4‘90. .

KI . VT“... . M
t - . 3
ﬁlm? » .,..
V . v v s ...
$1.3 _

4 a,

u
2

.’1.

if...)

.untﬂiottlu

:- dun
I <

L»

74:391..-.
1'11.le .. z ..
.. (Wyncnwd‘uu:
an...“ (_

ﬂu emu?“

: 36...”... .HﬂnWJ.

. . , . ..-. .
Ar «

 

J . 1|":
.1 . ..1? (
~ﬁ1m‘oh. J3

. .

W’ﬁﬂdt

“5‘31“;
‘[ 1"."

3,

*2

H

 

0‘

t. {L
A... 3.
a:

 

THES‘S

(A

Illllll'llllllllllllllllllllllllllll

3 1293 01712 7337

This is to certify that the

dissertation entitled

THE SUASORY EFFECT OF AFFECTIVE AND COGNITIVE MESSAGES:
A TEST OF CONFLICTING HYPOTHESES

presented by

Kenzie Alynn Cameron

has been accepted towards fulﬁllment
of the requirements for

Ph.D. degree in Communication

 

 

hdajor rofessor

Date 07 August 1998

 

MSU is an Afﬁrmative Action/Equal Opportunity Institution 0- 12771

 

 

LIBRARY

Michigan State
Unlverslty

 

 

 

PLACE IN RETURN Box
to remove this checkout from your record.
To AVOID FINES return on or before date due.

 

MTE DUE MTE DUE DATE DUE

 

 

NW 39 1559
, 1,; -

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

we acummu

THE SUASORY EFFECT OF AFFECTIVE AND COGNITIVE MESSAGES:
A TEST OF CONFLICTING HYPOTHESES
By

Kenzie Alynn Cameron

A DISSERTATION
Submitted to
Michigan State University
in partial ﬁilﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Communication

1998

ABSTRACT

THE SUASORY EFFECT OF AFFECTIVE AND COGNITIVE MESSAGES:
A TEST OF CONFLICTING HYPOTHESES
By

Kenzie Alynn Cameron

An experiment was devised to assess the suasory eﬂ‘ect of affective and cognitive
arguments on attitudes that are aﬂ‘ectively or cognitively based. Conﬂicting hypotheses
regarding such effects include the mismatch hypothesis, suggesting that a persuasive
message will be more effective if the arguments employed in the message are mismatched
to one’s attitude type, and the match hypothesis, which suggests that a persuasive message
will be more effective if it is matched to one’s underlying attitudinal base. Aﬂ'ectively
based and cognitively based attitudes toward a ﬁctitious attitude object (a Xenon light
bulb) were induced by means of written messages that were either aﬁ'ect- or cognition-
laden. Following initial responses to this attitudinal induction, independent groups of
subjects were exposed to one of four counterattitudinal messages: (1) an affective
message, (2) a cognitive message, (3) a message containing both affective and cognitive
arguments, or (4) a control message, discussing a topic irrelevant to the attitudinal object.
Analyses of variance and contrast analyses were employed to test the conﬂicting
hypotheses. Results indicate that neither the mismatch nor the match hypotheses were
consistent with the data. A three-way interaction was observed. In the absence of

cognitive arguments, there is a main effect for the presence (versus absence) of an

aﬂ‘ective counterattitudinal message, and in the presence of cognitive arguments, there is a
main eﬁ‘ect for the type of attitudinal induction and no eﬁ'ect of aﬂ‘ective arguments.
These results let to the construction and subsequent testing of the Cognitive Argument
Primacy (CAP) Model, which was found to be consistent with the data. Theoretical

explanations of the CAP model are discussed.

This dissertation is dedicated to my parents, Jule and E. Alan Cameron, my brother, Bain
Cameron, my Great Aunt, Jean Makepeace, and to the memory of my Great Uncle,
Charlie Makepeace. Their constant love and belief in me is a continual source of strength

and inspiration. Thank you for everything.

iv

ACKNOWLEDGMENTS

I would like to thank my adviser, Dr. Franklin I. Boster, who gave of his time, his
ideas, his insight, and his humor in countless ways during my years here at Michigan State.
He always seemed to know the right time to call me out of the blue, to give me the
encouragement that I needed, or to give me a brief therapy session, consisting of the
words “Kenzie, stop it!” when I was overwhelmed or desperately bent on ﬁnding the non-
existent perfect way of presenting something. I am truly honored to have been able to
study with such a great mentor, a man who demands, and deserves, the best from his
students. In addition, I thank Frank and Linda Boster for their ﬁiendship and their
support.

I am also indebted to Dr. Kim Witte, who, in addition to being a committee
member for my dissertation, was my Masters’ adviser. Working with Kim on her research
team for my ﬁrst two years in the program provided me with invaluable research
experience, and incredible exposure to the ﬁeld of Communication.

Another thank you goes to my committee members, Dr. Charles T. Salmon, Dr.
Sandi W. Smith, and Dr. Michael Burgoon. Their questions and comments were a
continual source of knowledge and insight, challenging me to think beyond the scope of
each project. In addition, I would like to thank my entire committee for their graciousness

and willingness to adapt their already demanding schedules to ensure that I could meet the

necessary deadlines.

Although the ﬁrst project intended to be my dissertation did not reach completion,
I want to acknowledge Dr. Jon A. Krosnick of Ohio State University, who provided me
with materials that he had used in his study.

Furthermore, there are numerous colleagues and ﬁ'iends who deserve special
recognition. I’ll always remember working at the coﬂ‘eeshops with Shelly Campo, as we
both grappled with our dissertations and with ﬁlling those lacunas that we could ﬁnd;
being fed by Esther Baker, whose meals seem to appear magically out of nowhere at just
the right times; having Chinese food with Gwen Wittenbaum, who always took the time to
ﬁnd out just how I was doing; and hunting Easter eggs with Julie Burgess and her family
as well as being forever grateful for Julie’s ﬁ'iendship. Additional thanks go to Alice
Chan, Lisa Cutler, Pam Lannutti, Mindy Mershon, Sandra Metts, and Don Moorhead,
who continually cheered me on ﬁom afar via e-mail. I also want to recognize my cohort
members: Shelly Campo, Hui-Jung Chang, Betty LaFrance, Janet Lillie, Ren-He Huang,
and Brian Downey. Completing the program here at MSU would not have been as
rewarding nor as much fun without them.

My research assistants, Craig Hanger and Phil Shipley, deserve a resounding round
of applause for all of their help during this project. Anne Hubbell also deserves a thank
you, for her ﬁiendship, as well as her willingness to hold Spots for me in her classes to
ensure that I would have subjects for this dissertation.

I am grateful for the assistance of Diane and Matt Golzynski, Jennifer Heisler, Rie
Ohashi, and Cynthia Zuckerman for their help in coding, tallying, numbering, and entering

data; I truly could not have completed this work without their selﬂess and willing help at

times when they were busy themselves. An additional thank you is extended to Diane
Golzynski. Her friendship, her help, and her belief in me and my ability to complete the
dissertation kept me going when it seemed as though things would never fall into place.

Deb Tigner, Marge Barkrnan, Ann Wooten, Rachel Miller, Cathy Gladstone,
Arnetta Cage, and Jenny Springs went above and beyond the call of duty over the years
and for that I am grateful.

In addition, I want to thank Dr. David Wright, Barbara Decker, Amy Hirshman,
Molly Lauck, and Julie Reyes of the University Committee on Research Involving Human
Subjects ofﬁce. They came to know me over the years because of my ﬁequent trips to
their ofﬁce, and they bent over backwards to help me when I needed it.

I also want to thank the Michigan State University Graduate School for awarding
me a University Distinguished Fellowship (1994-1998), which funded my graduate
studies. Speciﬁcally, I extend my thanks to Dean Gerard Crawley, Dr. Les Manderscheid,
Dean Karen Klomparens, and Jo Trahan. Cathie Allison, Debra Conlin, and Mikala
Keating are also recognized for their guidance regarding this ﬁnished product.

Finally, I offer my heartfelt thanks, and my love, to my family -- Mom, Dad, and
Bain. No matter the day, the hour, or the need, they are always there for me to give me

encouragement, support, ideas, assistance, and most of all love. Put simply, I would not

be where I am today if it were not for my family.

vii

TABLE OF CONTENTS

LIST OF TABLES ..................................................................................... x
LIST OF FIGURES .................................................................................... xi
CHAPTER 1

INTRODUCTION ...................................................................................... 1
CHAPTER 2

AFFECT AND COGNITION

IN ATTITUDE FORMATION AND CHANGE ......................................... 3
CHAPTER 3

METHODS ................................................................................................ l 7
CHAPTER 4

RESULTS .................................................................................................. 34
CHAPTER 5

DISCUSSION ............................................................................................ 53
APPENDIX A

ITEMS SELECTED FOR USE IN INDUCTION CHECK ANALYSES... 76
APPENDIX B

MESSAGES ............................................................................................... 77
APPENDIX C

INSTRUCTIONS FOR FOCUSED READING OF NIESSAGES ............... 80
APPENDIX D

ITEMS, MEANS, SD OF THE 15 ATTITUDE ITEMS
FOLLOWING THE INDUCTION AND PERSUASION MESSAGES ...... 81

APPENDIX E

RESULTS OF THE CONFIRMATORY FACTOR ANALYSES
FOR THE DEPENDENT VARIABLES .................................................... 83

APPENDIX F

CONTRAST COEFFICIENTS EMPLOYED TO TEST THE
PREDICTIONS OF THE MISMATCH HYPOTHESIS ............................. 87

viii

APPENDIX G
CONTRAST COEFFICIENTS EMPLOYED TO TEST THE

PREDICTIONS OF THE MATCH HYPOTHESIS .................................... 88
APPENDD( H

CONTRAST COEFFICIENTS EMPLOYED TO TEST THE FIT

OF THE COGNITIVE ARGUMENT PRIMACY MODEL ........................ 89
LIST OF REFERENCES ............................................................................ 91

ix

LIST OF TABLES

Table 1 - Study Design ............................................................................... 13
Table 2 - Predictions of the Mismatch Hypothesis ....................................... 14
Table 3 - Predictions of the Match Hypotheis .............................................. 15
Table 4 - Cells Central to the Test of Conﬂicting Hypotheses ...................... 16
Table 5 - Overall Ratings of Degree of Cognition and Aﬁ‘ect

for Messages (Induction and Persuasion) .................................................... 24
Table 6 - Perceived Strength of Messages ................................................... 25

Table 7 - Obtained Means, Standard Deviations, and Cell Sizes
for Difference in Attitude ............................................................................ 38

Table 8 - Obtained Means, Standard Deviations, and Cell Sizes
for Difference in Value/Worth ..................................................................... 40

Table 9 - Obtained Means, Standard Deviations, and Cell Sizes
for Difference in Dependability .................................................................... 41

LIST OF FIGURES

Figure 1 - Aﬂ‘ective and Cognitive Dimensions of Attitudes ....................... 5

Chapter 1

INTRODUCTION

In order to ascertain the suasory eﬂ‘ects of messages in attitude formation and
change, researchers often analyze the base or type of the attitude that the messages are
designed to inﬂuence. It has been suggested that inﬂuence attempts need to “make
contact with an attitude’s origin” (Edwards, 1990, p. 202) in order to obtain the maximum
possible attitude change. An attitude may have its origin in, or may be based upon the
ﬁrnction the attitude serves for the individual. Much work has been done focusing on the
underlying functions of attitudes, work that was spurred by that of Smith, Bruner and
White (1956) and Katz (1960). The functional bases of attitudes has spawned much
research as scholars have attempted to determine what ﬁrnction or functions that attitudes
we hold serve (c.f. Herek, 1987; Katz, 1960; Shavitt, 1990; Smith, Bruner, & White,
1956). Another path that researchers have traveled is one focused upon deﬁning an
attitude type as being either aﬁ‘ectively or cognitively based (Edwards, 1990; Millar &
Millar, 1990; Millar & Tesser, I986, 1989).

Researchers using the ﬁmctional approach, as well as those who deﬁne attitudes as
being aﬂ‘ectively or cognitively based, have concentrated on the match or mismatch of a
persuasive argument to the underlying attitude toward an object (see also Petty &
Wegener, 1998). Scholars have attempted to discern when and under what conditions an
affective or cognitive argument will affect one’s attitude toward an object. Researchers

have studied the effects of such messages based on cognitions about an object (i.e.,

cognitive, rational, or non-emotional messages) or based on feelings about or emotions
aroused by an object (i.e., affective or emotional messages) and have arrived at diﬁ‘ering
and conﬂicting conclusions (Edwards, 1990; Millar & Millar, 1990).

Current studies present both a mismatch hypothesis, proposing that a message will
be more persuasive if it is of a type other than the existing attitude the inﬂuence attempt is
trying to change (Millar & Millar, 1990), as well as a match hypothesis, which suggests
that a persuasive message will be more effective if it matches the underlying base of the
attitude in question (Edwards, 1990). This study was developed in order to explore the
eﬁ‘ects of affective and cognitive arguments on aﬂ‘ectively based and cognitively based

attitudes in order to test these conﬂicting hypotheses.

Chapter 2

AFFECT AND COGNITION IN ATTITUDE FORMATION AND CHANGE

When engaging in the study of attitudes, the results of numerous studies have been
consistent with a theoretical distinction between aﬁ‘ect and cognition. The distinction
between aﬂ‘ective and cognitive components of attitudes has been described such that the
emotions and feelings about an attitude object constitute the aﬂ‘ective components, and the
perception one has about the attitude object, derived from information and beliefs,
constitutes the cognitive component (Fleming, 1967; McGuire, 1969; Millar & Millar,
1990). Breckler (1984) and Breckler and Wiggins (1989) investigated affect and
cognition as separate elements of an attitude and found their data to be consistent with
such a distinction.1 Zajonc (1980) posits that affective judgments not only may be
independent of, but may actually occur prior to cognitive judgments regarding an attitude
object. Katz (1960) proposes that some attitudes may contain multiple components (i.e.,
affective, cognitive, conative), while others may be primarily composed of one of the
components.

It is important to note that, although there is considerable research consistent with
the existence of multiple dimensions of attitude, these dimensions exist in conjunction
with, and not in the absence of, each other. Becker (1963) warns of the simpliﬁcation of
these dimensions, and the resulting weakness of oversimplifying. Much research has used
messages that have been characterized as emotional and rational. In reading these Studies,

it is possible to view these characterizations as constituting a single continuum, anchored

on one end by emotion, and on the other end by rationality, or logic. Becker (1963)
argues that such a simpliﬁed distinction is a misnomer, for rational (as cognitive messages
are often characterized) and emotional (a common characterization of aﬂ‘ective messages)
do not lie at opposite ends of a single continuum. A message ﬁlled with emotion is not
necessarily devoid of logic. Similarly, a message that is heavily aﬁ‘ect-laden is not
necessarily without a cognitive element. Further, a message may lack both elements of
emotion and logic; a message that is not emotional is not automatically a logical message.
Although the messages and attitude types investigated in this study will be referred to as
being affective or cognitive, it is recognized that aﬂ‘ect and cognition, like emotion and
logic, are not inherently incompatible. “The distinction between aﬁ‘ect- and cognition-
based attitudes is not a dichotomous one. It is unlikely that we ever form pure affect-
based or pure cognition-based attitudes. More often than not, affect and cognition jointly
determine the course of attitude acquisition, albeit in varying degrees and scenarios”
(Edwards, 1990, p. 204).

In this study efforts were made to develop and use messages that were of high
affective content/low cognitive content, and high cognitive content/low affective content
(see Chapter 3). None of the messages are considered to be “pure” messages (i.e., purely
affective or purely cognitive). Figure 1 depicts a representation of the multidirnensionality
of an attitude or message, when viewed as having an affective and cognitive component.
Messages could be developed to ﬁt any of the four quadrants. Two quadrants were

selected to be represented in this study: the quadrant containing high affective content/low

cognitive content (quadrant II in Figure I) and the quadrant containing high cognitive

content/low affective content (quadrant IV in Figure 1).

Figure 1:

Affective and Cognitive Dimensions of Attitudes

High Cognitive Content
45

Low Affective High Affective

‘7

 

Content Content

 

.1

Low Cognitive Content

ConﬂictingHypotheses Regardinyhe Suasory Effect of Affective and Cogm'tive
Message; on Attitudes

Recent studies have investigated the interaction of argument type (affective or
cognitive) and attitude type (affectively based or cognitively based attitudes) and have
provided inconsistent results. Millar and Millar (1990) presented a mismatch hypothesis,
and found that their data were consistent with their hypothesis. Ifone holds an affective

attitude toward an object, one is persuaded to a greater degree when one encounters a

cognitive counterattitudinal message, and vice versa Contrary to Millar and Millar’s
(1990) ﬁndings, Edwards (1990) found her data to be consistent with a match hypothesis.
Speciﬁcally, if one has an affective attitude toward an object, one will be persuaded to a
greater degree when one encounters an aﬁ‘ective counterattitudinal message, than one
would were one to encounter a cognitive counterattitudinal message.

Millar and Millar (1990) proposed that the suasory eﬂ‘ectiveness of rational or
emotional appeals would be dependent on the base of the attitude these messages
attempted to inﬂuence. Speciﬁcally, Millar and Millar (1990) suggested that cognitively
based attitudes, those attitudes formed through consideration of attributes and beliefs
about the object, would be more susceptible to emotional arguments (that is, those
arguments that are aﬁ‘ectively based, using emotional appeals) than to rational arguments.2
Further, Millar and Millar (1990, p. 218) hypothesized that those attitudes that are
aﬂ‘ectively based (formed though the consideration of emotions and feelings about the
attitude object) would be more susceptible to rational arguments (informational attacks
focused on beliefs and judgments about the attitude object). Millar and Millar (1990)
proposed that this mismatch of attitude type and argument type would be more effective
because such a mismatch would lessen the counterarguing process that individuals engage
in when presented with new information. Speciﬁcally, if one has a cognitively based
attitude toward an object, a cognitive counterattitudinal message may well attack the very
way that one has thought about the object, which may motivate one to counterargue
(Brehm, 1966; Millar & Millar, 1990). Alternatively, if an emotional counterattitudinal

message is presented, the information contained in the emotional argument will not

threaten the way one has thought previously about the object and thus will not promote
counterarguing (Millar & Millar, 1990). Similarly, Millar and Millar (1990) hypothesized
that if one holds an affective attitude, one will be more persuaded by a cognitive
counterattitudinal argument than by an affective counterattitudinal argument.

To test their mismatch hypothesis Millar and Millar (1990) designed a series of
studies in which subjects’ attitude toward an object (in Studies 1 and 2, a beverage; in
Study 3, analytical problems) was assessed as being either aﬁ’ectively based or cognitively
based. Then, the subjects were presented with counterattitudinal messages about the
attitude objects. The counterattitudinal message contained either rational or emotional
arguments. Results from two of their three studies indicated that the data were consistent
with their hypothesis. In Study 1, when _S_s’ initial attitude was classiﬁed as aﬂ‘ective, the
rational counterattitudinal message produced more agreement with the counterattitudinal
argument (M = 2.97) than did the emotional counterattitudinal message (M = 2.39),
E(1,64) = 5.23, p = .023” When SS’ initial attitude was classiﬁed as cognitive, the
affective counterattitudinal message produced more agreement with the counterattitudinal
argument (M = 3.14) than did the rational counterattitudinal message (M = 2.68), E(1,64)
= 4.65, p = .03. In Study 2, when 55’ initial attitude was classiﬁed as affective, the
rational counterattitudinal message produced more agreement with the counterattitudinal
argument (M = 4.88) than did the affective counterattitudinal message M = 3.96), E(l,90)
= 10.41, p = .01. When SS’ initial attitude was classiﬁed as cognitive, the
counterattitudinal arguments did not inﬂuence the amount of agreement with the

counterattitudinal message (E >1).5 In the third study, where the attitude object was a

series of analytical puzzles, and the Ss’ initial attitude was induced (as opposed to
classiﬁed, as had been done in Studies 1 and 2), when Ss’ initial attitude was classiﬁed as
aﬁ‘ective, the rational counterattitudinal message produced more agreement with the
counterattitudinal argument (M = 4.68) than did the emotional counterattitudinal message
(M = 3.42), E(1,46) = 5.62, p = .02. When Ss’ initial attitude was classiﬁed as cognitive,
the aﬁ‘ective counterattitudinal message produced more agreement with the
counterattitudinal argument M = 4.41) than did the rational counterattitudinal message
M = 2.86), 1311,46) = 5.65, p = .02.

In Studies 2 and 3, Millar and Millar (1990) asked subjects to list the thoughts they
had while they viewed the counterattitudinal messages. In Study 2, upon analysis of the
valence of the thoughts listed, they found that when Ss with affective attitudes were
presented with emotional messages, SS listed more negative thoughts (M = 1.27) than
when they were presented with rational messages M = .71), E(1,90) = 4.56, p = .05.
When SS with cognitive attitudes were presented with rational messages, SS tended to list
more negative thoughts (M = 1.52) than when they were presented with emotional
messages (M = 1.04), F_(1,90) = 1.91, p = .17. In Study 3, when SS with affective
attitudes were presented with emotional messages, SS tended to list more negative
thoughts M = 1.95) than when they were presented with rational messages M = 1.28),
E(1,3 8) = 2.68, p = .12. When SS with cognitive attitudes were presented with rational
messages, SS listed more negative thoughts M = 2.32) than when they were presented

with emotional messages (M = 1.32), _E(1,3 8) = 6.84, p = .02.

Arguing from a cognitive response approach, Millar and Millar (1990) conclude
that when SS are exposed to arguments that do not match the base of their attitude
(cognitive or affective), the SS will counterargue less and will therefore exhibit more
attitude change, when measured by agreement with the counterattitudinal message. When
a counterattitudinal message emphasizes the same component (affective or cognitive) that
forms the base of the Ss’ attitude, Ss will be motivated to counterargue, will produce
more counterarguments, and thus will not be as inﬂuenced by the counterattitudinal
message.

Relating her arguments to the functional approach, Edwards (1990) argues that
there are various reasons for which people form and modify attitudes. She draws on
Zajonc (1980) and Zajonc and Markus (1982) to assert her position that aﬁ‘ect and
cognition combine to form an attitude toward an object, and, depending on the attitudinal
basis (affective or cognitive) created, the attitude may be differentially affected by aﬂ‘ective
and cognitive counterattitudinal messages. Speciﬁcally, She hypothesizes that affective
arguments will be more persuasive than cognitive arguments when the Ss’ initial attitudes
are affectively based (Edwards, 1990). Her underlying argument for this hypothesis arises
from Zajonc (1980), and she states that an implication of his view that affect may be
primary for speciﬁc preferences is “that attitudes with affective origins may be relatively
impervious to inﬂuence attempts that rely on rational argumentation and might be more
responsive to persuasive appeals that tap their affective bases” (Edwards, 1990, p. 203;

Zajonc, 1980). Thus, she presents what could be termed a match hypothesis, speciﬁcally

when viewing affectively based attitudes.

To test her hypothesis, Edwards (1990) conducted two studies in which SS were
exposed to both aﬁ‘ective and cognitive messages about an attitude object (in Study 1 the
object was a Chinese ideograph, in Study 2 the object was a sport drink called Power
Plus). She held the content of the induction and suasory messages constant, as She
exposed all S3 to both cognitive and affective messages. Her focus, then, was on the
variation of the sequence of the presentation of the cognitive and affective messages. In
Study 1 those S s in the affective induction condition ﬁrst were exposed to a subliminal
affective prime (either a happy or an angry face), then to a supraliminal image of a Chinese
ideograph, then were asked to read a Short passage about the ideograph (either favorable
or unfavorable information, consistent with the aﬂ‘ective induction), then to another
supraliminal image of the same ideograph, and ﬁnally were asked to complete ratings
about the ideograph. Those SS in the cognitive induction condition ﬁrst were presented
with a supraliminal image of the ideograph, were asked to read about the ideograph, then
were exposed to the subliminal affective prime, were exposed to another supraliminal
image of the ideograph, and ﬁnally were asked to complete ratings.

After the attitude induction, the SS were exposed to either affective or cognitive
persuasive conditions. AS in the attitude induction, all SS were exposed to both aﬂ‘ective
and cognitive messages, the order of these messages varying among conditions. The SS
were presented with counterattitudinal information about the ideographs that they had
viewed and read about in the induction trials.

Edwards (1990) computed difference scores as to the degree of liking the Ss had

for each of the ideographs. Consistent with her hypothesis, those SS with attitudes that

10

were aﬂ‘ectively based (SS who had ﬁrst been exposed to the subliminal slides), reported
more attitude change when the persuasion was aﬁ‘ective than when it was cognitive
_F_(1,60) = 6.6, p = .02.6 Those SS whose attitudes were cognitively based reported an
equal amount of attitude change under both conditions F (1,60) = 1.0, ns.

In her second study Edwards (1990) used supraliminal messages for both the aﬁ‘ect
and the cognition variables. Subjects in the affective induction condition ﬁrst tasted
Power Plus, the ﬁctitious sport drink that was the attitude object. They then read a
passage (cognitive message) about Power Plus, and were asked to rate the degree to
which they liked the product. Those SS in the cognitive induction condition ﬁrst read the
passage, then sampled the drink, and then provided their ratings of the beverage. For
those SS who were then exposed to Edwards’ (1990) affective persuasion appeal, they
smelled a liquid that they were told was Power Plus (the liquid had an unfavorable smell),
read another passage about the beverage (the passage was unfavorable toward the
beverage), and then completed their ratings. Those SS in the cognitive persuasion appeal
condition ﬁrst read the passage, then smelled the beverage, then completed their ratings.
Consistent with her hypothesis and the results of Experiment 1, Edwards (1990) reported
that those subjects in the affectively based attitude condition reported a greater amount of
change when they were exposed to affective persuasion than when they were exposed to
cognitive persuasion E (1,103) = 4.3, p < .05. Those subjects with cognitively based
attitudes reported an equal amount of change in both persuasion conditions.

Edwards (1990) notes that in her experiment the procedure used to classify

whether or not an attitude induction or means of persuasion was affective or cognitive was

11

based upon the sequence of the messages. Those messages that had an aﬁ‘ective element
presented ﬁrst followed by a cognitive element were classiﬁed as aﬁ‘ective
induction/persuasion; those messages that had a cognitive element presented ﬁrst followed
by an aﬁ‘ective element were classiﬁed as cognitive induction/persuasion. Thus, Edwards
(1990) varied the sequence of the elements, and not the absolute contribution of each
process (p. 210). “The design, therefore, provides a strong test of the importance of the
sequence of affect and cognition in attitude formation and change” (Edwards, 1990, pp.
210-211).

In this study an experiment designed to test the competing hypotheses of Millar
and Millar (1990) (identiﬁed in this paper as the mismatch hypothesis) and Edwards

(1990) (identiﬁed in this paper as the match hypothesis) is offered. Subjects Will be

 

exposed to one of two initial attitude inductions (cognitively based or affectively based
attitudinal induction) (see Table 1). That is, they will be presented with a message
regarding an object ( a new product to which the subjects have never been exposed) that

offers either a cognitively based description of this object (e. g., providing information

 

regarding the attributes and others’ beliefs about the product) $3 an affectively based
description of the product (e.g., a message that emphasizes feelings and emotions about
the product). This initial description of the product, whether cognitive or aﬁ‘ective, will be
positive. In addition, the instructions provided to the subjects will attempt to focus their
responses either cognitively or affectively (described ﬁrrther below).

Independent groups of each of these SS (cognitively based attitudinal induction,

affectively based attitudinal induction) will, after completion of a questionnaire, be

12

exposed to a second message (cognitive argument condition x aﬁ‘ective argument

condition). This second message will contain either the presence or absence of cognitive

and affective arguments (i.e., cognitive/aﬁ‘ective present, cognitive present/aﬁ‘ective

absent, cognitive absent/aﬁ‘ective present, cognitive absent/aﬁ‘ective absent). This second

message regarding the product, regardless of condition, will be a negative description of

the product and will therefore be counterattitudinal. The full design of the study is

presented in Table l.

 

 

 

 

 

Table 1
Study Desigr_i
Absence of Cognitive Presence of Cognitive
(unfavorable Arguments (unfavorable Arguments
Absence of Presence of Absence of Presence of
Affective Affective Affective Aﬁ’ective
(unfavorable) (unfavorable) (unfavorable) (unfavorable)
Arguments Arguments Arguments Arguments
Affective
Induction
(favorable)
Cognitive
Induction
(favorable)

 

 

 

 

 

The mismatch hypothesis predicts that cognitively based attitudes will be more

vulnerable to affective arguments and that affectively based attitudes will be more

vulnerable to cognitive arguments (Millar & Millar, 1990). Therefore, if the mismatch

hypothesis is consistent with the data, results Should indicate a main eﬁ‘ect for cognitive

arguments when attitudes are affectively based, such that the presence of cognitive

13

 

arguments produces attitude change and the absence of cognitive arguments results in

little to no attitude change. Similarly, results should indicate a main eﬁ‘ect for aﬁ‘ective

arguments when attitudes are cognitively based, such that the presence of aﬂ‘ective

arguments produces attitude change and the absence of affective arguments results in little

to no attitude change. In those conditions when SS are exposed to joint messages, that is,

messages containing both cognitive and aﬁ‘ective arguments, we would expect attitude

change because the SS are being exposed to the mismatched arguments. Table 2 illustrates

the predictions of the mismatch hypothesis.

Table 2

Predictions of the Mismatch Hypothesis

 

 

 

 

 

Absence of Cognitive Presence of Cognitive
(unfavorable Arguments (unfavorable Arguments
Absence of Presence of Absence of Presence of
Affective Aﬂ’ective Affective Aﬁ‘ective
(unfavorable) (unfavorable) (unfavorable) (unfavorable)
Arguments Arguments Arguments Arguments
Aﬂ‘ective No change Little to no Attitude Attitude
Induction (control group) attitude change change change
(favorable)
Cognitive No change Attitude Little to no Attitude
Induction (control group) change attitude change change
(favorable)

 

 

 

 

 

Altemately, the match hypothesis predicts that affectively based attitudes will be

more vulnerable to affective arguments (Edwards, 1990). For this Study, the match

hypothesis has been extended to include cognitively based attitudes, such that cognitively

formed attitudes are hypothesized to be more vulnerable to cognitive arguments.

l4

 

Therefore, if the data can be explained by the match hypothesis, results should indicate a
main eﬁ‘ect for affective arguments when attitudes are aﬂ‘ectively based, and a main eﬁ'ect
for cognitive arguments when attitudes are cognitively based. Extending the argument for
the match hypothesis as was done for the mismatch hypothesis, it would indicate that there
should be an absence of change when attitudinal bases and arguments are mismatched: we
should observe no attitude change when aﬁ‘ectively based attitudes are exposed to
cognitive counterattitudinal arguments, as well as a lack of effect when S5 with cognitively
based attitudes are presented with affective counterattitudinal arguments. Table 3

illustrates the predictions of the match hypothesis.

Table 3

Predictions of the Match Hypothesis

 

 

 

 

 

Absence of Cognitive Presence of Cognitive
(unfavorable Arguments (unfavorable Arguments
Absence of Presence of Absence of Presence of
Affective Affective Affective Affective
(unfavorable) (unfavorable) (unfavorable) (unfavorable)
Arguments Arguments Arguments Arguments
Affective No change Attitude Little to no Attitude
Induction (control group) change attitude change change
(favorable)
Cognitive No change Little to no Attitude Attitude
Induction (control group) attitude change change change
(favorable)

 

 

 

 

 

In this 2 x 2 x 2 design there are four cells that are central to testing the conﬂicting
hypotheses. Those four cells are the pure cells, that is, the cells where the SS are exposed

to _e_ith_er_ an affective g cognitive message during the attitude induction, and then are

15

 

 

exposed to either an affective gr cognitive persuasive message (see Table 4). When these

four central cells are viewed, one can see the direct contradiction of the two hypotheses.

Table 4

Cells Central to the Test of Conﬂicting Hypotheses“

 

 

Presence of Aﬂ‘ective Presence of Cognitive
Arguments Arguments
Affective induction M&M: Little to no M&M: Attitude change

attitude change
E: Little to no attitude

 

 

E: Attitude change change
Cognitive induction M&M: Attitude change M&M: Little to no
attitude change

E: Little to no attitude
change E: Attitude change

 

 

 

*Note: Predictions of the Mismatch hypothesis are preceded by “M&M” and are in bold
text; predictions of the Match hypothesis are preceded by “E” and are in regular text.

16

 

Chapter 3
METHODS

An experiment was conducted to test the conﬂicting match and mismatch
hypotheses of Edwards (1990) and Millar and Millar (1990), respectively. Speciﬁcally,
the eﬁ‘ects of matched versus mismatched counterattitudinal messages, provided to the
subjects after an initial favorable attitudinal induction using either a cognitive or aﬁ‘ective
message, was assessed by viewing the dependent variables of the difference in attitude,
and Ss’ conﬁdence in their judgment, as per Millar and Millar’s (1990) and Edwards’
(1990) studies. Pilot tests were undertaken to obtain induction check items to be included
in the ﬁnal questionnaire as well as to obtain the messages used in each condition.

The Xenon Light Bulb

For this study, an attitude toward an unfamiliar object was induced in the subjects.
The decision to induce an attitude, as opposed to identifying a preexisting attitude, was
made so that there would be control over what information and emotional content was
available to the subjects regarding the attitude object. Edwards (1990) noted the difficulty
in identifying the attitudinal base of a preexisting attitude as the origins of such an attitude
are often unknown (p. 204).

For this study, then, a relatively neutral object, a light bulb, was chosen as the
attitude object. In order to ensure that both types of messages (affective and cognitive)
presented to SS contained novel information, a new type of light bulb was invented, the
Xenon light bulb. The Xenon light bulb, to my knowledge, does not exist, but was

determined to be a believable attitudinal object, especially as subjects are often familiar

l7

with halogen bulbs and neon lights. Xenon is a gas similar to neon (a non-reactive element
with a ﬁll] outer shell of valence electrons) so that if any subjects were well acquainted
with the periodic table of elements, they would recognize this element as an actual element
and one that potentially could be used in the development of lighting sources. In the pilot
tests of the messages, a random sub-sample of subjects was asked to complete a Short
questionnaire that queried how important they thought light bulbs were to them, asked
them to comment on the study, and asked them to comment on Xenon light bulbs. The
respondents overwhelmingly noted that light bulbs were not a central focus in their lives,
although many noted that they use and purchase light bulbs for their dorm rooms and
homes, so that the messages were relevant to them. Of the 13 students surveyed, none
questioned the existence or the possibility of existence of a Xenon light bulb.
Pilot Tests

Two rounds of pilot tests were conducted. Both rounds consisted of two
questionnaires; one focused on assessing the cognitive and affective content of a variety of
statements (pilot testing of induction check items), another focused on assessing the
cognitive and affective content of short messages (pilot testing of messages to be used in
the ﬁnal experiment). All subjects received both questionnaires, although the second
questionnaire (the messages) differed between the ﬁrst and second rounds of pilot testing.
The two rounds of pilot tests were conducted because the results of the ﬁrst pilot test
indicated that the affective negative messages (described ﬁrrther below) needed to be
revised and retested. Reported below are the sample, data collection procedure,

instrumentation, and analysis procedure for each pilot questionnaire.

18

Pilot Testing of Inmrcﬁon Check Items

 

Participants. Participants consisted of 73 college students (28 males and 45
females) from a large, midwestem university.7 The students were enrolled in a variety of
communication classes, and they received extra credit for their participation in this study.
Responses were anonymous.

Procedure and instrumentation. A 35-item questionnaire containing statements
about light bulbs was distributed to the participants. The 35 items had been precoded by
two coders (kappa = 1.00) as being either positive or negative statements about the Xenon
light bulb. The researcher informed the students that they were to read carefully each
statement and to judge it both for cognitive content and for affective content. The SS
were instructed that something that is cognitive can be factual and provides one with
information or knowledge about a product (e.g., a light bulb), and something that is
affective is something that appeals to emotions, moods, or feelings. The subjects were
asked to rank each statement on b_ot_l_1 scales, and were provided with an example. Further,
SS were allowed to ask any questions to clarify the distinction between cognitive and
affective Statements. Subjects were informed that they were not judging the truth or
falsity of each statement; rather, they were to judge the degree to which they believed each
statement to be cognitive and the degree to which they believed each statement to be
aﬁ‘ective. Their responses were assessed on a Likert-type scale, where l was “not at all
cognitive (affective)” and 5 was “extremely cognitive (affective).” This portion of the
pilot testing procedure took SS approximately 5-10 minutes to complete. Following their

completion of this ﬁrst task, the SS were given a second task, discussed below.

19

Data analysis. Two sets of means were computed for each of the 35 items, a mean
indicating the degree to which S s thought the item was cognitive, and a mean indicating
the degree to which the S s believed the item was aﬁ‘ective. T-tests for paired samples
were computed for each item so that those items that were judged as high in cognitive
content and low in aﬁective content, as well as those items that were judged as high in
aﬁ‘ective content and low in cognitive content could be ascertained.

Results of pilot testing for cogm'tive and affective statements. A total of Sixteen
items were desired for inclusion on the ﬁnal questionnaire. These sixteen statements were
selected from the 35 items in the pilot test to ﬁt four categories: aﬁ‘ective positive,
aﬂ’ective negative, cognitive positive, and cognitive negative. The statements chosen were
ones that SS rated as being high in cognitive content and low in aﬂ‘ective content
(cognitive statements) and as being high in affective content and low in cognitive content
(affective statements). The selected items and their means on the cognitive and aﬂ‘ective
dimensions are reported in Appendix A. T-tests for each of the items between their
affective and cognitive ratings were performed. The t-tests of all of the items selected
indicated that their ratings on the two dimensions (affective and cognitive) were
Signiﬁcantly different. The t-test results for each item are also reported in Appendix A.

These 16 items were then included in the ﬁnal questionnaire. The section

comprised of these sixteen items was used as an induction check for the experimental

study.

20

Pilot Testing pf Affective and Cognitive Messages

Two rounds of this second pilot questionnaire were completed. The goal of this
sequence of pilot testing was to determine those messages to be used as the attitude
induction messages (one affective induction message, favorable toward the attitude object,
and one cognitive induction message, also favorable toward the attitude object) and the
counterattitudinal persuasion messages (one affective counterattitudinal message,
unfavorable toward the attitude object, and one cognitive counterattitudinal message, also
unfavorable toward the attitude object). Two rounds were necessary as the results ﬁ'om
the ﬁrst round of pilot testing indicated that the message options that were intended to be
aﬁ‘ective negative (the affective counterattitudinal messages) messages were not perceived
as such by subjects during the pilot testing. Therefore, a second round of pilot testing was
undertaken in order to test other messages to be used in this condition.

Message development. Messages were written so as to be favorable toward the
Xenon light bulb (induction, positive condition) or unfavorable toward the Xenon light
bulb (persuasion, negative condition). A total of eight messages were initially written for
pilot testing, two for each of the four conditions. AS noted later, additional pilot testing
was necessary; therefore, an additional four messages for the negative affective condition
were written and pilot tested. To prepare the messages for pilot testing, a random
numbers table was used to determine the order of each packet of messages to be given to
each subject. For the ﬁrst round of pilot testing a total of 56 packets were developed,
each of which had a unique, random order of the 8 messages. For the second round of

pilot testing, the preparation procedure was the same, that is, the messages were ordered

21

using a random numbers table so that the order in which the subjects read the messages
was random and varied by subject.

Participants - Round 1. The sample for round 1 consisted of 50 college students
ﬁom a large, midwestem university who were enrolled in communication classes. There
were 21 male and 29 female participants.

Pm'cipants - Round 2. The sample for round 2 consisted of 22 college students
from a large, midwestem university who were enrolled in a communication class. There
were 7 male and 15 female participants.

Procedure. Subjects were again reminded of the distinction between cognitive and
affective messages. They were informed that they would read a series of eight messages,
and complete a thought listing task and a short questionnaire after each message. Prior to
each message, S s were given a set of directions, asking them to think careﬁrlly about what
was being presented in each message. They were informed that they would be asked to
write down each argument that they read in the message, and to identify whether or not
the argument was a cognitive argument or an aﬁ‘ective argument. The SS were provided
time to ask any questions, and then they began the study. They read a series of eight
messages and listed the arguments made in each message as well as answered some
questions about each message. This part of the pilot testing took approximately 30-40
minutes. After completing the questionnaires, the SS were thanked, debriefed, and
dismissed.8 The same procedure was followed for round two of the pilot testing;
however, subjects were only presented with ﬁve messages. The second round of the pilot

testing took approximately 20-30 minutes to complete.

22

MMmentation/Measurement. Following their reading of each message, SS were
asked to write down any thoughts that crossed their mind as to whether the message that
they just read presented affective (emotional) arguments, or cognitive arguments. SS were
asked to write down the arguments they believed to have been presented in the message,
and were also asked to indicate whether or not they believed those arguments to be
positive, negative, or neutral. The S s were allowed to reread the message if they needed
to do so. Following this instructed thought-listing task, SS were asked to answer a series
of questions about the content of the message. They were asked to rate the message as a
whole, and, if they were forced to choose between labeling the message as an aﬂ‘ective
message or a cognitive message, to indicate in which category they would place the
message they had just read. Then, they were asked to evaluate the message on both an
aﬁ‘ective a_n_d a cognitive dimension. For this response, SS were asked to indicate the
degree to which they believed the message they had read was affective and the degree to
which they believed the message was cognitive. These items were answered on seven-
point Likert-type scales ranging ﬁ'om “message is not at all affective (cognitive)” to
“message is completely affective (cognitive).” SS responded to these items after each of
the eight messages with which they were presented. In round 2 of the pilot testing the
instrumentation was identical, but SS only were asked to read ﬁve messages about the
Xenon bulb.

Data analysis. These pilot data were analyzed by running frequencies on responses

to the overall message type and degree of affect in the message/degree of cognition in the

message.

23

 

Results of pilot testing forJaffectiveMan cognitive mesSaggs. Results indicated that
three of the four messages in the ﬁrst round of pilot testing satisﬁed the conditions for
inclusion in the study. The chosen cognitive positive, cognitive negative, and aﬁ’ective
positive messages were messages that an overwhelming majority of subjects identiﬁed as
being of the speciﬁc type (cognitive or aﬁ’ective) an_d messages that were not only rated as
being high in the primary dimension, but as being low in the secondary dimension. The
messages intended to ﬁll the affective negative condition were unsatisfactory, as SS were
divided as to whether or not they believed the messages to be cognitive or aﬁ'ective.

Thus, the fourth message (aﬁ‘ective negative condition) was chosen from those messages
tested in the second round of pilot testing. Table 5 presents Ss’ ratings as to the degree to

which they believed each of the messages to be cognitive or affective.

Table 5

Overall Ratings of Degree of Cogpition and Affect for Messages (Induction and
Persuasion)*

 

 

 

 

 

High Cognitive (Low High Affective (Low

Affective) Message Cognitive) Message
Positive message M: 1.96 (1.34) M= 6.30 (0.74)
(Induction) M = 6.10 (1.05) M = 1.92 (0.78)
Negative message M = 2.42 (1.62) M = 5.27 (1.78)
(Argument/Persuasion) M:5.94 (1.30) M = 2.91 (1.66)

 

*Note: Affective ratings are in regular text, cognitive ratings are in bold text. Standard
deviations are noted in parentheses.

Test of Perceived Message Strength

A separate sample of SS (H = 123) was used to perform a post hoc test of

perceived message strength. Forty-six males and 77 females at the same large midwestem

24

 

university participated in this test. Subjects ranged in age ﬁom 18 to 35 M = 22.29, SQ
= 4.12). The ethnic background of the SS is as follows: 5.7% Asian, 16.7% Black/Aﬁican
American, 0.8% Hispanic, 2.4% Native American, 66.7% White, and 5.7% indicated
“other.”

Independent groups were asked to complete a Short questionnaire that presented
one of the four messages and then asked SS to respond to three Likert-type items
assessing the perceived strength of the message. The three-item scale had a reliability of or
= .89.

Results indicated that the strength of the cognitive and affective messages did not
differ in either the positive (induction) or negative (persuasion) conditions.9 Table 6

presents the results of the assessment of perceived strength of the messages.

Table 6

Perceived Strength of Messages*

 

 

 

 

High Cognitive Message High Affective Message
Positive Message M = 14.24 M = 13.06
(induction) S1; = 3.94 S_D = 2.97
p = 29 p = 33
Negative Message M = 10.40 M = 10.61
(argument/persuasion) S_D = 4.45 S_D = 4.67
g = 30 p = 31

 

 

 

 

*Note: Means reﬂect the combined score on the three-item scale, where l = weak and
21 = strong.

ANOVA results indicated that there was a main effect for valence such that the
positive messages were perceived as being stronger than the negative messages (E =

18.25, p < .001). There was no interaction effect between argument type and valence (E ,

25

1.00, ns). The differential perception of the strength of the messages in the two valence
conditions is not a threat to this study as this study’s focus was on the persuasive eﬁ‘ects
of aﬁ’ective and cognitive arguments on affective and cognitive attitudes.

The four messages that were selected for the study, then, were those that met the
above criteria. The messages are presented in Appendix B. The two positive/favorable
messages were used as the attitude induction (message 1; SS received m of these two
messages). The two negative/unfavorable messages were used as the counterattitudinal
persuasive messages [message 2; independent groups of SS received messages that were
one of the following four types: (1) affective only, (2) cognitive only, (3) both affective
and cognitive, or (4) a ﬁller/control message that gave information on a non-related
product (chair fabric)].

The Study

Using the messages selected by the above means, a study was undertaken to test
the conﬂicting match and mismatch hypotheses regarding the effectiveness of aﬁ‘ective and
cognitive messages (arguments) on changing attitudes that are affectively or cognitively
based.

Resign

An independent groups design with three factors (attitude induction, cognitive
argument condition, and affective argument condition) was used. The attitude induction
category refers to the type of attitude induced by the ﬁrst message (either cognitive or

affective).

26

The second set of messages was presented to the Ss after they had been exposed
to the initial message and had completed the initial questionnaire. These messages
comprised counterattitudinal statements that contained either cognitive arguments,
affective arguments, a combination of the two, or absence of both. The cognitive
argument factor had two conditions: either the presence or absence of a cognitive
argument. The cognitive argument presented information addressing beliefs and
judgments about the attitude object. The affective argument factor had two conditions:
the presence or absence or aﬁective arguments. The aﬁ’ective argument focused on
feelings and emotions about the attitude object. Thus, the design is a 2 x 2 x 2
independent groups design.

Participants

The sample consisted of 607 college students at a large, midwestem university.
All students were enrolled in undergraduate communication courses. Fourteen SS were
deleted from the sample because they had participated in the pilot test (in another course).
One subject was deleted from the sample because he was suspicious of the research
hypotheses. Another seven SS were deleted, three because they turned in virtually blank
questionnaires, and four who were hostile SS, evident by their responses to the thought-
listing questions and other written remarks they made throughout the questionnaire. The
ﬁnal sample, then, was 585 Students. Two hundred and forty-ﬁve males and 326 females
participated in this study. The ages of participants ranged from 16 to 31 M = 19.92, SD
= 1.72). Thirty-four percent of the subjects were freshmen, 30.4% were sophomores,

18.8% were juniors, 12.8% were seniors, 2.1% were ﬁfth-year seniors, and 0.5% of the

27

participants were graduate students or those enrolled in lifelong education. The ethnic
background or race of the subjects is as follows: 3.1% Asian, 13.3% Black/Aﬁican-
American, 1.7% Hispanic, 0.2% Native American, 74.7% White, 2.7% indicated “other,”
and 4.3% left this item blank.

Procedure

Subjects were told that they were engaging in a study, the goal of which was to
determine how they viewed message readability, or how easy or hard they believed the
message was to read. The directions with which they were presented on the
questionnaires themselves varied as to the condition in which the subjects were placed.
Instnrctions that served to focus their reading of the message and their subsequent
responses to questionnaire items in either a cognitive or an affective direction were
presented. Those SS in the affective condition were instructed to “...analyze h_o_w_ you feel
about what you are reading. That is, go over in your mind how you are feeling while you
read the message.” Those SS in the cognitive induction condition were instructed to
“...analyze w you think the way you do about the product. That is, go over in your
mind what it is about the information that you are presented that makes you think why you

would like the product or not. AS you are reading, think of reasons that make you like or

 

dislike the product.” These instructions were designed to focus the SS’ responses in either

a cognitive or affective direction. Such instructions have been used successﬁrlly by others

to focus the direction of SS’ attitudes (Millar & Tesser, 1986, 1989; see Appendix C).
After reading these directions, the SS were presented with a short written message

that contained either cognitive Q_r_ affective information about the Xenon light bulb

28

(attitude induction). This ﬁrst message, regardless of attitude type, was favorable toward
the Xenon light bulb. After reading the message, the S S were again primed so as to
interpret the message either cognitively or affectively. The same directions as noted above
were used in the repeat prime. The SS were then instructed to list either their reasons for
liking or disliking the product (cognitive induction) or the feelings they experienced while
reading the message (affective induction). SS also were instructed to indicate the valence
of the reasons or feelings they listed, by circling either -, 0, or + signs.

Following this directed thought-listing procedure, SS were asked to respond to a
series of eight items that dealt with the cover story. The remainder of the questionnaire
included items to check the inductions, items to measure SS’ attitudes toward the Xenon
light bulb, and items to measure their conﬁdence in their judgment.

Following their completion of the ﬁrst part of the questionnaire, SS were again
given instructions to focus their reading, and were presented with a second message. This
second message either contained counterattitudinal information about the Xenon light bulb
(three variants) or was a control message. Speciﬁcally, the second message was
composed of one of the following: (1) affective information only (affective argument
present/cognitive argument absent), (2) cognitive information only (affective argument
absent/cognitive argument present), (3) affective and cognitive information (affective
argument present/cognitive argument present), or (4) a control message with information

about another, irrelevant attitudinal object (affective argument absent/cognitive argument

absent) (chair fabric, see Appendix B).

29

After reading the second message S S were presented with further instructions,
again intended to prime them to respond to the message (1) affectively, (2) cognitively, or
(3) both aﬁ‘ectively and cognitively. Those SS in the control group were merely instructed
to “write down any comments that you had as you were reading this message.” Following
completion of a thought-listing task identical to that after the ﬁrst (induction) message, SS
completed the questionnaire a second time. Those Ss in the control group were instructed
to complete the thought-listing procedure and the readability items thinking of the
message about chair fabric that they had just read. SS in the control group answered six
questions related to chair fabric, and were then instructed to think back to their reactions
to the Xenon bulb (without turning back in the questionnaire), and they then completed
the remainder of the questionnaire which focused on their attitudes toward the Xenon light
bulb and their conﬁdence in their judgments. The ﬁnal section of the questionnaire
contained demographic questions. Upon completion of the questionnaire, SS were
debriefed, thanked, pledged to secrecy about the study, and dismissed.
Mgrumentation/MeasMurement

Induction check. Those 16 statements that were chosen from the results of the
ﬁrst pilot test were presented to the subjects. Subjects were instructed to place a
checkmark next to the three statements that best represented their reaction to the Xenon

bulb.

Attitude. In order to assess subjects’ attitudes toward the attitude object, 15 items

 

intended to measure attitude were included in the questionnaire. The items that Millar and

Millar (1990) and Edwards (1990) used in their studies were included, as were additional

30

items that have been used to assess attitudes in attitudinal research (c.f., Osgood, Suci, &
Tannenbaum, 1957; Petty & Wegener, 1998). Likert-type response scales, as well as
semantic diﬁ‘erential items (Osgood, Suci, & Tannenbaum, 195 7) were included to assess
SS attitude toward the Xenon light bulbs. The same items were used after both messages.

Conﬁdence in judgment. As Edwards (1990) also hypothesized that those SS who
had aﬂ'ective attitudes would be more conﬁdent in their judgment than those SS with
cognitive attitudes, conﬁdence was also measured in this study. Four Likert-type items
were included in the questionnaire to measure SS’ conﬁdence in their judgment about the
Xenon bulbs. These conﬁdence items followed the attitudinal items in the questionnaire.

Counterarguing. Counterarguing was measured by coding the thoughts and
feelings listed by the SS after each message. Individual coders rated the thoughts as either
“message-relevant” or “message irrelevant.” Ss themselves indicated the valence of each
statement, coding statements that were favorable or positive toward the Xenon bulbs (+),
negative toward Xenon bulbs (-), or irrelevant or neutral toward Xenon bulbs (0). Those
thoughts that were coded as message relevant and were in opposition to the valence
advocated by the message were considered to be counterarguments. Thus, those thoughts
after the ﬁrst message, which was favorable toward the Xenon light bulb, that were
message relevant and negative were considered to be counterarguments. Similarly, those
thoughts listed after the second message that were message relevant and positive (as the
second message was unfavorable toward the Xenon bulb) were considered as

counterarguments. This procedure parallels that of Millar and Millar (1990).

31

Ari—allies

Data were analyzed using a number of methods. The thoughts listed by each
subject were coded twice, using two separate coding schemes and two separate sets of
coders. The ﬁrst coding scheme sought to identify the type of thoughts and emotions that
were listed by each subject. Two independent coders coded all of the thoughts as being
aﬁ’ective, cognitive, both, or neither. Initially, only the two categories of aﬂ‘ect and
cognition were to be included in the coding scheme. Coders were instructed to code
feelings and emotions as being affective (e.g., makes me feel warm and cozy, I’m bored)
and to code non-emotional information, reasons given for liking or disliking the objects as
cognitive (e. g., economical product, light bulb is recyclable). It was discovered, however,
that, as the statements listed by the subjects were brief and may have only consisted of one
word, there were occasions when it was not possible to determine whether the subject was
implying something that would be coded as aﬂ‘ective, or that would be coded as cognitive.
Such comments (e. g., suspicious, dangerous) were coded as being both. In addition, some
comments listed by the SS were either irrelevant to the study, or did not ﬁt in any of the
three categories described above. Thus, the category (to be used infrequently) “neither”
was added. The comments that were coded in this category included statements irrelevant
to the attitude object (e. g., extra credit is good, I like extra credit)”. Interrater reliability
for this coding task was calculated using Cohen’s kappa, and equaled .93. Disagreements
were resolved by discussion of the two coders.

The second coding scheme was established in order to determine the relevance of

each comment. Two independent coders coded each Statement as being either message

32

relevant or message irrelevant. Those statements coded as message relevant were
statements having to do with the attitude object, repetition of information provided in the
message, or statements that the coders judged would not have been listed had the SS not
read the message (e.g., suggesting the message was too long, remarking on their distaste
for the name of the ﬁctitious company, Reyovan, etc.). Message irrelevant statements
included those that were not directly linked to the message and were deemed to arise ﬁom
forces other than the message itself (e.g., I’m bored, I’m hungry). Interrater reliability for
this second coding task was calculated using Cohen’s kappa and equaled .87.
Disagreements were resolved by a third coder.

The results of the induction check analysis were analyzed using Millar and Millar’s
(1990) procedure. S s’ attitudes toward the Xenon bulb were classiﬁed as being either
affective or cognitive by a majority rule. Thus, if a subject checked two items (out of
three items checked) that were previously determined to be affective positive, then the
subject was considered to have a favorable affective attitude toward the Xenon bulb.

Conﬁnnatory factor analysis was used in order to test the ﬁt of the attitude and
conﬁdence scales. ANOVA procedures and contrast analysis were employed to test the

match and the mismatch hypotheses.

33

Chapter 4
RESULTS

Mgrview

Results will be presented in order to parallel the analyses. Factor analyses of the
attitude items are presented, followed by the results of the induction checks and cover
story analyses. AN OVA results then are presented, followed by results of the contrast
analyses employed to test the mismatch and match hypotheses. An alternate explanation,
the Cognitive Argument Primacy Model, is presented and tested with contrast analyses.
Finally, analyses of Edwards’ (1990) conﬁdence hypothesis and Millar and Millar’s (1990)
counterarguing hypothesis are presented.
Ppelimingy analyses

In order to test the dimensionality of the attitude factor, conﬁrmatory factor
analyses were employed (Hunter & Gerbing, 1982). Upon examination of the observed
correlation matrices and factor loadings, as well as the predicted correlation matrices and
subsequent error values, the 15 items intended to measure attitude were inconsistent with
a unidimensional factor solution (items, means, and standard deviations are listed in
Appendix D). Upon inspection of the observed correlation matrices and factor loadings,
three distinct, yet highly correlated factors emerged. The three factors identiﬁed using
conﬁrmatory factor analysis were labeled attitude, value/worth, and dependability.“ Each
subject responded to the 15 items twice, once after the induction message, and again
following the persuasion message. Thus, two conﬁrmatory factor analyses were

performed to ensure that the three-factor model held for both sets of responses.

34

Conﬁrmatory factor analyses indicated that the three-factor model was consistent with the
data after both messages. The inter-item correlations with their factor loadings on the
three-factors following each message, as well as the predicted and residual matrices, are
presented in Appendix E.

1mm. Attitude was measured using a S-item scale with a reliability of or = .92
following the induction message and a reliability of or = .97 following the persuasion
message. The ﬁve items composing the attitude scale were Likert-type scales, upon which
the subjects’ responses ranged from 1, indicating a dislike or lack of favor toward the
Xenon bulb, to 7, indicating a liking or favorability toward the Xenon bulb.

Value/Worth. Subjects’ perceptions of the value or worth of the Xenon light bulb
were measured on a ﬁve-item semantic differential scale with a reliability of or = .88
following the induction message and a reliability of or = .93 following the
counterattitudinal persuasion message. Subjects’ responses were assigned values ﬁom 1,

indicating a belief that the light bulb was bad and worthless, to 7, indicating that the light

bulb was good and valuable.

 

Dependajility. Subjects’ perceptions of the dependability of the Xenon light bulb
were measured with two semantic differential items with reliabilities of or = .81 and or =

.82. Subjects’ responses ranged from 1, indicating undependable or unreliable, to 7,
indicating dependable or reliable.
IMuction Chetg

In order to ensure that the messages induced the type of attitude (affective or

cognitive) that was intended, an item on the questionnaire asked SS to check the three (out

35

of a possible 16) statements with which they most agreed. For those SS in the aﬁ‘ective
(positive) condition, 83.3% indicated an affective positive attitude. For those SS in the
cognitive induction condition, 89.8% indicated their attitude conformed to a cognitive
positive attitude.

In addition, the mean number of aﬁ‘ective positive and cognitive positive thoughts
that were coded indicated success of the attitude induction. Those Ss receiving an
aﬁ‘ective induction (and primes in the instructions) listed more affective positive thoughts
M = 1.00, S_D = 1.3) than those in the cognitive induction M = .00, SD = .06), _E(1, 583)
= 173.31, p <.001, n2 = .23, _r_ = .48. Those Ss receiving a cognitive induction (and
cognitive primes) listed more cognitive positive thoughts M = 2.61, SQ = 1.31) than
those Ss in the affective induction condition M = .73, S_D_ = 1.04), E0, 583) = 356.30, p
<.001, n2 = .38, 1: = .62. Induction type did not produce any effect on the total number of
thoughts that the SS listed, E(1, 583) = 1.23, ns, 112 = .00, g = .05.

Cover Stogg Analysis of Induction Messages

Three items were used to measure the ease of comprehension of the message (a =
.93). When responding to items assessing the ease or difﬁculty of understanding the
message they read, SS in the cognitive induction condition rated that induction message as
being slightly more difﬁcult to understand or comprehend (M = 2.24, SD = 1.28, p = 294)
than did those SS who were exposed to the affective induction message M = 1.99, SD =

1.10, p = 291), 5(1, 583) = 6.55, p <.02, n2 = .01, g = .11.12

36

Initial Dependent Variable Analyses

Change scores were computed for each of the three factors identiﬁed by the
conﬁrmatory factor analysis. The attitudinal induction (message 1) was always favorable
and the persuasive message (message 2) was always unfavorable. The Ss’ ﬁrst evaluations
were subtracted ﬁom their second evaluations to obtain a difference score for the
difference in attitude, difference in perceptions of value/worth, and diﬁ‘erence in

perception of dependability. The absolute value of these change scores was used for all
further analyses.

Separate analyses of variance were performed on each of the three factors. For the
attitude factor, ANOVA indicated main eﬁ‘ects for induction type (cognitive or aﬁ‘ective),
E(1, 541) = 6.06, p<.02, n2 = .01, g = .08, such that those SS whose attitudes were
cognitively induced were more persuaded by the persuasive arguments than were those SS
whose attitudes were affectively induced; for the presence or absence of affective
arguments (in the persuasion message), E(l, 541) = 64.23, p<.001, n2 = .07, g = .27, such
that the presence of affective arguments was more persuasive than if there were no
affective arguments; and for the presence or absence of cognitive arguments (in the
persuasion message), E(l, 541) = 203.93, p<.001, n2 = .23, p = .48, indicating that the
presence of cognitive arguments was more persuasive than if there were no cognitive
arguments present. '3 These main effects were qualiﬁed by two interactions. Analyses
indicated an affective argument by cognitive argument interaction, E(1, 541) = 63.82,
p<.001, n2 = .07, g = .27, such that when cognitive arguments were absent, the presence

of affective arguments was more persuasive than if there were no affective arguments, and

37

when cognitive arguments were present, the amount of attitude change was equal,

regardless of the presence or lack of aﬁ’ective arguments. In addition, there was a

cognitive argument by attitude induction interaction eﬁ‘ect, E0, 541) = 8.31, p<.005, n2 =

.01 , g = .10, such that when cognitive arguments were absent, the amount of attitude

change was equal, regardless of the induction type, and when cognitive arguments were

present, they were more persuasive in the cognitive induction condition. Analysis of

variance indicated that the three-way interaction was neither signiﬁcant nor substantial,

E(1, 541) = .208, ns, 112 = .00, g = .02. Table 7 presents the obtained means, standard

deviations, and cell sizes for each condition on the attitude change variable.

Table 7

Obtained Means, Standard Deviations, and Cell Sizes for Difference in Attitude

 

Absence of Cognitive Persuasive

Presence of Cognitive Persuasive

 

 

 

 

 

 

Arguments (-) Arguments 9)
Absence of Presence of Absence of Presence of
Affective Affective Aﬂ‘ective Affective
Persuasive Persuasive Persuasive Persuasive
Arguments (—) Arguments (-) Arguments (-) Arguments (-)
Affective M: .14 M= 2.16 M= 2.52 M=2.55
Attitudinal S_D = .85 S_D = 1.59 SD = 1.52 E = 1.38
Induction (+) n = 69 p = 73 g= 63 p = 71
Cognitive M=.24 M= 1.97 M=3.18 M=3.15
Attitudinal S_ = 1.02 SD = 1.46 _S_1_)_ = 1.52 _S_D_ = 1.48
Induction (+) p = 65 [_r_ = 69 p = 66 p = 73

 

 

 

For the value/worth factor, ANOVA indicated no main eﬁ‘ect for induction type

(cognitive or affective), 5(1, 548) = 3.18, ns, n2 = .00, g = .06. Main effects were

observed for affective arguments, E(1, 548) = 104.96, p<.001, n2 = .12, g = .35, such that

38

 

the presence of affective arguments were more persuasive than the lack of aﬁ‘ective
arguments; and for cognitive arguments, E(1, 548) = 141.14, p<.001, n2 = .17, _r_ = .41,
such that cognitive arguments were more persuasive than the lack of cognitive arguments.
These main eﬁ’ects were qualiﬁed by an aﬁ‘ective argument by cognitive argument
interaction, E(l, 548) = 46.11, p<.001, n2 = .05, g = .23, such that when cognitive
arguments were absent, the presence of affective arguments caused greater change in
perceptions of value/worth than did the lack of affective arguments, and when cognitive
argument were present, the amount of change was equal, regardless of the presence or
absence of affective arguments. In addition, there was a cognitive argument by attitude
induction interaction effect, E(1, 548) = 6.23, p<.02, n2 = .01, _r_ = .09, such that when
cognitive arguments were absent the amount of value/worth change was equal, regardless
of the type of induction, but when cognitive arguments were present, they led to greater
change in perceptions when attitudes had been cognitively induced than when attitudes
had been affectively induced. Analysis of variance indicated that the three-way interaction
was neither signiﬁcant nor substantial, E(1, 548) = 2.22, ns, n2 = .00, g = .05. Table 8

presents the obtained means, standard deviations, and cell sizes for each condition on the

value/worth change variable.

39

Table 8

 

 

 

 

 

t'nedM tan ardDeviainsan Cell ° sforDiﬁ‘ en inV
Absence of Cognitive Persuasive Presence of Cognitive Persuasive
Arguments (-) Arguments 0)
Absence of Presence of Absence of Presence of
Aﬁ‘ective Aﬁ‘ective Affective Aﬁ‘ective
Persuasive Persuasive Persuasive Persuasive
Arguments (-) Arguments (-) Arguments (-) Arguments (-)
Aﬂ‘ective M = -.05 M = 2.34 M = 2.18 M = 2.52
Attitudinal S_D = .43 SD = 1.69 SQ = 1.66 SQ = 1.56
Induction (+) p = 67 n = 69 pf 68 g = 74
Cognitive M= .11 M= 1.99 M=2.58 M=3.14
Attitudinal S_D = .90 SQ = 1.49 SQ = 1.73 SD = 1.63
Induction (+) g: 64 ﬁni= 71 E7 2 = 76

 

 

 

 

 

For the dependability factor, ANOVA indicated main eﬂ‘ects for induction type

(cognitive or affective), _E(1, 561 = 8.12, <01, 2 = .01, r = .11, such that those SS
12 11

whose perceptions of dependability had been cognitively induced changed more than did

those SS whose perceptions had been affectively induced; for affective arguments, E(1,

561) = 33.03, p<.001, n2 = .05, g = .22, such that the presence of affective arguments

were more persuasive than the lack of affective arguments; and for cognitive arguments,

E(1, 561) = 42.45, p<.001, n2 = .06, g = .25, such that cognitive arguments were more

persuasive than the lack of cognitive arguments. These main eﬂ‘ects were qualiﬁed by an

affective argument by cognitive argument interaction, 5(1, 561) = 24.54, p<.001, n2 = .04,

1; = .19, such that when cognitive arguments were absent, the presence of affective

arguments caused greater change in the perceptions of dependability than did the lack of

affective arguments, and when cognitive arguments were present, the amount of change

was equal, regardless of the presence or absence of aﬁ‘ective arguments. In addition, there

40

 

was a cognitive argument by attitude induction interaction, E0, 561) = 6.48, p<.02, n2 =

.01, r = .10, such that there was no effect of the attitude induction when cognitive

arguments were absent, but when cognitive arguments were present, Ss’ perceptions of

the dependability of Xenon bulbs changed to a greater extent when cognitively induced

than when aﬁ‘ectively induced. Analysis of variance indicated that the three-way

interaction was neither Signiﬁcant nor substantial, 5(1, 561) = .27, ns, 112 = .00, g = .02.

Table 9 presents the obtained means, Standard deviations, and cell Sizes for each condition

on the dependability change variable.

Table 9

Obtained Mgans. Standard Deviptions. and Cell Sizes for Difference in Dependabilig

 

 

 

 

 

 

 

 

 

 

 

Absence of Cognitive Persuasive Presence of Cognitive Persuasive
Arguments (—) Arguments (‘)
Absence of Presence of Absence of Presence of
Affective Affective Aﬁ‘ective Aﬁ‘ective
Persuasive Persuasive Persuasive Persuasive
Arguments (-) Arguments (-) Arguments (-) Arguments (-)
Affective M=-.03 M: 1.55 M=1.28 M= 1.38
Attitudinal SQ = .57 S_D_ = 1.69 S_D = 1.79 _SQ = 1.72
Induction (+) p = 68 _pt= 74 p = 68 p = 75
Cognitive M = .14 M = 1.47 M = 1.99 M = 2.12
Attitudinal S_D = .89 S__ = 1.72 SQ = 2.09 SQ = 1.73
Induction (+) g = 66 p = 72 p = 69 p = 77
The Misngtthh Hypothesis

 

Attitude change Contrast analysis was employed to test Millar and Millar’s

(1990) mismatch hypothesis. They hypothesized that, if the attitude and argument type

were matched, in the absence of other messages, we would see little to no attitude change,

41

 

for subjects would tend to counterargue the messages that matched their attitudinal base.
Thus, Millar and Millar predict that there should be a main effect for cognitive arguments
when attitudes are aﬁ‘ectively based, and a main effect for aﬁ'ective arguments when
attitudes are cognitively based. Further, there should be a la_c_lg of eﬁ‘ect when the
attitudinal bases match the argument type. To examine the mismatch hypothesis, a
Statistical model was constructed, using contrasts to indicate the Speciﬁc predictions of the
mismatch hypothesis. These contrasts are presented in Appendix F.

Data were inconsistent with the constructed model; although there was a nontrivial
effect for the contrast, 3(1, 541) = 90.23, p <.01, n2 = .10, r = .32, the residual explained
was substantial E(6, 541) = 43.61, p<.01, n2 = .29. Predictions ofMillar and Millar’s
(1990) mismatch hypothesis accounted for merely 25.6% of the explained sum of squares.

As noted earlier, there was a main eﬁ‘ect of the induction such that those SS whose
attitudes were cognitively induced demonstrated a greater amount of attitude change than
those SS whose attitudes were affectively induced. Millar and Millar would neither have
expected nor hypothesized this main effect for induction. Thus, a second contrast analysis
was performed, allowing for this main effect of induction (see Appendix F for contrast
coefﬁcients). Again there was an effect for this contrast E(1, 541) = 100.82, p<.01,'r|2 =
.11, g = .34, yet the residual explained was substantial E(6, 541) = 41.84, p<.01, n2 = .28,
and a mere 28.7% of the explained sum of squares were explained using these adjusted

contrasts. Thus, Millar and Millar’s (1990) mismatch hypothesis was inconsistent with the

data.

42

Focusing on the four cells that are central to the test of the mismatch hypothesis
(i.e., the four pure cells where SS received (1) either an affective or cognitive induction
and (2) either an affective or cognitive persuasive argument), when attitude type and
argument type are matched, the mismatch hypothesis predicts little to no attitude change.
In this study there was attitude change in both of those cells M = 2.16 for aﬁ‘ective
induction/affective persuasion and M = 3.18 for cognitive induction/cognitive persuasion).
Indeed, when SS with a cognitively induced attitude received cognitive persuasive
arguments, their attitude changed to a gLatel extent M = 3.18) than in any other of the
pure categories. The attitude change in these cells, then, accounts for the large residual
and the failure of the contrast model developed ﬁom the mismatch hypothesis.

Difference in Evﬂation of Value/Worth. As conﬁrmatory factor analyses
indicated the existence of a value/worth factor, contrast analyses were performed on this
factor as well. ‘4 Contrast analysis was employed to test the precise hypothesis. Data were
again inconsistent with the constructed model. Contrast analysis indicated that although
there was a nontrivial effect for the contrast, _E(1, 548) = 94.56, p<.01, n2 = .11, _ = .33
(explaining 30.9% of the explained sum of squares), the residual explained variance was
substantial, £(6, 548) = 35.30, p<.01, n2 = .25. Thus, the predictions of the mismatch
hypothesis were inconsistent with the data for the difference in judgments of value and
worth of the Xenon bulb. Again, viewing the four pure cells, results indicate that in those
cells where the mismatch hypothesis predicted little to no attitude change, attitude change

was evident (M = 2.34 for affective induction/affective persuasion, and M = 2.58 for

cognitive induction/cognitive persuasion).

43

Difference in Evaluation of Dependability. The mismatch hypothesis was again
tested through contrast analysis for the third factor, that of the dependability of the Xenon
bulb. The etrect was minimal, E0. 561) = 22.48, p<.01, n2 = .03, r = .18 (explaining a
mere 19.4% of the explained sum of squares); the residual explained was substantial, _F_(6,
561) = 15.60, p<.01, n2 = .14. Contrasts allowing for the unanticipated main effect ofthe
induction were also tested and were inconsistent with the data: E(1, 561) = 30.97, p<.01,
n2 = .05, _r_ = .21(explaining 26.7% of the explained sum of squares), with residual
explained E(6, 561) = 14.19, p<.01, n2 = .13. Again, the four pure cells Show where the
predictions failed: those SS whose induced attitude type matched the persuasion type
reported a change in attitude where the mismatch hypothesis would have predicted little to
no attitude change (M = 1.55 for affective induction/aﬁ‘ective persuasion and M = 1.99 for
cognitive induction/cognitive persuasion).

Thus, on all three factors identiﬁed in this Study, difference in attitude, difference
in value/worth, and difference in dependability, predictions of the mismatch hypothesis
were inconsistent with the data. Therefore, the results of this study indicate that the
mismatch hypothesis, that one will be more persuaded by a message that is opposite in
argument type to one’s underlying attitude type, was not consistent with the data.

The Match Hypothesis.

Attitude change. Contrast analysis was employed to test the match hypothesis

proposed by Edwards (1990). She hypothesized that, in order to attain attitude change, it

is necessary that the attitude type and the argument type match, that is, if one were to hold

 

an affective attitude, a persuasive message, in order to be successful, would have to

include aﬁ‘ective arguments. The match hypothesis was extended in this study to predict
that if one holds a cognitive attitude toward an object, then it is necessary that the
persuasive message includes cognitive arguments in order for attitude change to occur. In
addition, there Should also be a lack of change when the attitude type and argument type
are mismatched, that is, an aﬁ‘ective argument alone will not be persuasive for one who
holds a cognitive attitude, and a cognitive argument alone will not be persuasive for one
who holds an aﬂ‘ective attitude.

To examine the match hypothesis, a Statistical model was constructed, using

contrasts assigned to ﬁt the predictions of the match hypothesis. The contrast coeﬁicients

used are give in Appendix G.

Results of contrast analysis indicate that the data were inconsistent with the
constructed match model. The test produced a strong effect for the contrast, E(1, 541) =
172.51, p<.01, n2 = .19, 1_' = .44; however, the residual explained was substantial, E (6,
541) = 29.89, p<.01, n2 = .20, and only 49% of the explained sum of squares from the
ANOVA analysis were explained. Contrasts were assigned to account for the unexpected
main effect of the induction (see Appendix G for contrast coefﬁcients). This test
produced a Strong effect for the contrast, E(1, 541) = 195.64, p<.01, n2 = .22, _r_ = .47, yet
the residual explained was substantial at E(6, 541) = 26.04, p<.01, n2 = .18, and only
explained 55.6% of the explained sum of squares.

Viewing the four pure cells allows us to ascertain where the predictions of the
match hypothesis and its corresponding model failed. When attitude type and argument

type are mismatched, the match hypothesis predicts little to no attitude change. In this

45

study there was attitude change in both of those cells M = 2.52 for affective
induction/cognitive persuasion and M = 1.97 for cognitive induction/affective persuasion).

Difference in Evaluatipn of Value/Worth. Edwards’ match hypothesis was also
tested on the factor that assessed the diﬁ‘erence in the evaluation of the value or worth of
the Xenon light bulb after being exposed to both messages (favorable and unfavorable).
Again a Strong eﬂ'ect for the contrast was produced, E(1, 548) = 159.59, p<.01, 112 = .19,
r = .43 (explaining 52.1% of the explained sum of squares), yet the residual explained was
substantial as well E(6, 548) = 24.45, p<.01, n2 = .17. The four pure cells again allow us
to see where the predictions of the match hypothesis were inconsistent with the data M =
2.18 for aﬂ‘ective induction/cognitive persuasion and M = 1.99 for cognitive
induction/ affective persuasion).

Difference in Evaluation of Depengalgim. The ﬁnal variable, the diﬂ‘erence in
dependability of the Xenon bulb, presented here as a change score, fared as did the other
factors. There was a nontrivial effect for the contrast E(1, 561) = 58.17, p<.01, n2 = .09, g
= .29 (explaining 50.1% of the explained sum of squares); however, the residual explained
was nontrivial as well, E(6, 561) = 9.65, p<.01, n2 = .09. Employing contrasts that
accounted for the unexpected main effect of induction did not change these results: the
effect of the contrast was nontrivial at _F_(l, 561) = 69.29, p<.01, n2 = .10, _r_ = .32
(explaining 59.7% of the explained sum of squares), yet the residual was nontrivial as well,
E(6, 561) = 7.79, p<.01, n2 = .07. The large residual can be explained by viewing the pure

cells and the amount of attitude change that occurred where the match hypothesis would

46

have predicted little to no attitude change M = 1.28 for affective induction/cognitive
persuasion and M = 1.47 for cognitive induction/affective persuasion).

Thus, on all three factors identiﬁed in this study, diﬁ’erence in attitude, diﬂ'erence
in value/worth, and diﬁ‘erence in dependability, predictions of the match hypothesis were
inconsistent with the data. Therefore, the results of this study indicate that the match
hypothesis, that one will be more persuaded by a message containing an argument that is
consistent with the one’s attitudinal basis, was not consistent with the data.

The Cpgnitive Argument Primgcy Model

AS neither the match nor the mismatch hypotheses were consistent with the data,
the means were inspected to determine if an alternate explanation of the data could be
found. Upon inspection, a three-way interaction was evident. Initial ANOVA analyses
had indicated the lack of a three-way interaction. Because the three-way interaction that
was observed in these data was not the crossover interaction that an ANOVA would have
predicted, a contrast analysis was employed to test the observed interaction.

When cognitive arguments were absent (in the second message the SS received)
there was a main effect for affective arguments such that those SS who received aﬁ‘ective
arguments indicated a greater amount of attitude change M = 2.07, S_D = 1.53, p = 142)

than those SS who did not receive affective arguments (M = .19, SQ = .93, p = 134), _F(1,

272) = 150.03, p <.001, n2 = .35, g = .60. In addition, when cognitive arguments were
absent there was no effect of the type of induction, such that S3 with an affectively

induced attitude indicated an equal amount of attitude change M = 1.18, SQ = 1.63, p =

47

142) as did those SS with a cognitively induced attitude M = 1.13, SQ = 1.53, p = 134),
5(1, 272) = .10, ns, n2 = .00, r =02.

When cognitive arguments were present, a main eﬁ‘ect for the type of induction
was observed such that those SS whose attitude was cognitively induced indicated a
greater amount of attitude change M = 3.16, SQ = 1.50, p = 139) than those SS whose
attitude was affectively induced M = 2.54, SQ = 1.44, p = 134), E (1, 269) = 12.33, p
<.002, 112 = .04, g = .21. In addition, when cognitive arguments were present there was no
effect for the presence or absence of affective arguments: SS indicated an equal amount of
attitude change when affective argument were present M = 2.85, SQ = 1.46, p = 144) as
they did when affective argument were absent M = 2.86, SQ = 1.55, p = 129), _E(1, 269)
= .001, ns, n2 = .00, 2 =00.

Such results indicate that cognitive arguments are more important and effective in
changing attitudes toward a novel object, particularly when the attitude has been
cognitively induced. This importance of the cognitive arguments overrode any eﬁ‘ect of
the affective arguments. When cognitive arguments were present there was no effect for
the presence or absence of affective arguments, regardless of induction condition.
Affective arguments were only effective in changing attitudes when they were presented in
the absence of cognitive arguments. When SS had no cognitive arguments presented to
them, they did attend to the affective arguments and were persuaded by the affective
arguments, to a lesser extent than by the cognitive arguments. Therefore, it is
hypothesized that cognitive arguments are primary in attitude change processes when SS

are presented with persuasive messages. Contrast coefﬁcients were assigned to test the

48

observed interaction, and the resulting model was named the Cognitive Argument Primacy
(CAP) model (contrast coefficients are presented in Appendix H). This test provided a
strong eﬁ‘ect for the contrast E(1, 541) = 343.13, p<.01, n2 = .38, r = .62, and the residual
explained was trivial, E(6, 541) = 1.46, ns, 712 = .01. This Cognitive Argument Primacy
Model explained 97.5% of the explained sum of squares.

Contrast analyses were also employed to test the factors of value/worth and
dependability, using the change scores calculated for the above analyses. Results were
consistent with those of the attitude factor. Contrast analysis indicated that the CAP
model was consistent with the data for the difference in value/worth, E(1, 561) = 114.39,
p<.01, n2 = .17, g = .41, with a trivial residual, E(6, 561) = .28, ns, 112 = .00. The CAP
model explained 98.5% of the explained sum of squares. The difference in dependability
(again, using the change scores) produced parallel results for the CAP model: E(1, 548) =
295.62, p_<.Ol, n2 = .35, 1: = .59; residual again was trivial, 13(6, 548) = 1.79, ns, 112 = .01.
The CAP model explained 96.5% of the explained sum of squares for the difference in
judgments of dependability of the Xenon light bulb after reading both messages.
Conﬁdence in Judgment.

The questionnaire completed by subjects also included items designed to assess
their conﬁdence in the judgments that they made. A four-item scale was used to assess
Ss’ conﬁdence in their judgments, or = .93. Edwards (1990) hypothesized that subjects
with affective attitudes would be more conﬁdent in their judgments. Of the two studies
she performed, one study produced results consistent with her hypothesis. Edwards’

conﬁdence hypothesis was tested using ANOVA. Analyses indicate that there was a main

49

effect for the attitude induction; however, it was in the opposite direction of that predicted
by Edwards (1990). Cognitively based attitudes were expressed with greater conﬁdence
M = 4.83, SQ = 1.24) than were affectively based attitudes M = 4.38, SQ = 1.30), 5(1,
580) = 18.27, p<.001, n2 = .03, g = .17 following the attitude induction. Thus, Edwards’
(1990) hypothesis regarding conﬁdence in judgment was inconsistent with these data. ’5
Countemgping Explanation

Millar and Millar (1990) indicated that their mismatch hypothesis was based upon
a cognitive response approach. In particular, they hypothesized that those subjects who
were exposed to counterattitudinal messages that matched their attitudinal base of the
attitude would be more likely to counterargue the message presented to them, and would
thus profess a lesser degree of attitude change. Millar and Millar (1990) assessed
counterarguments in their study by utilizing a thought-listing procedure, and tabulating the
positive and negative thoughts listed after the counterattitudinal message. A similar
process was employed here. As the attitudes in this study were induced, and not
preexisting, as they were in Millar and Millar’s study, the thoughts were coded to be not
only positive or negative (neutral was also an option), but were also coded message
relevant or message irrelevant. By using such a coding scheme, those thoughts that were
listed that were either outside the realm of the study (e.g., students writing down their
thoughts that they were Stressed for ﬁnals, that they were tired, etc.) or those that were
triggered by the Study but were not relevant to the messages provided (e.g., in the
aﬂ‘ective conditions, subjects were asked to list the feelings that they experienced while

reading the message, a number of students noted that they were hungry or tired) were

50

eliminated. Thus, the following analyses used those messages that were coded by
independent coders as being message relevant.

In order to assess the extent of counterarguing, those relevant thoughts listed that
were opposite in valence to the information presented in the message were tallied.
Analyses were performed following the ﬁrst message (attitude induction) in order to
determine a baseline of counterarguing, and following the second (persuasive) message.
Results indicated a lack of effect for the induction with regard to counterarguing. Those
SS receiving the affective favorable induction recorded an equal number of negative
thoughts M = .56, S_D = .88, p = 291) as did those receiving the cognitive favorable
induction (M. = .70, Sp =1.10,p = 294),:(1, 583)= 3.21, ns, n2 = .01, r =07.

In order to test Millar and Millar’s counterarguing hypothesis after the second
(persuasive) message, only those cells where there was a distinct match (affective
induction/affective persuasive argument, cognitive induction/cognitive persuasive
argument) or a distinct mismatch (affective induction/cognitive persuasive argument,
cognitive induction/affective persuasive argument) were assessed.“5

Those subjects with an affectively induced attitude produced more relevant
positive thoughts after receiving the negative affective counterattitudinal message (match
condition) (M = .26, S_Q = .58, p = 74) than those who were exposed to negative
cognitive counterattitudinal messages (mismatch condition) M = .10, S_Q = .35, p = 70), t
(142) = 1.92, p = .05. Those SS with cognitively induced attitudes 2_11_sg produced more
relevant positive thoughts after the negative affective counterattitudinal message

(mismatch condition) (M = .30, S_D = .64, p = 74) than those who were exposed to

51

negative cognitive counterattitudinal messages (match condition) M = .11, SQ = .46, p =
71), t(143) = 1.99, p<.05. Results indicate that those who were exposed to negative
aﬁecﬁve counterattitudinal arguments consistently produced more positive thoughts than
those exposed to negative cognitive counterattitudinal arguments, a ﬁnding that is
inconsistent with Millar and Millar’s (1990) hypothesis that subjects will consistently

counterargue more with argument types that match their attitudinal base.

52

Chapter 5
DISCUSSION

mm

This experiment served to test the conﬂicting hypotheses regarding the effect of
affective and cognitive arguments on affectively based and cognitively based attitudes.
Results indicated that neither the existing mismatch hypothesis of Millar and Millar (1990),
nor Edwards’ (1990) match hypothesis, were consistent with the data. Inspection of the
means led to the proposal of a Cognitive Argument Primacy Model, and contrasts were
assigned and tested and the proposed CAP model was found to be consistent with the data
when the differences in attitudes, perceptions of value and worth of the attitude object (the
Xenon bulb), and perceptions of the dependability of the bulb were assessed.
lhte Misrn‘atch Hypothest

Millar and Millar (1990), arguing ﬁom a cognitive response approach, postulated
that if a counterattitudinal argument of the same tyge (i.e., affective or cognitive) as one’s
attitudinal base is presented, this “counterattitudinal argument will challenge the adequacy
of the person’s appraisal of the object either by a direct contradiction of his or her beliefs
(feelings) or by pointing out aspects of the object that he or She has failed to consider”
(Millar & Millar, 1990, p. 217-218). Further, they suggest that this challenge will lead one
to counterargue the information that is presented. AS such, Millar and Millar (1990)
expect that if an argument threatens the way one has thought about an object, one will be
motivated to counterargue, and thus there would be little to no attitude change in such a

condition. Citing Petty, Ostrom, and Brock (1981), Millar and Millar (1990) suggest that,

53

if one is motivated to counterargue, the counterarguments generated may well overwhelm
the persuasive information presented.

The data in this study were inconsistent with Millar and Millar’s (1990) mismatch
hypothesis. In the absence of a mismatched argument type, Millar and Millar (1990)
hypothesize that there will be little or no attitude change. Nevertheless, data indicate that
when SS with aﬁ‘ectively induced attitudes were presented with aﬁ'ective counterattitudinal
persuasive messages, not only was there attitude change, contrary to Millar and Millar’s
prediction, but also the amount of this attitude change was equal to that observed when
the S s were presented with cognitive (mismatched) counterattitudinal arguments.

Similarly, when S s with cognitively induced attitudes were presented with
cognitive counterattitudinal persuasive messages, attitude change was observed again
Indeed, when cognitive counterattitudinal arguments were present, for SS who received a
cognitive induction, there was $2 difference between the amount of attitude change Ss
indicated when those cognitive counterattitudinal arguments were paired with an aﬁ'ective
counterattitudinal arguments and when the cognitive arguments were presented in the
absence of affective arguments. Further, when SS who held a cognitively induced attitude

were presented with an affective counterattitudinal argument only (in the absence of a

 

cognitive argument), there was Signiﬁcgntly less ch_apge_ than when those SS received
cognitive counterattitudinal arguments, a result that is directly contradictory to Millar and
Millar’s mismatch hypothesis. Finally, the amount of attitude change expressed by
subjects was e_qga_l in all of the following conditions: (1) aﬂ‘ective induction/affective and

cognitive persuasion (counterattitudinal arguments), (2) affective induction/cognitive

54

persuasion, (3) affective induction/affective persuasion, and (4) cognitive
induction/aﬂ‘ective persuasion (see Table 6 in Chapter 3 to review the cell means). When
SS received a cognitive induction and either (1) affective gpd cognitive persuasion, or (2)
cognitive persuasion only, their attitudes changed to an even greater extent than those four
options listed in the previous sentence. ‘7 Ergo, the mismatch hypothesis of Millar and
Millar (1990) is inconsistent with the data and, particularly because of the attitude change
in those cells for which Millar and Millar would have predicted no change, fails to provide
an explanation for these data.

Results were also inconsistent with Millar and Millar’s counterarguing hypothesis.
Although those subjects with an affective attitude did produce more relevant positive
thoughts following the negative affective counterattitudinal messages than did those SS
who were exposed to a negative cognitive counterattitudinal message, the same pattern
was found for those SS with a cognitive attitude (i.e., that, when presented with negative
affective counterattitudinal arguments, they listed a greater number of positive thoughts
than those SS presented with negative cognitive counterattitudinal arguments). Therefore,
the negative affective counterattitudinal arguments appear to elicit (or, perhaps, failed to
overwhelm) more relevant positive thoughts (coded as counterarguments) than did the
negative cognitive counterarguments. Analyses in this study were performed to parallel
those used by Millar and Millar (1990) in their assessment of counterarguments. Such a
categorization led to all statements that were message relevant and coded positive by the
subjects (following the persuasive message) being coded as counterarguments. For

example, after receiving the negative message, if a subject noted that he was “conﬁdent --

55

I have information to make a judgment” (ID# 133), this response was considered message
relevant (as it related to the message and the attitude object) and was positive. This
response is not necessarily a counterargument, but is counted as one per the coding
scheme. Categories that specify the type of thought listed more precisely would be able to
separate those comments that are message relevant and opposite in valence to the
information presented in the message ﬁom those comments that advocate a position
opposite to that advocated in the message.

A more developed coding scheme could include categories such as (1)
counterargument (i.e., a statement that clearly advocates a position in opposition to the
position presented in the message), (2) information seeking (e. g., wanting to know more
about the Xenon bulb), (3) message repetition/message agreement (i.e., statements that
either repeat information presented in the message or that indicate the Ss’ agreement with
the message, e. g., “bulb is long lasting,” “I’m scared”), (4) irrelevant/ambivalent (i.e.,
statements that are not directly related to the message, such as “I’m hungry,” or
statements indicating ambivalence, such as “I don’t care”), and (5) ambiguous/unknown
(e. g., “message was kind of corny”).

An important difference between this study and that which was undertaken by
Millar and Millar (1990) is that the development of the counterattitudinal argument
messages was substantially different between studies. Millar and Millar developed their
cognitive and affective counterattitudinal messages for Study 1 by selecting cognitive and
affective statements that were obtained during pretests, and incorporating these statements

into messages that suggested that a majority (or minority) liked (or disliked) a particular

56

beverage. No further pilot testing of these constructed messages is indicated, thus it is
impossible to determine how SS viewed the arguments. For Study 2, Millar and Millar
(1990) chose advertisements that had been published in the popular press, and had 17
students identify the advertisements as being ones that “primarily dealt with attributes of
the product” (rational) or “primarily dealt with feelings generated by the product”
(emotional) (p. 221). Again, no information is presented as to the Ss’ assessment of the
strength of the messages selected. In their third study, Millar and Millar used statements
that they identiﬁed as being either cognitive arguments or aﬁ‘ective arguments, but no
details as to how these statements were developed or categorized are provided.

In contrast, the counterattitudinal messages used in this Study were pilot tested to
ensure that SS perceived them to include either high affective/low cognitive content or
high cognitive/low affective content. Further, the strength of each message also was
evaluated in this study and results indicated that there was no difference in the strength of
the message across induction or argument type (i.e., affective and cognitive).

The Mgtch Hypothesis

Edwards (1990) presented a match hypothesis, speciﬁcally regarding affective
attitudes and the subsequent inﬂuence of affective counterattitudinal arguments on these
attitudes. She suggests that affective attitudes are likely to be resistant to persuasive
messages that tap a cognitive dimension, for such a message may not serve to alter the
underlying affective base of the attitude, much like a persuasive message focusing on a
ﬁlnction other that that which is the basis of one’s attitude (from a functional perspective)

may fail in an inﬂuence attempt (Katz, 1960; Shavitt, 1990; Smith, Bruner, & White,

57

1956). Although Edwards only tested and discussed her match hypothesis for aﬁ‘ectively
based attitudes, for this study, the match hypothesis was extended to include cognitively
based attitudes. Edwards (1990) also noted conditions where she would expect to see
little or no attitude change - those conditions where the attitude and argument type did not
match.

The data in this experiment were inconsistent with Edwards’ (1990) match
hypothesis. In the absence of a matched attitude and argument type, Edwards
hypothesizes that there will be little to no change in attitude. In this study, data indicate
when SS with affectively induced attitudes were exposed to cognitive counterattitudinal
arguments, not only was there attitude change, contrary to Edwards’ prediction, but this
attitude change was equal to the attitude change observed when S S were presented with
affective (matched) counterattitudinal arguments (see Table 6 in Chapter 3 to review the
means).

Similarly, when SS with cognitively induced attitudes were presented with aﬂ‘ective
counterattitudinal messages, they reported attitude change. More in line with the match
hypothesis (although Edwards did n_o_t extend her arguments to a match of cognitive
attitudes and arguments; rather, that extension was done in this Study), a greater degree of
change was observed when SS with cognitively induced attitudes were presented with
cognitive counterattitudinal messages than when they were presented with affective
counterattitudinal messages. Therefore, based upon the results of this experiment,

Edwards’ match hypothesis, although appearing to hold for the cognitively induced

58

attitudes condition, is inconsistent with the data, as results in the aﬁ’ectively induced
attitude condition are inconsistent with the match hypothesis.

Edwards (1990) also hypothesized that those S8 with an aﬁ‘ectively induced
attitude would hold that attitude with greater conﬁdence than those Ss with a cognitively
based attitude. Results were in direct contradiction to her hypothesis; those SS with
cognitively induced attitudes indicated greater conﬁdence in their judgments than did
those SS with affectively induced attitudes. Edwards (1990) argued for her conﬁdence
hypothesis on the basis that affectively based attitudes are often expressed with
“particularly great conviction” (Edwards, 1990, p. 204), drawing upon Zajonc’s (1980)
work that suggests that affectively based attitudes may, in some cases, be irrevocable. It is
possible that this same conviction was lacking in this study because the affective attitudes
that were induced were induced by means of a written message describing an attitude
object in emotional and affect-laden terms. Edwards’ affective attitudes were induced
either sublirninally (Study 1) or through direct experience (Study 2). Research indicates
that those attitudes formed through direct experience are held with more conﬁdence and
more conviction than those attitudes formed through indirect experience (F azio & Zanna,
1981; Wu & Shaffer, 1987). In Edwards’ Study 2, the only study where her conﬁdence
hypothesis was consistent with her data, her subjects had sampled a pleasant tasting drink
(the attitude object) and then smelled what was ostensibly the same drink, to ﬁnd that it
had a “mildly aversive smell” (Edwards, 1990, p. 209). Thus, Edwards’ SS had a greater

amount of affective direct experience with the attitude object in her Study than did the SS

59

in this study. The SS in this study relied on reading aﬁ‘ect-laden or cognitively laden
passages about the Xenon light bulb in order to form and change their opinion.

In both Edwards’ study and the one described here, attitudes toward fairly novel
objects were induced, albeit by diﬁ’erent methods. Edwards’ Study differed in another
way. Her subjects, in both the induction and persuasion processes, were exposed to Spur
aﬁ’ective and cognitive messages. In this study, only a subset of participants was exposed
to each parallel message, and this occurrence was constrained to the persuasion message.
Further study is needed to assess the impact of the sequence of cognitive and aﬁ‘ective
messages in inﬂuence attempts.18
The Cogm'tive Argpment Primag Model

Upon discovery that neither the match nor the mismatch hypotheses were
consistent with the data, the means were inspected to determine if there was an alternate
explanation for the results. This inspection of the means led to the construction of a
statistical model, named the Cognitive Argument Primacy Model. The contrasts assigned
took into consideration that, in the absence of cognitive arguments, there was a main
effect for affective arguments (and no effect of the attitude induction), and in the presence
of cognitive arguments, there was a main effect for the induction, with the cognitive
induction producing a greater amount of change on all three dependent variables than did
the affective induction, and the presence or absence of affective arguments not affecting
the amount of attitude change in either induction condition. The CAP model, when tested

using contrast analysis, was consistent with the data for each dependent variable,

60

explaining from 96.5% to 98.5% of the explained sum of squares, and the residual in each
case was trivial.

In this study, the measurement of attitude was more comprehensive than the
measurement of either Millar and Millar (1990) or Edwards (1990). All of their studies
used Single-item measures to assess attitude. In Millar and Millar’s studies this item was
the degree to which SS liked the attitude object (dislike -- like) and in Edwards’ studies the
item used was also a liking measure (did not like at all - liked very much). This study
measured attitudes toward the object using a ﬁve-item scale that assessed one’s
favorability toward and liking of the attitude object (01 = .92, .98). In addition, this study
also measured Ss’ perceptions of the value and worth of the Xenon bulb, as well as their
perceptions of the dependability of the bulb. The CAP model was consistent with the data
across all three variables.

Edwards (1990) noted that it is important to ascertain the underlying base of an
attitude if the goal is to alter that attitude. The results of this study are consistent with
that need, for there was a difference as to how SS responded to the attitude object
(measured by their amount of attitude change) whether they received an affective or a
cognitive induction. Although these data are inconsistent with Edwards’ match
hypothesis, there is agreement in the need to ascertain the attitudinal base of one’s attitude
in order to change that attitude more effectively. This study indicated that the W of
an argument, whether affective or cognitive, regardless of one’s attitudinal basis would

promote attitude change. Results also indicated that there were varying degrees of

attitude change.

61

In the absence of cognitive counterattitudinal arguments if one is presented with an

affective argument (when compared to a no argument control group) equal amounts of

 

attitude change toward a novel attitude object will be observed in both aﬂ’ective and
cognitive attitude induction conditions.

In the presence of cognitive counterarguments, attitude change will be observed in
both affective and cognitive attitude induction conditions. This attitude change, however,
will be greater if one holds a cognitive attitude (as opposed to an affective attitude).
Further, in the presence of cognitive counterattitudinal arguments, whether or not one is
also presented with affective counterattitudinal arguments appears to be inconsequential
when assessing attitude change.

The presence of cognitive counterattitudinal arguments appears to override or
even nullify any effect of affective counterattitudinal arguments when viewing diﬁ‘erences
in attitudes, perceptions of value and worth, and perceptions of dependability of a novel
attitude object. This presence of cognitive counterattitudinal arguments interacts with
attitude type (induction) such that greater persuasion exists when the induction type is
matched (cognitive induction/cognitive argument) than when it is mismatched (cognitive
induction/affective arguments). There are three potential explanations for these results:
(1) the cognitive induction serves to enhance the effects of the cognitive counterattitudinal
arguments, (2) the effect of the cognitive counterattitudinal arguments is attenuated by the

aﬁ‘ective induction, or (3) both processes (enhancement and attenuation) are occurring

simultaneously.

62

The next step needed is to include a control group that receives no attitudinal
induction. By including this control group, one could ascertain which of these three
potential explanations is consistent with the data. Ifthe cognitive induction serves to
enhance the effects of the cognitive counterattitudinal arguments, then the amount of
attitude change observed in the cognitive induction/cognitive persuasion condition should
be greater than that observed in a control group (no induction)/cognitive persuasion
condition. Altemately, if the effect of the cognitive counterattitudinal arguments is
attenuated by the affective induction, the amount of attitude change observed in the
aﬁ‘ective induction/cognitive persuasion condition should be less than that observed in a
control group/cognitive persuasion condition. Finally, if both processes of enhancement
and attenuation are occurring Simultaneously, then both of the above predictions should be
observed.

Processing Capacity gs a Function of Tasl_< Difficulty

One model that attempts to explain subjects’ processing of persuasive messages is
Kahneman’s (1973) Capacity Model of Attention [also referred to as the Elastic Capacity
Model (Stiff, 1986)]. This parallel processing model postulates that the capacity available
to process a given task (although having an upper limit) expands to ﬁt the difﬁculty of the
task (Kahneman, 197 3; Stiff, 1986). Further, this model suggests that people are able to
process multiple stimuli simultaneously, and that they will divide their overall processing
capacity between primary and secondary tasks. Those persons who are engaged in
processing a task of greater diﬁiculty will not only exert more effort, leading to a larger

available amount of processing capacity, but also will focus more of that capacity on the

63

primary task, leaving little capacity available to process a secondary task (Kahneman,
1973; Stiﬂ‘, 1986).19

Applying the premise of the Elastic Capacity Model (i.e., expansion of processing
capacity as a ﬁlnction of task difﬁculty) to the results of this study engenders a potential
explanation of these results. If the cognitive induction triggers subjects to believe that
they are encountering a task of greater diﬁculty than those S s in the affective induction
condition, the Elastic Capacity Model would propose that those subjects presented with a
cognitive induction would have a greater capacity available to them to process ﬁrture
messages about the attitude object. As the cover story for this study dealt with the
readability and the ease of comprehension of the messages, items are available to
determine whether or not SS perceived a diﬁ‘erence in the difficulty of the cognitive and
affective messages. As reported in the results, subjects in the cognitive induction
condition rated that induction message as being more diﬁcult to understand or
comprehend than did those subjects who were exposed to the affective induction message

In addition, a cognitive induction may trigger SS to attend primarily to ﬁlrther
cognitive information (i.e., consider the processing of cognitive information to be the
primary task, and the processing of affective information to be the secondary task).
Therefore, perhaps those SS with a cognitively induced attitude had a greater amount of
processing capacity available to them when they encountered the persuasive message
(counterattitudinal arguments). Those subjects who received cognitive counterattitudinal
arguments then may have used up their available processing capacity in evaluating the

cognitive arguments. If this scenario were true, then the presence of affective

counterattitudinal arguments in conjunction with cognitive counterattitudinal arguments
would not have altered the amount of attitude change (as was the case in this study)
because there would be no processing capacity left over to process those secondary
affective messages. Those subjects, then, with a cognitively induced attitude, when
exposed to aﬁ’ective arguments in the absence of cognitive arguments, will focus on the
aﬁ‘ective arguments, but have substantial processing capacity available to them and
subsequently determine that the affective counterattitudinal argument is not as persuasive
(as the cognitive argument).

When the attitudinal induction is affective, SS may be triggered to believe that the
task is less complex, and thus less processing capacity is allotted for future messages
about the attitude object. This decreased capacity may serve to place a ceiling eﬂ‘ect on
the processing of ﬁrture messages. Therefore, when S S with affectively induced attitudes
are presented with counterattitudinal arguments, they may process to the ﬁrll extent of the
capacity that has been allocated for processing. Perhaps this capacity is not sufﬁcient to
process all of the information they receive in the counterattitudinal message, and therefore
a ceiling is reached, resulting in equal amount of attitude change, regardless of the type
(affective or cognitive) of counterattitudinal arguments presented (as was observed in this
Study). The Elastic Capacity Model indicates that one reason for failure of a persuasive
message is that the demand for capacity exceeds the available processing capacity.
Perhaps those SS in the affective induction condition who were then exposed to cognitive

counterattitudinal arguments were unable to process these arguments completely, due to a

lack of available capacity.

65

Scrutiny pf Persuasive Messages

As cognition and aﬂ‘ect are considered to be separate dimensions (Becker, 1963 ),
and it has been argued that they may be independent of each other (Zajonc, 1980), perhaps
each induction is serving to prime Ss to be more attentive, or more motivated, to process
similar messages that follow. Petty and Wegener (1998) would suggest that this
difference can be attributed to variable levels of scrutiny. They assert that receipt of a
message that matches one’s attitudinal base serves to enhance message scrutiny (p. 227).
Speciﬁcally, they hypothesize that subjects will pay more attention to a message if it
matches their attitudinal base, and will consequently distinguish between strong and weak
matching messages. Therefore, a matched message is not necessarily more persuasive
than a mismatched message, as SS are more likely to scrutinize a message matching their
attitudinal base. Such scrutiny will uncover the Strength of the message; thus, if a matched
message is weak, it will be seen as such and will be less persuasive than a strong or weak
mismatched message that has received less scrutiny (Petty & Wegener, 1998). When a
message does not match one’s attitudinal base, then the effects of message strength are
not as pronounced.

Although Petty and Wegener (1998) assessed matches and mismatches of attitudes
and arguments using a ﬁlnctional approach (matching high self-monitors to image appeals,
and matching low self-monitors to quality appeals), in their discussion they extend their
theorizing to address the conﬂicting results of Millar and Millar (1990) and Edwards
(1990). Petty and Wegener’s (1998) ad hoc explanation of the conﬂicting results suggests

that the messages in Edwards’ study were Strong messages and were difﬁcult to

66

counterargue, whereas the messages used by Millar and Millar were weaker messages that
were easier to counterargue. Petty and Wegener conclude that message strength may
account for the conﬂicting ﬁndings, the strong messages of Edwards being persuasive in a
matched condition, and the weak messages of Millar and Millar being persuasive in a
mismatched condition (where they encountered less scrutiny).

This ad hoc explanation, although one that could explain the conﬂicting results of
prior research, fails to explain the results of this study. Post hoc tests of the perceived
strength of the cognitive and affective messages used in this study indicated that there was
no difference in the perceived message strength of the cognitive and affective inductions,
nor was there a difference in perceived message strength of the cognitive and aﬁ'ective
counterattitudinal persuasive (argument) messages. Thus, the scrutiny model, although it
provides an ad hoc explanation for the conﬂicting results of Millar and Millar (1990) and
Edwards (1990), fails to explain the data in this study.

Limitation

Attitudes in this study were formed by SS reading a message that was designed to
be either high in affective content and low in cognitive content or high in cognitive content
and low in affective content. This study cannot claim that these affective attitudes were
formed by the “primacy of affect in the formation of certain preferences” (Edwards, 1990,
p. 203; Zajonc, 1980). We cannot demonstrate that affect preceded cognition in this study
(Zajonc, 1980). The induction checks demonstrate that SS in the affective induction
conditions did emerge with an attitude that was primarily affective. AS has been argued,

affect and cognition are not at opposite ends of one continuum, and thus it is likely that

67

both are present to some extent during attitude formation. What this study demonstrates,
if not the primacy of affect, per Zajonc (1980) and Edwards (1990) in attitude formation
(induction) is the primary inﬂuence of affect (or cognition). When Ss were exposed to a
high affective content/low cognitive content message they emerged with an attitude that
was primarily based on aﬁ’ect, and when SS were exposed to a high cognitive content/low
aﬁ‘ective content message they emerged with an attitude that was primarily based on
cognition.
Practical Application

It is recognized that, when engaging in persuasive communication, a
communicator may be unaware of, or unable to ascertain, the underlying base of the
attitude (i.e., affective or cognitive) of the person s/he is attempting to persuade. Results
from this study indicate the primacy of cognitive arguments. Regardless of the type of
attitude held, cognitive arguments produced attitude change, and they produced a greater
amount of attitude change when presented to SS with cognitive attitudes than when
presented to SS with affective attitudes. Therefore, in the absence of knowledge of a
person’s underlying attitudinal type it is recommended that a communicator present
cognitive arguments, as opposed to affective arguments.
Future Directions

The CAP model presented in this study leads to ﬁlrther questions about the eﬁ‘ects
of affective and cognitive arguments upon attitudes that are either affectively or

cognitively based. Results from this study indicate the primacy of cognitive arguments,

68

but future study is needed in order to better understand the persuasive eﬁ‘ects of aﬁ‘ective
and cognitive messages.

Other questions that remain include whether or not this CAP model will be
consistent with the data when inﬂuence attempts are directed at preexisting attitudes, as
opposed to attitudes induced during the experiment. This study focused on a ﬁctitious
attitude object in order to eliminate effects arising ﬁ'om previous knowledge about or
experience with the attitude object. Research indicates differences in subjects’ reactions to
persuasive attempts when attitudes arise from direct experience or indirect experience
(Fazio & Zanna, 1981; Wu & Shaffer, 1987). Speciﬁcally, Wu and Shaffer (1987) found
that attitudes that were formed through direct experience were more resistant to change,
when exposed to counterattitudinal arguments, than were those attitudes formed through
indirect experience. Future research could seek to ascertain if the CAP model is

consistent with the data when attitudes are formed through either indirect or direct

experience.

Conclusion

The purpose of this investigation was to test conﬂicting hypotheses regarding the effect of
affective and cognitive persuasive messages on attitudes that are either affectively or
cognitively based. Speciﬁcally tested were Millar and Millar’s (1990) mismatch
hypothesis, which predicts that a persuasive message will be more effective if the
arguments employed in the message are counter (or mismatched) to one’s attitude type,
and Edwards’ (1990) match hypothesis, which predicts that a persuasive message will be

more effective if it is matched to one’s underlying attitudinal base. Contrast analyses

69

indicated that neither the mismatch nor the match hypotheses were consistent with the
data. Inspection of the means led to the construction of the CAP model, which was then
tested through contrast analyses and found to be consistent with the data. A three way
interaction was observed. In the absence of cognitive arguments, there is a main eﬁ‘ect for
the presence (versus absence) of an aﬂ‘ective counterattitudinal message, and in the
presence of cognitive arguments, there is a main effect for the type of attitudinal induction.
Theoretical explanations of the CAP model are discussed, including an explanation of the
results using premises of Kahneman’s (1973) Elastic Capacity Model. Future studies
should investigate whether or not the CAP model tested here is consistent with the data

when preexisting attitudes or those formed through direct or indirect experience are

employed as the initial attitude.

7O

ENDNOTES

1. Breckler (1984) also investigated the function of behavior in attitudes as well,
presenting the tripartite model of attitudes (p. 1191).

2. Millar and Millar identiﬁed affectively based attitudes as being emotional and
cognitively based attitudes as being rational. Their terms are used in this section to
describe their results.

3. Higher means indicate greater agreement with the counterattitudinal message.

4. Millar and Millar (1990) did not provide effect sizes.

5. Means were not reported for the cognitively based attitude condition.

6. Edwards (1990) did not provide effect Sizes.

7. The number of participants reported here is from both rounds of pilot tests, as all
subjects received the same questionnaire regarding the induction check items. Fifty
students were involved in the ﬁrst round of pilot testing; 23 students were involved in the
second round of pilot testing.

8. A random subsample of participants (3 = 13) in the ﬁrst round of pilot testing also
were asked to ﬁll out an open-ended questionnaire that asked for their reaction to the
study (the two pilot tests) as well as their reaction to the topic of study (the Xenon light
bulb). As noted earlier, none of the SS expressed any suspicion about the existence of a
Xenon light bulb. This open-ended questionnaire took students about 5 minutes to

complete. Those SS who participated were thanked, debriefed, and dismissed after they

had completed the Open-ended questionnaire.

71

9. A t-test of the means in the positive message condition was performed and
revealed that the difference between the means in the cognitive and affective conditions
was not signiﬁcant, t(60) = 1.32.

10. One group of SS who participated in the study were seated in a particularly warm
room with little ventilation. AS a result, many of these SS wrote statements such as “I’m
hot,” “This room is hot.” These statements also were placed in the neither category.

11. Three of the initial 15 items, were dropped from the ﬁnal three-factor model as
factor analysis indicated that the three items did not measure attitudes, value/worth, or
dependability.

12. Subjects responded to Likert-type items, where 1 indicated that the message was
“extremely easy to read/comprehend/understand” and 7 indicated that the message was
“extremely diﬁicult to read/comprehend/understand.”

13. Data were analyzed to determine if there were any order effects on the dependent
variables for those Ss who received both the affective and cognitive persuasion messages
(message 2). The only combination to approach statistical Signiﬁcance was when SS who
had a cognitively induced attitude were presented with the combined persuasive message:
Those SS who read the cognitive part of the message ﬁrst expressed a greater amount of
attitude change M = 3.48, S_D = 1.46, p = 37) than those who read the aﬂ‘ective part of
the message ﬁrst (M = 2.81, S_D = 1.45, p = 36), E0. 71) = 3.87, p<.06, r12 = .05, g = .23.
This effect did not alter the pattern of results described and was thus not a threat to the

ﬁndings of this study.

72

14. Millar and Millar did not include a value/worth factor in their analysis. It is
unknown as to whether or not Millar and Millar would have predicted the same results
when assessing value/worth of an object. The possibility that they would have
hypothesized the same effect is, however, not inconsistent with their argument. If
attitudinally inconsistent information is presented to you about an object, you are more
likely to counterargue that information that matches your evaluation of that object, that is,
information that is of the same type as your attitudinal base.

15. Recall that Edwards (1990) also induced either an affective or cognitive attitude in
her subjects, i.e., she did not measure conﬁdence in a preexisting attitude. As this study
also induced an attitude, such analyses regarding conﬁdence in judgment were warranted
and were parallel to those analyses performed by Edwards (1990).

16. The other four conditions were eliminated because two of them were control
conditions, whose Ss did not receive a counterattitudinal message, and the remaining two
conditions were those in which SS received SgtLl affective and cognitive counterattitudinal
arguments, so that a test of matching versus mismatching of attitude and argument type
was not possible.

17. This result approaches being described better by a match hypothesis for the
cognitive induction condition, although we have seen that the match hypothesis also failed
to explain the data sufﬁciently.

18. In this study, when SS in the affective induction condition were exposed to both

counterattitudinal messages, the order effect was neither substantial nor Signiﬁcant, E(1,

69) = 1.09, ns, n2 = .02, g = .12.

73

19. Involvement is also considered to be a variable in eﬁ‘ect when testing the Elastic
Capacity Model. As all S s in this experiment were exposed to messages about the same
object and undertook the experiment under the same conditions, involvement may be

considered equal across conditions.

74

APPENDICES

75

APPENDIX A

Items Selected for Use in Induction Check Analyses

 

 

Affective Cognitive
Dimension Dimension t-value
Item M (SQ) M (SQL (71)

Affective Ppsitive
Xenon light bulbs can be soothing. 3.90 (1.16) 1.90 (.95) 10.64
Xenon light bulbs are exciting. 4.08 (1.36) 1.60 (1.00) 11.46
Xenon light bulbs sound comforting. 3.47 (1.42) 1.89 (1.16) 8.05
Xenon light bulbs could be invigorating. 3.50 (1.40) 1.94 (1.03) 7.17
Cogm'tive Positive
Xenon light bulbs are recyclable. 2.07 (.56) 4.38 (.91) -11.31
Xenon light bulbs are economical. 2.04 (1 . 14) 4.03 (1.14) -9.10
Xenon light bulbs seldom break. 1.65 (.89) 4.07 (1.11) -13.84
Xenon light bulbs are long lasting. 1.66 (.93) 4.17 (1.09) -13.55
Affective Negative
Xenon light bulbs would make

me feel uneasy. 4.03 (1.36) 1.74 (1.05) 11.50
Xenon light bulbs scare me. 4.03 (1.34) 1.72 (1.12) 10.80
I’m aﬁ'aid to use Xenon light bulbs. 4.08 (1.26) 1.72 (1.02) 12.04
Xenon light bulbs make me feel anxious. 4.08 (1.27) 1.61 (.83) 13.25
C_ognitive Negative
Xenon light bulbs are hard to ﬁnd in

stores. 1.68 (.87) 3.74 (1.28) -10.39
You need Special equipment in order

to use a Xenon light bulb. 1.40 (.80) 4.04 (1.35) -13.52
Xenon light bulbs can only be purchased

in large quantities. 1.38 (.85) 4.10 (1.30) -14.57
Xenon light bulbs are extremely expensive. 2.10 (1.25) 3.80 (1.31) -7.20

*Note: In the pilot tests, these statements were presented to the SS without the word

“Xenon” included in the statement.

76

APPENDIX B

Messages

Affective Positive (Induction)

REYOVAN is marketing a new light bulb called the Xenon bulb. When installed in
lighting ﬁxtures, the Xenon bulb creates a warm and cozy atmosphere, providing a
delightﬁrl luminescence to the lighted area. The bulb emits a soft glow that is not hard on
the eyes, making it easy to Spend hours in its comforting glow. People who installed the
Xenon bulb in their home have commented on the charming radiance the bulb gives the

rooms. Furthermore, they note the welcoming warmth they feel upon entering a room lit

by the Xenon bulb.

anitive Positive (Induction)

REYOVAN is marketing a new light bulb called the Xenon bulb. The Xenon bulb lasts up
to 10 times as long as a regular incandescent light bulb, providing up to 10,000 hours of
illumination. Not only does the Xenon bulb last, but it has been Shown to be the most
energy efﬁcient bulb in its class, due to precise manufacturing standards at REYOVAN.
When the Xenon bulb expires, you can return it to the point of purchase, and it will be

returned to REYOVAN so that those reusable parts may be remanufactured for ﬁlture

bulbs.

77

Affective Negative (Argument/Persuasion)

A few people are worried about potential averse effects of the Xenon bulb. They feel
squeamish about installing the bulb in their homes because they are uncertain as to any
potential negative side effects that the bulb may cause over time. They are worried that
the bulb may have negative eﬁ‘ects, such as causing headaches or violent illness. These
people are scared that there may be harmful effects of the Xenon bulb on their small
household animals or their children. They feel apprehensive about using Xenon bulbs as

lighting in their homes.

Cognitive Negative (Argument/Penman

The Xenon bulb will only work in common household lights with a Special adapter, which
can be purchased only through REYOVAN for $169.99. When installing the Xenon bulb
in lights in your home, you Should be careﬁrl not to overload any one circuit, as there is a
possibility that the energy required of the Xenon bulb may dominate your voltage output,
which could lead to a blown ﬁlse. Care should be taken in installing the bulb and adapter

in your lighting sockets; be sure to follow the 20-page booklet of directions included with

each bulb.

78

Filler Message (absence affective/absence cognitive argpment condition)

HARLAN FABRIC, INC. is marketing a new chair fabric for release in early 1999. This
luxurious fabric is manufactured speciﬁcally to cover oﬁice chairs, restaurant chairs,
airline seats, easy chairs, and couches. HARLAN FABRIC, INC. intends to market this
new fabric to homeowners, professional restaurateurs, oﬁice supply stores, and the airline
industry. Industry executives have been told about the release of this new fabric. The
makers of this fabric intend to display fabric samples at upcoming trade shows across the

\

nation. A full description of the fabric will be released early this summer.

79

APPENDIX C

Instructions for Focused Reading of Messages (Millar & Tesser, 1986, 1989)

ngm'tive Focus

As you read this paragraph, analyze WHY you think the way you do about the
[attitude object]. That is, go over in your mind what it is about the information that you
are presented that makes you think why you would like the product or not. As you are

reading, think of reasons that make you like or dislike the product.

 

Affective Focus

As you read this paragraph, analyze HOW you feel while reading the paragraph.

That is, go over in your mind how you are feeling as you read through the paragraph.

80

APPENDIX D

Items, means, SD of the 15 attitude items following the Induction Message

 

Item Mea_n_(SD)_
1. After reading the message, what is your reaction

to the Xenon bulb?

(strongly unfavorable -- strongly favorable) 5.06 (1.45)
2. After reading the message, what is your reaction

to the Xenon bulb?

(strongly dislike Xenon bulb -- Strongly like Xenon bulb) 5.08 (1.39)
3. How much do you think that you would like

the Xenon bulb?

(not at all -- absolutely) 5.07 (1.27)
4. How do you feel about the Xenon bulb?

(strongly unfavorable -- strongly favorable) 4.98 (1.31)
5. How would you evaluate the Xenon light bulb?

(not at all favorably -- extremely favorably) 5.10 (1.23)
The Xenon light bulb is:
6. bad -- good 5.57 (1.26)
7. useful -- useless (R) 5.48 (1.52)
8. negative -- positive 5.54 (1.32)
9. beneﬁcial -— harmful (R) 5.50 (1.45)
10. worthless -- valuable 5.32 (1.37)
11. nice -- awﬁll (R) 5.55 (1.32)
12. pleasant -- unpleasant (R) 5.46 (1.40)
13. dependable -- undependable (R) 5-20 (1 ~40)
14. boring -- interesting 4-64 (158)
15. unreliable -- reliable 5.17 (130)

(R) Item was reﬂected for analysis. E = 570.
Attitude Factor: items 1, 2, 3, 4, 5.
Value/Worth Factor: items 6, 7, 8, 9, 10.
Dependability Factor: items 13, 15.

81

Items, means, SD of the 15 attitude items following the Persuasion Message

 

Item Mean (SD)
1. Aﬁer reading the message, what is your reaction

to the Xenon bulb?

(strongly unfavorable - strongly favorable) 3.04 (1.73)
2. After reading the message, what is your reaction

to the Xenon bulb?

(strongly dislike Xenon bulb -- strongly like Xenon bulb) 3.11 (1.69)
3. How much do you think that you would like

the Xenon bulb?

(not at all --- absolutely) 3.08 (1.76)
4. How do you feel about the Xenon bulb?

(strongly unfavorable -- strongly favorable) 3 .04 (1.70)
5. How would you evaluate the Xenon light bulb?

(not at all favorably -- extremely favorably) 3.11 (1.73)
The Xenon light bulb is:
6. bad - good 3.24 (1.84)
7. useful - useless (R) 4.06 (1.89)
8. negative - positive 3.25 (1.82)
9. beneﬁcial -- harmﬁll (R) 3.69 (1.86)
10. worthless -- valuable 3.90 (1.72)
11. nice - awful (R) 3.78 (1.72)
12. pleasant -- unpleasant (R) 3.71 (1.78)
13. dependable -- undependable (R) 3.99 (1.71)
14. boring -- interesting 4.16 (1.64)
15. unreliable -- reliable 3.92 (1.60)

(R) Item was reﬂected for analysis. E = 544.
Attitude Factor: items 1, 2, 3, 4, 5.
Value/Worth Factor: items 6, 7, 8, 9, 10.
Dependability Factor: items 13, 15.

82

APPENDIX E

Results of the Conﬁrmatory Factor Analysis for the Dependent Variables

The Observed Correlations and ﬁctor Loading Matrix for the Three-Factor Model
Following the Attitude Induction*

123456789101315 ATTV/WDEP

SS

63 SQ

60 65 ZS

69 68 79 S1

65 67 8O 79 ZS

56 56 64 62 69 S2

43 47 49 53 53 54 SS

56 6O 64 63 69 73 61 ZS

9 41 42 45 47 48 48 6O 56 42

10 51 54 58 59 64 64 57 71 59 SZ

13 42 45 51 49 56 54 58 59 58 63 SS
15 41 42 41 46 50 51 49 52 43 58 68 8

WQQM-hUJNr—o

ATT 75 77 86 90 89 74 59 75 53 69 58 53 100 85 67
V/W 64 67 72 73 78 77 74 86 70 82 75 65 85 100 85
DEP 50 53 56 58 64 64 65 67 61 73 82 82 67 85 100

*The numbers used to identify the items are ﬁom Appendix D. The underlined numbers in
the diagonals Show the reliabilities for each item (without decimals). The factor loadings
complete the matrix, bolded loadings indicate an item is part of a factor where ATT =
attitude, V/W = Value/Worth, and DEP = Dependability.

83

The Expected and Error Correlation Matrix for the Three-F actor Model Following the
Attitude Induction“

123456789101315

1 5 -5 l -2 7 -4 1 -4 -1 1 O
2 58 -l -1 -2 6 -1 4 -4 0 3 O
3 65 66 2 3 8 -5 1 -6 2 4 -6
4 686977 -1 3 -4-3 -7-40 -3
5 67 69 77 80 11-3 4 -5 2 7 l
6 49 50 56 59 58 -3 7 -6 1 O -3
7 474854575657 -3 8 -4 6 -3
8 55566366656664 -40 -l -8
9 45 46 51 54 53 54 52 6O 2 9 -6

10 52 54 60 63 62 63 61 71 57 6 l
13 41 42 47 49 49 54 52 6O 49 57 1
15 41 42 47 49 49 54 52 6O 49 57 67

*The numbers used to identify the items are from Appendix D. In this matrix, the lower
diagonal presents the expected/predicted correlations (without decimals) for items of the
four factors; the upper diagonal presents the residuals.

84

The Observed Correlations and Factor Loading Matrix for the Three-Factor Model
Following the Persuasive (Counterattitudinﬂ Argpment) Mesgge“

123456789101315 ATTV/WDEP

SS

89 S2

85 89 21

85 9O 94 2;

85 86 89 89 SS

81 85 84 83 83 SS

57 59 61 59 58 65 S1

80 84 84 84 84 82 89 65 S1

68 71 72 71 69 77 7O 77 Z3

10 58 61 64 62 60 69 71 7O 68 65

13 52 55 55 54 53 56 61 56 6O 61 7O
15 52 52 53 53 52 55 53 54 51 59 7O ZQ

\OOOQGUIAUJNH

ATT 91 94 95 96 93 89 63 88 75 65 57 56 100 89 68
V/W 81 85 86 85 83 89 78 90 86 81 69 64 89 100 80
DEP 62 64 65 64 63 66 68 66 66 72 84 84 68 80 100

*The numbers used to identify the items are ﬁ'om Appendix D. The underlined numbers in
the diagonals Show the reliabilities for each item (without decimals). The factor loadings
complete the matrix, bolded loadings indicate an item is part of a factor where ATT =
attitude, V/W = Value/Worth, and DEP = Dependability.

85

The Expected and Error Corﬂrtion Matrix for the Three-Factor Model Follom_ng' til;
Persuasive (Counterattitudinal Argpment) Message“

123456789101315

l 3-1-209-67-2800
2 86 0 0 -1 11-69 -l-71-2
3 8689 319-58-1-41-1
4 87 90 91 O 7 -8 7 -2-7-1-2
5 85 87 88 89 9 -7 8 -2 -7 O -l
6 7274757674 490-3—4-6
7 636566676569 ~5389l
8 73 75 76 77 748070 0 -3 -4-6
9 7O 72 73 73 7177 67 77 -2 2 -7
10 6668 68 69 67 72 63 7370 7 5
13 52545455 536052605854 1
15 5254 54 55 53 61 5260 585471

*The numbers used to identify the items are from Appendix D. In this matrix, the lower
diagonal presents the expected/predicted correlations (without decimals) for items of the
four factors; the upper diagonal presents the residuals.

APPENDIX F

Contrast coeﬁicients employed to test the predictions of the Mismatch hypothesis
(Millar & Millar, 1990)

 

Absence of Cognitive Persuasive

Presence of Cognitive Persuasive

 

 

 

 

Arguments (-) Arguments (‘)
Absence of Presence of Absence of Presence of
Aﬁ‘ective Affective Affective Aﬁ‘ective
Persuasive Persuasive Persuasive Persuasive
Arguments (-) Arguments (-) Arguments (-) Arguments (-)
Affective
Attitudinal -1 -1 +1 +1
Induction (+)
Cognitive
Attitudinal -1 +1 -1 +1
Induction (+)

 

 

 

 

 

Contrast coeﬁicients employed to test the predictions of the Mismatch hypothesis with
allowance for the main effect of attitude induction

 

Absence of Cognitive Persuasive

Presence of Cognitive Persuasive

 

 

 

 

Arguments (-) Arguments (')
Absence of Presence of Absence of Presence of
Affective Aﬁ‘ective Affective Aﬁ‘ective
Persuasive Persuasive Persuasive Persuasive
Arguments (-) Arguments (-) Arguments (-) Arguments (-)
Aﬁ‘ective
Attitudinal -1 -1 +1 +1
Induction (+)
Cognitive
Attitudinal -2 +2 -2 +2
Induction (+)

 

 

 

 

 

87

 

 

APPENDIX G

Contrast coemcients employed to test the predictions of the Match hypothesis

(Edwards, 1990)

 

Absence of Cognitive Persuasive

Presence of Cognitive Persuasive

 

 

 

 

Arguments (-) Arguments (')
Absence of Presence of Absence of Presence of
Affective Aﬂ‘ective Aﬁ‘ective Aﬁ‘ective
Persuasive Persuasive Persuasive Persuasive
Arguments (-) Arguments (-) Arguments (-) Arguments (-)
Aﬁ'ective
Attitudinal -1 +1 -1 +1
Induction (+)
Cognitive
Attitudinal -1 -1 +1 +1
Induction (+)

 

 

 

 

 

Contrast coefﬁcients employed to test the predictions of the Match hypothesis with
allowance for the main effect of attitude induction

 

Absence of Cognitive Persuasive

Presence of Cognitive Persuasive

 

 

 

 

Arguments (-) Arguments (')
Absence of Presence of Absence of Presence of
Affective Affective Affective Aﬂ‘ective
Persuasive Persuasive Persuasive Persuasive
Arguments Q) Arguments (-) Arguments (-) Arguments (-)
Affective
Attitudinal -1 +1 -1 +1
Induction (+)
Cognitive
Attitudinal -2 -2 +2 +2
Induction (+)

 

 

 

 

 

88

 

 

APPENDIX H

Contrast coeﬂicients employed to test the ﬁt of the Cognitive Argument Primacy Model

 

Absence of Cognitive Persuasive

Presence of Cognitive Persuasive

 

 

 

 

Arguments (-) Arguments (-)
Absence of Presence of Absence of Presence of
Aﬂ‘ective Aﬁ‘ective Aﬂ'ective Affective
Persuasive Persuasive Persuasive Persuasive
Arguments (-) Arguments (-) Argumentgq Arguments (-)
Aﬂ‘ective
Attitudinal -1 0 0 0
Induction (+)
Cognitive
Attitudinal -1 0 1 1
Induction (+)

 

 

 

 

 

89

 

LIST OF REFERENCES

LIST OF REFERENCES

Becker, S. (1963). Research on emotional and logical proofs. Southern Smh
JQLu'nal, 2_8, 198-207.

Breckler, SJ. (1984). Empirical validation of aﬁ‘ect, behavior, and cognition as

distinct components of attitude. Journal of Personalin and Social Psychology, 511, 1191-
1205.

Breckler, S.J., & Wiggins, EC. (1989). Aﬁ‘ect versus evaluation in the structure of
attitudes. Journal of Experimental Social chhology, M, 253-271.

Brehm, J.W. (1966). A theoly of psychological reactance. New York: Academic
Press.

Edwards, K. (1990). The interplay of affect and cognition in attitude formation and
change. Journal of Personalig and Social Paycholggy, 5_9_, 202-216.

Fazio, R.H., & Zanna, MP. (1981). Direct experience and attitude behavior

consistency. In L. Berkowitz (Ed), Advances in ex erimental social cholo (Vol. 14,
pp. 161 - 202). New York: Academic Press.

Fleming, D. (1967). Attitude: The history of a concept. Perspectives in Amprican
Histog, I, 287-365.

Herek, GM. (1987). Can functions be measured? A new perspective on the
ﬁlnctional approach to attitudes. Social Psychology anrterly, M, 285-303.

Hunter, I E & Gerbing, D.W. (1982). Unidirnensional measurement, second-
order factor analysis, and causal models. Research in Organizational Behavior, 5, 267-300.

Kahneman, D. (1973). Attention and effort. Englewood Cliﬁ‘s: NJ: Prentice Hall.

Katz, D. (1960). The ﬁlnctional approach to the study of attitudes. Public Opinion
Quarterly, 2_4 163-204.

McGuire, W.J. (1969). The nature of attitudes and attitude change. In G. Lindzey,

& E. Aronson (Eds). The handboogof social psychology (2nd. ed.). Reading, MA:
Addison-Wesley.

91

Millar, M.G., & Millar, K.U. (1990). Attitude change as a ﬁrnction of attitude type
and argument type. qurnal of Per§onality and Social Paychology, _5__9_, 217-228.

Millar, M.G., & Tesser, A. (1989). The eﬁ‘ects of aﬁ’ective-cognitive consistency

and thought on the attitude-behavior relation. J ottrnal of Egperimental Social nghplogy,
M, 189-202.

Millar, M.G., & Tesser, A. (1986). Effects of affective and cognitive focus on the
attitude-behavior relation. Journal of Personality and Social Paychology, 5_1, 270-276.

Miller, N., & Campbell, D.T. (1959). Recency and primacy in persuasion as a
function of ﬁmhg of speeches and measurements. Journal of Abnormal and Sog'ﬂ
Payphology, 5_9, 1-9.

Osgood, C.E., Suci, G.J., & Tannenbaum, PH. (1957). The mmment pf
Mg. Urbana: University of Illinois Press.

Petty, RE, Ostrom, T.M., & Brock, TC. (1981). Historical foundations of the
cognitive-response approach to attitudes and persuasion. In R.E. Petty, T.M. Ostrom, &
T.C. Brock, (Eds), CogLiitive responses in persuasion (pp. 5-30). Hillsdale, NJ: Erlbaum.

Petty, R.E., & Wegener, D.T. (1998). Matching versus mismatching attitude
ﬁlnctions: Implications for scrutiny of persuasive messages. Personalig and Social
Pachology Bulletin, 141, 227-240.

Shavitt, S. (1990).The role of attitude objects in attitude functions. Journal of
Experimental Social Psychology, M, 124-148.

Smith, M.B., Brunet, J .S., & White, R.W. (1956). Opinions and Personality. New
York: Wiley.

Stiff, J .B. (1986).Cognitive processing of persuasive message cues: A meta-
analytic review of the effects of supporting information on attitudes. Communication
Monogtaphs, 5_3_, 75-89.

Wu, C ., & Shaffer, DR. (1987). Susceptibility to persuasive appeals as a function

of source credibility and prior experience with the attitude object. Journal of Personality
and Social Psychology, M, 677-688.

Zajonc, RB. (1980). Feeling and thinking: Preferences need no inferences.
American Psychologist 35 151-175.

9—’

Zajonc, R.B., & Markus, H. (1982). Affective and cognitive factors in preferences.
Journal of Consumer Research, 2, 123-131.

92