1996). Predicting Risk Behaviors:
Development and Validation of a Diagnostic Scale. Journal of Health Communication, 1,
317-341.

Witte, K., Meyer, G., & Martel], D. (2001). Eﬁ’ective health risk messages: A step-by-step
guide. Thousand Oaks, Calif: Sage.

Witte, K., & Morrison, K. (1995). Using scare tactics to promote safer sex among
juvenile detention and high school youth. Journal of Applied Communication Research,
23, 128-142.

Witte, K., & Morrison, K. (2000). Examining the inﬂuence of trait anxiety/repression-
sensitization on individuals’ reactions to fear appeals. Western Journal of
Communication, 64(] -29).

Witte, K., Peterson, T., Vallabhan, S., Stephenson, M., Plugge, C., Givens, V., et al.
(1993). Preventing tractor-related injuries and deaths in rural populations: Using a
Persuasive Health Message (PHM) framework in formative evaluation research.
International Quarterly of Community Health Education, 13, 219-251.

Woloshin, S., & Schwatrz, L. (2006). Giving legs to restless legs: A case study of how
the media helps to make people sick. Public Library of Science Medicine, 3(4), 170.

Wood, M., Read, J ., Palfai, T., & Stevenson, J. (2001). Social inﬂuence processes and
college student drinking: the mediation role of alcohol outcome expectations. Journal of
Studies on Alcohol, 62(1), 32-43.

Wurtele, S. K., & Maddux, J. E. (1987). Relative contributions of protection motivation
theory components in predicting exercise intentions and behavior. Health Psychology, 6,
453-466.

Zimbardo, P. (1960). Involvement and communication discrepancy as determinants of
opinion conformity. Journal of A bnormal and social psychology, 60, 86-94.

-234-

Zuckerman, M. (1991). The psychobiology of personality. New York: Cambridge
University Press.

Zuckerman, M. (1994). Behavioral expressions and biosocial bases of sensation seeking.
Cambridge, UK: Cambridge University Press.

-235-