Content uploaded by Richard E Petty
Author content
All content in this area was uploaded by Richard E Petty on Jan 31, 2019
Content may be subject to copyright.
What's in a frame anyway?: A meta-cognitive analysis of the impact of one
versus two sided message framing on attitude certainty
Derek D. Rucker
a,
⁎, Richard E. Petty
b
, Pablo Briñol
c
a
Kellogg School of Management, Northwestern University, USA
b
Ohio State University, USA
c
Universidad Autónoma de Madrid, Campus Cantoblanco, Madrid 28049, Spain
Received 20 January 2008
Available online 14 March 2008
Abstract
The current research examines a potentially new strategy to increase attitude certainty: framing messages as two sided. That is, we explore the
consequences of articulating that others have considered both the positives and negatives of a message position, in the absence of any real
differences in substantive content presented. Although classic research and theory appear to assume no clear benefit for simply framing a message
as two sided, we develop and apply a meta-cognitive approach that predicts advantages for such messages with respect to attitude certainty and
attitude-behavior correspondence.
© 2008 Society for Consumer Psychology. Published by Elsevier Inc. All rights reserved.
One topic that has received considerable attention within the
persuasion literature is whether persuaders can enhance their
influence by presenting some drawbacks to their product or
proposal. This can be seen clearly in work on one-sided versus two-
sided appeals (e.g., Hovland, Janis, & Kelly, 1953). In the former,
only positive information is given, whereas in two-sided messages,
a source discloses not only positive information, but some negative
information as well (e.g., Bohner, Einwiller, Erb, & Sibler, 2003;
Kamins & Marks, 1987; Pechmann, 1992; Sawyer, 1973).
Although research on one-sided versus two-sided persuasion
easily occupies volumes, there is an interesting caveat to this body
of work. Prior investigations have not considered the effects of
separating whether a message is simply framed as one-sided versus
two-sided from the actual content of the message. That is, relative to
a one-sided message, a two-sided message is accompanied by
negative information, which gives it the label “two-sided,”though
sometimes this information is also discounted (e.g., Lumsdaine &
Janis, 1953). However, what if the same message is simply framed
as being based on the consideration of one or two sides, but presents
the same substantive information? For example, consider a movie
that receives only praise from movie critics Ebert and Roeper.
Should the movie be marketed by simply highlighting the positive
opinions of these critics, or might there be a benefit to reminding the
audience that Ebert and Roeper are critics who consider both the
positives and negatives of the films they view?
Similarly, there is clear variance in how websites provide
shoppers with other consumers' opinions about a product. At
Epinions.com, feedback about the product is separated into the
pros and cons for each user, making it clear that both the good and
bad points of the product have been solicited and taken into
consideration. Amazon.com lists general user feedback making it
less clear whether users were encouraged to provide both their
positive and negative responses. Consequently, two reviews can
contain substantively equivalent information but differ in how
salient it is that both positive and negative information about the
product has been considered. An interesting question for
consumer research is, if the same substantive product attributes
are to be presented, does it matter whether a consumer perceives
the message to be a result of considering one versus both sides?
We refer to how the message is presented, as opposed to the actual
content of the message, as one-sided versus two-sided framing.
To our knowledge, evidence for the effects of one-sided versus
two-sided framing of the same message is absent. This might be
due in part to a) an emphasis on differences in the content of one-
sided versus two-sided messages, b) a failure to consider that
consumers' perceptions of how others reached an attitude or
A
vailable online at www.sciencedirect.com
Journal of Consumer Psychology 18 (2008) 137–149
Journal of
CONSUMER
PSYCHOLOGY
⁎Corresponding author.
E-mail addresses: d-rucker@kellogg.northwestern.edu (D.D. Rucker),
petty.1@osu.edu (R.E. Petty), pablo.brinnol@uam.es (P. Briñol).
1057-7408/$ - see front matter © 2008 Society for Consumer Psychology. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.jcps.2008.01.008
opinion could influence persuasion, and c) the lack of any
theoretical framework for predicting differences due to this type of
message framing. Nevertheless, we hypothesized consumers are
sensitive to the format of messages with respect to one-sided
versus two-sided framing and this framing can play a crucial role in
persuasion. Furthermore, we propose that this type of message
framing can have important implications for the amount of
certainty people attach to their attitudes.
A meta-cognitive perspective on one versus two sided
framing
Advances in meta-cognition—thinking about thinking (e.g.,
Alba & Hutchinson, 2000; Lee, 2004; Petty, Briñol, Tormala, &
Wegener, 2007; Schwarz, 2004; Wright, 2002)—can potentially
shed light on how consumers might respond to one-sided versus
two-sided framing. Work on meta-cognition has begun to examine
consumers' beliefs about how they reached an attitude or opinion.
This work has shown that such beliefs can influence the underlying
certainty people express in their attitudes. Attitude certainty is a
meta-cognition, as it is a secondary cognition (i.e., “How certain
am I of my attitude?) about a primary cognition (i.e., “My attitude
towards this product is positive.”). For example, Tormala and Petty
(2002) found that individuals who successfully defended their
attitudes reported greater certainty in their attitudes when they
perceived themselves to have resisted a strong, as opposed to a
weak, attack. Rucker and Petty (2004) found that when persuaded
by a very strong message, people reported being more certain of
their attitudes when they had made an effort to consider the faults
as opposed to having processed the message in a more objective
fashion. Common to both paradigms is that consumers' beliefs
about how they reached their attitudes influenced the certainty they
attached to those attitudes. The inference process itself is meta-
cognitive in nature as it reflects consumers' perceptions of their own
cognitive processes in reaching their attitudes (Petty et al., 2007).
Since it matters how consumers think they formed their own
opinions, we argue that consumers might also consider how others
have reached an opinion. Specifically, making it salient to
consumers that both sides have been considered by someone else
(i.e., a two sided-frame) could lead consumers to make a conscious
and thoughtful inference that their attitudes are based upon greater
knowledge (i.e., they reached their evaluation by knowing not only
the positives, but also the potential negatives). We predicted that a
two-sided frame would influence how certain individuals would be
regarding the attitudes they reached. In particular, when it is clear
that a source has considered the negatives as well as the positives,
and the negatives presented are either absent or few and
inconsequential, we propose consumers may intuit that there
must be few remaining unknown negative attributes; therefore they
can be confident in the positive evaluation of the message position.
If there is no indication of consideration of negative factors, people
might intuit their attitudes are based on less complete knowledge,
and thus will be less certain of the attitude formed. We propose this
reflects a thoughtful inference process whereby people consider the
completeness of their knowledge in determining their certainty.
Indeed, recent research suggests that when a message only
presents one-sided attributes (positive or negative), people some-
times assume that there might be opposite attributes of which they
are unaware (Priester, Petty, & Park, 2007). In one study, for
example, Priester et al. varied the amount of univalent information
participants received such that either one or seven pieces of positive
or negative information was presented. Priester and colleagues
found that participants reported feeling more ambivalent/uncertain
when they received one piece of information as opposed to seven,
and this uncertainty was mediated by beliefs about the existence of
missing information of the opposite valence. Furthermore, when
people are directly prompted to consider potentially missing infor-
mation, they evince less confidence in their evaluations (Sanbon-
matsu, Kardes, & Herr, 1992). An alternative way to view these
findings is that if increasing perception of missing information
increases ambivalence or uncertainty, then any variable that reduces
perception of missing information should increase certainty.
Although we expected our two sided framing manipulation to
affect perceptions of certainty for the reasons outlined above, we
did not expect effects on the extremity of the attitude for three
reasons. First, and perhaps most obviously, in past research
varying knowledge about a topic, attitude extremity was tied
largely to the actual content of the information provided. For
example, when people were exposed to more positive attributes
about a product or issue, they were more knowledgeable and more
persuaded (e.g., Petty & Cacioppo, 1984). In our research, only
perceptions of knowledge will be varied since the one and two
sided messages present the same content to message recipients.
Second, prior research has suggested that when arguments are
unambiguously strong, mere perceptions of knowledge in the
absenceof actual knowledge differences are less likely to influence
the extremity of one's e valuation (e.g., Rucker & Petty, 2004). This
is because when arguments are clearly strong and are easily
processed, individuals are readily able to determine their
evaluative reaction. However, whether one can trust that
evaluation might require input from additional sources (e.g., “Do
I have all the information?”). Third, in general, perceptions of
knowledge have been shown to be more strongly associated with
the certainty of attitudes rather than the extremity of attitudes
(Krosnick, Boninger, Chuang, Berent, & Carnot, 1993). The logic
for this is that feeling one is better informeddoes not signal that one
should take a more extreme position but that one can be more
confident in the attitude reached. Thus, our effects were expected
to be found primarily on attitude certainty. The finding ofeffects on
attitude certainty in the absence of extremity differences is also a
methodological advantage because if our manipulation affected
both extremity and certainty, inferences of certainty could
plausibly stem from the extremity of the attitude itself.
In short, we argue that when a source indicates that negatives
have been considered, concern over possibly missing negative
information can be put aside and people can feel more
knowledgeable and therefore more certain. Put differently,
even though consumers' actual thoughts about a product might
be similar in one and two-sided framing conditions, leading to
similarly valenced attitudes toward the product, their perception
of how the message was developed might prompt them to reach
different conclusions about their knowledge regarding their
evaluation, which might lead them to become more or less
certain in the evaluation formed. Positive attitudes will be held
138 D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
with greater certainty when consumers can infer that their
attitudes are based not only on the positive information they are
aware of but also the fact there are likely to be few negative
features of a product of which they are unaware.
The study of attitude certainty is an important domain of
inquiry in the persuasion literature because of the numerous
consequences of certainty. Attitudes held with certainty exert a
stronger influence on behavior than attitudes held with less
certainty (Berger, 1992; Fazio & Zanna, 1978; Krishnan &
Smith, 1998; Rucker & Petty, 2004), are more likely to persist
over time (Bassili, 1996) and are more likely to be resistant to
attempts to change them (Petrocelli, Tormala, & Rucker, 2007;
Tormala & Petty, 2002). If a company wishes to create loyal
customers or politicians wish to create dedicated voters who are
less likely to be swayed by competitors, instilling attitudes with
certainty is one means of accomplishing this.
Of course, it first remains to be seen whether our reasoning is
correct. It could be, for instance, that consumers are insensitive to
manipulations of message framing involving whether others have
focused on one or both sides of an issue, or, even if sensitive,
framing might not affect people's assessments of their knowl-
edge and certainty. Similarly, given a manipulation of message
framing highlights what another person has done, or purported
to have done, a message recipient might not be inclined to use
this information as they would use information about their own
attempts to consider both sides (Rucker & Petty, 2004). Thus,
the first goal of our research was to examine if framing a
message as one or two sided influences attitude certainty.
Experiment 1 tests the proposition that framing a message as
two-sided enhances attitude certainty. Experiment 2 examines
the moderating role of a consumer's self-reported category
knowledge. Experiment 3 directly manipulates the perception
both sides have been considered and examines a potential
alternative interpretation. Experiment 4 examines whether the
effects are due to a simple cue or a more complex association
process. Finally, Experiment 5 tests whether two-sided framing
enhances attitude-behavior correspondence over one-sided
framing, a classic consequence associated with attitude certainty.
Pilot testing
Although we did not anticipate changes in the extremity of
people's evaluations as a result of our framing manipulation, we
conducted pilot-testing to further guard against changes in
extremity. Past research has shown that presenting a substantive
two-sided message (i.e., presenting arguments against one's
position) can increase attitude change by increasing the credibility
of the source (Kamins & Marks, 1987). Although it was not clear
whether simply framing a message as two-sided, where no
negative information is actually presented, would affect the
perceived credibility of the source, we nonetheless did two things
to mitigate this possibility. First, we attempted to hold credibility
constant by selecting sources that we felt would be perceived as
relatively constant in credibility regardless of message framing. In
particular, the source of the information for Experiments 1, 2, 3,
and 5 was portrayed as either product-testers for, or employees of,
Consumer Reports. In Experiment 4, the sources of the message
were students who were reported to have used the product for an
extended period of time. In bothof these cases we used individuals
who were portrayed as knowledgeable about the product and who
had little reason to be biased in their reporting of the information.
To further ensure our manipulation did not influence source
credibility, we conducted a pilot test in which we exposed two
separate groups of participants to each of our messages and asked
them to evaluate the credibility of the sources. Specifically,
participants were asked how credible and how trustworthy the
individuals providing the feedback were (7-point scales). These
items were highly correlated (r=.71 for student testers (Experi-
ment 4), r=.86 for product testers (Experiments 1, 2, 3, and 5) and
combined to form an aggregate measure of credibility. When a
message came from product testers (N= 33), participants
perceived the source of the feedback as equally credible regardless
of whether the message was framed as one-sided (M=4.00,
SD = 1.03) or two-sided (M= 3.50, SD = 1.72; t(31) = 1.00,
p=.32). For the student testers, (N =52), sources were also seen
as equally credible for both the one-sided (M=4.10, SD =1.52)
and two-sided framing conditions (M=3.83, SD =.94; t(50)= .77,
p=.44). Thus, our initial pilot testing confirmed that the message
conditions used in the present research were unlikely to influence
credibility of the source in a positive or a negative fashion.
Experiment 1
Experiment 1 tested the effect of one versus two-sided framing
on attitude certainty. In addition, we tested whether the effect
required that others find nothing wrong with the product or that
others made an effort to consider both sides. That is, consumers
might feel more certain following two-sided framing because they
perceive others to have considered both sides, or they might feel
more certain only if the others report finding absolutely nothing
wrong. Based on the logic that the key input into consumers'
perceptionof knowledge is the feeling that others have considered
both the potential pros and cons of a message, we suspect that the
findings will hold even when a slight negative is present. This
assumption is tested in Experiment 1.
Method
Participants and design
One-hundred and nineteen undergraduates from a Midwes-
tern university participated in exchange for partial course credit.
The design of the experiment was a 2 (message frame: one-
sided, two-sided) X 2 (degree of negativity: none, slight)
between-participants design.
Procedure
Participants were given a packet of materials and asked to
complete them as part of a study on advertising research.
Participants read a page explaining the nature of the task followed
by a presentation of the product description, a cordless telephone.
Participants were told all information in the present experiment
was taken from product testers in a study conducted byConsumer
Reports. Finally, we assessed participants' attitudes and attitude
certainty.
139D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
Independent variables
Message frame. In the one-sided framing condition partici-
pants read a brief blurb about the product that included mention
of the extent of positive reviewer feedback. In the two-sided
framing condition, participants learned of the positive reviewer
feedback, but also received information about the extent, if any,
of negative feedback.
Degree of negativity. All participants were told that there were
30 individuals who had used the product. In the absolutely no
negative condition participants learned that all individuals were
positive towards the product, whereas in the slight negative
condition they learned that 29 were positive. We intentionally
used only a slight amount of negative information in an effort to
maintain equivalency in attitudes across conditions. In the one-
sided framing condition participants were told that either all 30
reviewers were positive (no negatives condition), or that of the
30 reviewers, 29 were positive (slight negative condition). No
mention was made of negatives. In contrast, in the two-sided
framing condition they were told that 30 reviewers were positive
and 0 were negative (no negatives condition), or of the 30
reviewers 29 were positive and 1 was negative (slight negative
condition). Thus, all participants received the same substantive
information, but the fact that negative reviews were possible was
made salient in the two-sided framing condition.
Dependent variables
Attitudes. Attitudes were assessed using a composite of three
semantic differential scales (good-bad, favorable-unfavorable,
positive-negative; α= .92). All scales ranged from 1 to 7 with
negative descriptors anchored at 1 and positive descriptors
anchored at 7.
Attitude certainty. Attitude certainty was assessed with a
composite of two items (α=.81). Participants were asked, “How
certain are you of your attitude toward this product?”and “How
convinced are you that your attitude toward this product is
correct?”Both items were completed on 7-point scales with
1=“notatall”and 7= “extremely.”These items were adapted from
past research (see Fazio & Zanna, 1978; Rucker & Petty, 2004).
Results
Attitudes
The data were analyzed using a series of 2 (message frame:
one-sided, two-sided) X 2 (degree of negativity: none, slight)
ANOVAS. Individuals tended to like the product more when
there were no negatives (M= 5.51, SD = .94) as opposed to one
negative (M= 5.26, SD = .88), but this difference was not
reliable, F(1, 115)= 2.44, p= .12. This finding reinforces the
idea that the slight negative was, as intended, weak. Participants
had equally positive attitudes regardless of whether they
received the one-sided (M= 5.35, SD = .87) or two-sided frame
(M= 5.42, SD = .97), Fb1, ns. There was no message framing
by degree of negativity interaction, Fb1, ns.
Attitude certainty
On the certainty measure, only a significant main effect of
message framing emerged such that individuals who received
the two-sided frame held their attitudes with greater certainty
(M= 5.30, SD = .81) than individuals who received the one-
sided frame (M= 4.93, SD= 1.15; F(1, 115) = 4.08, pb.05).
There was no main effect of degree of negativity on attitude
certainty (Fb1, ns) or an interaction (F(1, 115) = 1.18, p= .28).
Discussion
Experiment 1 found that message frame (one versus two-
sided) significantly influenced consumers' attitude certainty.
Participants reported greater certainty in their attitudes when the
message highlighted that reviewers could have presented both
sides, suggesting that both the positives and the negatives of the
product were considered. Importantly, this was not simply a
manifestation of differences in attitudes as there were no reliable
differences on the attitude measure. Finally, the effect was not
modified by whether there were zero negatives or one negative
presented. This suggests the certainty effect does not rest solely
upon highlighting the absolute absence of negatives in others'
reports, but rather with the two-sided framing.
Experiment 2
Experiment 2 was designed to examine a potential moderator
based on our framework. If framing a message as having presented
both sides is influencing consumers' certainty via enhancing
consumers' more general perception of their own knowledge, then
our effects should be most likely to be observed when participants
do not already have reason to perceive themselves to have
knowledge. For instance, for those who perceive themselves to be
knowledgeable about the product category, their perceptions of
knowledge regarding their evaluation towards a particular
product, and thus their certainty, can be derived from their own
perceptions of knowledge about the category. Therefore, their
knowledge about their evaluation should be less influenced by
message framing. However, those who are low in general
perceived knowledge about the product category, and are likely
to need other inputs into how knowledgeable they are, should be
more prone to be affected by message framing. To examine this
possibility, we included an item to assess consumers' general self-
perception of knowledge about the product category. Importantly,
based on our perspective, it is not necessarily actual knowledge
that should be as important in deriving one's certainty as is
perceived knowledge. Finally, we used a bicycle as our target
category to increase the generalizability of our results. Otherwise,
the design of Experiment 2 was conceptually identical to
Experiment 1.
Method
Participants and design
Ninety-three undergraduates from a Midwestern university
participated in exchange for partial course credit. Participants
were randomly assigned to conditions in a 2 (message framing:
140 D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
one-sided, two-sided) X 2 (degree of negativity: none, slight),
with self-reported category knowledge assessed continuously.
Procedure
Participants arrived to take part in a project on evaluating
advertisements. All participants then received product informa-
tion for the “Davinci”Bike. The information featured a
description of the product purported to be supplied by Consu-
mer Reports. The information by Consumer Reports indicated
features of the product such as “lightweight titanium frame.”
Next, participants were asked to report their attitudes and
attitude certainty. Participants then engaged in approximately
15-minutes of unrelated tasks before finally being asked how
much they generally knew about bicycles. Participants were
then thanked and debriefed.
Independent variables
Message framing. The one-sided frame contained two col-
umns labeled with a single header, “product specifications.”The
two-sided frame also contained two columns but one was labeled
“product pros”and the other labeled “product cons.”A sample of
the one and two-sided message frame conditions for those who
received one negative, is provided in the Appendix.
Degree of negativity. In the absolute no negative condition
there was no negative information supplied about the product.
Thus, the column labeled “product cons”was left blank in the
two-sided frame condition. In the slight negative condition, the
message indicated the product did not come with a water bottle,
a feature pilot-testing indicated was slightly negative. In the
one-sided frame condition this was placed at the start of the 2nd
column under the overall heading of “product specifications.”
This information occupied the same physical space in the two-
sided frame condition, but the information fell under the
“product cons”column. Thus, both groups received slightly
undesirable information, but the two-sided frame condition
highlighted more blatantly that both sides had been considered
by Consumer Reports.
Self-reported category knowledge. To assess participants'
perception of how knowledgeable about the bike category they
were, we asked participants, “In general, how much do you
know about bicycles?”Participants' responses were assessed on
a 7-point scale anchored at 1 = I know very little about bicycles,
and 7 = I know a lot about bicycles.
Dependent variables
Attitudes (α= .93) and attitude certainty (α= . 80) were
assessed as in Experiment 1.
Results
Analyses were conducted with ANOVA or, when the effects of
self-reported category knowledge were examined, using regression
with variables mean centered. Regression analyses were conducted
in a hierarchical fashion, entering main effects in the first step, two-
way interaction in the second step, etc. First, because self-reported
category knowledge was being used as an independent variable,
we made sure this variable was not affected by the manipulations.
There was no main effect of message framing (F(1, 89) = 2.79,
p= .10) or degree of negativity (F(1, 89)= 1.43, p= .23) on reported
category knowledge. Furthermore, there was no interaction
between message framing and degree of negativity (Fb1, ns).
1
Attitudes
There was no effect of message framing on attitudes as
participants held positive attitudes in both the one (M=5.04,
SD= .84) and two-sided (M=5.37,SD= 1.28) framing conditions,
F(1, 89)= 2.24, p=.14. Furthermore, replicating Experiment 1,
the inclusion of a slight negative did not influence attitudes (Fb1,
ns). There were no interactions of message framing with attitudes,
message framing with knowledge, or a 3-way (all p'sN.40).
Attitude certainty
As in Experiment 1, participants reporting being more certain
when receiving the two-sided frame (M=5.29, SD =1.16)
compared to the one-sided frame (M=4.78, SD =1.28; F(1, 89) =
4.03, pb.05). There was no effect of degree of negativity on
attitude certainty or an interaction between message framing and
degree of negativity (F's b1, ns). However, as anticipated, reg-
ression analyses revealed the main effect of message framing
was qualified by a significant message framing by self-reported
category knowledge interaction (t(89) = 2.01, pb.05). The inter-
action, depicted in Fig. 1, indicates the effect of message framing
was more prevalent for those low in self-reported category
knowledge. In fact, simple slopes analyses, using the procedures
described by Aiken and West (1991), indicated the effect of
message framing on certainty was significant for those low in
self-reported category knowledge (−1 SD; t(89) = 2.60, p= .01),
but was not significant for those high in self-reported category
knowledge (+1 SD; t(89) = .26, p= .80). There was no signi-
ficant 3-way interaction (pN.90).
Discussion
Replicating Experiment 1, consumers were more certain when
presented with a two-sided frame. This effect was qualified by
consumers'perceptions of knowledge about the product category.
Those low in perceived category knowledge showed a differential
influence of message framing. Those high in perceived category
knowledge showed no effect of message framing suggesting they
did not rely on whether others had considered both sides to
determine their certainty, but presumably relied on the fact they
felt knowledgeable about the category.
Experiment 3
Experiment 3 was conducted to provide a more direct test of
the hypothesis that the first step in our proposed reasoning
1
Because the question asked about the product category rather than the
particular bicycle, we expected this measure to be more stable and unaffected
by the message frame or degree of negativity for the particular bicycle.
141D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
process is indeed the perception others have considered both
sides. Specifically, we manipulate the perception that others
have considered both sides by informing participants, in the
two-sided condition, that consumers had been explicitly asked
to consider both the positive and negatives of the product. This
manipulation allows us to test the notion certainty is increasing
from the fact others had considered both sides rather than any
other qualities of the message itself. In addition, we examine
whether our effects might be explained by the perception that
people view the source as less biased when both sides appeared
to have been considered. On the one hand, people might be more
certain of their evaluation because they view the source as less
biased when both sides are presented. On the other hand, given
sources in our study ultimately make the same recommendation,
people might not view any differences in the bias (or view them
as equal in bias). Indeed, our pilot testing of the framing
manipulation described earlier indicated that this manipulation
did not lead to differential perceptions of source credibility (i.e.,
the sources' expertise or trustworthiness). Nonetheless, Experi-
ment 3 examined source credibility and also included more
specific measures of perceived bias.
Method
Participants and design
Twenty-eight undergraduates from a private Midwestern
university participated in the experiment in exchange for partial
course credit. Participants were randomly assigned to a one-
sided frame or two-sided frame condition.
Procedure
Participants were informed they would be participating in a
market research study interested in examining consumers'
perceptions of new products that would soon be in their area.
Participants then received the product tester reactions that
consisted of a 177-word summary of the reaction of three
product testers for a new brand of toothpaste. The responses
were positive and consisted of statements such as “This is easily
one of my favorite toothpastes I have tried,”and “I was
extremely impressed with its whitening power.”Finally,
participants' attitudes, certainty, and potential mediators were
assessed.
Independent variable
Message framing. Participants who received the one-sided
frame were told that all feedback was provided by product testers
who had been asked to use the product for a trial period and
provide their thoughts about the product. In the two-sided frame
condition, participants were told that, during the trial period,
participants were explicitly asked to consider any positive as
well as any negative reactions they had to the product. That is, it
was made clear that, in the process of using the product, product
testers were encouraged to focus on both their positive and
negative reactions. In both conditions participants subsequently
received the same identical information from product testers.
Thus, the manipulation occurred outside of the delivery of the
message itself.
Dependent variables
Attitudes. Attitudes were assessed as in prior experiments
(α= .92).
Attitude certainty. Attitude certainty was measured as in prior
experiments (α= .73).
Manipulation check. To assess whether participants perceived
that both sides had been carefully considered, they were asked to
respond to three items. Specifically, participants were asked to
report their a greement on 7-point scale s (1 = Disagree strongly,
7 = Agree strongly) with the statements, “The product testers'
assessment was based on a consideration of both their potential
negative and potential positive reactions”;“The product testers
had the opportunity to list both their positive and negative
reactions”;“The product testers could have provided either
positive or negative reactions to the product.”These items were
combined to form an aggregate measure of participants'
perceptions that both sides have been considered (α= .78).
Source bias. To assess participants' perception of the source's
bias, participants were asked, “How biased do you think the
product testers are,”“How biased do you think the product testers
were in evaluating the product.”Both items were completed on 7-
point scales (1 =Not at all biased, 7= Extremely biased) and
combined to form an aggregate measure of bias (α=.60).
Source credibility. To assess source credibility, participants
were asked, “How credible are the product testers,”“How much
expertise do the product testers have,”and “How trustworthy do
you think the product testers are.”Items were all completed on
7-point scales (1 = not at all, 7 = extremely) and combined to
form an aggregate measure (α= .90).
Results
Manipulation check
As anticipated, two-sided framing led to greater perceptions
of both sides being considered relative to the one-sided framing,
F(1, 27) = 6.05, p= .02.
Fig. 1. Experiment 2 message framing by self-reported category knowledge on
attitude certainty.
142 D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
Attitudes
In line with the previous studies, participants' attitudes did
not differ as a function of receiving one (M= 5.69, SD = 1.07) or
two-sided framing (M= 5.67, SD = 1.10), Fb1, ns.
Attitude certainty
Replicating prior experiments, there was a main effect of
framing on attitude certainty such that participants reported greater
certainty from the two-sided framing (M=5.57, SD =1.19) com-
pared to the one-sided framing (M=4.54, SD =1.31), F(1,27) =
4.80, p=.04.
Source bias and credibility
Message frame did not affect perceived bias of the source in the
one-sided framing (M=4.57, SD=1.19) compared to the two-
sided framing (M=5.21, SD=1.01), F(1,27)= 2.37, p=.14. If
anything, the means were in the direction opposite to a perceived
bias account.Similarly, source credibility did not differ in the one-
sided framing (M=3.05, SD = 1.20) compared to the two-sided
framing condition (M=3.35, SD =1.70, Fb1, ns.
Discussion
Experiment 3 found manipulating framing outside of the
content of the message had a similar effect upon certainty. This
provides more direct evidence that the difference in certainty
resulted from the perception both sides had been considered
rather than some other aspect of framing. Furthermore, framing
did not affect the credibility or bias associated with the source.
Experiment 4
Experiment 4 examined the inference we postulated to in-
fluence knowledge and thus certainty. Because two-sided framing
reflects a situation where the message or sources contained within
the message have apparently considered not only the positives
associated with the message, but the potential negatives as well,
people are less likely to think that there are negatives of which
they are unaware. We hypothesized that participants' reasoning
would thus be captured in differential perceptions of knowledge
of negative features, but not knowledge of positive features since
attention to positive information was clear in both conditions.
Our framework suggests that the changes in certainty are not
stemming from differences in participants' message-relevant
thoughts or recall of the message they received, but from infer-
ences about their knowledge tied to information that was not
presented (i.e., how much negative information might there be
about the message topic). Nonetheless, to examine this possibi-
lity, we measured participants' actual message-relevant thoughts
and recall of information as these can both serve as measures of
extent of message elaboration (Petty & Cacioppo, 1986).
Finally, Experiment 4 tested two competing interpretations of
our results. Specifically, one explanation for the present effects is
that message framing operates as a simple cue. A second expla-
nation is that the framing of the message affects certainty in a more
thoughtful way such as by affecting inferences about missing
information. To test these accounts, the present research measured
participants' natural proclivity to engage in thinking assessed via
individual differences in need for cognition (NFC; Cacioppo &
Petty, 1982; Cacioppo, Petty, & Kao, 1984). If framing is used as a
simple cue to certainty, differences in certainty should occur
primarily for those low in NFC; however, if the effects are more
related to careful thinking and meta-cognition about the basis of
one's attitude, the differences in certainty should occur primarily
for those high in NFC (see Briñol, Petty, & Tormala, 2004;
Tormala & Petty, 2004; Tormala, Petty, & Briñol, 2002).
Method
Participants and design
Forty-three undergraduates from a Midwestern university
participated in the experiment in exchange for extra credit in their
courses. Participants were randomly assigned to a one-sided
frame or two-sided frame condition and NFC was assessed.
Procedure
Participants read about a portable DVD player called the
Praxis. Individuals first received a screen that contained a
picture of the DVD player along with a brief description of the
product and a note about its features (e.g., 9ʺLCD screen,
rechargeable battery, and stereo speakers). On subsequent
screens participants received three reviews of the product that
were ostensibly from students who had used the product for a
month. Finally, participants completed measures to assess their
attitudes, attitude certainty, perceptions of missing information,
actual thoughts, message recall, and NFC.
Independent variable
Message framing. In the one-sided frame condition partici-
pants read feedback under the general heading “feedback”from
three individuals who were very favorable towards the product.
For example, participants read feedback from one user who
said, “I really love this portable, with the jog circular pad, it
makes finding any portion of the DVD movie so easy.”
The remaining user feedback was also positive and
commented on aspects of the product such as the size and
clarity of the screen and the high quality of the battery life. In
the two-sided frame condition, participants received the same
positive feedback under the heading “positive feedback”but
also had a heading called “negative feedback.”The negative
feedback consisted only of a brief comment indicating the
person had no negative responses (e.g., “I have none.”). Thus,
the actual arguments for liking the product were held constant.
Need for cognition (NFC). Participants' NFC was assessed
using the short form of the NFC scale (α= .87; Cacioppo et al.,
1984). Scores on the scale ranged from 43 to 85, with a median
of 61.
Dependent variables
Attitudes. Attitudes were assessed as in prior experiments
(α= .94).
143D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
Attitude certainty. Attitude certainty consisted of the same
items as in prior studies (α= .82).
Perceptions of missing information. To assess participants'
perceptions of information not presented they were asked, “How
many negative features do you think there are about this DVD
player that were not mentioned?”and, “How many positive
features do you think there are about this DVD player that were
not mentioned? Both items were assessed on 7-point scales
(1 = Few negative/pos itive features, 7 = Many negativ e/positive
features). As expected, these items proved to be independent (r=
−.16, p= .30) and thus were analyzed separately.
Thoughts. Participants were asked to list the thoughts they had
about the product (see Cacioppo & Petty, 1981 for details on the
thought-listing procedure; see also Batra & Ray, 1986 and
Wright, 1973 for illustrative examples). Message-relevant
thoughts were coded as favorable or unfavorable by two
independent coders. Interrater agreement was above 90% and
disagreements were resolved by discussion. Finally, a valenced
thought index was created by subtracting the number of
unfavorable message-related thoughts from the number of
favorable message-related thoughts and dividing this difference
by the total number of message-related thoughts (see Briñol et
al., 2004). Participants' thoughts were also coded for references
to the source, potential missing information, the time the source
spent developing arguments, the source's credibility, and general
meta-cognition (e.g., “I wonder if this thought is correct?”).
Recall. Participants were instructed to recall as many of the
product features as they could remember. Specifically, partici-
pants were given seven boxes for listing any attributes of the
product contained in the product description or mentioned by
the users. Participants' total number of unique features recalled
was used to index recall.
Results
Given the continuous nature of NFC, results were analyzed
using regression procedures as discussed in Experiment 2.
Attitudes
There was no effect of message framing on attitudes, β= .07,
t(39) = .41, p= .68, nor was there an effect of NFC on attitudes,
β=.11; t(39)= .71, p= .48, or a message frame X NFC
interaction, β= .10; t(39) = .58, p= .57. Participants had gen-
erally positive attitudes in both the one-sided (M=5.79,
SD = .68) and two-sided (M= 5.92, SD= .93) condition s.
Attitude certainty
Replicating our previous experiments, there was a main
effect of message framing on attitude certainty, β= .35, t(39) =
2.39, p= .02, such that participants report ed greater certainty
when the message professed to portray both the positives and
the negatives. Importantly, a reliable message frame by NFC
interaction also emerged, β= .30, t(39) = 2.06, pb.05. As seen
in the top panel of Fig. 2, the difference in attitude certainty was
pronounced among those high in NFC, t(39) = 3.20, p= .003,
but not those low in NFC, t(39) = .27, p= .79.
Thoughts
There were no differences in participants' message-relevant
thoughts towards the product as a function of condition as
assessed via the thought favorability index, t(39) = .62, p= .54.
Participants had a positive thought index whether exposed to the
one-sided (M= .53,SD= .66) or two-sided frame (M= .67,
SD = .53), indicating generally favorable thoughts across
conditions. There were also no differences in participants'
total number of message-relevant thoughts, t(39) = 1.40, p= .17.
Furthermore, NFC did not interact with message framing to
influence the thought index or number of thoughts (p's N.25).
We also found no difference in participants' thoughts on any of
the dimensions examined (all p's N.10).
Recall
There were no differences in participants recall as a function
of message framing, NFC, or a message framing X NFC
interaction (p's N.20).
Perceptions of missing information
When it came to participants' perceptions of the amount of
missing positive information they did not receive there were no
effects of message frame, NFC, or a message framing X NFC
interaction (p‘sN.19). However, when it came to participant's
perception of the amount of missing negative information there was
a main effect of message framing such that participants reported
perceiving fewer negative pieces of missing information when they
received the two-sided framed message, β=−.32, t(39) = 2.10,
p= .04. There was no effect of NFC on participants' perception of
missing negative information, β=−.01, t(39) = .06, p= .96.
Fig. 2. Experiment 4 framing by NFC interaction on attitude certainty (Top
panel) and unmentioned negatives (Bottom panel).
144 D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
Importantly, there was a reliable message framing by NFC
interaction, β=−.33, t(39)= 2.13, p= .04. As seen in the bottom
panel of Fig. 2, high NFC individuals (i.e., +1 SD on the NFC scale)
perceived there to be a smaller amount of undisclosed negative
information available in the two-sided compared to the one-sided
condition, t(39) = 3.06, p= .004, but low NFC (i.e., −1SDonthe
NFC scale) individuals reported no differences, t(39) = .01, p=.99.
Mediated moderation analysis
We examined whether the observed differences in certainty
were mediated by the differences in the perceived amount of
negative information missing about the product. Given the
presence of moderation in the form of NFC, we conducted a
mediated moderation analysis as suggested by Baron and
Kenny (1986) using software provided by Preacher and Hayes
(2004). We have already shown there to be a significant direct
effect of the interaction (message framing x NFC) on both the
dependent measure (attitude certainty) and the proposed
mediator (perceptions of missing negative information). The
results of the simultaneous regression revealed that, controlling
for the mediator, the effect between the independent variable
(message framing x NFC) and the dependent variable (attitude
certainty) was no longer significant, β=.18, t(38) = 1.24,
p= .22. However, the relationship between the mediator
(undisclosed negative information) and the dependent variable
was, β=−.38, t(38) = 2.65, p= .01.
Finally, to conduct a formal test of the significance of the
indirect effect, the path through the mediator, we applied
Bootstrap procedures suggested by Shrout and Bolger (2002)
and used applications provided by Preacher and Hayes (2004).
Bootstrapping computes a confidence interval around the
indirect effect, and if zero falls outside of that interval the
indirect effect can be said to be significant. We used a 95%
confidence interval and found zero did indeed fall outside of the
provided range (95% CI = .003 to .089). Thus, the mediated
moderation analysis depicted in Fig. 3 was successful.
Discussion
Experiment 4 provided evidence that framing had its effect
by a relatively high thought process as only those high in NFC
showed the effect. Experiment 4 also provided evidence of
mediation in the form of participants' thoughts about undi-
sclosed negative information. Because perceptions of missing
positive information did not differ, this suggests our mediator
was uniquely sensitive to the specific inference about negative
information. Also, because participants' actual thoughts did not
differ as a function of message condition, differential elabora-
tion seems an unlikely alternative explanation for our results.
2
It
is important to note that although meta-cognition tends to
operate under high thinking conditions, this does not mean that
meta-cognitive factors are necessarily salient in people's
thoughts or the exact meta-cognitive processes relied upon are
explicitly articulated. For example, work on the ease of retrieval
effect (Tormala et al., 2002) suggests that examining partici-
pants' thoughts does not reveal an explicit mention or
discussion of ease or difficulty. However, the psychological
process of ease can be detected by asking participants direct
questions. Similarly, people's thought did not contain specific
mentions about the framing of the message, or missing
information, but this mechanism can be detected through direct
inquiry, as done in the present experiment.
Finally, some readers might view the present findings as
conflicting with those of Experiment 2. Specifically, given
amount of thinking is governed by both one's motivation (e.g.,
need for cognition) and ability (e.g., knowledge) to think (Petty
& Cacioppo, 1986), one might argue that having low knowledge
could reduce one's ability to process the message and foster low
elaboration conditions. Thus, at first blush, it might seem that the
effects should have been observed for high knowledge
individuals rather than low knowledge individuals in Experi-
ment 2. We submit, however, that Experiment 2 used a message
that we intentionally designed to be comprehensible by low
knowledge participants. Indeed, the fact that all participants
reported equally positive attitudes is partial evidence they were
all able to process the message. Thus, given all participants were
able to comprehend the message, the question is who is most in
need of relying on others to ascertain the certainty with which
they can place in their evaluation. According to our perspective it
should be those who cannot derive knowledge from themselves
(i.e., those low in knowledge). To better understand our
explanation, assuming a message can be understood by all, we
suspect those low in knowledge should be more likely to rely on
others to ascertain their knowledge and thus their certainty;
furthermore, this outcome should be more likely when those
with low knowledge are motivated to process the message
carefully. Future research could address this issue by examining
if our effects are most likely to occur for those low in knowledge
but high in NFC (or under conditions of situationally induced
elaboration).
Experiment 5
One consequence of increasing attitude certainty is to increase
attitude-behavior correspondence (Berger, 1992; Krishnan &
Fig. 3. Mediated moderation analysis experiment 4.
2
It is worth noting that prior research suggests that low and high NFC
individuals might sometimes display different thought patterns. However, in the
present research participants were all told to treat the task seriously. Apparently
this was sufficient to encourage equivalent processing, but not sufficient to
motivate low NFC individuals to engage in the higher level of processing
required for meta-cognitive activity.
145D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
Smith, 1998; Rucker & Petty, 2004). That is, if individuals hold
their attitudes with greater certainty, those attitudes are more
likely to guide and influence behavior. Experiment 5 was
designed to demonstrate that two-sided framing can increase
attitude-behavior correspondence over one-sided framing.
Importantly, in studying the influence of attitude certainty on
behavior (or behavioral intentions), what is crucial is the attitude-
behavior correspondence rather than mean behavior/intentions. For
example, if two consumers have equally positive attitudes towards
a product, but one consumer is more certain of his attitude, this
more certain consumer should be more inclined to act on this atti-
tude (i.e., make a purchase). This would lead to both greater
attitude-behavior correlations and greater purchase behavior by the
consumer certain of his evaluation. However, if the product is on
sale, consumers who are relatively uncertain of their attitude might
purchase the product as the sale serves as a catalyst for purchase. In
this situation, behavior might be equivalent (i.e., purchase beha-
vior) for those who are certain and uncertain of their attitudes, but
for different reasons (i.e., reliance on one's attitude versus the pro-
motion). In short, the more diagnostic test of the impact of attitude
certainty on behavior/intentions is the strength of the correlation
between attitudes and behavior rather than mean behavior/
intentions.
Finally, we did not measure certainty in the present experiment.
This allowed us to examine whether two-sided message framing
could have consequences without making certainty salient by
measuring it. If behavioral consequences depended on explicitly
alerting participants' to their certainty, the practical implications of
our findings would be more limited.
Method
Participants and design
Forty undergraduates at a Midwestern university participated
in return for partial course credit. Participants were randomly
assigned to receive a one or a two-sided message frame.
Procedure
Participants were exposed to a one or two sided- message for
a pain reliever, Kedinol-PR. We assessed participants' beha-
vioral intentions in addition to attitudes. Certainty was not
assessed to demonstrate that two-sided framing could have
consequences without explicitly forcing participants to consider
their certainty.
Dependent variables
Attitudes. Attitudes were assessed using the same 3-items as
in prior studies (α= .94).
Behavioral intentions. Participants were asked what they
intended to do when the product became available. Participants'
were asked, “Will you purchase Kedinol-PR,”and “Will you
use Kedinol-PR?”Both items were completed on 7-point scales
with 1 = “Definitely Will Not”and 7 = “Definitely Will.”These
items were combined to form a measure of behavioral intent
(α= .98). Behavioral intention items such as these are the single
best predictors of actual behavior (Fishbein & Ajzen, 1981;
Sheppard, Hartwick, & Warshaw, 1988).
Results
Attitudes
Results were analyzed with ANOVA. As in prior studies, there
were no differences in attitudes between one-sided (M=5.91,
SD =1.01) and two-sided frames (M=6.21, SD=.83; Fb1).
Behavioral intentions
Individuals receiving the two-sided frame reported a greater
willingness to purchase the product than those receiving the one-
sided frame. That is, individuals who formed favorable attitudes
as a result of the two-sided frame were more likely to report
intentions to behave in a favorable manner toward the product
(M= 5.60, SD = 1.23) than were individuals who formed equally
favorable attitudes as a result of a one-sided frame (M= 4.63,
SD = 1.38; F(1, 38) = 5.44, p= .03). More importantly to the
hypothesis of interest, an examination of the attitude-intention
correlations revealed attitudes were more predictive of beha-
vioral intentions following a two-sided frame (r= .74, pb.01)
compared to the one-sided frame (r= .27, p= .27). Furthermore,
a comparison of the correlations revealed a reliable difference in
magnitude (z= 1.98, pb.05).
Discussion
Experiment 5 demonstrated that attitudes following a two-
sided frame were more predictive of behavioral intentions than
attitudes following a one-sided frame. In addition, individuals
who received a two-sided frame were more inclined to report
being likely to act in accordance with their favorable attitudes
(i.e., more likely to purchase the product and stronger attitude-
behavior correspondence). This provides additional support for
our hypothesized effects of message framing, and suggests that
even without alerting consumers to their certainty, there are
indeed observable and practical consequences of this sort of
message framing for persuasion.
General discussion
The current research suggests a new strategy to increase
attitude certainty: framing messages as two sided. Across five
experiments, explicitly highlighting that both sides of an issue
were considered by others led to attitudes held with greater
certainty (Experiments 1–4) and attitudes that were more likely
to influence behavioral intentions (Experiment 5). Individuals
apparently do not make the inference that two sides were con-
sidered in the absence of an explicit statement. In accordance
with our hypothesis, these observed differences did not appear to
result from any difference in consumers' actual thoughts about
the product itself, but stemmed from consumers' inferences
regarding possessing greater knowledge about information not
directly provided (e.g., amount of negatives). Finally, modera-
tion of these effects was observed via both category knowledge
and individuals' propensity to engage in thinking (i.e., NFC).
146 D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
Implications for meta-cognition in persuasion
Prior research on meta-cognitive processes in persuasion has
focused on people's perceptions of their own thoughts or
thought processes (Alba & Hutchinson, 2000; Briñol et al.,
2004; Petty, Briñol, & Tormala, 2002; Rucker & Petty, 2004;
Wright, 2002). In contrast, in the current work we show that
consumers are sensitive to the type of information that has been
considered by others. This extends prior work and theorizing on
meta-cognition by demonstrating meta-cognitions (such as the
certainty with which one holds an attitude) do not seem to stem
only from a reflection of how a message is processed by oneself,
but can also be invoked by information stemming from the
thoughts of other people. Subsequent research could examine
other types of inferences tied to how consumers believe a source
arrived at a message. Such inferences might include whether a
source invested a little or a lot of thinking into considering the
message and whether the source is confident in the position
being presented. We think this offers a novel framework for
understanding source effects in consumer persuasion.
The present research might also hold implications for specific
work on meta-cognition, discussed earlier, by Rucker and Petty
(2004). Specifically, these researchers found individuals
instructed to process a message by considering the faults of the
message were more certain in their attitudes than individuals
instructed to process the same message by considering the
thoughts that naturally came to mind. In that research, the
message contained very strong arguments so that individuals
would be unsuccessful in finding fault. Unclear from that prior
work is whether the increases in certainty stemmed from
consumers' awareness that they had considered both sides or
from the complete inability to find fault. Based on the present
research, we believe it is likely the case that directing consumers
to find fault increases attitude certainty, at least in part, because of
the perception one has considered both sides (and thus there is
little missing information). This suggests the effects observed by
Rucker and Petty might be applicable in a wide variety of
situations, not solely those where individuals are unable to find
fault. Of course, this does not mean that the absence of negatives
did not further contribute to increases in certainty in that research,
only that this is unlikely to be required.
Implications for classic research on one-sided versus two-sided
communications
The present research can be used as a springboard for revisiting
classic research on one-sided versus two-sided communication
(e.g., Kamins & Marks, 1987; Hovland et al., 1953; Lumsdaine &
Janis, 1953; Sawyer, 1973). For example, when one-sided and two-
sided communications yield equivalent attitude change, there might
still be significant differences in the certainty of the underlying
attitudes. Two-sided communications might lead people to feel
their attitudes are based on more complete knowledge, and thus
produce greater certainty in their attitudes. This, as in the present
research, could have implications for behavior/intentions.
In fact, the present results and theorizing might well
explain some classic research on two-sided communications.
Lumsdaine and Janis (1953) found that people were less likely
to change their attitudes to a subsequent attack if they had
initially been given a two-sided message. It seems plausible
that individuals who received the initial two-sided message,
even after controlling for differences in actual knowledge, may
have perceived their attitudes to be based on greater knowl-
edge since the source had given them both sides. This in turn
would lead to greater attitude certainty as in the present
research. Certainty could then explain the differential resis-
tance of these attitudes to a subsequent message as certainty
has been postulated to give rise to defensive processing
(Gross, Holtz, & Miller, 1995) and to prevent attitude change
(Tormala & Petty, 2002). This is an interesting and potentially
fruitful possibility for future research.
Attitudes versus attitude certainty
The present research also speaks to the importance of
recognizing the distinction between attitude certainty and attitude
extremity. Collapsing across Experiments 1–4, as predicted,
there was a reliable effect of message framing on attitude
certainty, F(1, 282) =17.02, pb.001, but no significant effect on
attitudes, F(1, 282)= 2.33, p=.13. Furthermore, the effects of
certainty remained significant even after controlling for attitudes,
F(1, 281) = 14.53, pb.001, whereas attitudes remained non-
significant, F(1, 281) = .007, p= .93. Several past research
findings have documented the distinctiveness of the constructs
(e.g., Rucker & Petty, 2004; Tormala & Petty, 2002), and earlier
we reviewed reasons why this would be likely in the current work.
However, future research could more fully examine whether such
framing effects might sometimes affect the extremity of an
evaluation. For example, perhaps if framing alerted people to
negative or positive aspects of a message they normally would not
have paid attention to (e.g., when elaboration is low and they are
unmotivated to read information carefully), framing might
influence how positive or negative participants' evaluations are.
Similarly, if a message is more ambiguous, directing people's
attention to negative information might influence extremity.
These represent interesting directions for future research.
Practical implications
In closing, the present research offers some practical
implications as well. First, in cases where a product largely
has positive information associated with it and few negative
pieces of information, the present research suggests it would be
worth drawing consumers' attention to the fact that potential
negatives were considered in developing the message. For
example, advertisements employing testimonials might include
clips not only of what people enjoyed about the product, but also
highlight that these consumers found little they disliked as well.
Second, the present research has implications for the structure of
websites offering consumer feedback. Providing user feedback
that clearly outlines both the positives and the negatives
associated with the product might lead consumers to hold their
evaluations with greater certainty, which might facilitate their
decision process. Finally, even when ads foster attitudes of
147D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
equivalent valence, additional insight into the effectiveness of
ads can be gauged by also including measures of attitude
certainty. Ultimately, our work suggests that attitude certainty
can serve as a measure of hidden effects of persuasion.
Acknowledgments
The authors would like to thank Nidhi Agrawal and Brian
Sternthal as well as the members of the 2004–2005 Group for
Attitudes and Persuasion at Ohio State University and members
of the 2005–2006 Kellogg on Attitudes, Motivation, and
Processing group at Northwestern University for helpful
comments on this research. This research was supported in
part by funding from the National Institute of Mental Health
individual grant no. 5F31MH12849-03, a Kraft Research
Professorship, and an honorable mention award from the
Society for Consumer Psychology's Sheth dissertation proposal
competition, to the first author.
Appendix A
Experiment 2 Sample Messages: one-sided (top) and two-sided (bottom) message frames
148 D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
References
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting
interactions. Newbury Park, CA: Sage.
Alba, J. W., & Hutchinson, W. (2000). Knowledge calibration: What consumers
know and what they think they know. Journal of Consumer Research,22,
62−74.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction
in social psychological research: Conceptual, strategic, and statistical
considerations. Journal of Personality and Social Psychology,52, 1173−1182.
Bassili, J. N. (1996). Meta-judgmental versus operative indexes of psycholo-
gical attributes: The case of measures of attitude strength. Journal of
Personality and Social Psychology,71, 637−653.
Batra, R., & Ray, M. L. (1986). Affective responses mediating acceptance of
advertising. Journal of Consumer Research,13, 234−249.
Berger, I. E. (1992). The nature of attitude accessibility and attitude confidence: A
triangulated experiment. Journal of Consumer Psychology,1, 103−123.
Bohner, G., Einwiller, S., Erb, H. P., & Sibler, F. (2003). When small means
comfortable: Relations between product attributes in two-sided advertising.
Journal of Consumer Psychology,13, 454−463.
Briñol, P., Petty, R. E., & Tormala, Z. L. (2004). Self-validation of cognitive
responses to advertisements. Journal of Consumer Research,30, 559−573.
Cacioppo, J. T., & Petty, R. E. (1981). Social psychological procedures for
cognitive response assessment: The thought listing technique. In T. V. Merluzzi,
C. R. Glass, & M. Genest (Eds.), Cognitive Assessment (pp. 309−342).
New York: Guilford.
Cacioppo, J. T., & Petty, R. E. (1982). The need for cognition. Journal of
Personality and Social Psychology,42,116−131.
Cacioppo, J. T., Petty, R. E., & Kao, C. F. (1984). The efficient assessment of
“need for cognition”.Journal of Personality Assessment,48, 306−307.
Fazio, R. H., & Zanna, M. P. (1978). Attitudinal qualities relating to the strength
of the attitude-behavior relationship. Journal of Experimental Social
Psychology,14, 398−408.
Fishbein, M., & Ajzen, I. (1981). Acceptance, yielding, and impact: Cognitive
processes in persuasion. In R. E. Petty, T. M. Ostrom, & T. C. Brock (Eds.),
Cognitive responses in persuasion (pp. 339−359). Hillsdale, NJ: Erlbaum.
Gross, S., Holtz, R., & Miller, N. (1995). Attitude certainty. In R. E. Petty, &
J. A. Krosnick (Eds.), Attitude Strength Antecedents and Consequences
(pp. 215−245). Mahwah, NJ: Erlbaum.
Hovland, C. I., Janis, I. L., & Kelley, H. H. (1953). Communication and
Persuasion. New Haven, CT: Yale University Press.
Kamins, M. A., & Marks, L. J. (1987). Advertising puffery: The impact of using
two-sided claims on product attitude and purchase intention. Journal of
Advertising,16,6−15.
Krishnan, H. S., & Smith, R. E. (1998). The relative endurance of attitudes,
confidence and attitude-behavior consistency: The role of information
source and delay. Journal of Consumer Psychology,7, 273−298.
Krosnick, J. A., Boninger, D., Chuang, Y., Berent, M., & Carnot, C. (1993).
Attitude strength: One construct or many related constructs. Journal of
Personality and Social Psychology,65, 1132−1151.
Lee, A. Y. (2004). The Prevalence of Meta-Cognitive Routes to Judgment.
Journal of Consumer Psychology,14(4), 345−355.
Lumsdaine, A. A., & Janis, I. L. (1953). Resistance to “counterpropaganda”
produced by one-sided and two-sided “propaganda”presentations. Public
Opinion Quarterly,17,311−318.
Pechmann, C. (1992). Predicting when two-sided ads will be more effective than
one-sided ads: The role of correlational and correspondent inferences.
Journal of Marketing Research,29, 441−453.
Petrocelli, J. V., Tormala, Z. L., & Rucker, D. D. (2007). Unpacking attitude
certainty: Attitude clarity and attitude correctness. Journal of Personality
and Social Psychology,92,30−41.
Petty, R. E., & Cacioppo, J. T. (1984). The effects of involvement on response to
argument quantity and quality: Central and peripheral routes to persuasion.
Journal of Personality and Social Psychology,46,69−81.
Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of
persuasion. In Leonard Berkowitz (Ed.), Advances in experimental social
psychology (vol. 19, pp. 123–205). New York: Academic Press.
Petty, R. E., Briñol, P., & Tormala, Z. L. (2002). Thought confidence as a
determinant of persuasion: The self-validation hypothesis. Journal of
Personality and Social Psychology,82, 722−741.
Petty, R. E., Briñol, P., Tormala, Z. L., & Wegener, D. T. (2007). The role of
meta-cognition in social judgment. In E. T. Higgins, & A. W. Kruglanski
(Eds.), Social psychology: A handbook of basic principles (2nd ed.,
pp. 254–284). New York: Guilford Press.
Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating
indirect effects in simple mediation models. Behavior Research Methods,
Instruments, & Computers,36, 717−731.
Priester, J. R., Petty, R. E., & Park, K. (2007). Whence univalent ambivalence?
From the anticipation of conflicting reactions. Journal of Consumer
Research,34,11−21.
Rucker, D. D., & Petty, R. E. (2004). When resistance is futile: Consequences of
failed counterarguing for attitude certainty. Journal of Personality and
Social Psychology,86, 219−235.
Sanbonmatsu, D. M., Kardes, F. R., & Herr, P. M. (1992). The role of prior
knowledge and missing information in multiattribute evaluation. Organi-
zational Behavior and Human Decision Processes,51,76−91.
Sawyer, A. G. (1973). The effects of repetition of refutational and supportive
advertising appeals. Journal of Marketing Research,10,23−33.
Schwarz, N. (2004). Metacognitive experiences in consumer judgment and.
decision making. Journal of Consumer Psychology,14, 332−348.
Sheppard, B. H., Hartwick, J., & Warshaw, P. R. (1988). The theory of reasoned
action: A meta-analysis of past research with recommendations for
modifications and future research. Journal of Consumer Research,15,
325−343.
Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperi-
mental studies: New procedures and recommendations. Psychological
Methods,7, 422−445.
Tormala, Z. L., & Petty, R. E. (2002). What doesn't kill me makes me stronger:
The effects of resisting persuasion on attitude certainty. Journal of
Personality and Social Psychology,83, 1298−1313.
Tormala, Z. L., & Petty, R. E. (2004). Source credibility and attitude certainty: A
meta-cognitive analysis of resistance to persuasion. Journal of Consumer
Psychology,14, 426−441.
Tormala, Z. L., Petty, R. E., & Briñol, P. (2002). Ease of retrieval effects in
persuasion: A self-validation analysis. Personality and Social Psychology
Bulletin,28, 1700−1712.
Wright, P. (1973). The cognitive processes mediating acceptance of advertising.
Journal of Marketing Research,10,53−62.
Wright, P. (2002). Marketplace meta-cognition and social intelligence. Journal
of Consumer Research,28, 677−682.
149D.D. Rucker et al. / Journal of Consumer Psychology 18 (2008) 137–149
