Journal of Consumer Research Inc.

The Shaken Self: Product Choices as a Means of Restoring SelfView Confidence

Author(s): Leilei Gao, S.Christian Wheeler, Baba Shiv
Source: The Journal of Consumer Research, Vol. 36, No. 1 (June 2009), pp. 29-38
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/10.1086/596028 .
Accessed: 12/02/2011 18:12

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .
http://www.jstor.org/action/showPublisher?publisherCode=ucpress. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

The University of Chicago Press and Journal of Consumer Research Inc. are collaborating with JSTOR to
digitize, preserve and extend access to The Journal of Consumer Research.

http://www.jstor.org
The Shaken Self: Product Choices as a
Means of Restoring Self-View Confidence
LEILEI GAO
S. CHRISTIAN WHEELER
BABA SHIV*

The present research shows that when a confidently held self-view (e.g., I am an
exciting person) is temporarily cast in doubt, individuals are motivated to choose
products that bolster their original self-view (e.g., choosing brands with exciting
brand personalities). The findings across three studies suggest that subtle manip-
ulations can temporarily shake ones self-view confidence, resulting in an in-
creased propensity of choosing self-view-bolstering products in a subsequent
choice task. The consequences of the shaken self for product choices are ex-
amined in different self-domains. The findings also suggest that the effects of the
shaken self are attenuated when individuals have the opportunity to restore their
self-view confidence prior to the final choice task.

The self is not something ready-made, but by proposing that a key determinant of this congruency is
something in continuous formation through self-view confidence. Specifically, we show that threats to
choice of action. (John Dewey) an important self-view can momentarily shake ones con-

M arketing researchers have long been interested in the

relationship between the self-concept and consumer
choice. Most early research was based on the notion that
fidence in that particular self-view, resulting in the choice
of products that help restore confidence in that self-view.
Ironically, then, we propose that those subtly led to doubt
consumers are motivated to purchase products with images that they hold a self-characteristic may sometimes be the
that are congruent with their self-beliefs (e.g., Birdwell ones most likely to choose products congruent with it.
1968; Levy 1959). For example, Birdwell (1968) showed Our research shares some kinship with recent work show-
that automobile owners perceptions of their cars are es- ing that individuals take actions in response to threats to
sentially congruent with their self-perceptions. However, de- self-concept. For example, in a series of studies, Tetlock et
spite some promising findings such as these, extensive re- al. (2000) show that threats to sacred moral values result
views failed to find consistent evidence that people choose not only in moral outrage but also in attempts at taking
products congruent with their self-perceived characteristics actions that will restore those moral values (e.g., volun-
(Kassarjian 1971; Sirgy 1982). A wide number of potential teering to help). More recently, Zhong and Liljenquist
moderators have been examined, including levels of self- (2006) show that such restoration of the moral self can be
constructs (e.g., actual vs. ideal self), personality, social achieved not only through direct moral cleansing but also
class, and product conspicuousness, but these moderators, through physical cleansing, a phenomenon they term the
too, have failed to yield consistent results (Sirgy 1982). Macbeth effect. Note, however, that the research cited above
In this research, we sharpen our understanding of the has been restricted to the moral domain; our research prop-
relationship between the self-concept and product choices ositions apply to a wider array of domains. Specifically, in
this research, we extend recent work on the moral self by
*Leilei Gao is assistant professor of marketing, the Chinese University showing that the types of effects documented in that research
of Hong Kong (leilei@baf.msmail.cuhk.edu.hk). S. Christian Wheeler is stream occur not only when ones moral purity is threatened
associate professor of marketing (wheeler_christian@gsb.stanford.edu) and but also across a wider range of contexts in which any
Baba Shiv is professor of marketing (shiv_baba@gsb.stanford.edu) at the
Stanford Graduate School of Business, Stanford, CA 94305-5015. Cor- important self-view is threatened.
respondence: Leilei Gao. The authors thank the editor, associate editor, and Our research also extends recent work on the moral self
three reviewers for their help and guidance at various stages of this project. in other significant ways. We document a new set of specific
processes through which self-view threats can affect product
John Deighton served as editor and Susan Broniarczyk served as associate
editor for this article. choice. Specifically, we show that a threat to any important
self-view results in (1) a momentary loss in confidence with
Electronically published December 4, 2008
that particular self-view and (2) a broad array of behaviors
29

2008 by JOURNAL OF CONSUMER RESEARCH, Inc. Vol. 36 June 2009

All rights reserved. 0093-5301/2009/3601-0008$10.00. DOI: 10.1086/596028
30 JOURNAL OF CONSUMER RESEARCH

that are aimed at restoring confidence in that self-view. We going thoughts related to a particular self-view and, thus,
also document different strategies by which individuals cope leading to a reduced confidence in that self-view (cf. Schwarz
with a threatened self-view. Specifically, we show that the 2004).
effects of lowered self-view confidence on consumer choice The arguments presented above receive support from
can be eliminated by both direct self-view-bolstering strat- some research on metacognition. For example, Brinol and
egies (e.g., boosting confidence in the threatened self-view Petty (2003) instructed participants to write with their dom-
domain) as well as indirect strategies (e.g., affirming an inant or nondominant hands about their good or bad qual-
unrelated self-value). Finally, we show that in addition to ities. Regardless of whether they wrote about their good or
more blatant threats to the self (e.g., recalling instances of bad qualities, those who wrote with their nondominant hands
ones immoral behavior as in Zhong and Liljenquist [2006]), reported lowered confidence in their self-related thoughts,
very subtle actions (e.g., writing with ones nondominant despite generating content equivalent to those who wrote
hand about ones health consciousness) can also affect self- with their dominant hands. Brinol and Petty (2003) argued
confidence. that thoughts listed with the nondominant hand are more
In the next section we provide a more detailed exposition difficult to write and appear shaky, which contributes to
of the conceptualization underlying our proposed effects. lowered confidence in the content of the listed thoughts.
We then present three experiments that test our conceptu- Although our discussion thus far shares kinship with psy-
alization and rule out alternative accounts for our findings. chological processes underlying the Macbeth effect, a key
Finally, we discuss the conceptual as well as managerial difference needs to be noted. In research on the Macbeth
implications of our findings and provide some promising effect (Zhong and Liljenquist 2006), some participants were
directions for future research. instructed to write about immoral deeds, an act that pre-
sumably altered their thoughts about themselves as being
THEORETICAL BACKGROUND moral individuals. In this research we examine whether self-
views can be shaken even when the content of ones
We outline the basis for our predictions in two broad thoughts is held constant. For example, people who write
sections. First, we present our conceptualization related to about ways in which they are intelligent, but who do so
the shaken self, highlighting the importance of self-view with their nondominant hands, should have doubts about
confidence in giving rise to the predicted effects. We then their intelligence even though (and in fact, because) the
discuss the consequences of a shaken self, highlighting the content of their thoughts suggests that they are intelligent.
direct and indirect coping strategies that could be employed This is because the doubt activated by the unfamiliar task
for bolstering a shaken self-concept and the implications of (writing with ones nondominant hand) is attributed to the
these strategies for product choices. content of ones thoughts (e.g., I am intelligent), leading to
doubt in those thoughts (e.g., I have low confidence that I
The Shaken Self am intelligent). Stated differently, whereas research on the
Macbeth effect manipulated self-views by varying the con-
Self-view confidence refers to the certainty that one has tent of participants thoughts (related to being either immoral
in ones self-concept and in ones thoughts related to the or moral), we hold the content of participants thoughts
self (DeMarree, Petty, and Brinol 2007; Pelham 1991). Prior constant yet show that self-views can still be shaken by very
research suggests that individuals are likely to hold highly subtle factors that influence the confidence they have in
confident self-views if they can draw coherent evidence those thoughts. In the next section, we discuss various strat-
about their self-characteristics from the environment and egies that could be brought to bear as a means of coping
from their past experiences (Campbell 1990; Pelham 1991). with a momentarily shaken self and the implications of these
Confident self-views have typically been assumed to depend strategies for the choice of products that are aimed at bol-
on stable evidence about the self, but they are also subject stering self-view confidence.
to situational influences. For example, recalling an unethical
deed in ones past can lead to attempts to bolster ones moral The Shaken Self and Coping Strategies
purity, consistent with the idea that temporarily thinking
about how one is immoral can lower ones feelings of mo- Individuals generally desire to hold confident self-views,
rality (Zhong and Liljenquist 2006). and evidence suggests that a lack of such confidence can
In addition to these more blatant influences, we posit that be aversive, resulting in negative psychological outcomes
self-views can also be momentarily shaken by primes and such as low self-esteem (Baumgardner 1990; Campbell
other subtle situational factors, such as when one has to 1990), unhappiness, and anxiety (Rosenberg 1979). We
perform familiar and routinized tasks in unfamiliar ways. therefore posit that when ones self-view confidence is tem-
Examples of such situations include writing with ones non- porarily cast in doubt, one will be motivated to recover from
dominant hand, driving in the wrong lane (left lane) when the negative psychological outcomes and restore ones orig-
traveling in Britain, and using chopsticks for the first time. inal levels of confidence.
We argue that subtle features of the situation or of ones Prior research suggests that individuals can adopt a variety
actions can trigger metacognitions signaling that something of strategies to cope with their lowered self-confidence. For
is wrong, resulting in a reduced confidence with ones on- instance, in some circumstances, individuals have been
THE SHAKEN SELF 31

shown to seek objective information to resolve their uncer- more indirect methods (e.g., affirming an unrelated self-
tainty (Tiedens and Linton 2001). However, obtaining ob- value). Third, in line with work by Brendl et al. (2003) and
jective information about oneself may not always be fea- Zhong and Liljenquist (2006), we provide tests of our con-
sible. In such situations, individuals may instead attempt to ceptualization by arguing that if such self-restoration is al-
bolster their original self-views through either direct or in- lowed to occur prior to the final choice, then the final choices
direct means. We discuss each of these strategies in turn. would be no different than those for whom the self-concept
was never shaken in the first place. Finally, we show that
Direct Means of Recovering from a Shaken Self. It trait-specific confidence is the underlying mechanism of the
is now common knowledge that products are chosen not shaken-self effects.
only for their functional qualities but also for their symbolic
qualities (Belk 1988; Levy 1959; Solomon 1983). Con- EXPERIMENT 1
sumption of products can convey the possession of specific
self-characteristics both to others (Belk, Bahn, and Mayer Experiment 1 was designed to test the hypothesis that
1982) and to oneself (Bem 1972; Solomon 1983) and serve subtle manipulations can increase choice of self-view-bol-
as a means of coping with low confidence in specific self- stering products and that choosing such products can restore
views (Assael 1981; Piacentini and Mailer 2004; Wicklund the shaken self. In experiment 1 we relied on the previously
and Gollwitzer 1982). Thus, we posit that the consumption validated handwriting procedure employed by Brinol and
of symbolic products (e.g., reading sophisticated books) Petty (2003) to subtly affect confidence in a particular self-
could help restore a momentarily shaken self-view by pro- view (intelligence) self-confidence (or not). Specifically,
viding direct evidence about ones attributes (e.g., intelli- akin to Brinol and Petty (2003) participants in this experi-
gence). Stated differently, this symbolic function of products ment were instructed to write with either their dominant or
could make individuals with momentarily lowered self-con- their nondominant hand an essay about their intelligence,
fidence particularly prone to choosing products that support the idea being that writing about the particular self-view
the shaken self-view. Because products can be self-expres- with the nondominant hand can subtly shake self-view con-
sive, they can signal to oneself and to others who one is, fidence.
thereby helping to restore ones self-view confidence. Experiment 1 also served to test our conceptualization
The arguments presented above are consistent with a goal and reduce the viability of alternative accounts. To test our
activation account. Lacking confidence regarding a self-view conceptualization, some participants were provided the op-
could activate a self-view recovery goal. Kurt Lewin (1935) portunity for self-recovery prior to the final choice task, the
suggested that objects are valued to the extent that they idea being that the opportunity for self-recovery in an in-
satisfy a goal, and hence, a self-view recovery goal could tervening task ought to attenuate the choice of the self-view
lead people to prefer self-view-bolstering products. Once (intelligence) bolstering option in the final task.
the goal is satisfied and deactivated (Brendl, Markman, and
Messner 2003), however, consumers should no longer be Participants and Procedure
more likely to display product-self congruity.
Eighty-one participants were randomly assigned to one
Indirect Means of Recovering from a Shaken Self. In of four conditions in a 2 (handwriting: dominant hand vs.
addition to the more direct means of restoring the self-con- nondominant hand) # 2 (opportunity for self-recovery: yes
cept delineated in the previous section, we propose that vs. no) between-subjects design. The cover story was that
individuals can also adopt more indirect coping strategies. a group of experimenters was testing the validity of ancient
Self-affirmation theory proposes that individuals can cope graphological methods in predicting ones professional suc-
with threats to their selves through a fluid compensation cess. The experimenters were ostensibly interested in both
procedure (Steele 1988) such as affirming an important self- dominant hand and nondominant hand handwriting. Partic-
value (Shrira and Martin 2005). As Steele (1988, 267) wrote, ipants were first asked to write three personal characteristics
This fluidity of adaptation suggests the existence of a larger, that depicted them as being intelligent individuals. A pretest
ego-protective self-system not geared to resolving specific revealed that participants in the study sample population
self-concept threats, but geared to maintaining an overall believe that they are intelligent individuals and are confident
conception of self-integrity. Thus, individuals who have a about this belief. Some participants were assigned to write
shaken self-concept along one dimension could potentially the essay with their dominant hands; others were assigned
restore their overall self-views by affirming an unrelated to write the essay with their nondominant hands. Participants
self-value. then engaged in an intervening choice task in which some
Next, we present three experiments that accomplish sev- participants had the opportunity to restore self-view confi-
eral broad goals. First, we test the notion that subtle situ- dence. Specifically, participants were told that their name
ational factors that reduce self-view confidence will instigate would be entered in a lottery in which they had the chance
processes aimed at restoring the self-concept. Second, we of receiving a product in one of four categories (gift cer-
test the notion that this restoration can occur via multiple tificate, magazine, home game, or CD). One group of par-
means, including more direct methods (e.g., purchasing ticipants received a set of lottery choices that contained
products to restore the specific shaken self-dimension) and products related to intelligence (i.e., Borders gift certificate,
32 JOURNAL OF CONSUMER RESEARCH

National Geographic magazine, Rubiks cube, and Mozart FIGURE 1

CD); another group was presented with lottery choices that
CHOICE OF THE INTELLIGENT OPTION (FOUNTAIN PEN):
contained no products related to intelligence (i.e., AMC EXPERIMENT 1
movie theater gift certificate, People magazine, Magic 8
Ball, and Now 19 pop music CD). A pretest conducted with
a separate sample from the same population (n p 40) in-
dicated that the intelligent products in each category were
rated significantly more related to intelligence than the non-
intelligent products ( ps ! .001).
All participants then completed the 20-item PANAS (Pos-
itive and Negative Affect Schedule) mood scale (Watson,
Clark, and Tellegen 1988) and rated their self-esteem using
a one-item self-esteem scale (I have high self-esteem,
where 1 p not very true of me; 5 p very true of me; Rob-
ins, Hendin, and Trzesniewski 2001). These measures served
to reduce the viability of alternative accounts related to
mood and self-esteem.
Finally, as our key dependent measure, participants chose
between two gift optionsa fountain pen or a pack of
M&Ms candiesas an additional reward for having taken self-view confidence. Once participants in the nondominant
part in the study. A pretest conducted with a separate sample hand conditions had the opportunity to bolster their self-
from the same population (n p 49) indicated that a fountain view in an intervening task, they were no more likely than
pen was rated as significantly more related to intelligence those in the dominant hand conditions to choose the intel-
than was a pack of candy (F(1, 49) p 94.80, p ! .001) and, ligent gift.
thus, constitutes an intelligence-bolstering choice. The
candy and pen did not bear any similarity to the options Alternative Accounts. An alternative explanation to our
presented in the intervening task, and so the choice of items findings is that participants might have chosen exciting prod-
in the previous task should not affect the desirability of these ucts not as a means to restore self-confidence but instead
items, except for their common intelligence-related associ- as a means of either boosting their lowered self-esteem or
ations. To control for social and normative factors, which liberating themselves from a negative mood state as a result
could have otherwise influenced participants decisions (see of writing with the nondominant hand. We reduced the vi-
Ratner and Kahn 2002; West and Broniarczyk 1998), the ability of this account in two ways. First, the interactive
final choice task occurred completely out of sight and ear- pattern of results on participants final choices reduces the
shot of the other respondents and the experimenters. viability of the mood account. If the mood account had been
valid, one ought to have observed only a main effect of the
handedness factor on the dependent measure. Second, we
Results examined participants responses to the 20-item PANAS
A logistic regression revealed a main effect of the oppor- scale (Watson et al. 1988). We observed a marginal main effect
tunity-for-self-recovery factor (x 2 p 4.56, p ! .03). More of the handedness factor on negative affect (F(1, 76) p 3.62,
important, the analysis revealed a main effect of the hand- p p .06), but importantly, the handedness # opportunity-for-
edness factor (x 2 p 3.57, p p .05) such that writing with self-recovery interactions on both positive and negative af-
the nondominant hand about their intelligence increased par- fect were not significant ( ps 1 .30). Additionally, neither pos-
ticipants likelihood of choosing the fountain pen (65% vs. itive nor negative affect was correlated with choice ( ps 1 .20).
39% in the nondominant and dominant hand conditions, To examine whether an explanation related to self-esteem
respectively) as their take-home gift. Consistent with pre- could account for the findings, we examined participants
dictions, the logistic regression also revealed a marginally responses to the self-esteem measure. Participants self-es-
significant handedness # opportunity-for-self-recovery in- teem scores did not differ across the four conditions (ps 1
teraction (x 2 p 2.64, p p .10). A planned comparison .25), thereby reducing the viability of this alternative account.
showed that the choice of the fountain pen (the intelligent
option) was higher in the nondominant hand/no-self-restor- Discussion
ing-opportunity condition (78%) than in the other three con-
ditions (the average choice share was 45%; b p 1.49, p ! The findings of experiment 1 suggest that writing an essay
.02). Participants likelihood of choosing the fountain pen with ones nondominant hand about how intelligent one is
in the other three conditions did not differ from each other results in a greater likelihood of choosing intelligence-re-
( ps p NS; see fig. 1). These results suggest that writing lated products in a subsequent unrelated task. Because writ-
with the nondominant hand led to a higher likelihood of ing with ones nondominant hand has been previously
choosing self-view-bolstering products only when partici- shown to reduce participants self-view confidence (Brinol
pants did not previously have an opportunity to restore their and Petty 2003), these results provide initial support for our
THE SHAKEN SELF 33

conceptualization that momentary lapses in self-view con- (Shrira and Martin 2005). Participants were given 5 minutes
fidence will lead to a greater tendency to choose self-view- to write (with their dominant hands) about the most im-
bolstering products. portant value in their life, explain why it was important to
In addition, experiment 1 demonstrated the self-view re- them, and describe how they had acted consistently with it
covery function of product choice. After making the self- in the past and planned to act consistently with it in the
restoring choices, participants who suffered from a shaken future. (It is important to note that none of the participants
self (i.e., those who wrote with their nondominant hands) in the self-affirmation condition wrote about health con-
were no more likely to choose a pen (vs. M&Ms) at the end sciousness.) The other half of the participants spent an equiv-
of the experiment than were participants who did not so alent amount of time on a filler task (also with their dominant
suffer. These results suggest that choosing self-view-bol- hands). All participants then completed a set of measures
stering products can serve as a means of restoring self-view such as gender, mood (PANAS; Watson et al. 1988), and
confidence. self-esteem (Robins et al. 2001), many of which served as
covariates in the various analyses. Last, all participants
EXPERIMENT 2 chose between a healthy snack (apple) and an unhealthy
snack (M&Ms).
The purpose of experiment 2 was threefold. First, we
sought to explore whether an indirect self-bolstering strat-
egy, namely, affirming an important but unrelated self-value,
Results
could be another means of coping with a shaken self-con- Manipulation Check. We predicted that writing with
cept, thereby attenuating the effects of lowered self-confi- ones nondominant hand about ones self-view would de-
dence on product choice. According to self-affirmation the- crease self-view confidence. Consistent with this prediction,
ory, individuals can cope with a self-threat through a fluid a one-way ANOVA showed that compared to participants
compensation procedure (Steele 1988), such that a threat writing with their dominant hands (M p 8.48), those writing
in one self-domain can be dealt with through affirmation in with their nondominant hands were indeed less confident in
an unrelated domain. Consistent with this account, recent their written thoughts regarding their health consciousness
research suggests that affirming an important self-value can (M p 7.84; F(1, 124) p 3.95, p ! .05). This result suggests
increase confidence (Brinol et al. 2007). Thus, we expected that writing about health consciousness with the nondomi-
that effects of shaken self-confidence on product choice nant hand did shake participants confidence in their health
would be reduced if individuals affirmed an important but consciousness.
unrelated self-value prior to the final choice.
Another important goal of experiment 2 was to provide Choice Task. We expected participants writing with
direct measures of self-view confidence after the handwrit- their nondominant hands to be more likely to choose the
ing manipulation. Although writing with the nondominant health-conscious option (apple) but only when they did not
hand has been documented to cause lower thought confi- have the opportunity to previously self-affirm. Four partic-
dence (Brinol and Petty 2003), it was not directly measured ipants did not take a snack either because of dietary restric-
in experiment 1. In the current experiment, we included self- tions or a food allergy, which left us with 126 valid data
view confidence measures directly after the handwriting ma- points for analysis. In line with our prediction, a logistic
nipulation to provide support for the efficacy of our ma- regression revealed a significant handedness # self-affir-
nipulation. mation interaction (x 2 p 3.93 , p ! .05), in addition to a
Finally, experiment 2 extended our examination to an- significant main effect of the self-affirmation factor (x 2 p
other self-domainhealth consciousness. A pretest revealed 7.19, p ! .01).
that participants held strong beliefs that they were health- To provide further support for our conceptualization, we
conscious individuals and were confident about these beliefs. conducted planned comparisons using a logistic regression.
Consistent with our conceptualization and as shown in figure
Participants and Procedure 2, the choice of the apple (the healthy option) was higher
in the nondominant hand/no-affirmation condition (66%)
One hundred thirty participants were assigned to the con- than in the other three conditions (the average choice share
ditions of a 2 (handedness: dominant vs. nondominant) # was 41%; b p 1.11, p p .01). Participants likelihood of
2 (self-affirmation: yes vs. no) between-participants factorial choosing the apple in the other three conditions (nondom-
design. The manipulation was similar to that used in ex- inant hand/self-affirmation, dominant hand/self-affirmation,
periment 1. Participants were first asked to write about their dominant hand/no affirmation) did not differ ( ps 1 .40).
health-conscious behaviors with their dominant or nondom- Thus, writing with the nondominant hand led to a higher
inant hand. Directly following the handwriting task, partic- likelihood of choosing self-view-bolstering products only
ipants rated how confident they were in their written thoughts when participants did not previously have an opportunity to
(1 p not at all confident; 10 p extremely confident). write about an important but unrelated self-value. Once par-
Next, manipulated orthogonally to the handwriting ma- ticipants in the nondominant hand conditions had the op-
nipulation was self-affirmation, wherein half of the partic- portunity to affirm their most important self-value, they were
ipants wrote about the most important value in their lives no more likely to choose the healthy gift than were those
34 JOURNAL OF CONSUMER RESEARCH

FIGURE 2 product choices. It also suggests that a direct-bolstering strat-

CHOICE OF THE HEALTHY OPTION (APPLE): EXPERIMENT 2
egy as in experiment 1 need not be the only way individuals
cope with a shaken self-view; individuals can also recover
from a shaken self-view through an indirect-bolstering strat-
egy by affirming another important value related to the self.
One notable finding in experiment 2 was that the medi-
ation analysis conducted using participants thought confi-
dence (i.e., confidence toward their written thoughts) as the
mediator, although in the predicted pattern, was not signif-
icant. One possible reason for this is that, conceptually, con-
fidence in ones self-view, rather than confidence in ones
written thoughts, ought to be a stronger mediator and more
precise test of our hypotheses. Therefore, in the next exper-
iment, we measured both self-view confidence and thought
confidence and tested them separately as mediators.

in the dominant hand conditions. The results suggest that EXPERIMENT 3

an indirect self-view-bolstering strategy, namely, affirming
an important but unrelated self-value, could be another
means of coping with a shaken self-concept, thereby atten- In experiment 3, we sought to replicate the previous find-
uating the effect of lowered self-confidence on product ings by using a manipulation of self-view confidence that
choice. was different than those used in experiments 1 and 2. Spe-
cifically, we directly manipulated self-view confidence with
Mediation Analysis. To provide further support for a a priming task adapted from Petty, Brinol, and Tormala
trait-specific shaken-self account, we conducted a moderated (2002). Participants were instructed to write two essays. The
mediation analysis. Based on a recent work by Preacher, first essay was designed to engender confidence or doubt in
Rucker, and Hayes (2007, model 3), a moderated mediation oneself, at a general level. In the second essay, participants
hypothesis is supported if the path from the independent were asked to write about a particular self-view. Based on
variable to the mediator is significant, and the effect of the prior findings (Petty et al. 2002), we expected that partici-
mediator on the outcome variable depends on the level of pants would attribute the general level of confidence or
a moderator variable. Hence, we expected a significant effect doubt induced by the first essay to the confidence or doubt
of the handedness manipulation on thought confidence and in the specific self-view they wrote about in the subsequent
a thought confidence # self-affirmation interaction on choice. essay.
However, the moderated mediation analysis did not succeed. In our previous studies, participants were asked to choose
Although the confidence manipulation significantly predicted between products that were either relevant or irrelevant to
participants thought confidence (b p .646, p ! .05), the a previously written self-view. For example, participants in
interaction term was not significant (b p 1.40, p ! .2). We experiment 1 chose between a pen and candy after writing
elaborate on this result in the discussion of this experiment. about their intelligence. Although our pretest showed that
a pen is more relevant than candy to intelligence, one could
Alternative Accounts Related to Mood and Self-Es- argue that candy might have been seen as an intelligence-
teem. As in experiment 1, neither the mood nor the self- related product because it contains glucose, which helps in
esteem account seems to be a viable alternative account to mental tasks. To rule out any such potential alternative ac-
our findings. First, only a marginally significant effect of counts, experiment 3 adopted a fully crossed design such
the affirmation factor on positive affect was observed that participants first received a confidence manipulation in
(F(1, 120) p 3.84, p ! .06). Second, the handedness # self- one of two self-view domains and then chose between prod-
affirmation interactions on both positive and negative affect ucts relevant to one or the other self-view. Thus, a particular
were nonsignificant ( ps 1 .65). Additionally, neither posi- product is self-view bolstering when it matches the doubted
tive nor negative affect were correlated with choice ( ps 1 self-view and is nonbolstering when it mismatches it.
.35). Similarly, the independent variables had no significant Another important goal of the present study was to pro-
effects on self-esteem ( ps 1 .65). vide more direct evidence that choosing self-view-bolstering
products can help participants recover from a momentarily
Discussion shaken self-view. To this end, we explicitly measured par-
ticipants self-view confidence before and after the choice
Experiment 2 provides further support that subtle situa- task. We expected that a doubt prime would lead to a de-
tional factors such as writing with the nondominant hand crease in before-choice confidence, but it should not affect
can shake self-confidence and lead to self-view-bolstering participants after-choice confidence.
THE SHAKEN SELF 35

TABLE 1

CHOICE OF COMPETENT AND EXCITING OPTIONS: EXPERIMENT 3 (%)

Activated self-view Product category Confidence priming Doubt priming Choice share change

Competence Magazine 58.6 73.1 14.5

Computer 41.4 61.5 20.1
Gift 58.6 76.9 18.3
Excitement Magazine 38.9 56.3 17.4
Computer 44.4 56.3 11.9
Gift 38.9 43.8 4.9

Participants and Procedure car, shopping with a personal fashion stylist) and three com-
petent gift ideas (a palm pilot, a Seiko watch, and a Mont
One hundred twenty-three participants were randomly as- Blanc fountain pen).
signed to the conditions of a 2 (confidence prime: confidence Finally, all participants completed a purportedly unrelated
vs. doubt) # 2 (self-view activated: exciting vs. competent) self-assessment task in which they rated whether various
between-subjects design. Participants were told that they personality characteristics were self-descriptive (1 p very
were helping out with a research project on human expe- uncharacteristic of me; 7 p very characteristic of me) and
riences. They were asked to write about their behaviors and how confident they were in these self-ratings (1 p not at
self-characteristics in two unrelated domains and answer all confident; 7 p extremely confident). Embedded in these
some questions afterward. The first essay was a confidence- characteristics were our target traits, namely, excitement and
priming task designed to engender confidence or doubt at competence.
a general level. Specifically, one group of participants was
asked to list two experiences in which they felt a great deal
of confidence or certainty; the other group was asked to list Results
two experiences in which they felt a great deal of doubt or Manipulation Check. We expected that writing about
uncertainty. This procedure has been successfully used in ones confident or doubtful life experiences would lead
previous research to manipulate confidence (e.g., Petty et one to feel confident or doubtful about a subsequently
al. 2002). written self-view. Consistent with this prediction, a one-
Next, manipulated orthogonally to the first essay, one of way ANOVA examining before-choice self-view confidence
two self-views was activated through a second writing task. showed that compared to participants writing about confi-
In particular, some participants were asked to list two ex- dent experiences, those writing about doubtful experiences
periences in which they felt that they were an exciting per- were less confident that they were a(n) exciting/competent
son; others were asked to list two experiences in which they person (Mconf p 5.43; Mdoubt p 4.38; F(1, 121) p 11.60,
felt that they were a competent person. Directly following p ! .001). In line with this finding, a one-way ANOVA
the second essay, all participants rated how confident they showed that compared to participants writing about confi-
were that they were a(n) exciting/competent person (1 p dent experiences, those writing about doubtful experiences
not at all confident; 7 p extremely confident). They also were also less confident in the exciting/competent examples
rated how confident they were in their written thoughts con- they wrote (Mconf p 5.63; Mdoubt p 5.11; F(1, 121) p 4.28,
cerning excitement/competence (1 p not at all confident; p ! .05). Thus, the manipulation successfully decreased par-
7 p extremely confident). ticipants confidence in their before-choice self-view and in
Following this procedure was a purportedly unrelated their self-related thoughts.
choice task. Participants were told that they were helping a
group of researchers identify appropriate prize-drawing re- Product Choices. Our prediction was that participants
wards for an upcoming large-scale experiment. Participants would be more likely to choose self-view-bolstering prod-
were asked to indicate their decisions in five domains. A ucts after a doubt prime than after a confidence prime. Spe-
pretest was conducted to identify products with either ex- cifically, we predicted that a doubt prime would lead to more
citement or competence as their primary brand personality exciting product choices and less competent product choices
using Aakers (1997) 55-item Brand Personality Scale. Our when excitement was activated; however, we predicted it
target questionschoice of magazines, computers, and would lead to more competent product choices and less
giftswere the first, third, and fifth questions, respectively. exciting product choices when competence was activated.
Choices of computers included an exciting computer brand Consistent with our conceptualization and as shown in ta-
(Apple) and a competent computer brand (IBM). The mag- ble 1, a doubt prime increased the choice of competent
azine category included two exciting magazines (Rolling products by an average 17.6% when competence was ac-
Stone and Vanity Fair) and two competent magazines tivated, whereas a doubt prime increased the choice of ex-
(Newsweek and Time). Choice of gifts included three ex- citing products by an average of 11.4% when excitement
citing gift ideas (a skydiving tour, 1-day rental of a luxurious was activated. Using an ANOVA to examine the number
36 JOURNAL OF CONSUMER RESEARCH

of competent products that participants chose from the three TABLE 2

target choice sets (possible range p 03) revealed only a
SELF-VIEW CONFIDENCE RATINGS IN EXPERIMENT 3
significant interaction between confidence priming and acti-
vated self-view (F(1, 119) p 8.34, p ! .01). Planned com-
Self-view Condition Before choice After choice
parisons showed that when competence was activated, par-
ticipants chose more competent products in the doubtful Competence Doubt 5.38 6.16
condition (2.12) than in the confident condition (1.59; Confident 6.14 6.22
F(1, 119) p 5.59, p p .02); when excitement was acti- Excitement Doubt 3.56 6.01
vated, participants chose fewer competent products (and, Confident 4.74 6.09
hence, more exciting products) in the doubtful condition
(1.44) than in the confident condition (1.78; F(1, 119) p weak in experiment 2. It is also consistent with the idea that
2.85, p p .09). self-view confidence is a more proximal mediator of these
particular effects than is thought confidence.
Confidence Recovery. We predicted that choosing self-
view-bolstering products would boost participants self-view
confidence, especially for those in the doubt conditions. To Discussion
test this hypothesis, we conducted a repeated-measures
ANOVA comparing before-choice and after-choice self- The results of experiment 3, in which we used new op-
view confidence ratings within the excitement and compe- erationalizations of self-view confidence, were highly con-
tence conditions. The results indicated a significant main sistent with the shaken-self account and the prior studies.
effect of the time factor such that participants were more Compared to participants who were primed at a general level
confident about their written self-view after the choice task with confidence, those primed with doubt chose more ex-
(Fcomp (1, 53) p 8.06, p ! .01; Fexcite (1, 65) p 51.11, p ! citing products when excitement was activated and chose
.001). Moreover, we found a significant interaction between more competent products when competence was activated.
time and confidence manipulation (Fcomp (1, 53) p 5.32, p ! A moderated-mediation analysis further indicated that a mo-
.05; Fexcite (1, 65) p 4.63, p ! .05). Planned comparisons mentary lack of confidence in one self-domain leads to
showed that participants in the confidence and doubt condi- choosing products that are consistent with that specific self-
tions differed significantly in their before-choice confidence domain. Finally, a fully crossed design further ruled out the
ratings (Fcomp (1, 53) p 7.16, p p .01; Fexcite (1, 65) p 7.65, alternative hypothesis that the effect might be due to certain
p ! .01). However, no such difference was found when ex- characteristics of the handwriting manipulation or other
amining after-choice confidence ratings (Fs ! 1). Partici- qualities of the irrelevant products used in previous studies.
pants in the confidence and doubt conditions were equally
confident about their target self-view after the choice task GENERAL DISCUSSION
(see table 2). These findings provide additional support for
the hypothesis that choosing self-view-bolstering products Three studies showed that the confidence with which par-
can help participants recover from a shaken self-view. ticipants hold their self-view can be momentarily shaken by
subtle manipulations, which in turn can lead them to choose
Mediation Analysis. To provide further support for a self-view-bolstering products. In our studies, participants
trait-specific shaken-self account and to understand why we self-view confidence was subtly manipulated such that they
failed to find moderated mediation in experiment 2, we con- would misattribute the doubt originating from an irrelevant
ducted two separate mediation analyses, one using self-view source to the lack of confidence in a specific self-view. Our
confidence and the other using thought confidence as me- results suggest that such subtle manipulations were enough
diators. The results show that moderated mediation succeeded to shake participants self-view confidence and lead to self-
when self-view confidence served as the mediator. The con- bolstering product choices in different self-view domains.
fidence manipulation significantly predicted participants self- Further, we showed that the effects of these subtle confi-
view confidence (b p 1.052, p ! .001), and the effect of self- dence-shaking actions on choice can be eliminated when
view confidence on choice of competent products depended participants are first provided with the opportunity to bolster
on the self-view being activated (b p .235 , p ! .05 ). These their self-concept either directly (through choosing self-
results suggest that the indirect effect of the confidence ma- view-bolstering products) or indirectly (through self-affir-
nipulation on product choices through self-view confidence mation), suggesting a fluid and flexible system against
varies as a function of the self-view being activated, thus threats to ones self-views. These results highlight the im-
supporting a trait-specific shaken-self account. portance of self-view confidence for product choice and
When we used thought confidence as the mediator to illustrate the role of product choices in maintaining stable
conduct mediation analysis, the effect of confidence manip- self-views.
ulation on thought confidence was still significant (b p Prior research suggests that an individuals self-view con-
.528, p ! .05), but the conditional effect of thought confidence fidence is a relatively stable construct. This research shows
on choice was only marginally significant (b p .241 , p ! that it can be momentarily shaken by subtle manipulations.
.10). This accounts for why the moderated mediation was In addition to more blatant threats to ones self-view (e.g.,
THE SHAKEN SELF 37

reflecting on ones past immoral deeds; Zhong and Liljen- ing procedures suggested that participants had no awareness
quist 2006), very subtle actions can affect ones self-view that the confidence manipulation tasks affected their subse-
confidence. This has important implications for understand- quent choices, and so such choices did not appear to stem
ing consumers behavior. First, it substantially extends the from deliberate attempts at self-repair. Because the manipu-
range of instigators of self-view-bolstering consumption to lations in our experiments were rather subtle, it remains plau-
include seemingly innocuous events that slip below the ra- sible that people might consciously attempt to repair doubt
dar. Second, because these threats are so subtle, they are stemming from more blatant confidence-shaking events.
more difficult to counteract. For example, one can minimize Finally, how certain can we be that our manipulations of
the effects of blatant threats to ones self-views through self-view confidence did not affect overall self-esteem, as
strategies such as rationalizing the event (e.g., maybe that prior research might suggest (Brinol and Petty 2003)? Note
exam was nondiagnostic, so Im not so dumb after all), that we found no effects of the independent variables on the
minimizing the importance of the event (e.g., its only one measure of overall self-esteem. A possible reason for the
test, after all), and so forth. When self-concept threats lack discrepancy between our findings and those reported in prior
an overt cause, these strategies are unavailable. Furthermore, research lies in the number of traits that were written about
because such threats can be implicitly induced, marketers in the manipulation (three traits in Brinol and Pettys study
may be less likely to encounter resistance when employing vs. one trait in our studies). Alternatively, the discrepancy
such strategies (e.g., Friestad and Wright 1994). could be due to the different framing of the task. Because
Third, our research also supports the counterintuitive as- the phenomenon is metacognitively driven, the experienced
sertion that products symbolizing certain characteristics are doubt should be directed at the subjective content of the
sometimes chosen most by those who doubt that they pos- thought. Brinol and Pettys participants were asked to write
sess those characteristics. Although some previous research with their nondominant hands about several general positive
has documented how self-doubt can influence information or negative qualities and, hence, should doubt that they have
processing and behavior (e.g., Brinol, DeMarree, and Petty, such qualities, thereby affecting overall self-esteem. Our
forthcoming; Brinol, Petty, and Wheeler 2006), the present participants, however, were asked to write about a single
research shows how self-doubt can not only lead to efforts specific trait, which should (and did) lead to doubt only that
at greater self-certainty but also lead to greater consistency they possessed that specific trait. Future research could fur-
between ones self-view and ones behavior. ther examine when more localized versus more general ef-
We found the greatest levels of self-product congruence fects of self-view confidence arise.
among those with shaken self-views, but we are not sug-
gesting that lower self-view confidence will always be as- CONCLUSION
sociated with larger self-product congruency effects. The
participants in our studies held confident self-views that The John Dewey quote at the beginning of this article
were temporarily cast in doubt. We suspect that those who suggests that the self is in continuous formation through
chronically lack confidence in their self-views may not bol- choice of action. The present studies highlight the validity
ster these self-views through product choice. Research on of this view. Seemingly inconsequential actions can shake
self-verification shows that people often prefer information ones self-confidence, leading one to choose products that
that confirms their self-views but only if they chronically bolster this shaken self-view. By the same token, other seem-
hold the self-views with confidence (Marecek and Mettee ingly inconsequential actions, such as reflecting briefly on
1972; Swann and Pelham 2002). an important value, can restore the self-view and eliminate
Our findings provide several directions for future re- the effects on choice and consumption. Thus, ones self-
search. For example, will the self-view-bolstering effects views are highly susceptible not only to lifes more perni-
that we document under short-term conditions persist over cious slings and arrows but also to more pervasive and subtle
longer time periods? When and how would individuals at- threats, thereby, making continuous the formation and re-
tempt to restore self-view confidence? Note that our research formation of the self.
examined relatively low-involvement choices. Future re-
search could examine whether our findings extend to more
involving decision contexts. Also, note that in our research, REFERENCES
individuals were presented with opportunities to restore a
shaken self-view. Future research could examine when and Aaker, Jennifer (1997), Dimensions of Brand Personality, Jour-
how individuals will actively seek opportunities to reduce nal of Marketing Research, 34 (August), 34757.
the self-threat. Assael, Henry (1981), Consumer Behavior and Marketing Action,
Is the underlying process related to bolstering a shaken Boston: Kent.
Baumgardner, Ann H. (1990), To Know Oneself Is to Like One-
self-view conscious or nonconscious? Note that although self: Self-Certainty and Self-Affect, Journal of Personality
our participants were aware of the confidence that they had and Social Psychology, 58 (June), 106272.
in a particular self-view, we do not know whether they were Belk, Russell W. (1988), Possessions and the Extended Self,
aware that this confidence had been momentarily shaken by Journal of Consumer Research, 15 (September), 13968.
our subtle manipulations. Either way, our funneled debrief- Belk, Russell W., Kenneth D. Bahn, and Robert N. Mayer (1982),
38 JOURNAL OF CONSUMER RESEARCH

Developmental Recognition of Consumption Symbolism, tion in Teenagers Clothing Choices, Journal of Consumer
Journal of Consumer Research, 9 (June), 417. Behavior, 3 (March), 25162.
Bem, Daryl J. (1972), Self-Perception Theory, in Advances in Preacher, Kristopher J., Derek D. Rucker, and Andrew F. Hayes
Experimental Social Psychology, Vol. 6, ed. Leonard Ber- (2007), Addressing Moderated Meditation Hypotheses: The-
kowitz, New York: Academic Press, 162. ory, Methods, and Prescriptions, Multivariate Behavioral Re-
Birdwell, Al E. (1968), A Study of the Influence of Image Con- search, 42 (June), 185227.
gruence on Consumer Choice, Journal of Business, 41 (Jan- Ratner, Rebecca K. and Barbara E. Kahn (2002), The Impact of
uary), 7688. Private versus Public Consumption on Variety-Seeking Behav-
Brendl, Miguel C., Arthur B. Markman, and Claude Messner (2003), ior, Journal of Consumer Research, 29 (September), 24657.
The Devaluation Effect: Activating a Need Devalues Unrelated Robins, Richard W., Holly M. Hendin, and Kali H. Trzesniewski
Objects, Journal of Consumer Research, 29 (March), 46373. (2001), Measuring Global Self-Esteem: Construct Validation
Brinol, Pablo, Kenneth G. DeMarree, and Richard E. Petty (forth- of a Single-Item Measure and the Rosenberg Self-Esteem
coming), Processes by Which Confidence (vs. Doubt) Influ- Scale, Personality and Social Psychology Bulletin, 27 (Feb-
ences the Self, in The Uncertain Self: A Handbook of Per- ruary), 15161.
spectives from Social and Personality Psychology, ed. Robert Rosenberg, Morris (1979), Conceiving the Self, New York: Basic.
M. Arkin, Kathryn C. Oleson, and Patrick J. Carroll, New Schwarz, Norbert (2004), Metacognitive Experiences in Con-
York: Psychology Press. sumer Judgment and Decision Making, Journal of Consumer
Brinol, Pablo and Richard Petty (2003), Overt Head Movements Psychology, 14 (October), 33248.
and Persuasion: A Self-Validation Analysis, Journal of Per- Shrira, Ilan and Leonard L. Martin (2005), Stereotyping, Self-
sonality and Social Psychology, 84 (June), 112339. Affirmation, and the Cerebral Hemispheres, Personality and
Brinol, Pablo, Richard E. Petty, Isamael Gallardo, and Kenneth G. Social Psychology Bulletin, 31 (June), 84656.
DeMarree (2007), The Effects of Self-Affirmation in Non- Sirgy, Joseph (1982), Self-Concept in Consumer Behavior: A
threatening Persuasion Domains: Timing Affects the Process, Critical Review, Journal of Consumer Research, 9 (Decem-
Personality and Social Psychology Bulletin, 33 (11), 153346. ber), 287300.
Brinol, Pablo, Richard E. Petty, and S. Christian Wheeler (2006), Solomon, Michael R. (1983), The Role of Products as Social
Discrepancies between Explicit and Implicit Self-Concepts: Stimuli: A Symbolic Interactionism Perspective, Journal of
Consequences for Information Processing, Journal of Per- Consumer Research, 10 (December), 31929.
sonality and Social Psychology, 91 (January), 15470. Steele, Claude M. (1988), The Psychology of Self-Affirmation:
Campbell, Jennifer D. (1990), Self-Esteem and Clarity of the Self- Sustaining the Integrity of the Self, in Advances in Exper-
Concept, Journal of Personality and Social Psychology, 59 imental Social Psychology, Vol. 21, ed. Leonard Berkowitz,
(September), 53849. San Diego, CA: Academic Press, 261302.
DeMarree, Kenneth G., Richard E. Petty, and Pablo Brinol Swann, William B. and Brett Pelham (2002), Who Wants out
(2007), Self-Certainty: Parallels to Attitude Certainty, In- When the Going Gets Good? Psychological Investment and
ternational Journal of Psychology and Psychological Ther- Preference for Self-Verifying College Roommates, Self and
apy, 7 (2), 14170. Identity, 1 (JulySeptember), 21933.
Friestad, Marian and Peter Wright (1994), The Persuasion Knowl- Tetlock, Philip E., Orie V. Kristel, S. Beth Elson, Melanie C. Green,
edge Model: How People Cope with Persuasion Attempts, and Jennifer S. Lerner (2000), The Psychology of the Un-
Journal of Consumer Research, 21 (January), 131. thinkable: Taboo Trade-Offs, Forbidden Base Rates, and He-
Kassarjian, Harold H. (1971), Personality and Consumer Behav- retical Counterfactuals, Journal of Personality and Social
ior: A Review, Journal of Marketing Research, 8 (Novem- Psychology, 78 (5), 85370.
ber), 40918. Tiedens, Larissa Z. and Susan Linton (2001), Judgment under
Levy, Sidney J. (1959), Symbols for Sale, Harvard Business Emotional Certainty and Uncertainty: The Effects of Specific
Review, 33 (JulyAugust), 11724. Emotions on Information Processing, Journal of Personality
Lewin, Kurt (1935), A Dynamic Theory of Personality, New York: and Social Psychology, 81 (December), 97388.
McGraw-Hill. Watson, David, Lee A. Clark, and Auke Tellegen (1988), De-
Marecek, Jean and David R. Mettee (1972), Avoidance of Con- velopment and Validation of Brief Measures of Positive and
tinued Success as a Function of Self-Esteem, Level of Cer- Negative Affect: The PANAS Scales, Journal of Personality
tainty, and Responsibility for Success, Journal of Personality and Social Psychology, 54 (June), 106370.
and Social Psychology, 22 (April), 98107. West, Patricia M. and Susan M. Broniarczyk (1998), Integrating
Pelham, Brett (1991), On Confidence and Consequences: The Multiple Opinions: The Role of Aspiration Level on Con-
Certainty and Importance of Self-Knowledge, Journal of sumer Response to Critic Consensus, Journal of Consumer
Personality and Social Psychology, 60 (April), 51830. Research, 25 (June), 3851.
Petty, Richard E., Pablo Brinol, and Zakary L. Tormala (2002), Wicklund, Robert A. and Peter M. Gollwitzer (1982), Symbolic
Thought Confidence as a Determinant of Persuasion: The Self-Completion, Hillsdale, NJ: Erlbaum.
Self-Validation Hypothesis, Journal of Personality and So- Zhong, Chen-Bo and Katie Liljenquist (2006), Washing Away
cial Psychology, 82 (May), 72241. Your Sins: Threatened Morality and Physical Cleansing, Sci-
Piacentini, Maria and Greig Mailer (2004), Symbolic Consump- ence, 313 (September), 145152.