How people can become persuaded by weak messages presented bycredible communicators: Not all sleeper effects are created equal

Full text

How people can become persuaded by weak messages presented by
credible communicators: Not all sleeper effects are created equal☆
Dolores Albarracín
a
, G. Tarcan Kumkale
b,
⁎, Patrick Poyner-Del Vento
c
a
University of Illinois at Urbana Champaign, Department of Psychology, 603 E. Daniel St, Champaign, IL 61822, USA
b
Kadir Has University, Department of Psychology, Cibali, Istanbul, Turkey
c
Simon Fraser University, British Columbia, Canada
HIGHLIGHTS
•Introduces a new kind of sleeper effect never considered before; a sleeper effect for the source
•Identifies conditions under which the new sleeper effect takes place
•Identifies reasons for the difficulty in generating the traditional sleeper effect for the arguments
•Presents a framework in which both types of effects can be generated
abstractarticle info
Article history:
Received 25 October 2015
Revised 2 4 June 201 6
Accepted 29 June 2016
Available online 01 July 2016
The sleeper effect has been proposed to describe temporal changes in persuasion for messages associated with
noncredible sources. The present research introduces a new kind of sleeper effect denoting increases in persua-
sion for weak messages associated with credible sources. This effect of the source was hypothesized to derive
from attending to the message source rather than the message arguments and reconstructing delayed attitudes
primarilyon the basis of the source information. Findings from threeexperiments revealed that when the focusof
attention was the communicator, there was a sleeper effect for the source. Specifically, during the time between
an immediate follow up and a delayed follow up, persuasion increased when credible sources presented weak
arguments. In contrast, when the focus of attention was the message arguments, a traditional sleeper effect
emerged. Thatis, persuasion increased when strong arguments were presented by a noncredible communicator.
These effects were mediated by relative recall ofarguments versus source attributes and replicated withdifferent
message topics and lengths of delay.
© 2016 Published by Elsevier Inc.
Keywords:
Sleeper effect
Attitude stability
Persuasion
Persistence
Memory
Attitude change
1. Introduction
More than two millennia ago Aristotle outlined the ingredients of ef-
fective communications in his book The Artof Rhetoric. According to him,
one of the most important ingredients for successful persuasion is that
the communicator must becredible. Equally important is thatthe argu-
ments must be strong and based on the rules of logic. Hundreds of
studies indeed confirm that credible sources and strong arguments are
essential for persuading an audience (Johnson, Maio & Smith-
McLallen, 2005). What is less clear, however, is how durable the influ-
ence of credible communicators is and when this influence takes
place. Filling this gap in knowledge is essential to better understand
the dynamics of persuasion and social influence (see Prislin & Wood,
2005).
The sleeper effect is an interesting effect identified in relation to the
durability of persuasive influences (Hovland, Lumsdaine, & Sheffield,
1949) and denotes a delayed increase in theimpact of a persuasivemes-
sage. Traditionally, it has been considered as a possibility for initially
discounted communications. For instance, when the credibility of a
message source induces suspicions of invalidity, little or no persuasion
takes place immediately. Over time, however, if the reason for
discounting the messages becomes less accessible in memory, a sleeper
effect can take place assuming an otherwise strong message. This kind
of a sleeper effect could be reliably observed in the past under certain
circumstance (for reviews, Cook & Flay, 1978; Gruder et al., 1978;
Journal of Experimental Social Psychology 68 (2017) 171–180
☆This workwas facilitated bygrants K01 MH01861, R01NR08325, R01 MH094241,R56
AI114501and K02 MH075616 awardedto Dolores Albarracin, and by the Turkish National
Academy of Sciences GEBIP award to Tarcan Kumkale. The first two experiments
constituted parts of the doctoral dissertation research conducted by G. Tarcan Kumkale
under the direction of Dolores Albarracin. We would like to th ank Rick Brown, Pete r
Delaney, Ian Handley, Rolf Holtz, Benjam in Karney, Ece Kumkal e, Alan Sawyer, Barry
Schlenker, and Vincent Yzerbyt for unusually insightful suggestions concerning this
research.
⁎Corresponding author.
E-mail addresses: dalbarra@illinois.edu (D. Albarracín), tarcan.kumkale@khas.edu.tr
(G.T. Kumkale).
http://dx.doi.org/10.1016/j.jesp.2016.06.009
0022-1031/© 2016 Published by Elsevier Inc.
Contents lists available at ScienceDirect
Journal of Experimental Social Psychology
journal homepage: www.elsevier.com/locate/jesp

Kumkale & Albarracín, 2004). As we will argue, however, traditional
conceptualizations of the sleeper effect has been restricted and incom-
plete. For instance, the possibility of a sleeper effectguided by a delayed
increase in theimpact of source attributes has never been considered. In
the present research, we demonstrate this possibility that we term the
sleeper effect for the source.
Just as a message can be discounted for its association with a
noncredible source, it can be discounted for containing weak arguments
as well. Thus, when a credible source presents such weak arguments, a
source-based sleeper effect can be observed if the influence of the cred-
ible source remains and becomes apparent later in time. As we will
argue in the following sections, this dormant influence of the source
may surface more readily if the recipients attend to thesource attributes
(rather than the arguments) at the time of message exposure or when
the relevant information is retrieved to make a delayed judgment.
Our primary underlying assumption is that, different aspects of the
communication are available when the communication is first received
relative to later, when delayed attitudes are reported. Delayed differ-
ences in accessibility of the communication information are likely to de-
pend on the degree to which recipients attend to the communication
source vs. the arguments, and differences in attentional focus should
in turn produce different sleeper effects. This notion shares much with
the differential decay hypothesis in recognizing that some aspects of the
communication can be less memorable (Pratkanis, Greenwald, Leippe,
& Baumgardner, 1988) than others but proposes no memorability ad-
vantage for the message arguments. We simply apply general consider-
ations about online construction of attitudes and the influence of
information accessibility (Schwarz & Bohner, 2001; Wyer, 2007)to
both the message arguments and the source of a communication.
In the present research, we systematically manipulated recipients'
focus of attention on the communication source vs. the arguments,
both at the time of exposure and retrieval. These manipulations along
with systematic variations in not only the credibility of the source but
also the strength of the arguments in the message permitted testing
for both the traditional argument-based sleeper effect as well as the
source-based sleeper effect that we introduce in this research.
1.1. Sleeper effect for the source
A source-based sleeper effect could not be identified in the past, be-
cause predictions about the sleeper effect have been based on the pre-
mise that the influence of the message source decays more than the
influence of the message arguments (e.g., Hovland & Weiss, 1951;
Hovland et al., 1949). We reasoned that, there was no a priori reason
to take this assumption for granted. For example, the personal charac-
teristics of political candidates can be more important and memorable
than the issues discussed in political campaigns (for reviews, see
Iyengar & Ottati, 1994; Lodge, McGraw, & Stroh, 1989; Ottati, 2001;
Wyer & Ottati, 1993). If asked today, most middle-aged Americans
would probably remember that Michael Dukakis was a longtime gover-
nor of Massachusetts, but very few would remember his arguments on
prison reform. Thus, the attributes of the source may remain more
memorable than the arguments at times (e.g., Erb, Pierro, Manetti,
Spiegel, & Kruglanski, 2007; Kruglanski, Fishbach, Erb, Pierro, &
Mannetti, 2004; Kruglanski, Pierro, Manetti, Erb, & Spiegel, 2006;
Pierro, Manetti, Erb, Spiegel, & Kruglanski, 2005). Past research has
paved the way to this consideration by showing that the sleeper effect
disappeared when the information source was made memorable (i.e.,
making it salient at the delayed testing; Kelman & Hovland, 1953;orre-
peating who the source is several times, Weber, 1972).
Weak arguments are commonly used in everyday communication
contexts. In these situations, the initial influence of a credible communi-
cator may decrease with the realization that the arguments are weak
(i.e., reason for discounting). As a result, there may be no perceptible
initial attitude change when acredible communicator presents weak ar-
guments.As the reason for discounting (i.e., weak arguments) becomes
less accessible, however, a source-based sleeper effect can take place
when the influence of the credible communicator becomes more appar-
ent. Perhaps because none of the earlier sleeper effect studies examined
the influence of argument strength (for a review, see Kumkale &
Albarracín, 2004), the possibility for a source-based sleeper effect
could not be identified in the past. Consequently, earlier conceptualiza-
tions of the sleeper effect revolved around the argument-based sleeper
effect, which relied on influential, and by contemporary terms strong ar-
guments presented by noncredible sources.
1.2. Focusing attention on the source or the arguments
The sleeper effect takes place because people rarely retrieve and
consider everything they know about an issue. Often times, they consid-
er only a subset of the information as a way of efficiently satisfying their
cognitive goals (Albarracín,2002; Feldman & Lynch, 1988; Kruglanski et
al., 2004, 2006; Schwarz & Bohner, 2001; Webb & Sheeran, 2006;
Wilson, Lindsey, & Schooler, 2000; Wyer, 2007). Chaiken, Wood, and
Eagly (1996), for instance, suggested that people whomake a judgment
or decision first apply the criterion that comes to mind most quickly.
As there is no reason to assume that arguments will always receive
more attention than source attributes (Chaiken & Eagly, 1983; Kelman
& Hovland, 1953;alsoseeErb et al., 2007; Kruglanski et al., 2006), find-
ing a novel sleeper effect for the source vs. the traditional sleeper effect
for the arguments is likely to depend on the recipients' focus of atten-
tion at the time of exposure. In particular, the traditional sleeper effect
for the arguments should emerge when, at the time of receiving the
message, people concentrate on the (strong) arguments more than
the noncredible source. Correspondingly, the sleeper effect for the
source should emerge when the focus is on the credible source rather
the (weak) arguments. As the focus of attention is likely to determine
later recall, this focus should make a difference at the delayed follow up.
The principal postulated mechanism is that persuasion at the de-
layed follow up will depend on whether the arguments or the source
are accessible at that time. Moreover, as the initial focus of attention is
likely to determine later recall, this focus of attention will also deter-
mine persuasion at the delayed follow up. Forming a solid impression
of a credible communicator who happens to present weak arguments
should yield increased persuasion over time, as should focusing on
strong arguments that happen to be presented by a noncredible
communicator.
Decades of social cognition research support the idea that represen-
tations can be organized around the attributes of the communicator or
source of a persuasive message (Devine & Ostrom, 1985; Lingle, Geva,
Ostrom, Leippe, & Baumgardner, 1979; Lingle & Ostrom, 1979; Wyer,
2004, 2007; Wyer & Budesheim, 1987). In the context of presidential
elections, for example, recipients may be primarily motivated to form
impressions of the candidates (Iyengar & Ottati, 1994; Lodge et al.,
1989; Ottati & Wyer, 1993; Ottati, Wyer, Deiger, & Houston, 2002;
Wyer & Ottati, 1993). In these conditions, attitude-relevant information
may be organized around the identity of the candidates rather than the
candidates' arguments. For example, the agendas of electoralcandidates
often look so similar that the only difference between candidates is the
person. These conditions may naturally instill a focus on the personal at-
tributes of the candidates rather than their arguments. When the focus
of attention is on the source, the representation of the arguments may
become less accessible over time and result in a sleeper effect for the
source.
If the sleeper effect is likely to vary as a function of focus of attention,
then it is important to identify determinants of focus of attention. One
aspect that is likely to have induced argument focus in past research is
presenting the message arguments before the source of the communi-
cation, which is the only condition under which the traditional effect
has been generated (Kumkale & Albarracín, 2004). Outside of the sleep-
er effect literature, however, the most common manipulation of focus of
attention has been to vary the processing goals of message recipients. In
172 D. Albarracín et al. / Journal of Experimental Social Psychology 68 (2017) 171–180

research on political persuasion (e.g., Lodge et al., 1989), for example,
frequently assigned experimental goals include evaluating the suitabil-
ity of a candidate for public office (a source focus) or the desirability of
the candidate's policy (an argument focus). Another strategy may in-
volve asking questions about selected aspects of the communication
content (Feldman & Lynch, 1988; Fitzsimons & Williams, 2000; Killeya
& Johnson, 1998). For example, political attitudes of an audience can
shift when commentators ask biased questions during the course of a
campaign (Ottati et al., 2002). Other strategies may involve selective
repetition of either information about the source vs. the arguments or
varying the visual vividness of each piece of information. In the present
research, we relied on these strategies in manipulating recipients' focus
of attention.
1.3. Overview of the present research
We tested our hypotheses in three experiments. In Experiments 1
and 2, participants evaluatedthe campaign materials of a fictitious polit-
ical party running for the student government on campus. Each partici-
pant received two ads: One ad described the arguments of the party and
contained either strong or weak arguments; the other adintroduced the
attributes of the candidates running for the office, who were portrayed
as either credible or noncredible. These two ads allowed us to generate
two conditions, one with strong arguments presented by a noncredible
source and the other with weak arguments presented by a credible
source.
Critical to Experiments 1 and 2, the messages were crossed with ma-
nipulations of attentional focus on either the arguments or the source.
We predicted that the course of attitude change, including both types
of sleeper effects would depend on these manipulations—without hav-
ing any discernible effects on immediate attitudes: A source focus of at-
tention should produce a sleeper effect when a credible source
presented weak arguments, whereas an argument focus of attention
should produce a sleeper effect when a noncredible source presented
strong arguments. Each experiment involved a different focus manipu-
lation to assess the robustness of the effect. In Experiment 1, the focus
manipulation was introduced in between the immediate and delayed
follow ups using a directed questioning approach (Feldman & Lynch,
1988; Fitzsimons & Williams, 2000; Killeya & Johnson, 1998). Specifical-
ly, participants were asked several questions about either the argu-
ments or the source attributes (e.g., whether the agenda of the party
sounded useful, relevant or whether the party representatives looked
experienced, credible or trustworthy). In Experiment 2, The focus ma-
nipulation was more direct: Participants were asked to examine either
the argument-ad or the source-ad twice during message exposure,
once for content and another for style. The delay in these two experi-
ments was approximately 45 min as in Pratkanis et al., 1988. Experi-
ment 3 involved a different focus manipulation, a different topic, and a
longer delay between attitude measurements (i.e., two weeks).
1
2. Experiment 1
Experiment 1 tested the idea that the focus of attention during the
message close to the time of presenting the message influences the
type of sleeper effect that is observed. Participants first developed and
expressed an attitude after examining the campaign materials of the po-
litical party. After reporting their attitudes, participants answered addi-
tional questions about their perceptions of either the source or the
arguments. We reasoned that focusing on either the source or the argu-
ments close to the reception of the message would produce differential
recall and dictate which kind of sleeper effect occurs. Following a delay,
participants reported their attitudes one more time.
2.1. Method
2.1.1. Participants
Participants were 72 male and 143 female introductory psychology
students who received course credit for their participation. They were
randomly assigned to the conditions of a 2 × 2 design,in which Informa-
tion Set (strong arguments presented by the noncredible source vs.
weak arguments presented by the credible source) and Focus of Atten-
tion (focus on arguments vs. source). There were between 48 and 62
participants in each condition.
2
2.1.2. Procedures
Participants were introduced to the study with information that the
student government at the university was interested in examining per-
ceptions of the campaign materials of the political parties that would be
running for office the following semester. We added that a better under-
standing of these perceptions could help the student government cut
costs by funding only effective campaign materials. The cover story indi-
cated that the study would also provide invaluable knowledge about
formation of political attitudes. Under this pretense, asking participants
to focus on a selected piece of information (e.g., the piece that conveyed
the arguments vs. the source attributes) became easier. Thus, partici-
pants received the campaign materials of a fictitious political party.
Each participant received two print ads. The first ad described the plat-
form of theparty and contained four arguments that were either strong
or weak. The second ad introduced the attributes of the candidates,
portraying them as either credible or noncredible. A set of questions
right after the immediate attitude measures were introduced to manip-
ulate the attentional focus.
Participants first examined and evaluated the campaign materials of
the party. Then, for about 45 min, they worked on filler measures and
messages presented on the computer. The delayed measures were ob-
tained after the filler tasks as in Pratkanis et al. (1988). The study took
about an hour, and involved up to eight participants in separated
cubicles.
2.1.2.1. Information set. Under the pretense of helping the student gov-
ernment in assessing the campaign materials of the political parties
that wouldbe running for the office next semester, participants received
two print ads of a fictitious political party called the Gator Party. One of
the ads described the goals of the party and contained four arguments
that were either strong or weak. The second ad introduced the attri-
butes of the candidates and portrayed them as either credible or
noncredible.
Given that the sleeper effect becomes relevant when the evaluative
implications of the arguments and the source attributes are inconsistent
with each other (for reviews, see Cook & Flay, 1978; Cook, Gruder,
Hennigan, & Flay, 1979; Kumkale & Albarracín, 2004), the information
set given to the participants involved an inconsistency as follows:
1
The choiceof time intervalbetween messageexposure and delayed posttestshas been
a long-lasting question in persistence research. As notedby Cook and Flay (1978), time is
not a psychological construct; hence, it is important to design studies around processes
that mediate persistence effects rather than time per se, as “the same events can often
be packed into oneyear, one month, or even 1 day”(p.7). Thus, what isdone in an exper-
iment in terms of capturing processes that takeplace in time may be morecritical than the
duration between measurements. Indeed, in their meta-analysis, Kumkale and Albarracin
(2004) did not find evidence for the moderating role of time interval between measure-
ments (p. 163). Sleeper effects have been reported in studies involving delays as short
as an hour (e.g., Pratkanis et al., 1988) as well as in studies involving longer delays such
as a week (e.g., Priester et al., 1999).
2
Although we did not intentionally try to oversample a particular condition, the num-
ber of participantsacross conditionsended up unequalin this study (range= 48–62). This
unevenness could constitute a problem in the analyses if the sample size was small. The
current study, however, was significantly more powerful than the typical sleeper effect
study. The median sample size for the discounting-cue conditions was 27 on average in
the literature (for a review, see Kumkale& Albarracin, 2004; p. 159),with substantial var-
iance across studies (range = 9 to 80). Furthermore, standard deviations were similar
across conditions. Thus, we do not think that this unevenness might have brought about
significant bias in the results.
173D. Albarracín et al. / Journal of Experimental Social Psychology 68 (2017) 171–180

When the arguments received were strong, the associated source was
always noncredible; when the arguments were weak, however, the
source was always credible–representing the conditions related to the
traditional and the novel sleeper effects respectively.
The ad presenting the political platform of the party contained either
four strong arguments or four weak arguments. For example, one of the
strong arguments asserted that the party would pressure the school ad-
ministration to provide financial support for students performing well
academically. Another strong argument asserted that the party would
negotiate an increase in the salary of students involved in work-study
programs. In contrast, one of the weak arguments asserted that the
party would utilize unused tuition fees for student government ban-
quets held each semester. Another weak argument indicated that the
party would install online cameras in the registrar's office so that the
students could see how long the lines were before leaving home.
These arguments were selected from a pool of over 30 pretested ar-
guments; several of which were directly taken from the campaign
materials of actual student-led parties running for the office on cam-
pus. Furthermore, to bolster the cover story, we tried to make the
layout of the ads as similar as possible to the materials used by actual
student-led parties on campus. For instance, we boldfaced and
highlighted the Gator Party name on the top panel of the ad, and
compounded this header with images of alligator heads at the
bottom.
The identical layout (with the header and images) was used to con-
vey source credibility information in a separate print ad. The credible
candidates were introduced as experienced political science and law
students who were interested in pursuing careers in politics after grad-
uation. The noncredible candidates were introduced as open-minded,
part-time students who were undecided majors at the time. Each of
these pilot-tested ads contained four pieces of information, and were
about equalin length (Mwords =70; see supplementary materials on-
line). The validity of the argument strength and source credibility ma-
nipulations was established through extensive pilot testing.
3
The order
of the ads was counterbalanced. This factor had no main effect on any
of the dependent measures, nor did it interact with any of the other var-
iables. Therefore, its effects receive no further attention.
2.1.2.2. Focus of attention. Participants examined each ad twice before
reporting their immediate attitudes. Then, participants in the argu-
ment-focus conditions completed measures of beliefs about the argu-
ments, whereas participants in the source-focus conditions completed
similar measures about the source. Both measures included 12 items
rated along 11-point scales. For example, participants in the source-
focus conditions rated the extent to which the source was experienced,
trustworthy, credible, biased, strong, and influential. Participants in the
argument-focus conditions rated the extent to which the agenda of the
party included useful, relevant, favorable proposals that could serve the
interests of the students. In addition, before the delay, participants in
the argument-focus conditions were asked to recall the arguments,
whereas participants in the source-focus conditions were asked to recall
the attributes of thesource.They were told that another issue of interest
was the memorability of campaign materials. At that point, participants
were asked to recall as much as possible of either the arguments of the
party or the attributes of its candidates. The manipulation of focus was
therefore very close to the point of message reception but appeared
after the immediate attitude measure. Experiment2 produced a replica-
tion with the manipulation being introduced exactly at the point of
message reception.
2.1.3. Dependent measures
Immediate and delayed attitudes were assessed by asking par-
ticipants to rate “voting in favor of the party”along scales ranging
from −5to+5(is something that I would not like vs.like; would be
harmful vs.helpful;wouldbenegativevs.positive; would be a bad
idea vs.good idea). The reliability of the scale was high at each
time point (α
immediate
=0.98;α
delayed
= 0.98). For each time
point, responses were averaged to yield summary indices of
attitudes.
The critical dependent variable in the sleeper effect literature is the
amount of attitude change between the immediate anddelayed posttest
(for a discussion of the right dependent variable, see reviews by Cook &
Flay, 1978; Kumkale & Albarracín, 2004). Thus, an index was created by
subtracting immediate attitudes from delayed attitudes: Positive scores
represented increases in persuasion over time and hence possible abso-
lute sleeper effects, whereas negative scores indicated decay in persua-
sion over time.
2.2. Results
2.2.1. Immediate attitudes
Immediate attitudes did not differ significantly from the neutral
point of the scale, M=0.16(SD = 2.39), t=0.96,p=0.34.This
suggests that the communication was discounted as expected.
Most importantly, neither the main effects nor the interaction be-
tween Information Set and Focus of Attention were significant for
immediate attitudes, all Fsb1.1. Thus, conditions required to ob-
serve a sleeper effect were in place (Cook et al., 1979; Kumkale &
Albarracín, 2004).
2.2.2. Attitude change over time
We expected to observe different sleeper effects depending on the
focus of attention. As expected, the interaction between Information
Set and Focus of Attention on attitude change scores was significant,
F(1, 211) = 25.37, pb0.001. Noneof the main effects approached signif-
icance, Fb1.
As shown by the attitude change means in Table 1, the novel sleeper
effect for the source was verified when the recipients of weak argu-
ments presented by a credible source focused on the source (M
change
=
0.40, CI = 0.04/0.76) rather than the arguments (M
change
=−0.26,
CI = −0.58/0.06); contrast F(1, 211) = 7.31, pb0.01; d= 0.52, CI =
0.14/0.90. In contrast, the traditional sleeper effect for the arguments
was verified when the recipients of strong arguments presented by a
noncredible source focused on the arguments (M
change
= 0.55, CI =
0.20/0.91) rather than the source (M
change
=−0.53, CI = −0.86/−
0.21); contrast F(1, 211) = 19.29, pb0.001; d= 0.86, CI = 0.45/1.25.
Gender did not make a difference on any of these measures, neither in
this study nor in the subsequent experiments–consistent with past re-
search on the sleeper effect.
2.3. Discussion
Experiment 1 provided support for the idea that sleeper effects
could take place when either the source or the arguments are
made salient close to the time of receiving the message. Although
this study supported the notion that the attentional focus moderates
the type of sleeper effect that ensues, immediate attitudes were
measured before the introduction of the focus manipulation. There-
fore, it was desirable to replicate these findings while ensuring that
the immediate attitudes remained unaffected and that the effects
were solely due to how delayed attitudes were reconstructed at
the delayed follow up. This objective was accomplished in Experi-
ment 2.
3
In one of the pilot studies (N= 76), the ads were presented separately to different
groups of people. On a scale ranging from −5 to +5, participants found the strong argu-
ments to be more convincing than the weak arguments (M=2.37,SD =2.36vs.
M=−2.89, SD = 2.44, respectively; d= 2.20, pb0.001). The source credibility manipu-
lation was also successful (M= 3.28, SD = 1.50 vs. M=−0.68, SD = 2.20; d= 2.11,
pb0.001). Furthermore, the credible source was foundto possess more impressive char-
acteristics than the noncredible source, (d= 2.18, pb0.001).
174 D. Albarracín et al. / Journal of Experimental Social Psychology 68 (2017) 171–180

3. Experiment 2
3.1. Method
3.1.1. Participants
Participants were 74 introductory psychology students who re-
ceived course credit in exchange for participation (18 male; 56 female).
They were randomly assigned to the conditions of a 2 × 2 design with
the between-subjects factors Focus of Attention (focus on arguments
vs. source attributes) and Information Set (strong arguments presented
by the noncredible source; weak arguments presented by the credible
source). There were between 17 and 20 participants in each condition.
3.1.2. Procedures
The cover study and the materials of this experiment were identical
to the ones used in Experiment 1. Specifically,participants evaluated the
campaign materials of a fictitious political party running for the student
government on campus. The focus manipulation, however, was more
direct than the one used in Experiment 1. Specifically, to favor recall of
either the arguments or the source, we instructed participants to focus
on either the arguments or the source ad and examine it twice, once
for contentand another for style. Specifically, we told them: “Please ex-
amine this ad twice. First, read the ad for its content, then the second
time you read, please attend to how it is written, and underline the
key words in each part of the ad. Then, write below the ad what you
found most noticeable.”Participants read the other ad only once with-
out further work on it. As noted before, asking them to focus on a partic-
ular piece of the campaign materials was justifiable given the cover
story. Upon seeing that the order of ads did not makea difference in Ex-
periment1, we did not manipulate order this time and presented the ar-
gument ad before the source ad for everyone in the study.
Besides a change in the focus manipulation, Experiment 2 addition-
ally involved measures of recall obtained atthe end of the delayed post-
test. Finally, as described in the supplementary analyses section,
Experiment 2 included control conditions in which the participants re-
ceived evaluatively consistent communications where strong argu-
ments were associated with a credible source or weak arguments
were associated with a noncredible source.
3.1.3. Dependent measures
Participants reported their attitudes toward voting in favor of the
party immediately after exposure to the communications and also
45 min later. After the delayed measure of attitudes, participants com-
pleted a measure of free recall, which was intended to verify the effec-
tiveness of the focus manipulation.
3.1.3.1. Attitudes toward behavior. The attitude measures were identical
to the ones used in Experiment 1. The reliability of the scale was high
at each time point (α
immediate
= 0.95; α
delayed
= .98). For each time
point, responses were averaged to yield summary indices of attitudes.
As in Experiment 1, the index of attitude change was created by
subtracting immediate attitudes from delayed attitudes. Thus, positive
scores represented possible sleeper effects, whereas negative scores
represented decays in persuasion.
3.1.3.2. Recall of arguments and source attributes. At the end of the exper-
iment, participants were asked to write down all they could remember
from the campaign materials. We computed two separate recall scores
based on these responses, one for the arguments and one for the source
attributes. There were four pieces of information in each ad. Therefore,
recall scores ranged from 0 to 4 in each case. Undergraduate assistants
blind to the conditions of the study independently coded the protocols
with an acceptable level of agreement (r= 0.95). In addition, the first
author blindly coded the protocols twice with an interval of 10 weeks
between each coding (r=0.92).
3.2. Results
We expected that the likelihood of observing either type of sleeper
effect would depend on the focus of attention and the information re-
ceived. As in Experiment 1, these factors interacted in determining the
longitudinal course of change in attitudes. Before decomposing this in-
teraction, however, it is necessary to establish that the conditions neces-
sary for the emergence of the sleeper effect were met. This requires
examining immediate attitudes and attitude change over time separate-
ly (for a detailed review of issues related to analysis of sleeper effect
data, see Cook et al., 1979; Kumkale & Albarracín, 2004).
3.2.1. Immediate attitudes
As one of the conditions, it is necessary to verify that thereis little or
no persuasion initially. The presence of a discounting cue and its acces-
sibility at the time of exposure should restrict persuasion initially, pref-
erably almost entirely. Indeed, when people received strong arguments
from a noncredible source or when they received weak arguments from
a credible source, immediate attitudes did not differ significantly from
the neutral point of the scale ranging from −5to5(M= 0.38, SD =
2.28, t=1.44,p= 0.15.). Furthermore, as in Experiment 1, neither
the main effects nor the interaction between Information Set and
Focus of Attention were significant for immediate attitudes, all
Fsb1.06. Thus, an important condition required to observe a sleeper ef-
fect was met.
3.2.2. Attitude change over time
As in Experiment 1, focusing on aspects of the communication with
favorable implications brought about significant increases in persuasion
over time (see the third row of Table 2). The critical interaction between
Focus of Attention and Information Set was significant as in Experiment
1, F(1, 70) = 9.24, pb0.01. The source-based sleeper effect was verified
when the recipients of weak arguments presented by a credible source
focused on the source (M
change
= 0.50, CI = 0.02/0.98) but not when
Table 1
Attitudes and attitude change as a function of information set and focus of attention (Experiment 1).
Dependent measure Weak arguments-credible source Strong arguments-noncredible source
Focus on arguments
(n= 62)
Focus on source
(n= 48)
Focus on arguments
(n= 48)
Focus on source
(n= 57)
Immediate attitudes −0.17
(2.43)
−0.37
(2.64)
0.17
(2.35)
0.35
(2.16)
Delayed attitudes −0.43
(2.62)
0.03
(2.63)
0.73
(2.52)
−0.18
(2.43)
Attitude change −0.26
+
(−0.58/0.06)
0.40⁎
(0.04/0.76)
0.55⁎
(0.20/0.91)
−0.53⁎
(−0.86/−0.21)
Note. Immediate and delayed attitudes were measured using scales that ranged from −5 to 5.Values in parenthesesare standard deviations for immediate and delayed attitudes; for at-
titude change scores, they are 95% confidence intervals around the means.
+
pb0.06.
⁎pb0.05.
175D. Albarracín et al. / Journal of Experimental Social Psychology 68 (2017) 171–180

they focused on the arguments (M
change
=−0.46, CI = −0.93/0.06);
contrast F(1, 70) = 6.33, pb0.02; d= 0.86, CI = 0.19/1.53. The tradi-
tional, argument-based sleeper effect was verified when the recipients
of strong arguments presented by a noncredible source focused on the
source (M
change
= 0.72, CI = 0.23/1.21 vs. M
change
=0.04,
CI = −0.42/0.49); contrast F(1, 70) = 3.18, pb0.08; d=0.57,
CI = −0.09/1.23.
3.2.3. Recall of source attributes and arguments
Next, we examined whether these sleeper effects were mediated by
relative recall of arguments versus source attributes. Recipients' memo-
ry of the arguments and source attributes was measured using a free-re-
call task at the end of the delayed posttest (see the bottom panel of
Table 2). As expected, participants who initially focused on the argu-
ments recalled these arguments better than those who initially focused
on the source attributes (M=2.72,SD =0.84vs.M= 1.82, SD =1.00,
respectively). Only the main effect of focus of attention was significant
in a 2 × 2 ANOVA test with Information Set and Focus of Attention as
the between-subjects factors; F(1, 70) = 17.60, pb0.001. Similarly, par-
ticipants who initially focused on the source attributes later recalled
these attributes better than those who initially focused on the argu-
ments (M= 2.03, SD = 1.40 vs. M= 0.86, SD = 1.10, respectively).
Again, only the main effect of focus of attention was significant in a
2×2ANOVA;F(1, 70) = 14.44, pb0.001.
4
To examine whether the effect of focus on recall mediated the pat-
terns of change in attitudes, we followed the bootstrapping procedures
outlined by Preacher and Hayes (2004). The sleeper effects emerge only
when people focus on the piece of information that has favorable impli-
cations for judgment (e.g., strong arguments or the attributes of the
credible source). Therefore, we created a dummy-code for whether
the focus was on the positive or negative aspect of the communications
(1 = positive vs. 0 = negative). Then, we created an index for the me-
diator (i.e., biased recall) using the two indicants of recall summarized
in Table 2. Specifically, we converted the raw number of positive and
negative pieces of recalled information into proportions, and subtracted
the proportion of negative pieces of information from the proportion of
positive pieces of information recalled. Thus, a zero indicates lack of
bias, whereas positive scores indicate superior recall of the positive
piece of information and negative scores imply superior recall of the
negative piece of information. Consequently, scores on this index of bi-
ased recall ranged from −1to1.
To examine whether the indirect effect of focus of attention on atti-
tude change, via biased recall, was significantly different from zero, we
estimated the standard deviation of the indirect effect for 3000 boot-
strap samples (Preacher & Hayes, 2004). We found that the indirect ef-
fect was estimated to lie between 0.02 and 0.56 (Β=0.27,SE = 0.13).
Because zero is not in the 95% confidence interval, these data suggest
that the impact of focus of attention on attitude change was indeed me-
diated by biased recall. The effect of focus of attention on theamount of
change from the immediate to delayed posttest (B= 0.81, SE = 0.27,
pb0.01) became nonsignificant after controlling for the mediator
(B= 0.55, SE =0.31,pN0.08).
The bootstrap mediation data suggest that focus of attention may af-
fect course of change in attitudes by way of inducing biased recall. As an
alternative mediation possibility, it could be argued that people who
changedtheir attitudes as a function of focus might have retrieved infor-
mation from memory in a biased way to support these newly construct-
ed delayed attitudes. The same bootstrapping procedures supported
this possibility.
5
However, this alternative possibility does not consti-
tute a threat to the biased recall mechanism that we propose, as the re-
call measures in this study were obtained after the administration of
delayed attitude measures. Besides that, the latter possibility implies a
deliberative, strategic attempt to engage in biased retrieval of informa-
tion from memory to support a judgment. Such a deliberation was not
relevant in this context, as the participants did not know that they
would be asked to report a delayed judgment.
3.2.4. The effects of attentional focus under evaluative consistency: supple-
mentary analyses
Experiments 1 and 2 showed that focus of attention is critical in de-
termining attitude change over time when a communication conveys
evaluatively inconsistent pieces of information. Such a communication
is likely to induce an imbalanced attitude structure.Based on past theo-
rizing, wereasoned that focus of attention would matter less in absence
of such an inconsistency (Albarracin, Wallace, & Glasman, 2004; Cook &
Flay, 1978; Vallacher, Nowak, & Kaufman, 1994) and tested this idea
Table 2
Attitudes, attitude change, recall as a function of information set and focus of attention (Experiment 2).
Dependent measure
Weak arguments-credible source
(−,+)
Strong arguments- noncredible source
(+, –)
Focus on arguments
(n= 19)
Focus on source
(n= 18)
Focus on arguments
(n= 17)
Focus on source
(n= 20)
Attitudes
Immediate attitudes 0.40
(2.75)
0.93
(2.01)
0.34
(2.47)
−0.10
(1.89)
Delayed attitudes −0.05
(3.22)
1.43
(1.95)
1.06
(2.49)
−0.06
(3.21)
Attitude change −0.46
+
(−0.93/0.06)
0.50⁎
(0.02/0.98)
0.72⁎
(0.23/1.21)
0.04
ns
(−0.42/0.49)
Recall for message arguments and source attributes
Message arguments 2.74
(2.31/3.17)
1.61
(1.17/2.05)
2.71
(2.25/3.16)
2.00
(1.58/2.42)
Source attributes 0.90
(0.31/1.48)
2.00
(1.43/2.58)
0.82
(0.20/1.44)
2.00
(1.43/2.58)
Note. Immediate and delayed attitudes were measured using scales that ranged from −5 to 5.Values in parenthesesare standard deviations for immediate and delayed attitudes; for at-
titude change scores, they are 95% confidence intervals around the means.
Recall scores could range from 0 to 4 for both arguments and source attributes. Values in parentheses are 95% confidence intervals around the means.
+
pb0.06.
⁎pb0.05.
4
Clearly participants recalled the arguments better than source attributes. This effect
does not compromise the interpretation of our results but may be due to the fact that
the policies discussed in the arguments were known to participants whereas the informa-
tion about the party was not.
5
The indirect effect was estimated to lie between 0.02 and 0.63 (B= 0.29,SE = 0.16).
The effect of focuson biased recall (B= 0.82, SE =0.28,pb0.001) became nonsignificant
after controlling for the mediator, which was the amount of attitude change over time
(B= 0.53, SE =0.32,pN0.10).
176 D. Albarracín et al. / Journal of Experimental Social Psychology 68 (2017) 171–180

using the same Gator Party materials. In these evaluatively-consistent
conditions, participants received strong arguments associated with the
credible source or weak arguments associated with the noncredible
source. These conditions were not relevant to the sleeper effect directly.
However, they were included to show that focus of attention should
matter to a greater extent when communications convey evaluatively
inconsistent information. Thus, we expectedthat participants would de-
velop very favorable or unfavorable attitudes depending on the infor-
mation set received, and that these immediate attitudes would persist
over time regardless of focus.
Table 3 shows the data for these evaluatively consistent conditions.
As can be seen,when strong arguments were associated with a credible
source, participants formed very favorable attitudes, and these favor-
able attitudes remained stable overtime regardless of focus. Similarly,
when weak arguments were associated with a noncredible source, par-
ticipants formed very unfavorable attitudes, and these attitudes
persisted over time regardless of focus. On attitude change scores, nei-
ther the main effects nor the interaction wassignificant, Fs b1. Averaged
across all four conditions, there was clearevidence of stability over time
(M=0.00;SD =0.85,N= 72). This result is consistent with the pre-
mise that when the arguments and the source have the same level of
credibility/strength, attitudes can be reconstructed in the same fashion
regardless of whether the arguments or the source characteristics are
recalled.
3.3. Discussion
Experiment 2 provided further support for the hypothesis that the
sleeper effect for the source emerges when a credible source presented
weak arguments and the recipients focus on the source. Correspondingly,
the traditional sleeper effect for the arguments emerges when a
noncredible source presents strong arguments and the recipients focus
on the arguments. Importantly, this pattern was obtained even though
the focus of attention manipulation did not have an impact on immediate
attitudes, suggesting that the effect occurred at the point of reconstructing
attitudes during the delayed follow up. The effects on attitude change were
mediated by recall of the arguments and the source information.
Overall, Experiment 2 replicated the results observed in Experiment
1, with one exception:When the recipients focused on the attributes of
a noncredible communicator presenting strong arguments, delayed at-
titudes did not turn out to be more negative than the immediate atti-
tudes. Attitudes in this condition persisted over time (i.e., change =
0.04; see the fourth column of Table 2). Attitude stability in this condi-
tion could only take place in absence of biased recall. Indeed, the recall
data presented in Table 2 show that participants in this condition
could recall the arguments and the source attributes equally well inter-
estingly. Hence, there was no evidence of biased recall in this condition.
This attitude stability is probably due to some spontaneous integration
of the conflicting information. For example, participants might have
tried to explain why an otherwise weak source presented strong argu-
ments, leading to retrieving integrated argument and source informa-
tion at the delayed follow up. The more negative attitude change
observed in Experiment 1 might be due to differences in the nature of
the focus manipulation, which was forceful in Experiment 1 but rela-
tively more subtle in Experiment 2.
4. Experiment 3
Experiments 1 and 2 provided evidence that the likelihood of ob-
serving both types of sleeper effects as a function of the evaluative im-
plications of the arguments and the source as well as the focus of
attention. Providing evidence for the source-based sleeper effect wases-
pecially critical as it has never been shown before despite the interest
generated by these issues. In these experiments, however, the novel
sleeper effect was obtained within a single experimental session, with
a delay of 45 min in between the message presentation and the delayed
follow up (for similar procedures see Pratkanis et al., 1988; Waldum &
Sahakyan, 2012). Of course these conditions would only increase the
difficulty of observing an effect that relies on imperfect memory. How-
ever, for comparability with some of the prior work on the effect
(Kumkale & Albarracín, 2004), it seemed useful to replicate this novel
effect with a longer delay. Thus, in Experiment 3, delayed attitudes
were measured exactly one week after the message presentation.
The second goal of Experiment 3 was to replicate our findings in an at-
titude change rather than formation setting. In Experiments 1 and 2, par-
ticipants did not have existing attitudes toward the issues discussed in
the messages. In this experiment, however, participants were much likely
to have attitudes toward the issues discussed in the target message, which
was about the controversial issue of incorporating Turkey into the Europe-
an Union (EU). It has been a current controversial issue for several decades
in Turkey, where this experiment was conducted. Thus, Experiment 3 in-
cluded an additional no-message control condition.
The third goal of Experiment 3 was to highlight the importance of
the focus of attention by comparing the sleeper effect in a source-
focus condition to the trend in a no-focus control condition. In Experi-
ments 1 and 2, weak arguments presented by a credible source brought
about a sleeper effect when the focus was placed on the source attri-
butes. In absence of such a focus, there should be no reason to expect
a sleeper effect or decay over time, assuming that similarly-long source
and message passages receive similar attention. Thus, it was necessary
to verify that attitudes would remain relatively stable over time in the
no-focus control condition.
In summary, participants in Experiment 3 received a message con-
taining weak arguments about the incorporation of the country into
the EU and then reported their attitudes immediately after receiving
the message and then again, seven days after the message presentation.
As in Experiments 1 and 2, we expected to observe the novel sleeper ef-
fect only in the source-focus condition, but attitude stability in the no-
focus condition. As natural changes in attitudes were possible due to ac-
tive media coverage of the topic at the time, these trends were com-
pared to the trend in the baseline condition (for more detailed
information about defining the sleeper effect in relation to baseline
changes in attitudes, see Cook & Flay, 1978; Kumkale & Albarracín, 2004).
The fact that we expect to observe attitude stability rather than a
sleeper effect in the no-focus control condition requires an explanation.
Unlike past theorizing on the sleeper effect, we do not think that one
piece of attitude-relevant information should be necessarily more ac-
cessible, or relevant than another pieceinherently. If thesource descrip-
tion is detailed or salient as in our case, its representations can stay as
strong as the representations of the arguments over time and bring
about attitude stability rather than sleeper effects or decays. In making
this point, we concur with other researchers that cues and arguments
are better conceptualized as functionally equivalent pieces of informa-
tion (see e.g., Erb et al., 2007; Kruglanski et al., 2004). To bolster this
Table 3
Attitudes and attitude change in the evaluatively-consistent control conditions (Experi-
ment 2).
Dependent measure
Weak arguments-
noncredible source
(−,–)
Strong arguments-credible
source
(+, +)
Focus on
arguments
(n= 19)
Focus on
source
(n= 18)
Focus on
arguments
(n= 16)
Focus on
source
(n= 19)
Immediate attitudes −2.39
(1.94)
−1.88
(2.68)
2.66
(1.93)
2.09
(1.86)
Delayed attitudes −2.53
(1.85)
−1.96
(2.82)
2.69
(2.02)
2.28
(1.82)
Attitude change −0.13
ns
(−0.60/0.34)
−0.08
ns
(−0.56/0.40)
0.03
ns
(−0.48/0.54)
0.18
ns
(−0.56/0.40)
Note. Immediate and delayed attitudes weremeasured using scales that ranged from −5
to 5. Values in parentheses are standard deviations for immediate and delayed attitudes;
for attitude change scores, they are 95% confidence intervals around the means.
177D. Albarracín et al. / Journal of Experimental Social Psychology 68 (2017) 171–180

argument, we revisited the meta-analytic database and conducted addi-
tional analyses involving the salience of source descriptions. We rea-
soned that a sleeper effect in absence of a focus manipulation might
have taken place when the source descriptions were brief but not
when they were long. Indeed, as expected, there were slight increases
over time when the descriptions were brief (d= 0.13), but clear evi-
dence attitude stability when the descriptions were long (d= 0.00);
z = 2.34, pb.02. As noted by Kumkale and Albarracín (2004),detecting
the traditional sleeper effect has not been easy in the past. In the major-
ity of the studies that they reviewed, the default pattern was attitude
stability rather than a sleeper effect, which is consistent with our argu-
ment about the importance of focus.
4.1. Method
4.1.1. Participants
Participants were 144 introductory psychology students who received
course credit for taking part in a two-session experiment (61 females and
83 males). Participants were randomly assigned to one of the following
three conditions (source focus; no focus; no-message control).
4.1.2. Procedures
Participants signed up to a study involving tasks and measures os-
tensibly part of two separate studies—a consumer behavior study and
a text-readability study. The target message and the associated attitude
measures were presented as part of the readability study with a cover
story similar to the one used in Experiment 2.
In the first session, participants first completed various individual
difference measures as part of the consumer behavior study, which
was intended to identify individual difference correlates of discretion-
ary spending. This study involved some decision tasks unrelated to the
target issue. In the middle of the session, participants in the source-
focus and no-focus conditionsreceived the targetmessage. The message
contained three weak arguments stating that Turkey should become a
member of the European Union. It was attributed to a credible source.
In the no-focus condition (n= 42), participants reported their attitudes
toward Turkey's membership in the EU immediately after reading the
message. In the source-focus condition (n= 40), participants rated the
credibility and the expertiseof the message sourcefirst and then report-
ed their attitudes. The assessment of the message source along these di-
mensions was the only difference between these two experimental
conditions. Participants in both conditions reported their attitudes
again by coming to the lab exactly one week after message exposure.
Participants in the no-message control (n= 62) condition did not re-
ceive the target message but reported attitudes at both time points so
that potential changes in baseline attitudes could be assessed.
6
4.1.2.1. Communication. The message concerned Turkey's incorporation
into the EU and included extensively piloted weak arguments. One of
the arguments highlighted that membership would make it easier to
distinguish higher quality, environmentally-friendly products from
lower quality, harmful products. Another argument focused on changes
that would take place in the fishing industry and argued that adopting
the EU standards on size, quality, packaging, and labeling of sea-based
products would revolutionize the industry to the benefit of the consum-
er. None of the arguments, which describe real changes that would take
place in case of membership, had been heard by the participants before.
The message, presumably taken from a news agency specializing in EU
affairs, was presented with the title “How will the EU change our every-
day life?”and was 400-words long.
With respect to the communication source, the message was attrib-
uted to a professor presumably working as a member of the EU Harmo-
nization Commission. A 60-word description provided information
about the educational background of the professor as well as his scien-
tific achievements. This information was also established through pilot
testing, and found to be more credible than other types of sources
(e.g., a student and an experienced journalist).
4.1.2.2. Focus manipulation. Participants in the source-focus condition
rated the expertise and trustworthiness of the message source prior to
reporting their immediate attitudes, using two questions that were
typed with a slightly larger, bold-faced font. The idea was to have recip-
ients think about these dimensions in ways that may not occur sponta-
neously (Pratkanis et al., 1988). Participants in the no-focus condition
did not rate the source on any dimension.
4.1.3. Dependent measures
Attitudes toward Turkey's incorporation into the European Union
were assessed by asking participants to rate “Turkey's accession into
the EU”along scales ranging from −5to+5(would be very bad vs.
very good; is something that I would not want at all vs.Iwouldvery
much want; would be harmful vs.helpful; is very unnecessary vs.very nec-
essary). Responses were averaged to create summary indices of atti-
tudes (α= 0.96 for both posttests). The interval between the
immediate and the delayed posttest was 7 days.
4.2. Results
As in Experiments 1 and 2, we expected a sleeper effect for the
source when participants focus was on the attributes of the source.
Without such a focus, immediate attitudes would most likely remain
stable over time. To test this hypothesis, Experiment 3 included an addi-
tional no-focus control condition and the aforementioned baseline con-
dition. The means corresponding changes observed in these three
conditions to this study appear in Table 4.
Asshowninthefirst row of Table 4, immediate attitudes of partici-
pants who received the target message seemed to be more positive
than the attitudes of control group participants who did not receive a
message. Although these differences were not significant, the lack of
complete suppression of persuasion at the time of exposure could re-
duce the probability of observing a sleeper effect in the long run (as ar-
gued in Cook et al., 1979; Kumkale & Albarracín, 2004), F(2, 141) =
1.35, ns.pN0.25. Despite this constraint, expected differences were ob-
served at the delayed posttest, F(2, 141) = 3.53, pb0.04: Although at-
titudes remained relatively stable over time in both of the control
conditions, there wasa significant increase only in the source-focuscon-
dition, F(2, 141) = 3.65, pb0.03.
4.3. Discussion
The goal of Experiment 3 was to verify the source-based sleeper ef-
fect usinga longer delay between attitude measurements. Furthermore,
we wanted to see if we could generate the source-based sleeper effect in
an attitude change setting, as Experiments 1 and 2 involved formation
of new attitudes. Finally, we wanted to highlight the importance of
focus by comparing the situation to a situation in which the recipients
were not induced to focus on a particular piece of information. In that
case, attitude stability rather than a sleeper effect prevailed. Corroborat-
ing the results of the previous experiment, Experiment 3 demonstrated
the importance of the focus of attention in attitude persistence and
change: The source-based sleeper effect emerged when people focused
on the attributes of the credible source but not otherwise.
6
As in Experiment 1, sample sizes across conditions were not equal in Experiment 3.
The data for this study were collected by a research assistant in a large computer-lab,
where the URL address directing participants to the web-based questionnaire was
projected on the screen. There were three different web-based questionnaires (with dif-
ferent addresses) representing the condition of the study. Conditions were randomized
acrosssessions within thesame day. Thus, theimbalance in samplesizes across conditions
was due to running an additional session for the no-message control condition.
178 D. Albarracín et al. / Journal of Experimental Social Psychology 68 (2017) 171–180

5. General discussion
Building on the premise that message arguments and source attri-
butes can both influence persuasion, we investigated a new type of
sleeper effect never considered before—a sleeper effect guided by a de-
layed increase in the impact of source attributes rather than the argu-
ments. Moreover, the present research not only verified the source-
based sleeper effect, but also shed some light on the question of why
it has been difficult to replicate the effect in the past.
Given that individuals retrieve just enough information to make a
reasonable judgment, their judgments and behaviors can be dispropor-
tionately influenced by the evaluative implications of accessible atti-
tude-relevant information (Albarracín, 2002; Chaiken et al., 1996;
Feldman & Lynch, 1988; Schwarz & Bohner, 2001; Wilson et al., 2000;
Wyer, 2004,2007). Accordingly, representationsof attitude-relevant in-
formation could entail the arguments or the source, not just the argu-
ments as it has been traditionally assumed. Thus, when people first
see a presidential candidate, they may form a person impression that
is later accessible to recall. If they later think about a candidate in a pres-
idential race, they may recall personal attributes of this candidate rather
than her arguments aboutspecific issues (Iyengar & Ottati, 1994). In this
case, representations of the issues supported by the candidate may be
integrated as well, but may only be accessed along with other informa-
tion about the candidate as a person. In general then, these conditions
would facilitate easier recall of information about the source and more
difficult recall of the arguments presented by the source.
Although the sleeper effect literature evolved independently of the
literature on source monitoring, it is possible to see the sleeper effect
as a special case of failure of source monitoring (for a review, see
Johnson, Hashtroudi, & Lindsay, 1993). Source monitoring failures
arise when aspects of the original learning context and the learned in-
formation are represented in different memory systems. For instance,
the arguments conveyed in a message may be represented in semantic
memory, whereas information suchas when, where, and how that mes-
sage was received may be represented in episodic memory. As semantic
recall is more resource depleting than episodic recall (Tulving, 2002),
the arguments contained in a message may be more easily retrieved
than the message's context, including its source (for recent relevant
work, see Ladowsky-Brooks & Alcock, 2007). Moreover, conditions
that lead to semantic/episodic dissociations should increase the likeli-
hood of observing sleeper effects. For example, the use of different
codes, such as implicit, subtle cues along with explicit propositional ar-
guments may produce sleeper effects of greater magnitude than those
observed in the present and past work.
Another future direction in this line of research concerns individual
differences that can moderate the sleeper effect. Thus far, only need
for cognition has been shown to facilitate the effect (Priester,
Wegener, Petty, & Fabrigar, 1999), but other factors may play a role as
well. For example, the present research suggests that the sleeper effect
should be most likely for people who are high in need for cognition
but low in need to evaluate (Jarvis & Petty, 1996). People who are
high in need for cognition are likely to have paid attention to the focal
element (i.e., the arguments or the source) more than people who are
low in this trait. People who are high in need to evaluate tend to form
attitudes spontaneously, and so may have immediate attitudes that re-
duce the need to construct attitudes at the delayed follow up. Thus, fu-
ture research should investigate whether these two traits interact to
produce sleeper effects.
Another direction for future research is studying attitude persistence
and change in message dense, dynamic environments. In this line of re-
search, the effects of a single message are typically considered in a vac-
uum. In everyday life people receive competing and repeated messages
from various points of view. For instance,in a typical political campaign,
there are accusations, defenses against those accusations, and com-
ments of third parties. In the meantime, people's affective states such
as moods may change as well. Hence, it may be timeto consider the lon-
gitudinal course of attitudes in dynamic contexts where there are mul-
tiple messages from multiple sources. A better understanding of how
attitudes evolve in such dynamic settings, along with more measures
of change, will enhance the applicability of attitude research.
In closing, we found that sleeper effects for the source do emerge
when message recipients form impressions of a credible communicator
who presents weak arguments. The conditions for ignoring the informa-
tion presented by a communicator entail a focus on the communicator,
and may naturally unfold when personal characteristics are highly sa-
lient. For example, personal contact with a communicator may produce
situations in which the actual arguments countless as time goes by.
Therefore, far from concluding that credibility does not matter in the
long run, one must conclude that credibility may sometimes be the
only piece of information that endures the passage of time.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.jesp.2016.06.009.
References
Albarracín, D. (2002). Cognition in persuasion: An analysis of information processing in
response to persuasive communications. In M. P. Zanna (Ed.), Advances in experimen-
tal social psychology (pp. 61–130). San Diego, CA: Academic Press.
Albarracin, D., Wallace, H. M., & Glasman, L. R. (2004). Survival and change of attitudes
and other social judgments: A model of activation and comparison. In M. P. Zanna
(Ed.), Advances in experimental social psychology (pp. 251–315). San Diego, CA: Aca-
demic Press.
Chaiken, S., & Eagly, A. H. (1983). Communication modality as a determinant of persua-
sion: The role of communicator salience. Journal of Personality and Social Psychology,
45,241–256.
Chaiken, S., Wood, W., & Eagly, A. H. (1996). Principles of persuasion. In E. T. Higgins, & A.
W. Kruglanski (Eds.), Social psychology: Handbook of basic principles (pp. 702–742).
New York: Guilford Press.
Cook, T. D., & Flay , B. R. (1978). The p ersistence of experimentally induced attitude
change. In L. Berkowitz (Ed.), Advances in experimental social psychology (pp. 1–57).
New York: Academic Press.
Cook, T. D.,Gruder, C. L., Hennigan, K. M., & Flay,B. R. (1979). Historyof the sleeper effect:
Some logical pitfa lls in accepting the null hypothesis. Psychological Bulletin,86,
662–679.
Devine, P. G., & Ostrom, T. M. (1985). Cognitive mediation of inconsistency discounting.
Journal of Personality and Social Psychology,49,5–21.
Table 4
Attitudes at the immediate posttest, after 7 days, and change as a function of focus of attention (Experiment 3).
No-message control
(n= 62)
No-focus control
(n= 42)
Focus on source
(n= 40)
Immediate attitudes 1.67
(2.38)
2.15
(2.17)
2.42
(2.39)
Delayed attitudes 1.58
(2.66)
2.18
(2.17)
2.86
(2.07)
Attitude change −0.09
ns
(−0.37/0.19)
0.02
ns
(−0.24/0.29)
0.44⁎
(0.14/0.73)
Note. Attitudes were measured using 11-pointrating scales ranging from −5 to 5, where higherscores indicatedgreater agreement withthe advocated position.Numbers in parentheses
for immediate and delayed attitudes are standard deviations, and confidence intervals for change scores reported in the last row.
⁎pb0.05.
179D. Albarracín et al. / Journal of Experimental Social Psychology 68 (2017) 171–180

Erb, H. P., Pierro, A., Manetti, L., Spiegel, S., & Kruglanski, A. W. (2007). Biased processing
of persuasive information: On the functional equivalence of cues and message argu-
ments. European Journal of Social Psychology,37,1057–1075 .
Feldman, J. M., & Lynch, J. G. (1988). Self-generated validity and other effects of measure-
ment on belief, attitude, intention, and behavior. Journal of Applied Psychology,73,
421–435.
Fitzsimons, G. J., & Williams, P. (2000). Asking questions can change behavior: Does it do
so automatically or effortfully? Journal of Experiment al Psychology: App lied,6,
195–206.
Gruder, C. L., Cook, T. D., Hennigan, K. M., Flay, B. R., Alessis, C., & Halamaj, J. (1978). Em-
pirical tests of the absolute sleeper effect predicted fromthe discounting-cue hypoth-
esis. Journal of Personality and Social Psychology,36,1061–1074.
Hovland, C. I., & Weiss, W. (1951). The influence of source credibility on communication
effectiveness. Public Opinion Quarterly,15,635–650.
Hovland, C. I., Lumsdaine, A. A., & Sheffield, F. D. (1949). Experiments on mass communica-
tion. Princeton, NJ: Princeton University Press.
Iyengar, S.,& Ottati, V. (1994). Cognitive perspective in political psychology. In R.S. Wyer,
& T. K. Srull (Eds.), Handbook of social cognition (pp. 143–187). Hillsdale, NJ:Erlbaum.
Jarvis, W. B. G., & Petty, R. E. (1996). The need to evaluate. Journal of Personality and Social
Psychology,70,172–194.
Johnson, B. T., Maio, G. R., & Smith-McLallen, A. (2005). Communication and attitud e
change: Causes, processes, and effects. In D. Albarracín, B. T. Johnson, & M. P. Zanna
(Eds.), The handbook of attitudes (pp. 617–670). Mahwah, NJ: Erlbaum.
Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993). Source monitoring. Psychological
Bulletin,114,3–28.
Kelman, H. C., & Hovland, C. I. (1953). Reinstatement of the communicator in delayed
measurement of opinion change. Journal of Abnormal and Social Psychology,48,
327–335.
Killeya, L. A., & Johnson, B. T. (1998). Experimental induction of biased systematic pro-
cessing: The directed-thought technique. Personality and Social Psychology Bulletin,
24,17–33.
Kruglanski, A. W., Fishbach, A., Erb, H. P., Pierro, A., & Mannetti, L. (2004). The parametric
unimodel asa theory of persuasion.In G. Haddock, & G. R. Maio (Eds.), Contemporary
perspectives on the psychology of attitudes. NY: Psychology Press.
Kruglanski, A. W., Pierro,A., Manetti, L., Erb, H. P., & Spiegel, S. (2006). Persuasion accord-
ing to the unimodel. Journal of Communication,56,105–122.
Kumkale, G. T., & Albarracín, D. (2004). The sleeper effect in persuasion: A meta-analytic
review. Psychological Bulletin,130,143–172.
Ladowsky-Brooks, R., & Alcock, J. E. (2007). Semantic-episodic interactions in the neuro-
psychology of disbelief. Cognitive Neuropsychiatry,12,97–111.
Lingle, J. H., & Ostrom, T. M. (1979). Retrieval selectivity in memory-based impression
judgments. Journal of Personality and Social Psychology,37,180–194.
Lingle, J., Geva, N., Ostrom, T., Leippe, M., & Baumgardner, M. (1979). Thematic effects of
person judgments on impression formation. Journal of Personality and Social
Psychology,37,674–687.
Lodge, M., McGraw, K. M., & Stroh, P. (1989). An impression-driven model of candidate
evaluation. American Political Science Review,83,399–420.
Ottati, V. (2001). The psychological determinants of political judgment. In A. Tesser, & N.
Schwarz (Eds.), Blackwell handbook of social psychology: Intraindividual processes
(pp. 615–634). Oxford: Blackwell.
Ottati,V., & Wyer, R. S. (1993). Affect and political judgment. In S. Iyengar, & W. J. McGuire
(Eds.), Explorations in political psychology (pp. 296–320). Durham, NC: Duke Press.
Ottati, V., Wyer, R. S., Deiger, M., & Houston, D. (2002). The psychological determinants of
candidate evaluation and voting preference. In V. Ottati (Ed.), The social psychology of
politics (pp. 3–28). New York: Springer.
Pierro, A., Manetti, L., Erb, H. P., Spiegel, S., & Kruglanski, A. W. (2005). Informational
length and order of presenta tion as determinants of persuasion. Journal of
Experimental Social Psychology,41,458–469.
Pratkanis, A. R., Greenwald, A. G., Leippe, M. R., & Baumgardner, M. H. (1988). In search of
reliable persuasion effects: III. The sleeper effect is dead: Long live the sleeper effect.
Journal of Personality and Social Psychology,54,203–218.
Preacher, K. J., & H ayes, A. F. (2004). SPSS and SAS procedures for estimating indirect ef-
fects in simple mediation models. Behavi or Research Methods, Instruments &
Computers,36,717–731.
Priester, J., Wegener, D., Petty, R. E., & Fabrigar, L. (1999). Examining the psychological
process underlying the sleeper effect: The elaboration likelihood model explanation.
Media Psychology,1,27–48.
Prislin, R., & Wood, W. (2005). Social influence: The role of social consensus in attitudes
and attitude change. In D. Albarracin, B. T. Johnson, & M. P. Zanna (Eds.), Handbook
of attitudes (pp. 671–706). Hillsdale, NJ: Erlbaum.
Schwarz, N., & Bohner, G. (2001). The construction of attitudes. In A. Tesser, & N. Schwarz
(Eds.), Blackwell handbook of social psychology: Intrapersonal processes (pp. 436–457).
Oxford, UK: Blackwell.
Tulving, E. (2002). Episodic memory. Annual Review of Psychology,53,1–25.
Vallacher, R. R., Nowak, A., & Kaufman, J. (1994). Intrinsic dynamics of social judgment.
Journal of Personality and Social Psychology,67,20–34.
Waldum, E.R., & Sahakyan, L. (2012). Putting congeniality effects into context:Investigat-
ing the role of context in attitude memory using multiple paradigms. Jou rnal of
Memory & Language,66,717–730.
Webb, T. L., & Sheer an, P. (2006). Does changing behavioral intentions engenderbehavior
change? A meta-analysis of the experimental evidence. Psychological Bulletin,132,
249–268.
Weber, S. J. (1972). Opinionchange as a function of the associative learning of the content
and source factors. Unpublished doctoral dissertation, Northwestern University, Ev-
anston, IL.
Wilson, T. D., Lindsey, S., & Schooler, T. Y. (2000). A model of dual attitudes. Psychological
Review,107,101–126.
Wyer, R. S. (2004). Social comprehension and judgment. Mahwah, NJ: Erlbaum.
Wyer, R. S. (2007). Principles of mental representation. In A.W. Kruglanski, & E. T. Higgins
(Eds.), Social psychology: Handbook of basic principles (pp. 285–307). New York:
Guilford Press.
Wyer, R. S., & Budesheim, T. L. (1987). Person memory and judgment: The impact of in-
formation one is told to disregard. Journal of Personality and Social Psychology,53,
14–29.
Wyer, R. S., & Ottati, V. (1993). Political information processing. In S. Iyengar, & W. J.
McGuire (Eds.), Explorations in pol itical psychology (pp. 264–295). Durham, NC:
Duke Press.
180 D. Albarracín et al. / Journal of Experimental Social Psychology 68 (2017) 171–180