Content uploaded by Nicholas Epley
Author content
All content in this area was uploaded by Nicholas Epley on Nov 06, 2018
Content may be subject to copyright.
1159
Mirror, Mirror on the Wall:
Enhancement in Self-Recognition
Nicholas Epley
University of Chicago
Erin Whitchurch
University of Virginia
(Dunning & Cohen, 1992). Negative stereotypes about
others are selectively activated and applied to make one-
self look better (Kunda & Sinclair, 1999). Flattering infor-
mation about the self is accepted readily, whereas
threatening information is evaluated more critically and
ultimately derogated (Ditto & Lopez, 1992; Ditto, Munro,
Apanovitch, Scepansky, & Lockhart, 2003; Ditto,
Scepansky, Munro, Apanovich, & Lockhart, 1998). These
and other deliberate reasoning strategies (Chambers &
Windschitl, 2004) are the tools that frequently enable
people to form a more desirable image of their traits and
abilities than reality might allow.
We sought evidence in this research, however, for a
more direct and potentially automatic version of
enhancement, using a novel method to study biases in
the evaluation of relatively objective self-relevant stim-
uli. In particular, we investigated whether people would
show evidence of enhancement in the recognition of their
own facial image. Do people recognize themselves—their
very own faces—as being more desirable than they actu-
ally are? In addition, are people more calibrated evalu-
ating the attractiveness of strangers about whom they
know relatively nothing?
Authors’ Note: We thank Julia Crowell, Nicholas Josephowitz,
Megan Kolasinski, Julie Ragsdale, Aram Seo, and Audrey Tse for
assisting with data collection; Justin Kruger, Jasmine Kwong, Allen
McConnel, Gillian Rhodes, and Bernd Wittenbrink for helpful com-
ments; and National Science Foundation Grant SES0241544, the
University of Chicago Graduate School of Business, and the
Templeton Foundation for financial support. Please address corre-
spondence to Nicholas Epley, Graduate School of Business, University
of Chicago, 5807 South Woodlawn Avenue, Chicago, IL 60637; e-
mail: epley@chicagogsb.edu.
PSPB, Vol. 34 No. 9, September 2008 1159-1170
DOI: 10.1177/0146167208318601
© 2008 by the Society for Personality and Social Psychology, Inc.
People’s inferences about their own traits and abilities
are often enhancing. A series of experiments suggests
that this enhancement extends to more automatic and
perceptual judgments as well, such that people recognize
their own faces as being more physically attractive than
they actually are. In each experiment, participants’ faces
were made more or less attractive using a morphing pro-
cedure. Participants were more likely to recognize an
attractively enhanced version of their own face out of a
lineup as their own, and they identified an attractively
enhanced version of their face more quickly in a lineup of
distracter faces. This enhancement bias occurred for both
one’s own face and a friend’s face but not for a relative
stranger’s face. Such enhancement was correlated with
implicit measures of self-worth but not with explicit mea-
sures, consistent with this variety of enhancement being a
relatively automatic rather than deliberative process.
Keywords: self-enhancement; recognition; implicit attitudes;
automatic evaluation; self-esteem
People tend to evaluate their own traits and abilities
more favorably than is objectively warranted.
Although debate exists about the magnitude and exact
nature of such biases across both individuals and cul-
tures (Heine, Lehman, Markus, & Kitayama, 1999;
Kwan, John, Kenny, Bond, & Robins, 2004; Sedikides,
Gaertner, & Toguchi, 2003), a large literature docu-
ments the variety of clever ways in which people can
create flattering images of themselves (Dunning, 1999).
Such self-enhancement effects are not simply mindless
attempts for people to “see” what they want to see but
rather represent a more thoughtful—albeit biased—
processing of self-relevant information. Ambiguous traits
are defined in ways that enable favorable self-evaluations
We predicted that people would indeed show an
enhancement bias in self-recognition based on the
emerging consensus that people’s automatic and intu-
itive associations to the self tend to be objectively posi-
tive (see Koole & DeHart, 2007, for a review). People,
for instance, tend to like the letters in their own name
more than letters that are not in their name (Koole,
Dijksterhuis, & van Knippenberg, 2001; Nuttin, 1985)
and are faster to identify positive words and slower to
identify negative words following a self-relevant sub-
liminal prime than a self-irrelevant prime (Spalding &
Hardin, 1999). As a result of such positive associations,
people tend to unknowingly prefer jobs, spouses, and
even grades that bear some resemblance to the self
(Nelson & Simmons, 2007; Pelham, Mirenberg, &
Jones, 2002). Because liking tends to be used as a cue
for familiarity, with liked others being recognized as
more familiar than disliked others (Monin, 2003), we
reasoned that such automatic liking for the self would
likewise bias recognition of one’s own facial image. The
more positive one’s intuitive and automatic assessment
of the self, the more they should recognize positively
enhanced images of their face as their actual face. Such
a bias should not, however, extend to relative strangers
that do not provoke intuitively positive assessments.
People are therefore likely to be more calibrated when
evaluating others’ images than when evaluating their
own image.
Such results—self-enhancement compared to both
reality and evaluations of others—would share many
hallmarks of existing self-enhancement results. We
believe, however, that such a bias in self-recognition
would not only be widely interesting but also scientifi-
cally important for at least three reasons. First, self-
enhancement in the recognition of one’s own image may
represent a distinct form of self-enhancement, one pro-
duced by relatively implicit or automatic psychological
mechanisms rather than by the existing list of deliberate
information-processing mechanisms. Visual recogni-
tion, including facial recognition, is a textbook example
of an automatic psychological process (Grill-Spector &
Kanwisher, 2005), occurring rapidly and intuitively
with little or no apparent conscious deliberation (e.g.,
Thorpe, Fize, & Marlot, 1996). As a result, enhance-
ment biases in self-recognition should be correlated
with automatic or implicit measures of self-worth that
gauge such automatic associations to the self rather
than with explicit and deliberate measures of self-
worth. This relationship with implicit measures would
contrast sharply with existing enhancement results that
are related to explicit measures of self-worth (e.g.,
Brown, 1986; Gramzow & Willard, 2006; Kobayashi
& Brown, 2003) and suggest there may be two rela-
tively independent sources of self-enhancement. One
source would be reflected in the automatic and implicit
associations people possess about themselves, and
another would be reflected in the more explicit and
deliberate attitudes people hold about themselves that
may distort the recruitment and evaluation of ambigu-
ous evidence. This research would therefore suggest that
both automatic and controlled mechanisms can under-
lie objective self-enhancement. In so doing, it would
also advance the developing and expanding literature
on the consequences and importance of implicit atti-
tudes in general, and the consequences and importance
of implicit self-esteem more specifically.
Second, many existing self-enhancement biases
emerge on relatively ambiguous stimuli that must be
constructed or recalled at the time of judgment—from
one’s relative standing on ambiguous traits (Dunning &
Cohen, 1992), to the quality of one’s relationships
(Murray, Holmes, & Griffin, 2004), to the liking of
one’s initials (Koole et al., 2001), to the validity of self-
threatening information (Ditto et al., 2003). There is no
easy method for objectively identifying one’s true lead-
ership ability, the true quality of one’s romantic rela-
tionship, the actual attractiveness of the letter E, or the
true implications of negative performance feedback for
one’s self-evaluation. Other research has demonstrated
that people recall past behavior on objectively verifiable
stimuli in a self-serving fashion (such as prior SAT or
GPA scores; Gramzow & Willard, 2006), but such
results appear to stem from the effortful reconstruction
of memory or deliberately exaggerated self-reports
rather than from a distorted perception of a stimulus
that appears right before people’s eyes. Self-enhanced
recognition of one’s own face, in contrast, would
emerge on a concrete and verifiable feature of the self
that most people see multiple times every day of their
lives. Few aspects about the self are more readily
observable and objectively verifiable than one’s own
facial image. Nevertheless, top-down processes based
on motivations, goals, or expectations have been shown
to influence perception of objective and readily observ-
able stimuli. In one experiment, for instance, partici-
pants judged the slope of a hill to be steeper when they
were encumbered by a heavy backpack than when they
were not so encumbered (Bhalla & Proffitt, 1999). In
another, participants’ motivations to see a particular
object on a screen led them to actually perceive an
ambiguous stimulus in a manner consistent with their
motivated desires (Balcetis & Dunning, 2006). These
findings are consistent with the “new look” approach
to perception that has emphasized the ways in which
top-down processes can influence what people “see”
when they look at the external world. This research
would be the first within this tradition to look at how
top-down processes can influence people’s perceptions
1160 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
of their very own image. It would therefore be the first
to demonstrate that people, quite literally, see them-
selves as more desirable than they actually are.
Third, our experiments involve having people recog-
nize themselves as well as a relative stranger and may
therefore provide additional demonstrations that
people’s evaluations of others can be more calibrated
than evaluations of themselves. For instance, people in
one set of experiments tended to overestimate the likeli-
hood that they would engage in a desirable behavior at
some point in the future, from buying a flower to bene-
fiting a charity to voting in a U.S. presidential election,
but were remarkably well calibrated when predicting
the behavior of a relative stranger (Epley & Dunning,
2000, 2006). The same results also emerged in predic-
tions of relationship longevity, in which those in the
midst of a relationship were more miscalibrated (in the
predictable direction) than were observers (see also
Gagne & Lydon, 2004). Similar miscalibration in this
experiment would perhaps be even more interesting
given that people so clearly have more exposure to, and
experience with, their own image than they do to others.
The miscalibration, however, does not come from the
lack of information about the self compared to others but
rather from the positively distorted nature of the infor-
mation and associations people have about themselves
that they do not have about relatively unknown others.
We tested our main hypotheses—that people would
show enhancement in the recognition of their own facial
image; that such enhancement would be correlated with
implicit, but not explicit, measures of self-worth; and
that people would not show such biases in the recogni-
tion of relative strangers—in three experiments using a
morphing procedure to create more and less attractive
versions of participants’ own facial image. Participants
were then asked to identify their actual face out of a
lineup (Experiments 1 and 2) and that of a relative
stranger and a close friend (Experiment 2), or to recog-
nize their own image from a series of distracters as
quickly as possible (Experiment 3). We predicted that
participants would recognize their own image as being
more desirable than it actually is, that this bias would
also emerge for well-liked friends but not for relative
strangers, and that people would also be faster to iden-
tify positively distorted images of themselves than to
recognize their actual image.
EXPERIMENT 1: SELF-RECOGNITION
Method
Caucasian participants (N=27, 18 female) posed for
a photograph at the end of an unrelated experiment.
The experimenter instructed participants to remove
their glasses and facial piercings and to pull back their
hair if it fell onto their faces. Participants were also
instructed to form a neutral expression for the picture.
These images were then cropped and subjected to a pro-
cedure designed to systematically alter their facial
attractiveness, namely, by morphing their photograph
with a highly attractive or unattractive face.
In this averaging procedure, participants’ facial images
were morphed in 10% increments (up to 50%) with a highly
attractive or unattractive same-gender target face. The
attractive targets (male and female) were composite
images of several dozen faces (obtained at http://www
.uni-regensburg.de/Fakultaeten/phil_Fak_II/Psychologie/
Psy_II/beautycheck/english/durchschnittsgesichter/
durchschnittsgesichter.htm), and the unattractive targets
were individuals suffering from craniofacial syndrome
(obtained at http://www.craniofacial.net/). This morphing
was done by matching up points on the participants’ face
with the identical points on the target’s face (corners of
mouth, shape of eyes, etc.). A 50% morph with the
attractive target would therefore represent a face that was
halfway between their own and the target’s on all of these
critical points. This procedure produced 11 faces, 5 mor-
phed with the attractive target (up to 50%), 5 morphed
with the unattractive target, and the actual photograph
(see Figure 1).1
Participants returned 2 to 4 weeks after taking their
photograph, were seated at a computer, reminded of the
photograph taken several weeks earlier, and told that
they would be presented with a series of images modi-
fied from their original picture. Participants were then
presented with a randomly ordered lineup containing all
11 faces and asked to identify their actual image by
pointing to it on the computer screen. We defined actual
image to participants as the image taken of them in the
prior session. This served as our primary measure of
self-recognition.
Participants then completed a second recognition
measure in which they were shown each picture in iso-
lation on the computer screen (in a random order) and
asked to estimate the likelihood that each image was
their actual image, on a scale ranging from 0% (not at
all likely) to 100% (certain). Finally, participants were
again shown each imagine in isolation (in the same ran-
dom order) and asked to indicate how much they liked
each image on a scale ranging from –5 (bad) to 5
(good).
Participants next completed an implicit measure of
self-worth (Spalding & Hardin, 1999) in which they
were asked to classify words as positive or negative fol-
lowing subliminal (15 ms) presentations of self-relevant
(me, myself) or self-irrelevant (two, manner) primes (see
also Hetts, Sakuma, & Pelham, 1999). The extent to
which self-relevant primes, compared to irrelevant primes,
Epley, Whitchurch / ENHANCED RECOGNITION 1161
speeds the identification of positive words and slows
identification of negative words is taken as an index of
implicit self-worth. Participants then completed the stan-
dard measure of explicit self-esteem—the Rosenberg
(1965) Self-Esteem scale—and a second measure of
implicit self-esteem—the name-letter effect (Nuttin,
1985). Liking one’s own initials more than different ini-
tials, compared to others with different initials, is taken as
a measure of implicit self-esteem (see Baccus, Baldwin, &
Packer, 2004, for a description of the scoring procedure).
We collected two measures of implicit self-esteem, because
of a lack of an accepted standard, to collapse them into a
more reliable composite measure of the underlying latent
variable (Cunningham, Preacher, & Banaji, 2001).
Results
All relevant means and correlations for this experiment
are presented in Table 1.
Face recognition and self-esteem. As predicted, par-
ticipants tended to select one of the faces morphed with
an attractive target as their own out of the lineup more
often than faces morphed with an unattractive target,
t(26) =2.63, p =.02, d=1.01. A frequency distribution
for this measure is presented in Figure 2.
To help identify the source of this bias in self-recognition,
we correlated the image participants recognized as their
own with the implicit and explicit measures of self-
worth. Because the two measures of implicit self-esteem
were significantly correlated with each other, r(25) =
.52, p=.01, we standardized each measure and then
averaged them into a single composite measure. As pre-
dicted, participants’ self-recognition was significantly
correlated with the composite measure of implicit self-
esteem, r(25) =.47, p =.01, but neither positively nor sig-
nificantly correlated with explicit self-esteem, r(25) =
–.26, p=.19. The stronger participants’ automatic
1162 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
Figure 1 Example of the averaging procedure.
Epley, Whitchurch / ENHANCED RECOGNITION 1163
Figure 2 Percentage of participants selecting each face as their own image in the averaging procedure from Experiment 1, Experiment 1
follow-up, and Experiment 2.
0%
5%
10%
15%
20%
25%
30%
35%
40%
50% 40% 30% 20% 10% Actual 10% 20% 30% 40% 50%
Percentage of Participants Selecting Face
Exp. 1
Exp. 1 Follow-Up
Exp. 2
Unattractive Morph Attractive Morph
TABLE 1: Average Enhancement in Self-Recognition and Its Relation to Implicit and Explicit Self-Esteem Across Experiments
Experiment 1 Experiment 1 Follow-Up Experiment 2
M (SD)M (SD)M (SD)
Enhancement in recognition
Averaging procedure
Self 6.30% (12.45)* 12.00% (19.71)* 13.33% (16.33)*
Friend 10.09% (19.00)*
Stranger (experimenters)
Female 1.3% (18.72)
Male 3.3% (29.44)
Correlation with self-recognition
Implicit self-esteem
Name-letter preference 0.44* 0.15 0.44*
Automatic evaluation 0.30 0.52* 0.25
Composite average 0.47* 0.40 0.53*
Explicit self-esteem
Rosenberg Self-Esteem scale −0.26 0.11 −0.03
NOTE: Enhancement in recognition (the top half of the table) is reported as the mean morph level selected by participants in the experiments,
with positive percentages indicating a mean selection morphed in the direction of the positive target (e.g., 10% indicates that participants, on
average, selected the 10% morph with the attractive target). Negative numbers indicate a mean selection morphed in the direction of a negative
target. Standard deviations are reported in parentheses.
*p< .05.
positive associations to the self (indexed by implicit self-
esteem) was, the more enhanced an image participants
recognized as their own.
Likelihood and liking. Participants’ ratings of the
likelihood that each face, rated in isolation, was their
own showed the same conceptual pattern as the self-
recognition measure already reported, with participants
indicating that the more desirable morphs were more
likely to be their own. As can be seen in the top panel of
Figure 3, the average likelihood judgments across the 11
faces showed a roughly bell-shaped curve, with the peak
shifted toward the attractive morphs. Participants indi-
cated that the 20% attractive morph, for instance, was
most likely to be their own (M=65.93%, SD =20.24),
a figure significantly higher than the likelihood ratings
for their actual image (M=54.07%, SD =26.93), paired
t(26) =2.17, p=.04, d=.85. The average estimated like-
lihood assigned to the five attractive morphs was signifi-
cantly higher (M=36.15, SD =15.23) than that assigned
to the five unattractive morphs (M=19.56, SD =12.48),
paired t(26) =6.06, p<.001, d=2.38.
In addition to making these likelihood estimates, par-
ticipants also saw each image in isolation and indicated
how much they liked each image. As can be seen in the
bottom panel of Figure 3, participants tended to like the
images morphed with the attractive faces more than
those morphed with the unattractive faces. More impor-
tant, the average correlation for each participant
between these likelihood and liking ratings was signifi-
cantly positive (M=.42, SD =.30), t(26) =7.38, p<
.001, d=2.89, as was the overall correlation across par-
ticipants between the average likelihood estimate
assigned to each of the 11 images and the average liking
of those images, r(26) =.58, p<.001. These results are
consistent with our proposed mechanism that auto-
matic liking of one’s image is used as a guide for self-
recognition and are consistent with the significant
1164 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
0%
10%
20%
30%
40%
50%
60%
70%
50% 40% 30% 20% 10% Actual 10% 20% 30% 40% 50%
Estimated Likelihood That Face Is One's Own
Unattractive Morph Attractive Morph
–5
–4
–3
–2
–1
0
1
2
3
50% 40% 30% 20% 10% Actual 10% 20% 30% 40% 50%
Liking for Face
hpr
o
M evit
c
arttAhproM
e
vitcar
tt
anU
Figure 3 Average estimated likelihood and liking for each image presented in isolation (Experiment 1).
NOTE: Likelihood estimates were made on a scale ranging from 0% to 100%. Liking estimates were made on a scale ranging from –5 (bad) to
5 (good).
implicit self-esteem correlation observed on the direct
self-recognition measure.
Discussion
As predicted, participants tended to recognize their
own facial image as being more desirable than it actu-
ally was. What is more, this bias was correlated with
implicit measures of self-worth in two procedures, but it
was not correlated with explicit measures of self-worth.
This suggests that automatic positive associations to the
self guide the enhancement observed in this experiment
rather than more deliberate and controlled assessments
of the self, a finding further corroborated by the signifi-
cant relationship between assessed likelihood that an
image is one’s own and liking for that image.
Although consistent with our hypotheses, readers
might be concerned that this experiment used partici-
pants’ actual photograph rather than their mirror image,
and they may entertain alternative interpretations of our
results based on this feature (Mita, Dermer, & Knight,
1977). We see no theoretical reason to expect this to be
an important boundary condition for self-enhanced
recognition. To examine this generalizability directly, we
simply conducted a small follow-up with 15 partici-
pants from the University of Chicago using their mirror
image in the same morphing procedure described in
Experiment 1. As before (see Table 1), participants
tended to select an enhanced image as their own actual
image, t(14) =2.36, p=.03, d =1.26, with an effect size
very similar to that observed in Experiment 1 on this
procedure. This image selection was again positively
correlated (with similar effect size, but nonsignificantly
given the smaller sample size) with the composite mea-
sure of implicit self-esteem, r(12) =.40, p=.15, but not
with explicit self-esteem, r(12) =.11, p=.71 (see Table
1 for all relevant means and correlations for this exper-
iment). Using participants’ actual photograph rather
than their mirror image does not seem to explain the
results of Experiment 1.
EXPERIMENT 2: FRIENDS AND STRANGERS
One hallmark of enhancement biases is that one’s
close friends and relatives get caught up in the act as
well. People in satisfying relationships, for instance, rate
their romantic partners more favorably than those very
partners rate themselves (Murray et al., 2004) and
sometimes even rate their friends more desirably than
people rate themselves (Kenny & Kashy, 1994). Given
such findings for liked others, we expected that the pat-
terns of enhancement in recognition we observed in
Experiment 1 would not be restricted just to one’s own
image, but would extend to close friends as well. It
should not, however, extend to relative strangers. We
tested this prediction directly in Experiment 2.
Method
Harvard University undergraduates (N=24, 17
female) participated in exchange for course credit. As
part of an unrelated study, participants were asked to
bring along a close friend. This two-part experiment
was conducted by three experimenters, one male and
one female in the first session and a new female experi-
menter for the second session. Participants completed
this experiment individually (each participant com-
pleted the first part with either the male or female
experimenter and then finished the study with the
remaining experimenter). At the end of this unrelated
experiment, participants posed for a photograph
against a white wall. Photographs of both participants
as well as both experimenters were then morphed using
the averaging procedure described for Experiment 1.
Participants were invited to return to the lab alone 3
to 4 weeks later by a new experimenter. In this second
session, participants completed the same recognition
procedure used in Experiment 1 for their own image,
for their friend’s image, and for both of the experi-
menters’ images’ (in random order). Finally, partici-
pants completed the three self-esteem measures from
Experiment 1.
Results
Because of computer malfunctions, the composite
self-esteem scores for 2 participants are based only on
the name-letter measure.
As can be seen in Table 1, participants showed evi-
dence of enhancement when identifying their own
image, t(23) =4.12, p <.001, d =1.72, as well as their
friend’s image, ts(23) =2.69, p =.01, d=1.12, but not
when identifying either of the experimenters’ images,
both ts <1. Averaging across the two experimenters’
images, this difference across targets was reliable, F(2,
20) =3.65, p=.04, η2=.27. Follow-up contrasts
revealed that the observed recognition bias did not dif-
fer between self and friend, F<1; was significantly dif-
ferent between self and the average of the
experimenters’ images, F(1, 21) =5.47, p=.03, η2=
.21; and was marginally significant between the friend’s
image and the average of the experimenters’ images,
F(1, 21) =3.95, p=.06, η2=.16.
As in Experiment 1, self-enhancement was signifi-
cantly correlated with the composite measure of implicit
self-esteem, r(19) =.53, p <.01, but not with explicit
self-esteem, r(19) =–.03, ns. Given that the self was the
target of these measures (rather than the friend or the
Epley, Whitchurch / ENHANCED RECOGNITION 1165
experimenters), it is not surprising that neither implicit
nor explicit measures of self-esteem correlated system-
atically with enhancement of one’s friend or the experi-
menters (ps >.4).
Discussion
Experiment 2 again provides support for our main pre-
diction that people would show evidence of an enhance-
ment bias in self-recognition, a bias that was related to
implicit or automatic measures of self-worth rather than
to explicit or deliberate measures. Experiment 2 also
demonstrates that such enhancement bias extends to
familiar and liked others as well, in this case to one’s close
friends, but does not extend to unfamiliar strangers.
Consistent with other existing research, participants
tended to show evidence of enhancement when evaluating
their close friends to the same extent as when evaluating
themselves (Kenny & Kashy, 1994).
There is one lower level ambiguity that this experi-
ment does not completely address, namely, whether the
enhancement in recognition observed in this experiment
was due to liking or familiarity. People tend to like both
themselves and their friends, but they are also more
familiar with both themselves and their friends (com-
pared to strangers). These two variables are close trav-
eling companions in daily life, as people tend to feel
they are familiar with liked others (Monin, 2003) and
simple exposure to others tends to increase liking
(Moreland & Zajonc, 1982). Although the simple com-
parison between friends and strangers in this experi-
ment cannot distinguish between these two mechanistic
accounts, the relationship between implicit self-esteem
and self-recognition clearly suggests that liking is play-
ing a larger role in the effects observed in these experi-
ments than simple familiarity (see also Note 1).
Existing research demonstrates that the positive associ-
ations indexed by these implicit self-esteem measures are
more than simple measures of exposure or familiarity, and
instead index favorable conditioned associations toward a
target (in this case, toward the self; Baccus et al., 2004;
Dijksterhuis, 2004; Jones, Pelham, Mirenberg, & Hetts,
2002). Distinguishing the impact of familiarity versus lik-
ing in enhanced recognition for others would require mea-
suring automatic evaluative associations to these others or
identifying targets that are equivalent in terms of liking
but vary in familiarity or who vary in liking but are equal
in familiarity. The self-recognition results from these
experiments, and their relation to implicit measures of
self-worth, suggest that implicitly disliked targets, for
instance, should show evidence of derogation in recogni-
tion. Indeed, one recent experiment using a procedure
nearly identical to Experiment 2 showed that people
who tend to dislike themselves—namely, those with body
dysmorphic disorder—tend to recognize themselves as
being less attractive than they actually are (Clerkin &
Teachman, in press). Whether implicitly disliked targets
show evidence for negative distortion is an interesting
avenue for further research and may shed light on likely
sources of variability and boundary conditions for recog-
nition biases in judgment.
EXPERIMENT 3: SPEED OF RECOGNITION
Although we find the influence of familiarity on
recognition biases to be interesting, we felt it more pru-
dent in the scope of the present research to provide
stronger evidence for the phenomenon of enhanced self-
recognition itself. In particular, all of the results pre-
sented thus far have relied on self-reported recognition
as an index of actual recognition. People are willing to
report that an attractively enhanced image is truly their
own, but we sought more evidence in Experiment 3 that
people truly believe what they report. Although such
self-report biases would not explain the pattern of cor-
relations with implicit and explicit self-esteem, they
might serve as an alternative account for the overall
enhancement bias observed in self-recognition. In par-
ticular, we sought evidence for self-enhancement from a
less reactive measure that uses the speed with which
people can identify an image as an indication of their
recognition. In this experiment we asked participants to
identify, as quickly as possible, their own, an attrac-
tively enhanced, or an unattractively enhanced photo of
themselves out of a series of lineups composed of dis-
tracter faces. We reasoned that people would recognize
a face that they truly believed to be their own more
quickly than a face that they did not truly believe was
their own. Given the results of Experiments 1 and 2, we
predicted a linear trend—that participants would iden-
tify the attractively enhanced version the fastest, their
actual image more slowly, and their unattractively
altered version the slowest.
This new methodology also allows us to test the
plausibility of an alternative interpretation of the results
of Experiments 1 and 2 based on the similarity between
the extreme targets used in the morphing procedure and
participants’ actual faces. In particular, it is possible that
participants’ actual faces resemble the highly attractive
target more than they resemble the highly unattractive
target. If one’s actual image is more similar to the attrac-
tive morph than it is to the unattractive morph, the range
of morphs produced with the attractive target may sim-
ply be smaller than the range produced with the unat-
tractive target. Any error in recognition is therefore
likely to lead to biased responses in the direction of the
attractive target. This alternative interpretation would
1166 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
presumably apply to all faces rather than simply one’s
own, and the difference in recognition for self versus the
relative stranger in Study 2 therefore seems inconsistent
with this alternative. So too are the likelihood estimates
reported in Figure 2 inconsistent with this interpretation,
as the range of likelihood estimates is directionally larger
across the attractive faces than it is across the unattrac-
tive faces, showing a clear ability to differentiate
between the images. Nevertheless, the new procedure in
Experiment 3 allows us to test this alternative directly
by measuring, through reaction times, participants’
ability to distinguish among the morphed images. If the
range of attractive morphs is small and is therefore hard
to distinguish among them, participants should show a
smaller difference in recognition speed between the
attractive morph and their actual face (or perhaps no
difference at all) than between their unattractive morph
and their actual face.
Method
University of Chicago undergraduates (N=20, 17
female) participated in exchange for $6. As in the pre-
vious experiments, participants arrived in the labora-
tory for an unrelated experiment and had their
photograph taken before leaving. Participants’ facial
images were then morphed using the averaging proce-
dure from Experiments 1 and 2. After returning to the
lab, participants were seated in front of a blank com-
puter screen, told that they would be shown an array of
12 photographs (1 of which was their own), that they
were to identify their own image as quickly as possible,
and that they would see a total of 12 such arrays. Each
array included 1 image of the actual participant (in a
random location)—either their actual image, the 20%
attractive morph, or the 20% unattractive morph—and
11 same-gender distracter faces (taken from 33 partici-
pants in previous experiments). Each of the three target
images (attractive, actual, unattractive) was therefore
presented four times in a predetermined random order.
Participants were told to press the space bar as soon as
they spotted their own image, and to then point to their
image on the screen that they recognized to be their
own. We used the speed with which participants pressed
the space bar as an indication of their recognition speed.
When finished with all 12 arrays, participants were
thanked, paid, and dismissed.
Results and Discussion
Reaction times more than 3 SD from the mean of a
particular trial were removed on an item-by-item basis
(3.5% of observations in total). Participants saw each of
the target images four times, and we calculated the average
speed with which they recognized each of the target
images.
A repeated measures ANOVA on participants’ aver-
age reaction times to recognize the three target faces
revealed a significant effect consistent with our pre-
dicted linear trend, F(2, 18) =4.43, p=.03, η2=.33.
Participants recognized their attractive morph most
quickly (M=1,892.09 ms, SD =442.53), their actual
image more slowly (M=2,068.44 ms, SD =559.47),
and their unattractive morph the most slowly (M=
2,133.12 ms, SD =460.11). Follow-up contrasts
revealed that participants were marginally faster to rec-
ognize the attractive morph than their actual image,
F(1, 19) =3.35, p=.08, η2=.15, and were significantly
faster to recognize the attractive morph than the unat-
tractive morph, F(1, 19) =6.47, p=.02, η2=.25. The
difference in reaction times for participants to recognize
their actual image and the unattractive morph was non-
significant, F<1. Notice that this result is not only
inconsistent with what an alternative interpretation
based on range restrictions between participants’ actual
image and the morphed targets would predict, it is pre-
cisely the opposite of what such an alternative would
predict. People recognize objects more quickly when
those objects match their mental representations, and
these results clearly suggest that people’s mental repre-
sentations of their own facial image are, on average,
more desirable than their image actually appears in
reality.
GENERAL DISCUSSION
These experiments demonstrate that people tend to
show evidence of self-enhancement in self-recognition,
that such self-enhancement extends to close friends but
not relative strangers, and that it is related to automatic
and implicit measures of self-worth rather than deliber-
ate and explicit measures. These experiments demon-
strate that self-enhancement can emerge during the
relatively automatic and perceptual process of face
recognition, and the pattern of correlations with
implicit measures of self-worth suggest a version of self-
enhancement potentially distinct from existing demon-
strations of positive illusions that are consistently
correlated with explicit measures.
The methodology used to derive these conclusions is
novel and immediately seems to raise at least two inter-
esting issues. First, the methodology in these experiments
bears some relation to measures of implicit self-esteem
and raises questions about whether these results reflect
another measure of implicit self-esteem or a consequence
of implicit self-esteem. We believe it is a consequence,
Epley, Whitchurch / ENHANCED RECOGNITION 1167
rather than another measure, of implicit self-esteem.
Psychological measures of latent constructs are designed
to tap the central and defining features of the construct
itself. Implicit self-esteem is theoretically defined as the
positivity of automatic or intuitive associations to the self
(Koole & DeHart, 2007) and is therefore indexed by
measures of liking for things associated with the self
(such as liking for the initials of one’s name; Dijksterhuis,
2004; Koole et al., 2001) or by the valence of associations
automatically activated when primed with the self-con-
cept (e.g., Hetts et al., 1999; Spalding & Hardin, 1999).
Because liking is used as a cue for recognition (Monin,
2003), we believe that enhancement bias in self-recognition
is one causal step removed from the more basic underly-
ing measure of implicit self-regard and therefore is better
conceptualized as a psychological phenomenon rather
than as a psychological measure.
Second, novel methodologies such as the one used
here raise concerns about procedural artifacts that may
provide alternative interpretations of an experiment’s
results. Although we took great care to avoid obvious
confounds, readers may be concerned that the images
may have somehow differed in luminance, blurriness, or
some other aspect systematically across the experi-
ments. We have no evidence to suggest such procedural
artifacts but notice that any such procedural concerns
would need to account for the breadth of findings
across the entire set of experiments—not simply the
mean-level bias in recognitions observed for self and
friends (Experiments 1 and 2), for one’s actual image
and one’s mirror image (follow-up to Experiment 1), or
in the speed of recognition (Experiment 3), but also for
the consistent pattern of correlations with implicit ver-
sus explicit measures of self-worth (and the symmetry
procedure mentioned in Note 1).
One alternative interpretation that might systemati-
cally be able to explain this entire pattern, and one we
therefore believed required additional empirical investiga-
tion, involves people’s actual attractiveness. People who
are actually more attractive may have more difficulty dis-
tinguishing between images morphed with the attractive
target simply because the objective difference between an
attractive person and the attractive target used in these
experiments would be smaller. Attractive participants
might therefore false alarm to the attractively enhanced
images more often. If attractive people also have higher
implicit self-worth, such an alternative could account for
the pattern we obtained. Although we mentioned that a
similar alternative interpretation does not seem to account
for the pattern of results observed in Experiment 3 across
all participants, our concern here has more to do with the
individual-level correlations between participants.
We examined this alternative by having separate
groups of participants (between 20 and 24 for each
experiment) rate the actual attractiveness of each origi-
nal image used in the experiments. Raters evaluated the
participants from only one of the experiments included
in this article, with images of the original participants
appearing in a random order. No consistent relationship
emerged between average ratings of actual attractive-
ness and the photo participants originally selected as
their own (rs =–.37, –.15, and .29, for Experiment 1,
Experiment 1 follow-up, and Experiment 2, respec-
tively, all ps >.05) or in the speed with which they iden-
tified any of the images in Experiment 3 (ps >.3).
Average ratings of attractiveness were also not consis-
tently related to the composite measure of implicit self-
esteem across the three experiments (ps >.2). Differences
in actual attractiveness do not seem to account for the
results we observed.
Although these results are generally consistent with
the general picture of self-enhancement that emerges
from the social psychological literature, they appear at
odds with empirical results regarding body satisfaction
more generally. In particular, both men and women tend
to show consistent dissatisfaction with their body image
(Powell, Matacin, & Stuart, 2001) and may even hold
negatively distorted perceptions of their own bodies.
Women (especially those with eating disorders) asked to
represent their waist circumference with a rope, for
instance, tend to create a circle that may be as little as
half the size of their actual waist (for a review, see Cash
& Deagle, 1997). The resolution, we suspect, may come
in the automatic and implicit versus relatively con-
trolled and deliberate nature of the measures used in
these differing experiments. The self-enhancing results we
report here emerge on recognition measures that are pro-
totypic examples of automatic psychological processes,
whereas body dissatisfaction results emerge in more
deliberate and explicit measures. Similar dissociations
between implicit and explicit self-views have emerged in
cross-cultural studies of self-enhancement between
Asian and Western Europeans. Although explicit mea-
sures show consistent evidence of self-criticism and dep-
recation among East Asians compared to Western
Europeans (Heine et al., 1999), their implicit self-assess-
ments are predictably favorable (Hetts et al., 1999;
Kitayama & Karasawa, 1997; Yamaguchi et al., 2007).
Whether a similar dissociation would emerge in percep-
tions of body image more generally is an interesting
avenue for further research.
More generally, the relation between self-enhance-
ment and implicit self-worth in these experiments sug-
gests a potentially distinct version of self-enhancement
from existing demonstrations of positive illusions that
are consistently correlated with explicit measures.
Implicit attitude measures tend to reflect consistent and
repeated associations in the environment, producing
1168 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
automatic associations through classical conditioning
that can then extend to related but unconditioned stim-
uli. People structure their lives in ways that tend to
increase the likelihood of positive associations to the
self (or close friends and loved ones), and stimuli asso-
ciated with the self are therefore automatically evalu-
ated positively (Beggan, 1992; Pelham, Carvallo, &
Jones, 2005). The experiments here suggest that such
associations can create positive distortions of the self as
well, a top-down influence that not only influences how
people think about themselves but also what people see
when they look directly at themselves. It is perhaps of
little wonder, then, that people so rarely seem to like the
photographs taken of themselves. The image captured
by the camera lens just does not match up to the image
captured in the mind’s eye.
NOTE
1. Participants also completed a second procedure originally
intended to allow a conceptual replication of the correlation with
implicit versus explicit self-esteem. In particular, we created a series of
morphed images increasing in symmetry. Existing research demon-
strates that people tend to like symmetrical images of themselves
(Little & Jones, 2003), and symmetry is generally perceived as being
attractive (Rhodes, 2006). In this symmetry procedure, participants’
facial images were split in half vertically, the right side flipped to cre-
ate a perfectly symmetrical face, and participants’ actual image was
morphed with this symmetrical image in 10% increments up to 90%,
resulting in 11 total images. Because this procedure only increases
symmetry rather than decreases it, mean biases in the recognized
image cannot be taken as clear evidence of self-enhancement, as any
random error in the selection of faces can only lead to a selection of
an enhanced face. We included this procedure only to serve as a pos-
sible conceptual replication for the relation between recognition and
implicit versus explicit self-esteem.
Recent evidence, however, suggests that this chimeric procedure
can slightly decrease the apparent attractiveness of the face due to
anomalies introduced by asymmetrical facial features (for a review,
see Rhodes, 2006). We tried to eliminate these concerns by removing
three people who had obvious facial marks that would appear
bizarre with this procedure, but we cannot rule out artifacts alto-
gether. This procedure, unbeknown to us at the time, therefore con-
tains some ambiguities that may make it less than ideal for our
purposes.
That said, results from this symmetry procedure were at least con-
sistent with our predictions. As with the averaging procedure, partic-
ipants’ tendency to select the more symmetrical self-images was
significantly correlated with the composite measure of implicit self-
esteem, r(22) =.40, p =.05, but not with explicit self-esteem, r=.18,
p=.40. Participants tended to select a face that was significantly more
enhanced than their own (M=26.25% morph with perfectly sym-
metrical image), t(23) =4.68, p<.001, d=1.95.
REFERENCES
Baccus, J. R., Baldwin, M. W., & Packer, D. J. (2004). Increasing
implicit self-esteem through classical conditioning. Psychological
Science, 15, 498-502.
Balcetis, E., & Dunning, D. (2006). See what you want to see:
Motivational influences on visual perception. Journal of
Personality and Social Psychology, 91, 612-625.
Beggan, J. K. (1992). On the social nature of nonsocial perception:
The mere ownership effect. Journal of Personality and Social
Psychology, 62, 229-237.
Bhalla, M., & Proffitt, D. R. (1999). Visual-motor recalibration in geo-
graphical slant perception. Journal of Experimental Psychology:
Human Perception and Performance, 25, 1076-1096.
Brown, J. D. (1986). Evaluations of self and others: Self-enhancement
biases in social judgments. Social Cognition, 4, 353-376.
Cash, T. F., & Deagle, E. A., III (1997). The nature and extent of body
image disturbances in anorexia nervosa and bulimia nervosa.
International Journal of Eating Disorders, 22, 107-125.
Chambers, J. R., & Windschitl, P. D. (2004). Biases in social compar-
ative judgments: The role of nonmotivated factors in above-aver-
age and comparative optimism effects. Psychological Bulletin,
130, 813-838.
Clerkin, E. M., & Teachman, B. A. (in press). Perceptual and cogni-
tive biases in individuals with body dysmorphic disorder symp-
toms. Cognition and Emotion.
Cunningham, W. A., Preacher, K. J., & Banaji, M. R. (2001). Implicit
attitude measures: Consistency, stability, and convergent validity.
Psychological Science, 12, 163-170.
Dijksterhuis, A. (2004). I like myself but I don’t know why:
Enhancing implicit self-esteem by subliminal evaluative condition-
ing. Journal of Personality and Social Psychology, 86, 345-355.
Ditto, P. H., & Lopez, D. F. (1992). Motivated skepticism: The use of dif-
ferential decision criteria for preferred and nonpreferred conclusions.
Journal of Personality and Social Psychology, 63,568-584.
Ditto, P. H., Munro, G. D., Apanovitch, A. M., Scepansky, J. A., &
Lockhart, L. K. (2003). Spontaneous skepticism: The interplay of
motivation and expectation in responses to favorable and unfa-
vorable medical diagnoses. Personality and Social Psychology
Bulletin, 29, 1120-1132.
Ditto, P. H., Scepansky, J. A., Munro, G. D., Apanovich, A. M., &
Lockhart, L. K. (1998). Motivated sensitivity to preference-incon-
sistent information. Journal of Personality and Social Psychology,
75, 53-69.
Dunning, D. (1999). A newer look: Motivated social cognition and
the schematic representation of social concepts. Psychological
Inquiry, 10, 1-11.
Dunning, D., & Cohen, G. L. (1992). Egocentric definitions of traits
and abilities in social judgment. Journal of Personality and Social
Psychology, 63, 341-355.
Epley, N., & Dunning, D. (2000). Feeling “holier than thou”: Are
self-serving assessments produced by errors in self or social
prediction? Journal of Personality and Social Psychology, 79,
861-875.
Epley, N., & Dunning, D. (2006). The mixed blessings of self-knowledge
in behavioral prediction: Enhanced discrimination but exacerbated
bias. Personality and Social Psychology Bulletin, 32, 641-655.
Gagne, F. M., & Lydon, J. E. (2004). Bias and accuracy in close rela-
tionships: An integrative review. Personality and Social
Psychology Review, 8, 322-338.
Gramzow, R. H., & Willard, G. (2006). Exaggerating current and
past performance: Motivated self-enhancement versus reconstruc-
tive memory. Personality and Social Psychology Bulletin, 32,
1114-1125.
Grill-Spector, K., & Kanwisher, N. (2005). Visual recognition: As
soon as you know it is there, you know what it is. Psychological
Science, 16, 152-160.
Heine, S. J., Lehman, D. R., Markus, H. R., & Kitayama, S. (1999).
Is there a universal need for positive self-regard? Psychological
Review, 106, 766-794.
Hetts, J. J., Sakuma, M., & Pelham, B.W. (1999). Two roads to posi-
tive regard: Implicit and explicit self-evaluation and culture.
Journal of Experimental Social Psychology, 35, 512-559.
Jones, J. T., Pelham, B. W., Mirenberg, M. C., & Hetts, J. J. (2002).
Name letter preferences are not merely mere exposure: Implicit ego-
tism as self-regulation. Journal of Experimental Social Psychology,
38, 170-177.
Kenny, D. A., & Kashy, D. A. (1994). Enhanced coorientation in the
perception of friends: A social relations analysis. Journal of
Personality and Social Psychology, 67, 1024-1033.
Epley, Whitchurch / ENHANCED RECOGNITION 1169
Kitayama, S., & Karasawa, M. (1997). Implicit self-esteem in Japan:
Name letters and birthday numbers. Personality and Social
Psychology Bulletin, 23, 736-742.
Kobayashi, C., & Brown, J. D. (2003). Self-esteem and self-enhancement
in Japan and America. Journal of Cross-Cultural Psychology, 34,
567-580.
Koole, S. L., & DeHart, T. (2007). Self-affection without self-reflec-
tion: Origins, models, and consequences of implicit self-esteem. In
C. Sedikides & S. Spencer (Eds.), The self in social psychology
(pp. 36-86). New York: Psychology Press.
Koole, S. L., Dijksterhuis, A., & van Knippenberg, A. (2001). What’s
in a name: Implicit self-esteem. Journal of Personality and Social
Psychology, 80, 614-627.
Kunda, Z., & Sinclair, L. (1999). Motivated reasoning with stereo-
types: Activation, application, and inhibition. Psychological
Inquiry, 10, 12-22.
Kwan, V. S. Y, John, O. P., Kenny, D. A., Bond, M. H., & Robins, R. W.
(2004). Reconceptualizing individual differences in self-enhance-
ment bias: An interpersonal approach. Psychological Review, 111,
94-110.
Little, A. C., & Jones, B. C. (2003). Evidence against perceptual bias views
for symmetry preferences in human faces. Proceedings of the Royal
Society of London Series B–Biological Sciences, 270, 1759-1763.
Mita, T. H., Dermer, M., & Knight, J. (1977). Reversed facial images
and the mere-exposure hypothesis. Journal of Personality and
Social Psychology, 35, 597-601.
Monin, B. (2003). The warm glow heuristic: When liking leads to
familiarity. Journal of Personality and Social Psychology, 85,
1035-1048.
Moreland, R. L., & Zajonc, R. B. (1982). Exposure effects in person
perception: Familiarity, similarity, and attraction. Journal of
Experimental Social Psychology, 18, 395-415.
Murray, S. L., Holmes, J. G., & Griffin, D. W. (2004). The benefits of
positive illusions: Idealization and the construction of satisfaction
in close relationships. In H. T. Reis & C. E. Rusbult (Eds.), Close
relationships: Key readings (pp. 317-338). Philadelphia: Taylor &
Francis.
Nelson, L. D., & Simmons, J. P. (2007). Moniker maladies: When
names sabotage success. Psychological Science, 18, 1106-1112.
Nuttin, J. M. (1985). Narcissism beyond Gestalt and awareness: The
name letter effect. European Journal of Social Psychology, 15,
353-361.
Pelham, B. W., Carvallo, M., & Jones, J. T. (2005). Implicit egotism.
Current Directions in Psychological Science, 14, 106-110.
Pelham, B. W., Mirenberg, M. C., & Jones, J. K. (2002). Why Susie
sells seashells by the seashore: Implicit egotism and major life deci-
sions. Journal of Personality and Social Psychology, 82, 469-487.
Powell, J. L., Matacin, M. L., & Stuart, A. E. (2001). Body esteem:
An exception to self-enhancing illusions? Journal of Applied Social
Psychology, 31, 1951-1978.
Rhodes, G. (2006). The evolutionary psychology of facial beauty.
Annual Review of Psychology, 57, 199-226.
Rosenberg, M. (1965). Society and the adolescent self-image.
Princeton, NJ: Princeton University Press.
Sedikides, C., Gaertner, L., & Toguchi, Y. (2003). Pancultural self-
enhancement. Journal of Personality and Social Psychology, 84, 60-79.
Spalding, L. R., & Hardin, C. D. (1999). Unconscious unease and self-
handicapping: Behavioral consequences of individual differences in
implicit and explicit self-esteem. Psychological Science, 10, 535-539.
Thorpe, S., Fize, D., & Marlot, C. (1996). Speed of processing in the
human visual system. Nature, 381, 520-522.
Yamaguchi, S., Greenwald, A. G., Banaji, M. R., Murakami, F., Chen,
D., Shiomura, K, et al. (2007). Apparent universality of positive
implicit self-esteem. Psychological Science, 18, 498-500.
Received July 7, 2007
Revision accepted February 1, 2008
1170 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN