“#288

MO

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the
dissertation entitled

A SPINOZAN MODEL OF PERSUASION

presented by

CHRISTOPHER J. CARPENTER

has been accepted towards fulfillment
of the requirements for the

PhD. degree in Communication

 

 

QMJ M

 

V 1 Major Professor’s Signature

 

Date
July 22, 2010

MSU is an Affirmative Action/Equal Opportunity Employer

 

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

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5108 KIProj/AccarPreleIRC/Dateouejndd

A SPINOZAN MODEL OF PERSUASION
By

Christopher J. Carpenter

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
Communication

2010

ABSTRACT
A SPINOZAN MODEL OF PERSUASION
By

Christopher J. Carpenter

The Spinozan model of persuasion (SMP) was advanced to explain the cognitive
processes involved in memory-based versus on-line processing. A 3 (processing
instructions: on-line, memory-based, or none) X 2 (weak v. strong arguments) X 2
(easy v. difficult to understand arguments) fully crossed, independent groups design
experiment (N = 360) was employed to test hypotheses derived from the model. The
effect of the inductions on message acceptance was consistent with the SMP given that
the Ss in the memory-based condition held attitudes more consistent with message
recommendations than those in the on—line condition. The data were not consistent with
the SMP prediction that the Ss’ attitudes in the memory-based condition would not be
affected by argument strength. The finding that the relationship between attitudes and
message recall was substantial in the memory-based condition but not on the on-line

condition was consistent with predictions.

Acknowledgments
I would like to thank my adviser Frank Boster for providing invaluable advice and
assistance in developing and executing this project. I also want to thank my committee
members Tim Levine, Joe Cesario, and Dan Bergan for their insights and
recommendations as well. Additionally, I am grateful to my parents for their years of
support. Finally, I would like to thank Erin Spottswood and Jill Staunton for their

encouragement and moral support.

iii

TABLE OF CONTENTS

LIST OF TABLES ................................................................................ v
CHAPTER 1
INTRODUCTION ................................................................................. 1
CHAPTER 2
METHOD ......................................................................................... 12
CHAPTER 3
RESULTS ......................................................................................... 19
CHAPTER 4
DISCUSSION .................................................................................... 38
APPENDICES .................................................................................... 56
REFERENCES ................................................................................... 78

iv

LIST OF TABLES

TABLE 1 Correlation matrix of dependent variables ........................................ 65
TABLE 2 Effects of the Inductions on Perceived Argument Strength ..................... 66
TABLE 3 Effects of the Inductions on Perceived Difficulty of Understanding

the message. ................................................................................................................. 67
TABLE 4 Effects of the Inductions on Reported Message Elaboration ................... 68
TABLE 5 Effects of the Inductions on Attitudes .............................................. 69
TABLE 6 Effects of the Inductions on Recall ................................................. 70
TABLE 7 Effects of the Inductions on Recognition .......................................... 71
TABLE 8 Effects of the Inductions on Attitude Certainty .................................. 72

TABLE 9 Effects of the Inductions on the Attitude Certainty- Attitude Correlation. . .73
TABLE 10 Effects of the Inductions on the Recognition-Attitude Correlation ......... 74

TABLE 11 Effects of the Inductions on the Perceived Difficulty- Attitude
Correlation ......................................................................................................... 75

A Spinozan Model of Persuasion

Gilbert (1991) proposed that understanding is believing. He asserted that
most people assume that when they consider an argument or piece of information,
they are able to suspend judgment of its veracity until they have evaluated it. He
called this perspective the Cartesian view because the philosopher Descartes
supported it. Gilbert suggested that this perspective is in error and instead claimed
that when people comprehend a piece of information, they accept it as true until
they have evaluated it. If they do not evaluate the information as they are exposed
to it, they are substantially more likely to recall any information that they remember
as true. He referred to this proposition as the Spinozan view because the
philosopher Spinoza supported this perspective.

Gilbert (1991) described a library metaphor to explain these contrasting
processes. In the Cartesian view, every new book that arrived in the library was set
aside until it could be tagged fiction or non-fiction. A book that remained untagged
was not considered either fiction or nonfiction until it was tagged. If the librarian
were asked what type of book the untagged book was, the librarian would be unable
to classify the book because it had no tag. In the Spinozan view, the library has
adopted an efficiency measure. Because most of the books that enter the library are
non-fiction, only fiction books receive any kind of tag and all books with no tag are
considered non-fiction. If a book enters the library and is set aside before it is
tagged, anyone picking it up will assume it is non-fiction because it remains
untagged. Gilbert proposed that the human mind operates like the Spinozan library

where information that does not get tagged is assumed to be true. The analog of

fiction versus nonfiction is in perceiving information to be false or true. Unless the
mind specifically concludes that a piece of information is false, that piece of
information will likely be treated as true.

Several studies have directly tested this view (Gilbert, Krull, & Malone,
1990; Gilbert, Tafarodi, & Malone, 1993). In these tests, Gilbert and colleagues
provided people with information that they either labeled as true or false either
while distracted or not distracted by being asked to perform an additional task. In
their first set of studies (Gilbert et al., 1990), when the subjects were distracted they
consistently made errors such that they often recalled information as true even
thought it had been explicitly labeled as false in previous exposures. They were
much less likely to make the opposite error such that they recalled true information
as false. In their second set of studies (Gilbert et al., 1993) they found that not only
did distracted people incorrectly recall false information as true, they applied that
information to jury decisions and to an evaluation of how much they liked a
hypothetical individual. As a group, these studies were consistent with Gilbert’s
(1991) argument that the Spinozan view more accurately represents human
information processing than the Cartesian view.

Other studies have demonstrated the consistency of the Spinozan effect.
Newhagen (1994) demonstrated that the veracity of verbal information seen in a
news broadcast can be similarly disrupted. Also, older people are more susceptible
to disruptions causing them to forget negations (Chen & Blanchard-Fields, 2000).
This effect has also been demonstrated in the processing of advertising claims

(Cowley, 2006; Koslow & Beltramini, 2002) including the disruption of warnings

about an advertising claim being false (Skurnik, Yoon, Park, & Schwarz, 2005).
The Spinozan hypothesis has been found to be a consistent explanation of
information processing outcomes.

Taken together, these studies are consistent with the Spinozan model
predictions. This research has pointed toward variables that are required to model
these processes. Insights from work on memory-based versus on-line processing
(Hastie & Park, 1986) will help develop a model of these processes.
Memory-Based Versus Orr-line Processing

Hastie and Park (1986), explained on-line processing when they wrote,

...information for the operator follows a path from the stimulus environment

external to the subject into working memory and directly to the judgment

operator. We call tasks of this type on-line judgment tasks because the
subject is forming the judgment “on-line” as evidence information is

encountered (p. 261).

Opposed to on—line processing is memory-based processing. In memory-based
processing, the information skips the judgment operator and goes straight into
memory. The result is that after on-line processing, a judgment about the
information has already been formed during message exposure and the audience can
quickly report their attitude toward the object they were receiving information
about. After memory-based processing, the audience has not formed a judgment and
they will have to search their memory for information concerning the object and

draw their conclusions from that information. Memory-based processing tends to

produce stronger correlations between subjects’ attitudes and the valence of the
information they recall.

The work of Mackie and Asuncion (1990) demonstrated the effects that on-
line versus memory-based processing have on persuasion. They tested several
memory-based processing inductions such as instructing the audience to find the
verbs in the message, proofreading the message, or memorizing the message. They
then compared the responses of the audiences in memory-based processing
conditions to an audience who were given on—line instructions that asked the
subjects to pay careful attention to the strength of the arguments in the message. In
the memory-based processing conditions, they found that there was a much stronger
correlation between the amount of positive information versus negative information
recalled from the persuasive message and the audience’s attitude toward the subject
of the message than in the on-line processing conditions. Similar patterns have been
found in advertising research (Beattie & Mitchell, 1985; Lichtenstein & Srull,
1985)

Hastie and Park (1986) argued that when people form a judgment using
memory-based processing, they evaluate whatever information they can recall and
make their judgment based on the recalled information. Their description of
memory-based processing assumes a Cartesian view of cognitive processing such
that information is stored but is neither accepted nor rejected. Spinozan principles
would predict that the reason that the recall-attitude correlation is high is because
the audience who processed using memory based processing accepted that

information as true. When asked for their attitude, they produced their attitude

based on their impression that what they heard was true. If people really evaluated
the. information in a Cartesian way while determining their attitude after memory-
based processing of the information they heard, their attitude would be based on
their cognitive responses to the information they recalled instead of the amount of it
they could remember. Cartesian processing would depress the recall-attitude
correlation to a size similar to what is found in on-line processing. The following
model integrates Gilbert’s (1991) Spinozan processing principles and Hastie and
Park’s distinction between memory-based and on-line processing into a model of
cognitive processing of persuasive messages.
Spinozan Model of Persuasion

Gilbert (1991) suggested that his model of Spinozan processing could be
used to understand the persuasion process. Other than suggesting this possibility, he
did not elaborate on how to model those processes. The Spinozan model of
persuasion (SMP) is a model of the effects of Spinozan processing on the
persuasiveness of messages. Under memory-based processing when people are
exposed to a given piece of information and if they are able to comprehend the
information they will accept that the information is persuasive. Their judgments and
attitudes will reflect this acceptance of the information to the extent that belief in
the validity of that information is related to their attitudes about an object. The
predictors of whether or not a piece of information will affect their attitudes is the
difficulty they have comprehending and remembering it and how much effort they

put into understanding it.

If the audience is forming or updating their attitude on-line based on a
message, the persuasiveness of the message will be based on an interaction between
the quality of the argument, the audience’s difficulty in comprehending it, and the
effort they put into that evaluation. Specifically, the effects of persuasive messages
during on—line processing are predicted to reflect the principles of cognitive
processing that Kruglanski and Thompson (1999) described in their unimodel. They
argued that when an audience is processing an argument evaluatively, they base
their attitude on the strength of the argument(s) that they are able to comprehend. If
they are putting forth a great deal of effort, they will be able to evaluate the strength
of information that is either difficult or easy to understand. This information could
include processing complex arguments or simple credibility cues. When the
audience does not devote much effort to comprehending the information, only
information that is easy to comprehend will affect their attitude such as simple
credibility cues. Kruglanski and Thompson presented evidence consistent with their
prediction that argument strength affects the persuasiveness of the message even
when the audience is processing shallowly if the arguments are very simple. They
also presented evidence consistent with their prediction that complex credibility
cues could affect the persuasiveness of a message when the audience was carefully
considering the message and not when they were shallowly processing it.

For example, someone might read a billboard that suggests that drinking
beer is fun. If they do not attempt to evaluate the information’s veracity and if they
comprehend the billboard’s message, they will be likely to believe that beer

drinking is fun. The next time they require an attitude about beer, they will recall

that beer drinking is fun and their attitude will move toward a more favorable
impression of beer consumption based on the extent to which they value fun things.
If the subject evaluates the billboard, they will decide whether to maintain their
belief that beer drinking is fun or reject it. The result of this on-line evaluation will
be based on the strength of the argument in the message.

Two of the SMP’s key variables that predict Spinozan processing are
processing type and difficulty of understanding the message. These will be
examined in more detail next.

Type of Processing

Memory-based processing requires that the audience understand the
message arguments but fail to evaluate the message and can happen due to a variety
of causes. First, the audience can have little to no expectation that they will need to
form an attitude about the subject of the message. If they feel that they will not need
an attitude about the described attitude object, they may not evaluate the
information in the message on-line. One reason they may not feel that they need an
attitude is because they lack involvement in the issue (e.g., Haugtvedt & Petty,
1992). Memory-based processing can also be due to a task induction in which they
are given a different judgment to make (e. g., to judge the dynamism of the speaker,
Mackie & Asuncion, 1990) or because they are distracted and therefore unable to
devote the cognitive resources necessary to do more than simply learn the messages
(e.g., Gilbert et a], 1993). In order for an audience to engage in memory based
processing they must either be unmotivated to engage in online processing that

evaluates the arguments, be unable to do so, or both.

When an audience is processing a persuasive message on-line, they are
forming an evaluation of that message as they read or listen to the message. Hastie
and Park (1986) argued that for most situations on-line processing is the default.
Given that evaluations can be made with very little effort (Zajonc, 1980) it is not
surprising that memory-based processing is uncommon. Hastie and Park noted that
most attitude change studies almost always induce on-line processing because when
the subjects know that they are in a persuasion study they know they are going to
have to report an attitude. Though traditional low involvement inductions may not
motivate an audience to care about the issue enough to carefully evaluate
arguments, if there is some simple cue to use to form an evaluation, the uninvolved
audience will still process that simple cue online and form an attitude during
message exposure (Petty, Priester, & Wegener, 1994).

Some have argued that on-line processing may be similar to the central
processing postulated by the elaboration likelihood model (ELM; Petty et al., 1994).
Under their conceptualization, on-line processing that forms an attitude about the
message requires substantial cognitive effort whereas memory-based processing
requires minimal effort. This suggestion is contradicted by the countless ELM
studies demonstrating that people who are not processing centrally form a judgment
using simple cues like simple descriptions of source credibility (Kruglanski &
Thompson, 1999) or by the sheer number of arguments encountered (Petty &
Cacioppo, 1984). These audiences were able to make evaluations on-line without
putting forth much cognitive effort. The possibility of making evaluations on-line

using minimal cognitive effort is also consistent with Zajonc’s (1980) research

findings demonstrating that “preferences need no inferences” (p. 151). Furthermore,
memory-based processing could occur when the audience is devoting substantial
resources to understanding the message as long as they are not evaluating it. Some
information requires a great deal of cognitive effort just to be understood (e. g. dense
academic writing). What is important to memory-based versus on-line processing is
not how much cognitive effort the audience is spending on the message but whether
or not the audience is forming an attitude based on the message during message
exposure.
Difliculty of Understanding

The second key variable in the SMP is how difficult it is to understand a
message. There are three sources of difficulty of understanding: message related,
external, and internal. Message-related sources of difficulty of understanding are
variations in the message itself and include how complex the language in the
message is and how long the message is (e. g., Hafer, Reynolds, & Obertynski,
1996; Lowrey, 2008; Regan & Cheng, 1973). External sources cause the message to
be harder to understand but are not part of the message itself and include
distractions like showing slides during an audio recording of a message (Rosenblatt,
1966), nonverbal proximity violations (Stacks & Burgoon, 1981), loud beeping
sounds of varying frequencies (Romer, 1979), etc. Some research suggests that
people vary in how easily they are distracted by these external sources (Forster &
Lavie, 2007). Internal sources are individual differences in how difficult it is for
particular individuals to process a message. These can include general variations in

ability like intelligence or specific variations such as the amount of prior knowledge

about the topic the audience has. Any of these can contribute to how difficult it is
for an individual to process a given message.
The Current Study

In order to test the predictions of the SMP a 3 (processing instructions:
memory-based, on-line, or none) X 2 (comprehension difficulty: low v. high) X 2
(argument strength; weak v. strong) fully crossed between subjects experiment was
designed. Cognitive effort was held constant and high by inducing high response
involvement in all S3. The SMP predicted that the message acceptance scores of the
group in the memory-based condition will only be affected by the difficulty of
understanding the message induction such that those exposed to easier to
understand arguments will be more persuaded than those exposed to difficult ones.
In the on-line processing condition, argument strength and the difficulty of
understanding the message were predicted to combine non-additively such that the
strong arguments will be more persuasive than weak arguments but that this effect
will be larger in the simple message conditions. Given that Hastie & Park’s (1986)
research suggests that on-line processing is the default type of processing, the group
given no processing instructions is expected to produce results similar to the group
given on-line instructions.

In addition to the primary SMP predictions, previous findings concerning
memory-based versus on-line processing were predicted to be replicated. Hastie and
Park (1986) noted that the hallmark of memory-based processing is a strong,
positive relationship between the Ss recall of positive information from the message

and their attitudes whereas there is a weak relationship during on-line processing. In

10

this study it is expected that there will be a strong relationship between the recall
scores and the message acceptance scores of the Ss in the memory-based condition
but not the on-line condition. Furthermore, the Ss in the memory-based condition
will score higher on a recognition of facts from the message scale than the Ss in the
on-line condition due to their focusing on learning the information. The Ss in the
on-line processing condition will recall more of the message arguments than the Ss
in the memory-based condition due to their focus on evaluating the arguments.
Bizer et al., (2006) found that $5 who formed on-line attitudes reported that
they were more certain about their attitudes than those who formed memory-based
attitudes. This effect was predicted to be replicated such that the attitude certainty
scores will be higher in the on-line conditions than in the memory-based conditions.
Furthermore, attitude certainty scores are predicted to be substantially and

positively correlated with need to evaluate scores.

11

Method

Sample

Ss were 360 undergraduate students at a large Midwestern university. Two
were dropped because they admitted on the recall measure that they had not read
the messagethey were assigned. There were 174 male and 183 female Ss. They
were recruited from undergraduate communication courses and they were given
course credit in exchange for their participation. Their mean age was 20.29, SD =
2.15 and ranged from 18-47.
Procedures

53 logged onto an online data collection website that randomly assigned
them to one of the twelve conditions. They first viewed an informed consent form.
They were then presented with the cover story and their processing induction
instructions. Next, they were presented with the persuasive message. The message
was followed by the attitude scale, an argument recall measure, a recognition
measure, an attitude certainty measure, a measure of perceived argument strength
(La France & Boster, 2001), a measure of perceived difficulty of reading the
message, a measure of message elaboration (Reynolds, 1997), and the need to
evaluate scale (Jarvis & Petty, 1996).
Design

This experiment employed a 3 (processing instructions: memorization,
argument evaluation, or none) X 2 (comprehension difficulty: low v. high) X 2
(argument strength: weak v. strong) fully crossed independent groups design. Ss

were assigned randomly to conditions with 30 Ss per cell. Need to evaluate (Jarvis

l2

& Petty, 1996) was also measured as an alternative indicator of their processing
type as previous research has found that need to evaluate scores and processing
inductions combine additively to affect processing outcomes (Tormala & Petty,
2001).

Materials

In order to keep involvement constant and high in all conditions, all 83 were
told that the Communication Department was assisting the Education Department in
developing tests of college students’ aptitude. Their instructions stated that experts
at a later time would view their responses, and that their names would be associated
with their responses in case the experts needed to ask follow up questions.

The processing type inductions consisted of three different sets of
instructions. The first type was the memorization instructions, which were designed
to induce memory-based processing. The Ss in this condition were told that they
should read the message that followed very carefully and try to learn the message as
well as they could because they were going to be quizzed on their ability to recall
the information in the article. The argument evaluation instructions were designed
to induce on—line processing. The Ss in this condition were told that they needed to
carefully evaluate the quality of the arguments in the message. They were told that
they would be asked to write an essay discussing their reasons for their overall
impression of the proposal in the message. The no processing instructions did not
seek to induce any particular processing type and merely told the Ss that they would

be asked some questions about the message.

13

The message concerned an imaginary political candidate for the President of
Ukraine though some of the arguments were based on real proposals from the
candidates in the 2010 Ukrainian election (see Appendix A for messages). This
topic was chosen in order to prevent Ss in the memory-based conditions from
simply recalling a previously held attitude rather than constructing one based on
what they could recall from the message. Furthermore, the topic minimized
variation in previous knowledge about the topic that might affect the Ss difficulty of
comprehending the message.

The two levels of comprehension difficulty were induced by starting with a
moderately complex version of the arguments. Then, using recommendations from
Lowrey (2008), more complex language and structure were added to create the
difficult to comprehend messages. The language and structure were simplified to
create easy to comprehend messages. The two levels were carefully balanced by
examining changes to the Flesch Reading Ease score (Flesch, 1948) such that
higher scores indicated passages that are easier to read (Appendix A contains the
Flesch Reading Ease scores for each message).

Argtunent strength was induced by giving the strong arguments sound logic
with strong support and giving the weak arguments logical errors with weak
support. The strong arguments contained well-established evidence from qualified
sources whereas the weak arguments contained weakly supportive evidence from
dubious sources.

A pre-test was conducted to determine if comprehension difficulty and

argument strength were successfully induced. A 2 (low argument strength v. high

14

argument strength) X 2 (difficult to understand v. easy to understand) independent
groups design was employed. The data was collected using an online data collection
website. All SS were instructed that they were going to evaluate both the strength of
the argument and how difficult it was to read. The 83 were undergraduate students
taking Communication classes (mean age = 20.13, SD = 1.98, 76 female, 24 male).
Argument strength and perceptions of difficulty were measured using the same
items used in the main study.

There was a substantial and statistically significant main effect for the
argument strength induction on perceptions of argument strength, F (1 , 94) = 16.30,
p < .001 , r = .39, r’ = .40 such that strong arguments (M= 4.25, SD = .86) were
perceived as stronger than weak arguments (M = 3.50, SD = .97). The difficulty
induction and the interaction were both insubstantial and not statistically significant
predictors of perceptions of argument strength. There was a substantial and
statistically significant main effect for the difficulty induction on perceptions of
difficulty, F (1, 96) = 47.74, p < .001, r = .57, r ’= .60 such that the arguments that
were difficult to understand (M = 4.85, SD = 1.01) were rated as more difficult to
understand than the arguments that were easy to understand (M = 3.39, SD = 1.12).
Neither the argument strength induction nor the interaction produced a substantial
or statistically significant effect on perceptions of difficulty. Based on this pretest,
the message inductions were used in the main study.

Instrumentation
A correlation matrix of all measures can be found in Table 1. The first set of

questions measured attitudes toward the proposed candidate by asking the Ss to rate

15

the proposed policy on five, 7-point semantic differential items anchored at wise/
foolish, beneficial/ harmful, desirable/ undesirable, good/ bad, and advantageous/
disadvantageous. Hunter and Gerbing’s (1982) centroid method of confirmatory
factor analysis found that the scale was internally consistent based on the low
RMSE = .02. The 53 scores on these items were averaged for each S. The mean and
standard deviation on this scale was M = 4.59, SD = 1.45 with adequate reliability or
= .95. The distribution was slightly negatively skewed.

The 53 were then asked to recall as many arguments as they could from the
message. Their responses were coded for the number of accurately recalled message
elements by two trained coders. The two coders’ inter-coder reliability was
estimated using Ebel’s (1951) coefficient, EC = .90 and was acceptable. Their
scores were averaged together to form an estimate of how many arguments the Ss
correctly recalled. The number of correctly recalled message elements ranged from
zero to 7.5 with M = 2.86, SD = 1.65 and approximated the normal distribution.

The Ss were then given seven multiple-choice questions measuring their
ability to recognize the factual information contained in the messages (see
Appendix B for items). Confirrnatory factor analysis was consistent with the
hypothesis of a unidimensional measure (RMSE= .05). The scale did, however, have
disappointingly low reliability, or = .53. There was no combination of items
uncovered that increased the reliability so the scale was maintained. Given the low
reliability, the measure should be interpreted cautiously. The number of items they

correctly responded to was summed for each S to create their recognition score,

16

which ranged from 0 to 7 and was skewed slightly negatively (M = 4.73, SD =
1.63).

Attitude certainty was measured using five Likert items with 7-point
response scales. Two of the items were adapted from Bizer et al.’s (2006) attitude
certainty scale and three new ones were written to increase scale reliability (see
Appendix C for items). Confirmatory factor analysis found that the scale was
internally consistent given that an acceptably low RMSE = .06 was obtained. The Ss
scores on these items were averaged. The scale approximated the normal
distribution and the mean and standard deviation of this scale was M = 4.18, SD =
1.23 with adequate reliability or = .88.

Perceived argument strength was measured next by asking Ss to respond to
seven semantic differential items with 7 points (La France & Boster, 2001 ).
Confirmatory factor analysis was consistent with the hypothesis of internal
consistency, RMSE= .03. The scale was slightly negatively skewed, M = 4.55, SD =
1.32. The scale was reliable, or = .94.

The next scale measured perceived difficulty of comprehending the
arguments using five semantic differential items with 7 points. The items were
anchored at easy/ hard, simple/ difficult, effortless/ demanding, clear/ confusing,
and understandable/ incomprehensible. The acceptable RMSE = .06 produced by
confirmatory factor analysis is consistent with the hypothesis that the measurement
model fit the data. The distribution was somewhat leptokurtic with M = 3.61, SD =

1.42. The scale was adequately reliable, or = .92.

17

The Ss next responded to Reynolds’s (1997) elaboration measure. The scale
was composed of twelve items with 7-point response scales. Confirmatory factor
analysis of the scale found somewhat weak support for Reynolds’s one-factor
solution, RMSE = .13. Examination of the error matrix did not show that any
particular item was contributing to a large portion of the error and no combination
of the items reduced the error. Given that Reynolds found a one-factor solution and
that the scale was reliable (or = .87), the scale was maintained. The scale was
distributed normally, M = 4.20, SD = .93.

Finally, the Ss responded to the 16-item need to evaluate scale using 5-point
likert response scales (Jarvis & Petty, 1996). Confirmatory factor analysis of the
entire scale was not consistent with good fit as demonstrated by the unacceptably
high average error, RMSE = .13. Examination of the residual error matrix revealed
that the reverse coded items were contributing much of the error. Removal of the
six reverse coded items left a ten item scale that demonstrated adequate fit, RMSE =
.06. The Ss scores on the remaining ten items were averaged for each S. The scale
had a mean of 3.1 (SD = .74) and approximated the normal distribution. Reliability

was adequate, a = .87.

18

Results

The report of the data analysis will proceed by first examining the induction
checks to determine if the three inductions had substantial effects on their target
variables and relationships among variables. Then the effects of the inductions on
the dependent variables will be examined in order to test the primary hypotheses
advanced by the SMP. This analysis will be followed by examination of the effects
of the inductions on the continuous variables’ relationship with attitudes. Then the
data concerning the hypothesis that NTE would act as an independent predictor of
on—line processing will be examined.
Induction Checks

Argument strength. First, the argument strength induction was predicted to
affect perceived argument strength such that strong arguments would be perceived
as stronger than weak arguments. A 3 (processing instructions: on-line, memory, no
instructions) X 2 (strong arguments v. weak arguments) X 2 (easy to understand v.
difficult to understand) ANOVA was conducted to test this hypothesis (see Table 2
for means and standard deviations). The data were consistent with this hypothesis as
strong arguments (M = 5.19, SD = 1.02) were perceived as stronger than weak
arguments (M = 3.95, SD 1= 1.29). This difference was substantial and statistically
significant, F(1,331) = 100.27,p < .001, r = .47, r ’= .48. The mean in the weak
argument condition was at the midpoint of the scale, which may indicate that the
induction was limited to causing strong arguments to be perceived as particularly

strong and weak arguments to be perceived as moderately strong.

19

There was also a small but statistically significant effect for processing type

on perceived argument strength, F (2, 331) = 3.33, p = .04, n2 = .02. The effect was

such that the highest perceived argument strength was in the memory-based
condition (M = 4.76, SD = 1.20), the second highest was in the no instructions
condition (M = 4.51, SD = 1.28), and lowest in the on—line condition (M = 4.39, SD
= 1.46). Pairwise tests indicated that the difference between the memory-based
condition and the on-line condition was statistically significant but modest in size,
t(229) = 2.11, p = .04, r = .14, r’ = .14. The differences between the no-instructions
condition and the other two processing conditions were statistically significant. This
effect is consistent with the SMP because Ss in the on-line condition are expected to
be critical and thus less persuaded whereas the Ss in the memory-based condition
are expected to recall any arguments as persuasive. Given the mean in the no
instructions condition was within sampling error of the other two conditions, no
conclusions can be drawn from this analysis about which type of processing they
were using to form their attitudes.

Though there was no substantial main effect for the difficulty of
comprehension induction on perceived argument strength, there was a small but
statistically significant interaction between argument strength and the difficulty of
comprehension, F(1, 331) = 5.24, p = .02, r = .10, r’ = .10. The interaction was such
that when the arguments were strong, there was a small difference between the
means of the easy to understand condition (M = 5.28, SD = 1.08) and the difficult to
understand condition (M = 5.11, SD = .97). There was a larger difference between

the means when the arguments were weak such that in the easy to understand

20

condition the mean (M = 3.75, SD = 1.43) was lower than in the difficult to
understand condition (M = 4.15, SD = 1.10). Given that this interaction was not a
crossover interaction, the effect size derived from the 2-way ANOVA is
underestimated. A contrast analysis was applied such that the weak, easy to
understand condition was coded -2, the weak difficult to understand condition was
coded -1, the strong easy to understand condition was coded 2, and the strong,
difficult to understand condition was coded 1. The result was a more accurate
estimate of the interaction effect, F (1, 331) = 104.12, p < .01 , r = .48, r‘ = .50.
There were no other substantial or statistically significant interactions.

Difficulty of comprehension. The difficulty of comprehension induction was
predicted to affect perceived difficulty of comprehending the message such that the
difficult condition was predicted to be perceived as harder to comprehend than the
easier condition. The full factorial ANOVA was run with difficulty of
comprehension as the dependent variable to test this hypothesis (see Table 3 for
means and standard deviations). The data were consistent with this hypothesis as
the message in the difficult to understand condition was perceived to be harder to
understand (M = 4.14, SD = 1.35) than the message in the easy to understand
condition (M = 3.07, SD = 1.27). This difference was substantial and statistically
significant, F(1, 338) = 59.78,p < .001, r = .38, r’ = .40.

There was also a small but statistically significant interaction between

processing type and argument strength on perceived difficulty F (2, 338) = 3.87, p =

2 . . .
.02, n = .02. The rnteractron was such that when the Ss were instructed to process

on-line, the weak arguments (M = 3.39, SD = 1.65) were perceived as easier to

21

understand than the strong arguments, r = .13, r’ = .14 (M= 3.80, SD = 1.51). When
the Ss were instructed to process in a memory-based fashion, the finding was
reversed such that the strong arguments (M = 3.21 , SD = 1.37) were perceived as
easier to understand than the weak arguments, r = -.21, r’ = -.22 (M = 3.74, SD =
1.09). The difference between the effect sizes of argument strength on difficulty
perceptions in these two processing conditions was larger than could be attributed
to sampling error, 2 = 2.52, p = .01. When the 55 were given no specific processing
instructions, the means were nearly identical (M = 3.75, SD = 1.26 for the weak
arguments and M = 3.74, SD = 1.51 for the strong arguments).

Processing Type. There were several predicted indicators of a successful
processing induction. First, attitudes were predicted to be substantially correlated
with the number of arguments the Ss recalled in only the memory-based condition.
Among the Ss in the memory-based condition, there was a substantial and
statistically significant relationship between the number of arguments they recalled
and their attitudes, r = .32, p = .001, r’ = .33. The relationship between attitudes and
recall were insubstantial in the on-line condition (r = .06, r’ = .07) and the no
instructions condition (r = .10, r’ = .11). The difference between the size of the
correlation in the memory-based condition and the no instructions condition was
marginally statistically significant, 2 = 1.80, p = .07.

Also, attitudes were predicted to be correlated with the number of arguments
the Ss recognized on the recognition test, but in only the memory-based condition.
Among those in the memory-based condition, there was a substantial relationship

between recognition and attitudes, r = .33, p < .001, r’ = .47. The relationship

22

between recognition and attitudes was insubstantial in the on-line condition (r = .02,
r’ = .03). In the no-instructions condition there was an unpredicted substantial and
statistically significant relationship between the number of arguments correctly
identified and the Ss attitudes, r = .31, p = .001, r’ = .44. The difference in the size
of the correlation between the no instructions condition and the on-line condition
was statistically significant, 2 = 2.30, p = .02.

Finally, it was predicted that the Ss’ perception of the difficulty of
understanding the message should be negatively related to their attitudes only in the
memory-based condition. There was a substantial and statistically significant
negative relationship between attitudes and the difficulty the SS reported having
comprehending the arguments r = -.32, p < .001, r’ = -.34. The relationship between
perceived difficulty of comprehending the message and attitudes was not substantial
for the Ss in either the on-line condition (r = .08, r’ = .09) or the no-instructions
condition (r = -.10, r’ = -.11). The difference between the size of the relationship
between difficulty and attitudes in the memory-based condition and the no
instructions condition was marginally statistically significant, 2 = 1.78, p = .08.

The evidence is consistent with a successful induction of memory-based
processing as the number of arguments recalled, the number of arguments
successfully identified, and the difficulty of comprehending the arguments were all
substantially related to the Ss attitude scores in the memory-based processing
condition. Also as predicted, none of these measures were substantially related to
attitudes when the subjects were given on-line processing instructions. The type of

processing that occurred in the no-instructions condition remains unclear as recall

23

and difficulty of understanding the message were not substantially related to
attitudes but recognition scores were associated with attitudes.

Processing Depth. Finally, the Ss in all conditions were asked to process the
message carefully and were told that they might be contacted by experts about their
responses. It was predicted that all of the 85 would be engaging in a great deal of
cognitive elaboration. The full factorial ANOVA of all the inductions was run with
reported elaboration as the dependent variable (see Table 4 for means and standard
deviations). There were no substantial or statistically significant effects of the
inductions or their interactions based on the 3-way AN OVA. Examination of the
means in Table 4 did suggest that the mean in the on-line group (M = 4.30, SD =
1.00) was higher than in the no instructions group (M = 4.04, SD = .90). This
difference, although modest in size, was statistically significant, t(218) = 1.96, p =
.05, r = .13, r’ = .14 The mean for the memory-based condition (M = 4.23, SD =
.86) was not outside of sampling error of either of the other two processing
conditions. Given that the difference between the means of the cognitive
elaboration measure in the memory-based condition and the on-line condition was
insubstantial and within sampling error of zero, any effect the processing
instructions have on the dependent variables cannot be attributed to differences in
cognitive elaboration.

The difference between the mean for the entire sample (M = 4.20, SD = .93)
and the midpoint of the scale (4) was statistically significant, t(328) = 3.82, p < .001

although it is only about a fifth of a standard deviation above the midpoint. This

24

finding is consistent with the prediction that the Ss would be consistently processing
deeply regardless of processing instructions.
Eflects of the Inductions on the Dependent Variables

Attitudes. It was predicted that in the on-line condition, strong arguments
would be more persuasive than weak arguments, but that this difference would be
larger when the arguments were easier to understand than when they were difficult
to understand. In the memory-based condition it was predicted that argument
strength would not affect attitudes but that the arguments that were easier to
understand would be more persuasive than the arguments that were harder to
understand. The results in the no-instructions condition were expected to be similar

to the results in the on—line condition.

A 3 (processing instructions: on—line, memory-based, or none) X 2 (weak v.
strong arguments) X 2 (easy v. difficult to understand arguments) factorial ANOVA
was calculated with attitude scores as the dependent measure (see Table 5 for means

and standard deviations). The main effect for type of processing was statistically

significant but modest in size F (2, 339) = 5.50, p = .004, n2 = .02. The Ss in the

memory-based condition (M = 4.90, SD = 1.34) were more persuaded than those in
the on-line condition (M = 4.33, SD = 1.55) or the no instructions condition (M =
4.54, SD = 1.40). Examination of the pair-wise comparisons reveals that the
difference between the mean in the memory based condition and the on-line
condition was statistically significant, t(231) = 3.02, p = .003, r = .19, r’ = .19 as
was the difference between the memory-based condition and the no instructions

condition t(229) = 1.99, p = .05, r = .13, r’ = .13. The difference between the means

25

in the on-line condition and the no instructions condition was insubstantial and not
statistically significant, t(208) = 1.1 1, p = .27, r = .08, r’ = .08. This finding is
consistent with the hypothesis that when Ss are asked to memorize the arguments,
they will be more persuaded by them than if they are trying to evaluate the
arguments as they were instructed to do in the on-line condition and predicted to do

in the no instructions condition.

The main effect for argument strength was substantial and statistically
significant F(1, 339) = 102.18,p < .001, r = .48, r’ = .49. The Ss who read strong
arguments (M = 5.28, SD = 1.18) adopted attitudes more consistent with the
message than those who read weak arguments (M = 3.91, SD = 1.37). These main
effects were qualified by two higher order interactions. Specifically, the interaction

between argument strength and difficulty of comprehending the arguments was
statistically significant though modest in size, F (1 , 339) = 6.63, p = .01, n2 = .01.

The expected three-way interaction was also statistically significant but modest in
size, F(2, 339) = 3.04, p = 05,112 = .01.

The interaction between argument strength and comprehension difficulty
was examined within each processing type separately. First, the results in the on-
line processing condition were examined. There was a substantial and statistically
significant main effect for argument strength on attitude scores, F (1, 114) = 30.54, p
< .001, r = .46, r’ = .47 but the difficulty of understanding the arguments main
effect and the interaction were not substantial or statistically significant. This

finding is inconsistent with the hypothesis that there would be an interaction

26

between argument strength and difficulty of comprehension in the on-line
condition.

Next, the attitude scores of the Ss in the memory-based processing condition
were examined. Again, there was a substantial and statistically significant main
effect for argument strength, F(1, 114) = 49.52,p < .001, r = .55, r’ = .56 such that
strong arguments were more persuasive than weak. Also, there was no substantial
or statistically significant main effect for difficulty of comprehension or an
interaction. This finding is not consistent with the hypothesis that there would be a
main effect of difficulty in this condition and no effect for argument strength.

In the condition where the audience was not given processing instructions
there was a main effect for argument strength on attitudes, F (1, 112) = 26.59, p <
.001, r = .42, r’ = .43. This main effect was qualified by a substantial and
statistically significant interaction between argument strength and difficulty of
comprehension F(1, 112) = 12.39,p = .001, r = .28, r’ = .29. When the arguments
were strong, the easy to understand arguments were more persuasive than the
difficult to understand arguments, r = .32, r’ = .33. When the arguments were weak,
the difficult to understand arguments were more'persuasive than the easy to
understand arguments, r = -.31, r’ = -.32. The difference between these two effect
sizes is statistically significant, 2 = 3.41, p < .001. This finding is consistent with the
prediction that the Ss in the no instructions condition would process the message
similarly to how the Ss in the on-line condition were predicted to respond.

Recall. It was hypothesized that the difficulty induction would make it more

difficult for the $5 to learn the message arguments. Therefore, there should be a

27

main effect for the difficulty of understanding the message on the Ss ability to recall
the arguments accurately. It was also hypothesized that the Ss in the on-line
condition would recall the message arguments better than those in the memory-
based condition. A 3 (processing instructions: on—line, memory-based, or none) X 2
(weak v. strong arguments) X 2 (easy v. difficult to understand arguments) factorial
ANOVA was calculated with recall scores as the dependent measure (see Table 6
for means and standard deviations). The predicted main effect for difficulty was
substantial and statistically significant, F(1, 346) = 23.34, p < .001, r = .25 such that
difficult to understand arguments (M = 2.46, SD = 1.61) were not as well recalled as
the easy to understand arguments (M = 3.26, SD = 1.60).

There was also an unpredicted statistically significant though modest in size

interaction between processing type and arguments strength on the number of

accurately recalled arguments, F (2, 346) = 5.64, p = .004, n2 = .03. The effect was

such that in the on—line condition, more arguments were accurately recalled in the
weak argument condition (M = 3.32, SD = 1.47) than in the strong argument
condition, r = .25 (M = 2.50, SD = 1.74). The effect was reversed in the memory-
based condition as strong arguments (M = 3.25, SD = 1.62) were more easily
recalled than weak arguments, r = -.17 (M = 2.69, SD = 1.64). The difference
between these two effects was statistically significant, 2 = 3.28, p < .001. In the no
instructions condition, the means in the two argument strength conditions were
nearly identical (M = 2.77, SD = 1.60 for the weak arguments and M = 2.64, SD =
1.70 for the strong arguments). The simple effects in this interaction parallel the

effect of the inductions on perceptions of how difficult to understand the messages

28

were. There were no other substantial or statistically significant main effects or
interactions. The lack of a clear main effect for processing type is inconsistent with
the processing type hypothesis concerning recall.

Recognition The comprehension difficulty induction was predicted to reduce
recognition of the arguments. There was also a main effect for processing type
prediction such that the Ss in the memory-based condition were predicted to
recognize more of the information from the messages than those in the on-line
condition. A 3 (processing instructions: on-line, memory-based, or none) X 2 (weak
v. strong arguments) X 2 (easy v. difficult to understand arguments) factorial
ANOVA was calculated with recognition scores as the dependent measure (see
Table 7 for means and standard deviations). There was a small and marginally
statistically significant effect of the difficulty induction on recognition scores, F (1 ,
335) = 3.35,p = .07, r = .10, r’ = .14. The effect was such that easy to understand
arguments (M = 4.88, SD = 1.60) were recognized more accurately than difficult to
understand arguments (M = 4.57, SD = 1.66). This finding was consistent with the
hypothesis that difficult to understand arguments would be harder to recognize.
When corrected for measurement error, the size of the effect of the difficulty
induction on recognition was comparable to the effect of the difficulty induction on
recall.

There was also an unexpected small but statistically significant effect of the
argument strength induction on recognition F (1 , 335) = 4.92, p = .03, r = .12, r’ =
.16. The effect was such that when the arguments were strong (M = 4.92, SD =

1.59) they were more easily recognized than when they were weak (M = 4.54, SD =

29

1.66). This effect was particularly surprising because the recognition test was the
same for both conditions and only covered the factual parts of the arguments that
were not varied between argument strength conditions (e. g. the name of the 2005
revolution).

The argument strength effect on recognition was qualified by a small but

statistically significant interaction between processing type and argument strength,

F(2, 335) = 3.34, p = .04, n2 = .02. The interaction was such that the effect of

argument strength was largely confined to the memory—based condition.
Specifically, in the memory-based condition, more arguments were accurately
recognized in the strong arguments condition (M = 5.32, SD = 1.47) than in the
weak arguments condition (M = 4.34, SD = 1.67). This difference was substantial, r
= .30, r’ = .41. In the on-line condition there was little difference in recognition
scores based on argument strength (M = 4.93, SD = 1.58 for the weak arguments
and M = 4.83, SD = 1.50 for the strong arguments, r = -.03, r’ = -.04). There was
also little difference in recognition scores based on argument strength in the no
instructions condition (M = 4.33, SD = 1.67 for the weak arguments and M = 4.61 ,
SD = 1.73 for the strong arguments, r = .08, r’ = .11). The difference between the
size of the effect of argument strength on recognition in the memory-based
condition and the no instructions condition was marginally statistically significant, 2
= 1.71 , p = .09. There were no other substantial or statistically significant main
effects of the inductions or their interactions on recognition scores. The lack of a

main effect for processing type is inconsistent with the prediction that those in the

30

memory-based condition would score higher on the recognition test than those in
the on-line condition.

Attitude Certainty. It was predicted that Ss in the on-line condition would
report that they were more certain of their attitudes than the Ss in the memory-based
condition. A 3 (processing instructions: on-line, memory-based, or none) X 2 (weak
v. strong arguments) X 2 (easy v. difficult to understand arguments) factorial
ANOVA was calculated with attitude certainty scores as the dependent measure
(see Table 8 for means and standard deviations). There was a small but statistically
significant main effect for the difficulty of understanding the arguments induction
F(1, 339) = 7.62,p = .03, r = .15, r’ = .16 such that difficult to understand
arguments (M = 4.00, SD = 1.17) produced lower attitude certainty scores than easy
to understand arguments (M = 4.36, SD = 1.28). Although the main effect for
processing type was statistically insignificant, a planned comparison found that the
mean for the on-line condition (M = 4.33, SD = 1.22) was somewhat higher than the
mean for the memory-based condition (M = 4.04, SD = 1.19). This difference was
marginally statistically significant, t(232) = 1.89, p = .06, r = .12, r ’= .13. This
difference, though small, was in the direction predicted. The mean attitude certainty
scores of the Ss in the no instructions condition (M = 4.15, SD = 1.28) was within
sampling error of the other two processing conditions. There were no other
substantial or statistically significant effects on attitude certainty.

The Effects of the Inductions on the Dependent Variables’ Relationship with

Attitudes

31

Recall. Examining the association between recall and conformity to message
recommendations indicates that there are substantial differences among conditions.
Although the expected effect of the processing induction on the attitude-recall
correlation was found (see induction checks), recall also combines non-additively
with the some of the other experimental treatments to affect attitudes.

Specifically, a stronger recall-attitude correlation was found in the strong
argument conditions (.61 , r’ = .63) than in the weak argument conditions (.17, r’ =
.17). The difference between these two correlations was larger than could be
attributed to sampling error, 2 = 4.97, p < .001. The difference between these two
correlations cannot be attributed to differential variance as the standard deviation of
the mean of attitudes was lower in the strong arguments condition (SD = 1.18) than
in the weak arguments condition (SD = 1.3 7). Although the standard deviation of
the mean of the recall scores was slightly higher in the strong arguments condition
(SD = 1.69) than in the weak arguments condition (SD = 1.57), the difference was
trivial.

Furthermore, a stronger recall-attitude correlation was found in the difficult
to understand condition (.33, r’ = .34) than in the easy to understand condition (.02,
r’ = .02). The difference between these two correlations was larger than could be
attributed to sampling error, 2 = 2.99, p = .003. The difference in the size of these
correlations cannot be attributed to differential variance as the standard deviation of
the mean of the attitude measure in the difficult condition (SD = 1.37) was lower
than it was in the easy condition (SD = 1.53). The standard deviation of the mean of

the recall scores was higher in the difficult condition (SD = 1.60) than the easy

32

condition (SD = 1.55) but the difference was trivial.

The argument strength X recall interaction and the comprehension difficulty X
recall interaction must be qualified by a powerful recall X argument strength X
comprehension difficulty interaction. As Table 9 indicates, when arguments were
strong a substantial recall-attitude correlation was observed at both levels of
difficulty of understanding. When, however, arguments were weak, difficult to
understand arguments produced no evidence of a recall-attitude association. On the
other hand, when arguments were both weak and easy to understand, negative
association was observed. That is, the more Ss recalled the less they conformed to
message recommendations. The difference between the two correlations in the weak
conditions was statistically significant, 2 = 2.09, p = .04.

Recognition. Examination of the correlation between recognition of the
arguments and the Ss attitudes reveals substantial variation in the size of this
association among the conditions. As noted previously when the induction checks
were being examined, recognition was substantially correlated with attitudes in the
memory-based condition and the no instructions condition but not the on-line
condition. There was also an effect of the argument strength induction on the
attitude-recognition association. Similar to the attitude-recall correlation, attitudes
and recognition were substantially correlated in the strong arguments condition
(.42, r’ = .58) but not the weak arguments condition (-.O3, r’ = .04). In all of the
strong argument conditions, the association is positive and in most of the weak
argument conditions the association is within sampling error of zero (see Table 10

for the correlations in every condition). The exception to this pattern is that in the

33

no processing instructions condition when the arguments were weak and difficult to
understand, the relationship between recognition and attitudes is substantial and
positive.

Perceived Difficulty of Understanding the Message. As noted previously in
the induction checks section of the results, the relationship between the Ss perceived
difficulty of understanding the message and attitudes was substantial in the
memory-based processing condition but not in the other processing conditions. The
relationship between perceived difficulty and attitudes varies substantially as a
result of the other inductions’ interaction with processing type as well. Specifically,
the relationship between perceived difficulty and attitudes is substantial when the Ss
were in the on-line processing condition and were reading easy to understand,
strong arguments (see Table 11 for the correlations in every condition). In both the
memory-based condition and the no instructions condition, the relationship between
perceived difficulty and attitudes was only ample when the arguments were both
difficult to understand and strong. The correlations between perceived difficulty and
attitudes were insubstantial and with in sampling error of zero in the other
conditions.

Cognitive Elaboration. The relationship between elaboration and attitudes is
affected by the processing type induction such that the relationship is modest in the
memory-based processing condition (.20, r’ = .22) but insubstantial in the on-line
condition (-.04, r’ = -.04) and the no instructions condition (-.03, r’ = -.O3). The
difference between the strength of this relationship in the memory-based condition

and the on-line condition is marginally statistically significant, 2 = 1.77, p = .08.

34

The relationship between elaboration and attitudes was also qualified by argument
strength such that the correlation was positive when the arguments were strong (.35,
r’ = .38) and negative when they were weak (-.20, r’ = -.22).

The impact of argument strength on the elaboration-attitude correlation
varies with processing type (see Table 12 for the correlations in every condition).
Specifically, in both the on-line processing condition and the memory-based
condition, when the Ss read strong arguments, the correlation was consistently
positive (.39, r’ = .43 & .31, r’ = .34 respectively). When the Ss read weak
arguments the correlation was negative in the on-line condition (-.23, r’ = -.25) and
near zero in the memory-based condition (.05, r’ = .06). The difference between
these two correlations is not, however, larger than could be attributed to sampling
error, 2 = 1.47, p = .14. This pattern is consistent with the Spinozan hypothesis
because more cognitive elaboration did not affect the attitudes of the Ss in the
memory-based condition who read weak arguments but it did cause the Ss in the on-
line condition to be less persuaded by the arguments. The no-instructions condition
reveals a more complex interaction. Specifically, in the strong argument conditions,
the relationship between elaboration and attitudes was positive in both the easy to
understand condition and the difficult. On the other hand, when the arguments were
weak, there was a substantial negative relationship between elaboration and attitude
when the arguments were easy to understand but almost no relationship when the
arguments were difficult to understand.

Need to Evaluate

35

NTE was predicted to be an independent indicator of on-line versus
memory-based processing such that higher scores on the NTE were predicted to be
associated with a greater likelihood of on-line processing. First, given that on-line
processing was predicted to be associated with greater attitude certainty, NTE was
predicted to be correlated with attitude certainty. The data were consistent with this
hypothesis, r = .20, p < .001, r’ = .23.

In order to explore the effects of NTE on attitudes, the relationship between
NTE and attitudes was explored under different conditions of the inductions. First,
the relationship between NTE and attitudes was affected by the argument strength
induction such that when the arguments were weak, NTE and attitudes were
uncorrelated (r < .001) but when they were strong there was a modest correlation
between NTE and attitudes (r = .2, p = .01, r’ = .22). The difference in the strength
of this relationship cannot be attributed to differential variance in NTE as the
difference between the standard deviations was trivial (.02). When the arguments
were easy to understand, there was again almost no correlation between attitudes
and NTE (r = -.03, r’ = .03) but there was a modest correlation when the arguments
were difficult to understand (r = .18, p = .02, r’ = .20). This difference in the
strength of the relationship between NTE and attitudes across levels of difficulty
cannot be attributed to differential variance as the difference between the standard
deviations of the mean of NTE in the two levels of difficulty was trivial (.04).
Examination of the relationship between NTE and attitudes under different
conditions of both difficulty and argument strength reveals that NTE is positively

related to attitudes when the arguments are strong and difficult to understand but the

36

relationship is within sampling error of zero in the other three conditions of
argument strength and difficulty of understanding the message (see Table 13 for the
correlations at different levels of argument strength and difficulty of
understanding). This finding suggests that increases in the likelihood of on-line
processing are helpful primarily in detecting strong arguments when the message is

difficult to understand.

37

Discussion

In order to investigate the effects of memory-based versus on-line
processing on persuasion, a 3 (processing instructions: on-line, memory-based, or
none) X 2 (weak v. strong arguments) X 2 (easy v. difficult to understand
arguments) independent groups design experiment was employed. The effects of
these inductions on attitudes, recall of the arguments, recognition of the message
information, perceived difficulty of comprehending the arguments, and cognitive
elaboration were examined. The results will be discussed by first examining if the
data were consistent with the hypotheses that were derived from the SMP.
Hypotheses Testing

Attitude-recall relationship First, the SMP predicted that the recall-attitude
correlation would be stronger in the memory-based condition than in the online
condition and the no instructions condition. The evidence was consistent with this
hypothesis. There were, however, unanticipated effects. Specifically, in all
processing conditions there was a substantial correlation between recall and
attitudes when the arguments were strong. When the arguments were weak and
difficult to understand, there was no substantial relationship between recall and
attitudes. When the arguments were weak and easy to understand, there was a
negative relationship between recall and attitudes. It appears that in general, when
arguments are strong, the more of them the audience can recall, the more persuaded
they will be. When the arguments are weak, the relationship depends on how
difficult they are to understand. If the arguments are easy to understand, recalling

more of them allows the audience to see the weakness of the position advocated. If

38

the arguments are difficult to understand, the audience’s attitude is unaffected by
recalling the weak arguments, possibly because they do not understand what they
recall or the implications of what they recall.

Attitude-recognition relationship. Although previous research had not
examined the relationship between recognition of message elements and attitudes, it
was predicted that, like recall, recognition would be correlated with attitudes in the
memory-based condition but not the on-line or no-instructions condition. As
predicted, there was a substantial relationship between recognition and attitudes in
the memory-based condition, but not in the on-line condition. The substantial
relationship between recognition and message acceptance in the no instructions
condition was not predicted. The no instructions condition was expected to produce
results similar to the on-line condition as on-line processing was assumed to be the
default mode of processing. It appears that when the Ss are left to decide how they
will process a persuasive message, some of their results point toward on-line
processing (a low attitude-recall correlation), and some toward memory-based
processing (a substantial attitude-recognition correlation that was within sampling
error of the size of the correlation in the memory-based condition).

The finding that recognition was substantially correlated with attitudes in
the no instructions condition but recall was insubstantially correlated with attitudes
in that condition may indicate that recognition scores do not function the same way
as recall scores in this context. Some research suggests that recognition tests
measure the degree to which an audience was exposed to a message and recall tests

measure how well they learned a message (Shapiro, 1994). In this study, the

39

recognition test may have been measuring different aspects of processing in the
different processing conditions. In the on-line condition, recognition scores may
have been lowered by some of the Ss focusing on the arguments and not the factual
information. These Ss who scored lower on the recognition test may have been
reading the arguments very carefully despite their low recognition scores. In the
memory—based condition and the no instructions condition, low scores on the
recognition test may be related to an overall decrease in carefully reading the
message. The Ss in the memory-based condition and the no instructions condition
had no reason to focus on memorizing the arguments instead of the factual
information so the recognition test may be measuring how well they learned the
message in general. In these conditions, recognition may act as a general message
exposure measure. If an audience does not learn the message they cannot be
persuaded by it (McGuire, 1968). The positive relationship between recognition and
attitudes in the no instructions and memory-based conditions may be the result of
minimal persuasion occurring when the audience did not read the message carefully
enough to understand it and more persuasion occurring when they did carefully read
the message.

Argument strength also affected the strength of the relationship between
recognition and attitudes. In all of the strong argument conditions there was a
consistently substantial and positive relationship between recognition and attitudes.
In most of the weak argument conditions, there were only insubstantial
relationships between recognition and attitudes. The exception to this pattern was

the condition in which Ss were not given processing instructions and the arguments

40

were weak but difficult to understand. In this case, the relationship was substantial
and positive. The general pattern suggests that the more the audience is processing
carefully enough to learn the factual information in the article, the more persuaded
they would be by strong arguments. On the other hand, processing carefully enough
to recognize more information does not have a reciprocal negative effect on
attitudes when the arguments are weak.

Attitude-difficulty of comprehension relationship. One of the unique
predictions of the SMP was that the difficulty of understanding the message should
have a strong direct effect on the persuasiveness of the message when the 83 were
engaging in memory-based processing but not in on-line processing. This prediction
was made because in memory-based processing, the ability of the S8 to learn the
message is the primary predictor of its persuasiveness whereas in the on-line
conditions the Ss ability to learn the message is predicted by the interaction between
argument strength and the difficulty of understanding the arguments. The effect of
the processing induction on the difficulty- attitude correlation was consistent with
this hypothesis. In the memory-based condition the more difficult the arguments
were perceived to be to understand, the less the Ss held message-congruent
attitudes. The relationship between difficulty of understanding and attitudes was
insubstantial in the other two processing conditions. This finding was, however,
further moderated in unpredicted ways.

Specifically, in the memory-based and the no instructions conditions, the
negative relationship between attitudes and difficulty of understanding was only

ample when the arguments were both difficult to understand and strong. When the

41

Ss were assigned to the on-line condition with arguments that were easy to
understand and strong, the relationship was also negative and substantial. These
findings were not predicted by the SMP. Given that the sample sizes are relatively
small in these individual conditions, compared to the comparison based solely on
processing type, replication is required before rejecting the more parsimonious
explanation offered by examining only the effect of the processing instructions on
the attitude-difficulty correlation.

Attitude-cognitive elaboration relationship. Reynold’s (1997) measure of
cognitive elaboration was designed to measure the kind of cognitive elaboration that
is usually associated with evaluating message arguments. It was surprising then,
that in the memory-based condition, this measure would be positively associated
with attitudes. Examination of Reynold’s items reveals that most of them do not
contain an evaluative component. Many of them simply refer to cognitive effort or
thinking and many of the Ss assigned to the memory-based condition may have
interpreted them to be referring to the effort they put into memorizing the
arguments. This interpretation helps makes sense of the positive correlation
between elaboration and attitudes in the memory-based condition because putting
more effort into memorizing the arguments was predicted by the SMP to have a
direct positive effect on attitudes. When the arguments were not processed on-line,
they were predicted to be recalled as persuasive. The insubstantial correlations
between elaboration and attitudes in the on-line and no instructions conditions were
as predicted given that elaboration should interact with argument strength to affect

attitudes when the Ss are processing evaluatively.

42

The predicted interaction between argument strength and cognitive
elaboration did emerge in the on-line condition. When the Ss read weak arguments,
more elaboration was negatively associated with message acceptance. For these Ss,
expending more cognitive effort makes it easier to see how weak the arguments are
and reject them. For the Ss in the on-line condition who read strong arguments,
more cognitive elaboration was associated with more positive attitudes towards the
message advocacy. They were better able to appreciate the strong arguments with
increased cognitive elaboration.

The relationship between elaboration and attitudes was also affected by
argument strength in the memory-based condition. Specifically, there was a strong,
positive relationship between elaboration and attitudes in the strong arguments
condition. On the other hand, there was almost no relationship between cognitive
elaboration and message acceptance in the weak arguments condition. This outcome
suggests that although Ss who produce their attitudes on-line or use a memory-
based process are both more persuaded by strong arguments when they exert more
cognitive effort, only the Ss in the on—line condition are more likely to reject weak
arguments due to exerting more cognitive effort. The Ss in the memory-based
condition were less likely to evaluate the message when they expended more
cognitive effort because they expended their cognitive effort on memorizing the
arguments rather than evaluating the weakness in the arguments.

In the no processing instructions condition, the relationship between
elaboration and attitudes was more complex. Like the other processing conditions,

when the arguments were strong, the relationship between elaboration and message

43

acceptance was positive regardless of the difficulty of understanding condition they
were assigned. On the other hand, when the arguments were weak, the relationship
was affected further by the difficulty of understanding induction. For the Ss in the
no instructions condition who read weak and easy to understand arguments,
increased message elaboration resulted in less message acceptance. Yet, when the
arguments were weak and difficult to understand, there was no substantial
relationship between elaboration and attitudes. It appears that although the Ss in the
no instructions condition were likely to realize strong arguments were strong when
they put effort into analyzing them, effort only helped them recognize weak
arguments were weak when they were easy to understand.

Eflects of the inductions on attitudes. The SMP predicted that the attitudes
of the Ss in the memory-based condition would not be affected by argument
strength but would instead be affected by the difficulty of processing induction. The
data were not consistent with this hypothesis as the Ss in the memory-based
condition produced more message-consistent attitudes when the arguments were
strong than when they were weak. They were unaffected by the difficulty induction.
The Ss in the on-line condition were similarly affected by argument strength. On the
other hand, the Ss in the memory-based condition were, in every condition, more
persuaded by the message than the Ss in the comparable on-line condition. This
pattern of findings suggests that although Spinozan processes are not powerful
enough to make weak arguments as persuasive as strong ones, Spinozan processes
may provide strong and weak arguments a persuasive boost that they would not get

if they were being processed on-line.

44

In the on-line condition there was an interaction predicted between
argument strength and difficulty of processing such that strong arguments were
predicted to be consistently more persuasive than the weak arguments, but that this
difference would be larger when the arguments were easier to read. Although
examination of the means for the Ss in the on-line condition in Table 5 does show
that the weak arguments were slightly less persuasive when they were harder to
understand than when they were easier to understand, this difference was not larger
than could be attributed to sampling error. It may be that the response involvement
induction was so strong that the Ss in the on-line condition made the effort to
evaluate carefully the strength of the message arguments, even when they were
difficult to understand.

Stiff (1986) argued for an elastic capacity model of processing such that
when the audience is motivated to process carefully they will have an excess of
processing capacity if the message does not require much effort to be evaluated. He
argued that in such conditions, the audience would be able to evaluate both the
strength of the message and other cues such as source credibility. These conditions
might be similar to the on-line condition with easy to understand arguments.
Applying Stiff‘s model would predict that the Ss in the on-line, easy to understand
arguments condition would be affected by peripheral cues like source credibility as
well as argument strength. While none were induced in this study, future work
should examine this possibility. Stiff argued that when the processing demands are
greater, there will not be enough left over processing capacity to attend to the

peripheral cues. The on-line condition with difficult to understand arguments might

45

be a case where this effect occurs. Stiff’s model would predict minimal impact of
peripheral cues in such a condition.

The Ss in the no instructions condition were predicted to react the same way
as those in the on-line condition given Hastie and Park’s (1986) claim that on-line
processing is the default. The data from the Ss in this condition were consistent with
the prediction of an interaction between difficulty and argument strength. When the
arguments were difficult to understand, the Ss were slightly more persuaded by
strong arguments than weak. When the arguments were easy to understand, they
were much more persuaded by strong arguments than weak. According to the
results of the elaboration measure, the Ss in the no instructions condition were
elaborating somewhat less than those in the on-line condition. Perhaps the Ss in the
on-line condition were not as affected by the difficulty induction because they were
elaborating more than the Ss in the no instructions condition. When the Ss were
elaborating moderately in the no instructions condition, they may have been more
susceptible to the effects of the difficulty of understanding induction. It is possible
that in the no instructions, difficult to understand condition, the processing demands
of evaluating the arguments exceeded the amount of cognitive resources the 55 were
willing to devote to processing the message. This finding is consistent with the
unimodel (Kruglanski & Thompson, 1999) prediction that at a moderate level of
elaboration, argument strength and difficulty of understanding the message interact
such that argument strength will have a substantial effect when the arguments are

easy to understand but not when they are difficult to understand.

46

Effects of the inductions on attitude certainty. Based on Bizer et al.’s (2006)
findings, it was hypothesized that the attitudes of the Ss in the on-line condition
would be held with greater certainty than those in the memory-based condition. The
data were consistent with this hypothesis as the Ss in the on-line condition did
report slightly higher attitude certainty than those in the memory-based condition.
This finding provides a counterpoint to the finding that memory-based processing
increases persuasion. As Bizer et a1. noted, attitudes held with more certainty are
more resistant to counter-persuasion. Thus, although memory-based processing may
make arguments more persuasive, the attitudes that are formed may not remain
consistent with the message when the audience encounters opposition arguments.
Future research is needed to explore the effects of memory-based versus on-line
processing longitudinally to determine if attitudes formed using memory-based
processing are less resistant to change than those formed on-line.

There was also an unanticipated main effect of difficulty of understanding
on attitude certainty such that the Ss who read the easier to understand messages
reported higher attitude certainty than those who read the harder to understand
messages. One way to interpret this finding is that people feel less confident in their
new attitude when they are not as confident that they understood all of the
information upon which it is based.

Effects of the inductions on recall and difliculty of understanding. It was
hypothesized that due to the Ss in the on-line condition focusing on the arguments
in order to evaluate them, they might recall more of them than those in the memory-

based condition. There was no evidence of a main effect for processing instructions

47

on recall. Mackie and Asuncion (1990) found that more arguments were recalled in
the on-line condition than the memory-based condition in their first study but not
their second. It may be a relatively small effect that requires more power to be
detected (their study 1 had 21 Ss per cell for this comparison, their study 2 had 26
Ss per cell for this comparison). Also, Mackie and Asuncion did not induce their
arguments to be particularly strong or particularly weak. In this study, the on-line 85
only recalled more arguments than the memory-based subjects when the arguments
were weak. The finding was reversed when the arguments were strong. Perhaps
Mackie and Asuncion’s arguments were perceived to be weak in their first study.
They did not indicate if the arguments they used were strong or weak and they did
not report what those arguments were. In general, the attitude scores in their studies
did change in a direction consistent with the messages their Ss saw, but they usually
did not change past the mid-point of the scale. Although the finding that the attitude
change was in the direction advocated by the message is consistent with the
possibility that the arguments were strong, the finding that most of their Ss did not
change their overall stand on the topic is consistent with weak arguments. Future
work examining message recall under different processing inductions should induce
high and low argument strength to determine if the pattern of recall found here can
be replicated.

This interaction might be explained by the possibility that when Ss are
trying to learn arguments, making them weak makes them harder to understand than
if they are strong. Weak arguments have poor logical structure and may therefore be

more difficult to comprehend. The 55 reading these weak arguments may have had

48

to expend more cognitive effort on sense-making than those who read strong
arguments. This may have reduced the amount of cognitive resources available for
memorization. This possibility may explain why the weak arguments were harder to
recall than strong arguments in the memory-based condition. On the other hand,
when the audience is trying to evaluate the arguments on-line, weak arguments may
be recalled more easily because they are so easy to reject; whereas strong arguments
are harder recall because their lack of flaws results in Ss spending less time thinking
about them. This interpretation is consistent with the identical interaction between
processing type and argument strength with difficulty of processing as the
dependent variable. Ss in the memory-based condition reported that the weak
arguments were harder to understand and than the strong arguments. Those in the
on-line condition reported that the weak arguments were easier to understand than
the strong arguments. When focusing on how much of the message the audience
learns, the data are consistent with an interpretation such that weak arguments
makes evaluative processing easier but strong arguments make memorization
easier.

Effects of the inductions on recognition. The data were not consistent with
the hypothesis that the Ss in the memory-based condition would accurately
recognize more parts of the message than those in the on-line condition. There was
an unexpected interaction that demands further examination. In the memory-based
condition, scores on the recognition test were higher in the strong arguments
condition than in the weak. Given that the audience perceived the weak arguments

as more difficult to understand than the strong, it is unsurprising that the weak

49

arguments were less likely to be recognized accurately. Perhaps the on-line Ss did
not show a parallel finding because the recognition test focused on the factual
information of the message and the on-line Ss were more heavily focused on the
arguments concerning why the proposed candidate would be successful. Given the
low reliability of the recognition test, null findings should be interpreted with
additional caution.

Eflects of NT E. Given that NTE is thought to be a personality variable that
predicts on-line processing, it was expected to be associated with attitude certainty.
As predicted, it was modestly and positively associated with attitude certainty.
Although, examination of Table 1 shows that the correlation between attitude
certainty and cognitive elaboration (.37) is stronger than the correlation between
NTE and attitude certainty (.22). The difference between these two correlations is
statistically significant, 2 = 2.01 , p = .05. The likelihood of the audience devoting
substantial cognitive resources to elaboration of a message seems to be a stronger
predictor of attitude certainty than their tendency to evaluate messages. It is also
worth noting that NTE and cognitive elaboration were not substantially correlated.
One may have a tendency to engage in evaluative thinking, but it is not necessarily
translated into extensive elaboration of a given message. This result is consistent
with the SMP claim that evaluative processing does not necessarily require
extensive cognitive elaboration. One can make an evaluative judgment using simple
heuristics or careful processing.

NTE was also measured as an alternative index of on-line processing. Given

that NTE is not necessarily associated with greater cognitive effort it was predicted

5O

that there would be an interaction between the effects of the argument strength
induction, the difficulty of processing induction, and NTE on attitudes. Specifically,
when the arguments are strong, there should be a positive relationship between NTE
and attitudes because as NTE scores increase, Ss are expected to be increasingly
likely to evaluate the strength of the arguments. When the arguments are weak, the
relationship should be negative as higher NTE people are more likely to engage in
evaluative processing and thus reject weak arguments. Given that NTE does not
necessarily cause more cognitive elaboration, these relationships should be stronger
when the arguments are easy to understand than when they are difficult. The data
were not consistent with these hypotheses as the only condition in which there was
a substantial relationship between NTE and attitudes was when the arguments were
strong but difficult to understand. It may be the case that when the arguments are
very easy to understand, most people will engage in evaluative processing sufficient
to notice that the arguments are strong or weak. Yet, when the arguments are
difficult, only those who chronically engage in evaluative processing will be able to
appreciate the strong arguments.
Summary of Evidence Consistent with the SMP

At the heart of the Spinozan interpretation of memory-based processing of a
persuasive message versus the traditional, Cartesian approach (e.g. Hastie & Park,
1986; Lichtenstein & Srull, 1985) is the question of what happens when Ss who are
producing their attitudes in a memory-based fashion are asked to report them. The
traditional approach suggests that they recall the information they have on the

subject and then evaluate it immediately to produce their attitude. The SMP argues

51

that they do not evaluate that information and that they recall it as persuasive
because they did not evaluate the information at the time of information exposure.
The data were not wholly consistent with either approach. The robust effect of the
argument strength induction under conditions of memory-based processing suggests
that the information was evaluated by the Ss before they reported their attitude. On
the other hand, the Ss in the memory-based conditions did report consistently more
message acceptance than the Ss in the on-line conditions. If the audience merely
recalls information and then evaluates it as the traditional model suggests, it is
unclear why they should be more persuaded by that information than if they had
evaluated it during message exposure.

What this combination of findings suggests is a weaker form of the SMP.
Although Spinozan processes do provide a boost to the persuasiveness of the
recalled arguments, they are not strong enough to overwhelm how obvious it was
that the weak arguments were truly specious. Gilbert’s (1991) original research
focused on beliefs, not attitudes. He did find evidence that misremembered beliefs
were used to form an evaluation (Gilbert et al., 1993), but he found that the change
in evaluation was caused by an increase in how believable the information was. It is
possible that the Spinozan effects of learning information without evaluating it are
limited to the believability of the information rather than the general persuasiveness
of an argument. Future research needs to examine the believability of each piece of
recalled information to determine if the persuasiveness boost that learned, but not
evaluated, arguments get is mediated by believability judgments.

Limitations

52

One possible limitation of this study is that the topic was chosen to reduce
any pre-existing attitudes in order to examine the effects of the processing
inductions without any interference from previously held attitudes. In particular,
previously held attitudes could limit the impact of recall on attitudes in memory-
based conditions. It is easier for Ss to simply recall and report their old attitude
rather than try to update their attitudes with the new information they just learned. It
is uncertain then, if memory-based processing would provide an increase in
message acceptance if the audience had pre—existing attitudes that would have to be
changed rather than shaped by the message.

Mackie and Asuncion’s (1990) studies used topics for which the audience
would have had pre-existing attitudes and they found a strong attitude-recall
correlation in their memory-based condition and not their on-line condition. There
were not, however, any differences in overall attitude change between the memory-
based condition and the on—line condition. Their work is consistent with the
possibility that memory-based processing does not increase the persuasiveness of a
message when there are pre-existing attitudes. On the other hand, their study had
very low power to detect the modest sized effect for processing type on attitudes
found here. This study had 120 Ss per cell to detect the main effect for processing
type whereas Mackie and Asuncion had 26 Ss per cell and 21 Ss per cell in their two
studies. Future research with adequate sample sizes is needed to explore the effect

of pre-existing attitudes on memory-based processing.

53

Future Directions

There are a number of studies needed to explore some of the more intriguing
findings uncovered here. First, the unpredicted variation in the attitude-recall
correlation based on argument strength and difficulty of processing needs to be
replicated and probed further. Due to the focus on cognitive responses by most
persuasion researchers, the attitude-recall relationship has been neglected and the
finding of strong attitude-recall correlations in some of the on—line conditions
demands further examination. Future research would benefit by using long
messages such as those used here to allow sufficient variation in comprehension.

The way people construct their attitudes in situations of memory-based
processing requires further examination. One possible way to examine this type of
processing would be to use a thought-listing task, but to ask S3 to report the
thoughts they had while making a decision on the first attitude scale item rather
than while they read the message. It is possible that recalling unevaluated
information produces more positive thoughts during attitude formation based on
Spinozan processes. This type of research would also help distinguish between the
proposed SMP explanation and the traditional explanation of how attitudes are
constructed during memory-based processing.

This study and previous research (Lichtenstein & Srull, 1985; Mackie &
Asuncion, 1990) have demonstrated that Ss will process a persuasive message in a
memory-based fashion when an experimenter instructs them to. The next step is to
determine what kinds of messages can be included in persuasive experiments to

induce the audience to avoid evaluating the arguments and process in a memory-

54

based fashion without an experimental instruction. Such a message treatment would
have to maintain the audience’s focus on the message enough for them to learn the
message but at the same time discourage evaluation of it. In advertising, the
message might begin by asking the audience to think of ways that they use a type of
product and then inform them of the benefits of their brand. The focus on
generating uses might prevent evaluation of the advertisement’s claims.

An unexpected finding from Mackie and Asuncion’s first study suggests
another possibility. That study had four conditions. They had an on-line condition
with no source information, a memory-based condition with no source information,
a high expertise source condition, and a low expertise source condition. They found
that when a high expertise source was used, the relationship between recall and
attitudes was substantial even though the audience had not been given memory-
based processing instructions. Perhaps when the source has high expertise on the
topic, the audience is less likely to try to evaluate the message using counter-
arguing and instead focus on simply understanding what the speaker is saying.
Further research is needed to determine if the attitude-recall relationship with a high
expertise source replicates and if it can be explained as a type of memory-based
processing.

Conclusion

The Spinozan model of persuasion was proposed and tested. The data were
inconsistent with some aspects of the model and a revised model was proposed.
Future research will be able to determine if it will have utility in understanding the

persuasion process.

55

Appendix A
Strong and Difficult to Understand Arguments (Reading Ease = 10.1)

The superlative candidate for President of Ukraine is Varoslav Khmelnyt. He has
numerous socio-political policies that will be of assistance to superceding his

aspiration of progressing Ukraine from the 85th in the hierarchy of developed
nations to the top 30.

Exponentially accelerating the disjointing of Ukraine from Russia is advocated by
Khmelnyt. Ukrainians are fatigued by Russia’s interference in Ukrainian
geopolitical maneuvering. The popular Orange Revolution of 2005 endeavored to
frustrate the efforts of Russian-friendly Ukrainian politicians who had undertaken to
illicitly engineer the election in their favor, thus maintaining close ties with Russia.
The Orange revolution was spearheaded by Khmelnyt; he aided in the restoration of
unprejudiced and impartial elections to Ukraine, and promoted the independence of
Ukraine from Russia. Denser integration with Europe is now promoted by
Khmelnyt by his support of the admittance of Ukraine into the European Union, a
progression generally lauded as necessary to the advancement of expansion in
Ukraine’s economic sectors.

During the unforgiving winter of 2009, the flow of natural gas into Ukraine was
discontinued by Russia. Khmelnyt demonstrated outstanding leadership by
undertaking to travel to Russia and negotiating the recontinuation of an abundant
supply of natural gas into Ukraine. Presently, he has expanded beyond this
stratagem, favoring vigorous energy independence in pursuance of frustrating
additional attempts by Russia to dominate Ukraine as an “energy hostage”
recurrently in the future. It is contended by the United Nations panel on Eastern
European Energy that his proposition to provide energy independence from Russia
is pivotal in order to reach the ambition of stabilizing the Ukrainian economy.

A sustained multi-pronged campaign in opposition to corruption in the Ukrainian
parliamentary system of government was originated by Khmelnyt. Dissatisfaction is
increasingly germinating amongst many Ukrainians concerning the rampant and
impetuous corruption amongst their elected officials. Khmelnyt is a proponent of
constituting an independent anti-corruption bureau that would be modeled after
analogous successful agencies in other countries. Oxford political scientist Dr.
Richard Billingsworth expounded that such a bureau would be a key first step
towards strengthening the people’s faith in democracy and expanding the efficacy
in which the Ukrainian government functions.

The economic circumstances in Ukraine are dire, potentially calamitous. Levels of
unemployment were reached in Ukraine that were even more excessively elevated
than during the break-up of the Soviet Union. After extensive consultations with
preeminent economists, a sweeping jobs package was proposed by Khmelnyt. The
manufacturing foundation of Ukraine is largely based on steel production, and

56

without diversifying the economy it will be difficult for Ukraine to convalesce from
the global economic stagnated recession. In particular, a policy that Ukraine
substantially augments their green technology manufacturing is advocated by
Khmelnyt. Ukraine retains many industrial factories that could uncomplicatedly be
adapted for this proposition, and which would put people to work very
expeditiously in employment situations that disburse a very reasonable living wage.
The market for electricity generating windmills and other green energy
commodities is expanding precipitately in Europe due to commitments to reduce
greenhouse gases. This expansion will help diversify their economy, make Ukraine
less susceptible to fluctuations in the price of steel exports, and so augment the
economy with an undeniably robust boost. Khmelnyt’s support of increased trade
with Europe will safeguard that if Ukraine expands into green technology,
Ukrainian manufacturers will have access to those important markets.

Ukraine also has a deteriorating Acquired Immunodeficiency Syndrome (AIDS)
dilemma. It is estimated by The World Health Organization (WHO) a department of
the United Nations that they have the most accelerated growth in their human
immunodeficiency virus (HIV) infected population in the world. Despite the
magnitude of the predicament, previous Ukrainian leaders, including those
competing with Khmelnyt for the Presidency, have consistently rejected any
systematic proposals to initiate widespread sex education programs. Conversely, a
comprehensive HIV education and prevention program endorsed by WHO has been
proposed by Khmelnyt. The implementation of expanded safe sex education
programs throughout Ukraine’s schools and introduction expansive campaigns
targeted to the general public are also advocated by Khmelnyt. Based on programs
that have been shown to be effective in several similar neighboring countries in
Eastern Europe, this program has the additional benefit that the intervention was
adapted to be sensitive to local cultural beliefs. Khmelnyt believes that prevention is
the key to decelerating the expansion of HIV in Ukraine, and ultimately eliminating
the scourge of AIDS. This program also received the enthusiastic support of the
Red Cross’s panel on AIDS prevention.

Strong and Easy to Understand Arguments (Reading Ease = 53. 4)

Varoslav Khmelnyt would be the ideal President of Ukraine. He has good ideas for

Ukraine. These will help him reach his goal of moving Ukraine from the 85th most
developed nation to the top 30.

First, he wants a bigger split of Ukraine from Russia. Ukrainians tire of Russia’s
getting involved in their affairs. The popular Orange Revolution of 2005 stopped
the efforts of Russian-fiiendly Ukrainian politicians who tried to rig the election.
Those politicians had wanted to keep close ties with Russia. Khmelnyt helped lead
the Orange revolution to restore fair elections to Ukraine. He also promoted the
independence of Ukraine from Russia. Today he favors stronger ties with Europe by
supporting the entrance of Ukraine into the European Union. This idea is necessary
to improve economic growth in Ukraine.

57

Second, Russia cut off the natural gas flow into Ukraine during the bad winter of
2009. Khmelnyt showed strong leadership by going to Russia and getting natural
gas flowing into Ukraine again. He has since moved beyond this plan. He now
favors stronger energy independence to stop Russia from holding Ukraine as an
“energy hostage” again. The United Nations panel on Eastern European Energy
thinks that his plan to provide energy independence from Russia is key to
stabilizing the Ukrainian economy.

Third, Khmelnyt is fighting corruption. Many Ukrainians are angry about the
corruption of many of their leaders. Khmelnyt wants to make an independent anti-
corruption bureau. It would be based on agencies in other countries that worked
well. Oxford political scientist Dr. Richard Billingsworth says that such a bureau
would be a key first step towards strengthening the people’s faith in democracy and
helping the Ukrainian government work well.

Fourth, the economy of Ukraine is weak. Unemployment is higher than it was
during the break-up of the Soviet Union. Afier talking with top economists,
Khmelnyt proposed a big jobs plan. Ukraine’s industries are mostly making steel.
Without diversifying the economy it will be hard for Ukraine to recover from the
global recession. Khmelnyt wants Ukraine to expand their green tech production.
Ukraine has many factories that could easily be adapted to green tech production. It
would put people to work fast in jobs that pay a good living wage. Europe is greatly
expanding their efforts to reduce greenhouse gases. Their efforts are quickly
growing the market for electricity generating windmills and other green energy
products. If Ukraine moves into this market it will help diversify their economy.
Ukraine would rely less on the price of steel exports. This change would give the
economy a much needed boost. Khmelnyt’s support of increased trade with Europe
will help make sure that if Ukraine expands into green tech, they will have access to
those important markets.

Fifth, Ukraine has a growing AIDS problem. The World Health Organization
(WHO) found that they have the fastest growing HIV infection rate in the world.
Despite the size of the problem, previous Ukrainian leaders, as well as other
candidates for President, have not started sex education programs. Khmelnyt, on the
other hand, supports an HIV education and prevention program endorsed by WHO.
Khmelnyt wants to start safe sex education programs in schools. He also wants
safe-sex campaigns for the general public. This program will be based on others
that have worked in similar countries in Eastern Europe. The programs also have
the added benefit that they were adapted to be sensitive to local cultural beliefs. He
thinks that prevention is the key to slowing the spread of HIV in Ukraine, and of
ultimately stamping it out. This program has also received the strong support of the
Red Cross’s panel on AIDS prevention.

58

Weak and Difficult to Understand Arguments (Reading Ease = 16.1)

The superlative candidate for President of Ukraine is Varoslav Khmelnyt. He has
numerous socio-political policies that will be of assistance to superceding his

aspiration of progressing Ukraine from the 85th in the hierarchy of developed
nations to the top 30.

Exponentially accelerating the disjointing of Ukraine from Russia is advocated by
Khmelnyt. Ukrainians are fatigued by Russia’s interference in Ukrainian
geopolitical maneuvering. Khmelnyt promotes the distancing Ukraine from Russia
has been promoted by Khmelnyt because he supports requiring that only the
Ukrainian language be spoken on the entirety of Ukrainian television stations. It has
also been favored by Khmelnyt that Ukrainian universities implement guidelines
such that all Ukrainian Universities only administer entrance examinations in
Ukrainian. This change in linguistic requirements will obstruct Russia from
interfering in Ukrainian affairs by reducing the underpinning of support for
Russians living in Ukraine. When the popularity of Russians is reduced in Ukraine,
Ukraine is able to develop faster. It is believed by many members of Khmelnyt’s
political party that such policies will effectively keep Ukraine independent.

During the unforgiving winter of 2009, the flow of natural gas into Ukraine was
discontinued by Russia. Khmelnyt has consistently articulated the position that
Russia will be unlikely to discontinue the natural gas distribution into Ukraine
again. He contends that even under the circumstances where Russia does
discontinue the continuous flow of natural gas into Ukraine again, Ukraine should
accumulate their savings of currency reserves to make whatever payments Russia
demands. He envisions that exploring and searching for other sources of energy
reserves or alternative energy sources would be a misapplication of the
govemment’s efforts. Ukraine has always received its natural gas allotment from
Russia so they should experience minimal difficulties in obtaining natural gas in the
future.

Khmelnyt is also endeavoring to lead the fight against corruption. A sustained
multi-pronged campaign in opposition to corruption in the Ukrainian parliamentary
system of government was originated by Khmelnyt. Dissatisfaction is increasingly
germinating amongst many Ukrainians concerning the rampant and impetuous
corruption amongst their elected officials. Propositions have been put forth by
Khmelnyt supporting the creation of a committee to create a statement on possible
resolutions to the corruption problem. Many intractable problems have been solved
by examining potential solutions therefore this committee is guaranteed to
successfully find a resolution to the corruption difficulties. Some of the wealthiest
businesspeople in Ukraine have provided preferentiality to this solution to the
bribery and corruption problem.

59

The economic circumstances in Ukraine are dire, potentially calamitous. Levels of
unemployment were reached in Ukraine that were even more excessively elevated
than during the break-up of the Soviet Union. An extensive an economic plan has
been constructed by Khmelnyt for Ukraine. It has been proposed by Khmelnyt that
the preeminent stratagem for stimulating the economy would be to enlarge taxes on
small businesses. By implementing this policy, the government would expand their
currency reserves to prepare for the contingency of the economy substantially
declining further. When businesses are faced with increased taxation, they work
increasingly rigorously to sustain profitability. Therefore, these harder working
small businesses will consequentially encourage economic growth and substantially
increase the number of available employment opportunities for the unemployed and
the underemployed. Khmelnyt has also chosen to champion the restriction of
exports of Ukrainian products to other countries. He has confidence that this policy
of restricted trade will provide valuable assistance to Ukrainian businesses by
encouraging them to focus on the distribution and marketing of products that
Ukrainians find appealing. It has been argued by Khmelnyt that this policy will help
Ukraine build a strong domestic market because only when people in a country buy
domestically produced products will the entirety of the economy expand.

Ukraine also has a deteriorating Acquired Immunodeficiency Syndrome (AIDS)
dilemma. It is estimated by The World Health Organization (WHO) a subdivision
of the United Nations that Ukraine has the most accelerated grth in their human
immunodeficiency virus (HIV) infected population in the world. Despite the
magnitude of the predicament, a comprehensive HIV prevention program has not
been constructed in Ukraine. A proposition has been constructed by Khmelnyt for
the implementation of more draconian penalties for heroin abusers who inject the
drug intravenously using needles. Given a portion of those suffering from AIDS are
intravenous injection heroin addicts, reducing their numbers will eliminate the
AIDS crisis. Khmelnyt argues that one is either for strengthening and expanding
anti-drug laws or one is in favor of spreading AIDS via intravenous drug users. The
Ukrainian Society for Stricter Drug Laws and the leadership of the Orthodox
Church have enthusiastically campaigned for the prevention of AIDS by this
strategy.

Weak and Easy to Understand Arguments (Reading Ease = 60. 9)

Varoslav Khmelnyt would be the ideal President of Ukraine. He has several ideas

for Ukraine. These will help him reach his goal of moving Ukraine from the 85th
most developed nation to the top 30.

First he favors a bigger split of Ukraine from Russia. Ukrainians tire of Russia’s
getting involved in their affairs. The popular Orange Revolution of 2005 tried to
stop the efforts of Russian-friendly Ukrainian politicians who had tried to rig the
election in their favor. Khmelnyt wants to distance Ukraine from Russia by making
a law saying that only the Ukrainian language can be spoken on all Ukrainian TV
stations. He also favors a policy such that all Ukrainian Universities only give

60

entrance exams in Ukrainian. This will keep Russia from getting involved in
Ukrainian affairs by reducing support for Russians who live in Ukraine. When
Russians are less popular in Ukraine, Ukraine can develop faster. Many members of
his political party think that such laws will keep Ukraine independent.

Second, Russia cut off the natural gas flow into Ukraine during the bad winter of
2009. Khmelnyt believes that Russia will not cut off the natural gas flow again. He
says that even if Russia does cut off the gas flow, Ukraine should put aside more
money to pay off Russia. He feels that finding other sources of energy would be a
waste of the government’s time. Ukraine has always gotten its natural gas from
Russia so they should always be able to get it in the future.

Third, Khmelnyt is fighting corruption. Many Ukrainian people are growing angry
about all the corruption amongst their leaders. This is weakening faith in
government. Khmelnyt wants to creation a committee to make a report on some
solutions to the corruption problem. Many problems have been solved by trying to
make solutions so this committee is sure to succeed. Some of the richest people in
Ukraine like this answer to the bribery and corruption problem.

Fourth, the economy of Ukraine is weak as unemployment is as high as it has been
since the break-up of the Soviet Union. Khmelnyt has made an economic plan for
Ukraine. He thinks that the best way to grow the economy would be to raise taxes
on small businesses. This way the government could save more money in case the
economy gets worse. When businesses pay more taxes, they work harder to make
money. The harder working small businesses will in turn help economic growth and
raise the number of jobs available. Khmelnyt also wants to restrict exports to other
countries. He thinks this will help Ukrainian businesses focus on selling products
that Ukrainians like. He says that this will help Ukraine build a strong local market.
This is because when people in a country buy their own products, the entire
economy grows.

Fifth, Ukraine also has a big AIDS problem. The United Nations says that they have
the fastest growing number of people with HIV in the world. Despite this problem,
Ukraine does not have a complete HIV prevention program. Khmelnyt wants
stricter penalties for heroin users who use needles. Some of the people who have
AIDS are needle using drug addicts. If there are less of them it will solve the AIDS
problem. He says that one is either for strong anti-drug laws or one wants to spread
AIDS via drug users. The Ukrainian Society for Stricter Drug Laws and the leaders
of the Orthodox Church like this proposal a lot.

61

Appendix B
Recognition Items

1. What was the candidate’s name?

a. Yushchenko
b. Khmelnyt

c. Yanukovych
d. Kmelnovich

2. What office is the candidate running for?

a. Prime Minister

b. President

c. Head of Parliament

(1. Parliamentary Executive

3. The candidate described in the message wanted to move Ukraine’s ranking as a
developed nation from what to what?

a. 60th to the top 20
b. 85th to the top 10
0. 85th to the top 30
d. 60th to the top 5

4. What was the name of the revolution in 2005?

a. Yellow Revolution
b. Green Revolution

c. Orange Revolution
(1. Velvet Revolution

5. During which winter did Russia temporarily cut off the flow of natural gas to
Ukraine?

3.2007
b.2008
c.2009
d.2010

6. According to the message, unemployment is higher in Ukraine now than it was
during what other period?

a. World War II

62

b. The Cultural Revolution
c. The break-up of the Soviet Union
d. Right after 9/1 1/2001

7. According to what organization does Ukraine have the fastest growing AIDS
population?

a. Red Cross

b. Centers for Disease Control
c. World AIDS Forum

d. World Health Organization

63

Appendix C
Attitude Certainty Scale
1. I am sure that my attitude towards Khmelnyt for President of Ukraine is correct.
2. I feel confident that my attitude towards Khmelnyt for President of Ukraine is the
most accurate attitude possible.
3. I feel very certain that my attitude about Khmelnyt for President of Ukraine is
correct.
4. I believe that if someone challenged my views on Khmelnyt for President of
Ukraine, I would be able to easily defend my point of view.
5. I do not think that my attitude towards Khmelnyt for President of Ukraine is

going to change.

64

Table 1

Correlation Matrix 1C ontinuous Variables

1 2 3 4 5 6 7 8

 

1Attitude 0.16 0.32 0.14 0.83 -0.13 0.01 0.03
2Reca|| 0.16 0.78 0.28 0.07 -0.47 0.53 0.09
3Recognition 0.23 0.57 0.23 0.28 -0.49 0.50 0.03
4Attitude Certainty 0.13 0.26 0.16 0.12 -0.30 0.42 0.25

Perceived Argument

SStrength 0.78 0.07 0.20 0.11 -0.06 0.00 -0.02
6Perceived Difficulty -0.12 -0.45 —0.34 -0.27 -0.06 -0.37 -0.07
7Elaboration 0.01 0.49 0.34 0.37 0.00 -0.33 0.20

8Need to Evaluate 0.03 0.08 0.02 0.22 -0.02 -0.06 0.17

 

Note. The correlations below the triangle are the raw correlations. Those above the
triangle have been corrected for measurement error.

65

Table 2

Effects of the Inductions on Perceived Argument Strength

 

 

 

Difficulty
Argument Strength Easy Difficult
On-line

Strong 5.02 (1.19) 5.23 (.96)
Weak 3.47 (1.61) 3.95 (1.20)

Memory-Based
Strong 5.65 (.93) 5.25 (1.01)
Weak 4.08 (1.07) 4.20 (.98)

No Instructions
Strong 5.18 (1.02) 4.86 (.91)
Weak 3.69 (1.52) 4.31 (1.11)

 

66

Table 3

Effects of the Inductions on Perceived Difficulty of Understanding the Message

 

 

Difficulty
Argument Strength Easy Difficult
On-line
Strong 3.23 (1.42) 4.35 (1.41)
Weak 2.73 (1.46) 4.06 (1.57)
Memory-Based
Strong 2.60 (1.23) 3.81 (1.33)
Weak 3.26 (.86) 4.24 (1.09)

 

No Instructions
Strong 3.28 (1.46) 4.20 (1.43)

Weak 3.28 (1.07) 4.21 (1.28)

 

67

Table 4

Mects of the Inductions on Reported Message Elaboration

 

 

Difficulty
Argument Strength Easy Difficult
On-line
Strong 4.18 (.76) 4.25 (.94)
Weak 4.41 (1.27) 4.35 (1.03)
Memory-Based
Strong 4.31 (.93) 4.36 (.92)
Weak 4.35 (.80) 3.93 (.75)

 

No Instructions
Strong 3.99 (.98) 3.98 (.87)

Weak 4.26 (.95) 3.95 (.80)

 

68

Table 5

Effects of the Inductions on Attitudes

 

 

 

Difficulg
Argument Strength Easy Difficult
On-line

Strong 5.02 (1.18) 5.07 (1.18)
Weak 3.41 (1.61) 3.87 (1.49)

Memory-Based
Strong 5.84 (.94) 5.50 (1.11)
Weak 4.23 (1.16) 4.02 (1.19)

No Instructions
Strong 5.52 (1.06) 4.74 (1.31)
Weak 3.56 (1.27) 4.37 (1 .22)

 

69

Table 6

Effects of the Inductions on Recall

 

 

 

Difficulty
Argument Strength Easy Difficult
On-line

Strong 2.72 (1.86) 2.28 (1.62)
Weak 3.85 (1.15) 2.77 (1.57)

Memory-Based
Strong 3.29 (1.40) 3.21 (1.84)
Weak 3.05 (1.69) 2.33 (1.52)

No Instructions
Strong 3.38 (1.70) 1.88 (1.35)
Weak 3.27 (1.59) 2.27 (1.47)

 

70

Table 7

Effects of the Inductions on Recognition

 

 

 

Difficulg
Argument Strength Easy Difficult
On-line

Strong 4.79 (1.68) 4.87 (1.33)
Weak 5.35 (1.56) 4.48 (1.50)

Memory-Based
Strong 5.15 (1.61) 5.48 (1.33)
Weak 4.48 (1.60) 4.20 (1.75)

No Instructions
Strong 4.92 (1 .60) 4.26 (1.83)
Weak 4.57 (1.48) 4.10 (1.84)

 

71

Table 8

Efi'ects of the Inductions on Attitude Certainty

 

 

 

Difficulty
Argument Strength Easy Difficult
On-line

Strong 4.35 (1.11) 4.17(1.12)
Weak 4.53 (1.43) 4.28 (1.21)

Memory-Based
Strong 4.50 (1.19) 3.87 (1.21)
Weak 4.03 (1.31) 3.79 (.98)

No Instructions
Strong 4.56 (1.32) 3.95 (1.15)
Weak 4.17 (1.28) 3.92 (1.34)

 

72

Table 9

Effects of the Inductions on the Attitude C ertainty- Attitude Correlation

 

Difficulfy
Argument Strength Easy Difficult
Strong 0.52 (.55) 0.68 (.73)
Weak —0.29 (-.31) 0.02 (.02)

 

Note. Correlations corrected for measurement error (r ’) are in parentheses.

73

Table 10

Effects of the Inductions on the Recognition- Attitude Correlation

 

 

 

Difficulty
Argument Strength Easy Difficult
On-line

Strong .35 (.38) .24 (.26)
Weak -.12 (-.13) -.09 (-.10)

Memory-Based
Strong .39 (.43) .54 (.59)
Weak -.18 (-.20) .13 (.14)

No Instructions
Strong .36 (.39) .50 (.55)
Weak -.15 (-.16) .46 (.50)

 

74

Table 1 1

Effects of the Inductions on the Perceived Difficulty- Attitude Correlation

 

 

 

Difficulty
Argument Strength Easy Difficult
On-line

Strong -.43 (-.46) -.02 (-.02)
Weak .04 (.04) .18 (.19)

Memory-Based
Strong -.16 (-.17) -.56 (-.60)
Weak .20 (.21) -.26 (-.28)

No Instructions
Strong -.26 (-.28) -.52 (-.56)
Weak .14 (.15) .22 (.24)

 

75

Table 12

Effects of the Inductions on the Elaboration- Attitude Correlation

 

 

 

Difficulty
Argument Strength Easy Difficult
On-line

Strong .33 (.36) .44 (.48)
Weak -.19 (-.21) -.29 (-.32)

Memory-Based
Strong .24 (.26) .39 (.43)
Weak -0.002 .07 (.08)

No Instructions
Strong .37 (.41) .35 (.38)
Weak -.49 (-.53) -.05 (-.06)

 

76

Table 13

Effects of ALgument Strength and Difficulty of Understanding on the N TE- Attitude

 

Correlation
Difficulty
Argument Strength Easy Difficult
Strong .11 (.12) .26 (.29)
Weak -.14 (-.15) .14 (.15)

 

77

References

Beattie, A. E., & Mitchell, A. A. (1985). The relationship between advertising recall
and persuasion: An experimental investigation. In L. F. Alwitt & A. A.
Mitchell (Eds.), Psychological processes and advertising effects: Theory,
research, and applications (pp.129-155). Hillsdale, NJ: Lawrence Erlbaum
Associates.

Bizer, G. Y., Tormala, Z. L., Rucker, D. D., & Petty, R. E. (2006). Memory-based
versus on-line processing: Implications for attitude strength. Journal of
Experimental Social Psychology, 42, 646-653.

Chen, Y., & Blanchard-Fields, F. (2000). Unwanted Thought: Age differences in
the correction of social judgments. Psychology and Aging, 15, 475-482.

Cowley, E. (2006). Processing exaggerated advertising claims. Journal of Business
Research, 59, 728-734.

Ebel, R. L. (1951). Estimation of the reliability of ratings. Psychometrika, 16, 407-
424.

Flesch, R. (1948). A new readability yardstick. Journal of Applied Psychology, 32,
221-233.

Forster, S. & Lavie, N. (2007). High perceptual load makes everybody equal:
Eliminating individual differences in distractibility with load. Psychological
Science, 18, 377-381.

Gilbert, D.T. (1991). How mental systems believe. American Psychologist, 46, 107-
119.

Gilbert, D.T., Krull, D.S., & Malone, PS. (1990). Unbelieving the unbelievable:
Some problems in the rejection of false information. Journal of Personality
and Social Psychology, 59, 601-613.

Gilbert, D.T., Tafarodi, R.W., & Malone, PS. (1993). You can’t not believe
everything you read. Journal of Personality and Social Psychology, 65, 221-
233.

Hafer, C. L., Reynolds, K. L., & Obertynski, M. A. (1996). Message
comprehensibility and persuasion: Effects of complex language in

counterattitudinal appeals to laypeople. Social Cognition, 14, 317-337.

Hastie, R., & Park, B. (1986). The relationship between memory and judgment

78

depends on whether the judgment task is memory-based or on-line.
Psychological Review, 93, 258-268.

Haugtvedt, C. P, & Petty, R. E. (1992). Personality and persuasion: Need for
cognition moderates the persistence and resistance of attitude changes.
Journal of Personality and Social Psychology, 63, 308-319.

Hunter, J. E., & Gerbing, D. W. (1982). Unidimensional measurement, second
order factor analysis, and causal models. Research in Organizational
Behavior, 4, 267-320.

Jarvis, W. B. G., & Petty, R. E. (1996). The need to evaluate. Journal of Personality
and Social Psychology, 70, 172-194.

Koslow, S., & Beltramini, RF. (2002). Consumer skepticism and the “waiting room
of the mind”: Are consumers more likely to believe advertising claims if
they are merely comprehended? Advances in Consumer Research, 29, 473-
479.

Kruglanski, A. W., & Thompson, E. P. (1999). Persuasion by a single route: A view
from the unimodel. Psychological Inquiry, 10, 83-109.

La France, B. H. & Boster, F. J. (2001). To match or mismatch? That is only one
important question. Communication Monographs, 68, 211-234.

Lichtenstein, M., & Srull, T. K. (1985). Conceptual and methodological issues in
examining the relationship between consumer memory and judgment. In L.
F. Alwitt & A. A. Mitchell (Eds.), Psychological processes and advertising
effects: Theory, research, and applications (pp.113-128). Hillsdale, NJ:
Lawrence Erlbaum Associates.

Lowrey, Tina M. (2008). The case for a complexity continuum. In E. F.
McQuarrie & B. J. Phillips (Eds.), Go figure: New directions in advertising
rhetoric (pp.159-177). Arrnonk, NY: M. E. Sharpe.

Mackie, D. M., & Asuncion, A. G. (1990). On-line and memory-based modification
of attitudes: Determinants of message recall-attitude change
correspondence. Journal of Personality and Social Psychology, 5 9, 5-16.

McGuire, W. J. (1968). Personality and susceptibility to social influence. In E.
Borgatta & W. Lambert (Eds.), Handbook of personality theory and
research (pp. 1130-1187). Chicago: Rand McNally.

Newhagen, J .E. (1994). Effects of censorship disclaimers in Persian Gulf War
television news on negative thought elaboration. Communication Research,
21, 232-248.

79

Petty, R. E., Priester, J. R., & Wegener, D. T. (1994). Cognitive processes in
attitude change. In R. S. Wyer & T. K. Srull (eds.), Handbook of social
cognition (2n ed., Vol. 2, pp. 69-142). Hillsdale, NJ: Lawrence Erlbaum.

Regan, D. T., & Cheng, J. B. (1973). Distraction and attitude change: A resolution.
Journal of Experimental Social Psychology, 9, 138-147.

Reynolds, R. (1997). A validation test of a message elaboration measure.
Communication Research Reports, 14,269-278.

Romer, D. (1979). Distraction, counterarguing, and the internalization of attitude
change. European Journal of Social Psychology, 9, 1-17

Rosenblatt, P. C. (1966). Persuasion as a fiinction of varying amounts of distraction.
Psychonomic Science, 5, 85-86.

Shapiro, M. (1994). Signal detection measures of recognition memory. In A. Lang
(Ed.), Measuring psychological responses to media messages mp. 133-148).
Hillsdale, NJ: Lawrence Erlbaum Associates.

Skurnik, 1., Yoon, C., Park, D.C., & Schwarz, N. (2005). How warnings about false
claims become recommendations. Journal of Consumer Research, 31, 713-
724.

Stacks, D. W., & Burgoon, J. K. (1981). The role of nonverbal behaviors as
distracters in resistance to persuasion in interpersonal contexts. Central
States Communication Journal, 32, 61-73.

Stiff, J. B. (1986). Cognitive processing of persuasive message cues: A meta-
analytic review of the effects of supporting information on attitudes.
Communication Monographs, 53, 75-89.

Tormala, Z. L., & Petty, R. E. (2001). On-line versus memory-based processing:
The role of “need to evaluate” in person perception. Personality and Social
Psychology Bulletin, 2 7, 1599-1612.

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

80