Empathy neglect: Reconciling the spotlight effect and the correspondence bias

Abstract

When people commit an embarrassing blunder, they typically overestimate how harshly they will be judged by others. This tendency can seem to fly in the face of research on the correspondence bias, which has established that observers are, in fact, quite likely to draw harsh dispositional inferences about others. These seemingly inconsistent literatures are reconciled by showing that actors typically neglect to consider the extent to which observers will moderate their correspondent inferences when they can easily adopt an actor's perspective or imagine being in his or her shoes. These results help to explain why actors can overestimate the strength of observers' dispositional inferences even when, as the literature on the correspondence bias attests, observers are notoriously prone to drawing those very inferences.

Full text

Empathy Neglect: Reconciling the Spotlight Effect
and the Correspondence Bias
Nicholas Epley
Harvard University
Kenneth Savitsky
Williams College
Thomas Gilovich
Cornell University
When people commit an embarrassing blunder, they typically overestimate how harshly they will be
judged by others. This tendency can seem to fly in the face of research on the correspondence bias, which
has established that observers are, in fact, quite likely to draw harsh dispositional inferences about others.
These seemingly inconsistent literatures are reconciled by showing that actors typically neglect to
consider the extent to which observers will moderate their correspondent inferences when they can easily
adopt an actor’s perspective or imagine being in his or her shoes. These results help to explain why actors
can overestimate the strength of observers’ dispositional inferences even when, as the literature on the
correspondence bias attests, observers are notoriously prone to drawing those very inferences.
Nearly everyone has experienced the fear of social evaluation
following an embarrassing mishap. One may worry about being
forever labeled a klutz following an ungraceful turn on the dance
floor, as disheveled if caught wearing mismatched socks, or as
thoughtless after attending a birthday party without bringing a gift.
These fears typically have relatively mild consequences, such as
worry, self-doubt, and occasionally awkward attempts to repair
face (Goffman, 1959). At other times, however, these fears can
give rise to more debilitating problems, including shyness (Zim-
bardo, 1990), social anxiety (Clark & Arkowitz, 1975), and para-
noia (Fenigstein & Vanable, 1992). Indeed, concerns over lost face
have been linked to teenage suicide (Shafer, 1974, 1988), domestic
violence (Cohen, Nisbett, Bowdle, & Schwartz, 1996), homicide
(Daly & Wilson, 1988), and even accelerated HIV progression
(Cole, Kemeney, & Taylor, 1997).
These fears are clearly important, but are they justified? Previ-
ously, we have shown that they are not—that people typically
overestimate how harshly they will be judged for their failures and
social blunders (Savitsky, Epley, & Gilovich, 2001). For example,
people who imagined accidentally tripping a security alarm in the
presence of others thought they would be judged more negatively
than they actually were. So too with people who imagined showing
up at a party without a gift, people who failed a particularly
difficult test in front of others, and people who were introduced to
a stranger as someone who experiences “occasional difficulties
with bed wetting.”
This miscalibration is further compounded by a related tendency
for people to overestimate the extent to which others notice and
attend to their appearance and behavior in the first place, a ten-
dency dubbed the spotlight effect (Gilovich, Medvec, & Savitsky,
2000). Of course, people cannot be judged harshly if their seem-
ingly obvious transgressions go unnoticed. Collectively, these
findings suggest that people’s fears over others’ harsh recrimina-
tions are generally exaggerated. Not only are people less inclined
than we think to see the worst in us, they are also less inclined to
see us at all.
At first blush, this conclusion may strike the readership of the
Journal of Personality and Social Psychology as implausible.
After all, more than a generation of attribution research testifies to
observers’ readiness to draw strong dispositional inferences on the
basis of even the weakest behavioral evidence (Gilbert & Jones,
1986; Gilbert & Malone, 1995; Jones & Harris, 1967; Ross, 1977;
Ross, Amabile, & Steinmetz, 1977), a tendency so common and
reflexive that it has been labeled the fundamental attribution error
(Ross, 1977). If observers are so eager to jump from acts to
dispositions, is it plausible to contend that actors nonetheless
overestimate observers’ dispositionalist tendencies?
Recent evidence suggests that it is indeed. Van Boven, Kamada,
and Gilovich (1999) asked individuals whose actions were con-
strained by experimental instructions to anticipate the inferences
Nicholas Epley, Department of Psychology, Harvard University; Ken-
neth Savitsky, Department of Psychology, Williams College; Thomas
Gilovich, Department of Psychology, Cornell University.
This research was supported by National Science Foundation Research
Grant SBR9809262. Portions of this research were presented at the annual
meetings of the American Psychological Society, Denver, Colorado, June
1999 and the Society for Experimental Social Psychology, St. Louis,
Missouri, October 1999.
We thank Laura Currin, Ellyn Poltrock, Debbie Share, and Kevin Van
Aelst for their help in collecting these data, Dennis Regan for commenting
on a version of this article, and Leaf Van Boven for helpful suggestions
throughout this research.
Correspondence concerning this article should be addressed to Nicholas
Epley, Department of Psychology, William James Hall, 33 Kirkland Street,
Harvard University, Cambridge, Massachusetts 02138. E-mail:
epley@wjh.harvard.edu, ksavitsk@williams.edu, or tdg1@cornell.edu
Journal of Personality and Social Psychology Copyright 2002 by the American Psychological Association, Inc.
2002, Vol. 83, No. 2, 300–312 0022-3514/02/$5.00 DOI: 10.1037//0022-3514.83.2.300
300

made about them by observers. These were then compared with
observers’ actual inferences. As in previous research, the observers
drew dispositional inferences despite full awareness of the actors’
constraints. But actors nevertheless overestimated the magnitude
of these correspondent inferences.
In Van Boven et al.’s (1999) research, the actor’s behavior was
constrained, the observer was aware of the constraints, and the
actor knew the observer was aware of the constraints. In situations
in which the constraints are known, observers typically reign in
their inferences, albeit insufficiently (Gilbert, 1989), leaving room
for actors to overestimate. But what about when the behavior is not
constrained? When people forget an acquaintance’s name, go
blank in the middle of a presentation, or fall flat while exiting the
ski lift, the behavior in question is rarely constrained in the usual
sense of that term. If people are quick to take such behavior at face
value, why might individuals who suffer these unconstrained mis-
fortunes nevertheless overestimate the harshness of others’
impressions?
We contend that this puzzle can be solved by the conjunction of
three psychological truths. First, as mentioned, research on the
correspondence bias suggests that people give short shrift to situ-
ational influences, not that they fail to consider them at all. People
do adjust their impressions to accommodate situational factors,
even though this adjustment is often insufficient (Gilbert & Jones,
1986; Jones & Harris, 1967).
Second, research has demonstrated that such adjustment is par-
ticularly pronounced when observers can empathize with an actor
or imagine being in his or her shoes (Regan & Totten, 1975;
Storms, 1973). Such an empathic orientation leads people to soften
the otherwise harsh dispositional inferences they would make
about someone caught in an embarrassing moment. This orienta-
tion may be particularly likely when a person has experienced the
same or a similar difficulty as the actor. Bystanders may readily
recall times when they have forgotten a friend’s name, gone blank
during a talk, or stumbled through any number of uncoordinated
athletic moves, leading them to soften their assessments of others
experiencing similar mishaps.
Yet people may easily lose sight of an observer’s empathic
orientation when they are in the midst of an embarrassing moment
because of a third psychological truth: People are fundamentally
egocentric and have difficulty getting beyond their own perspec-
tive when anticipating how they will be judged by others (Gilovich
& Savitsky, 1999; Griffin & Ross, 1991; Ross & Ward, 1996).
Indeed, one reason people overestimate how harshly they will be
judged after a blunder is that they focus egocentrically on the
blunder itself and neglect to consider other “nonfocal” factors that
might influence observers’ impressions of them (Savitsky et al.,
2001). This focusing illusion (Schkade & Kahneman, 1998; Wil-
son, Wheatley, Meyers, Gilbert, & Axsom, 2000) can lead people
to overlook others’ empathic orientation, causing them to miss
times when people will soften their dispositional inferences.
These considerations lead us to hypothesize that people overes-
timate the harshness of an observer’s assessment of them when-
ever the observer can empathize or imagine being in the actor’s
shoes. When observers have “been there, done that,” in other
words, they are likely to adopt the actor’s perspective—to empa-
thize with the actor
1
—and thus exhibit considerable inferential
charity. The actor’s fears of harsh recriminations in such situations
are thus unlikely to be confirmed.
2
In contrast, when the observers
find it hard to put themselves in the actor’s shoes, they are likely
to jump from acts to dispositions in the manner demonstrated so
frequently in the attribution literature. In these cases, the actor’s
fears of harsh recriminations are quite likely to be confirmed.
The distinction can be illustrated by a pair of thought experi-
ments. Imagine that you have just dropped a tray in a crowded
dining facility, sending food, dishes, and silverware careening
across the floor. Like most people, you would probably find this to
be an embarrassing turn of events, and you would likely entertain
worries that others would see you as something of a klutz. How-
ever, because most people have suffered a similar fate at some time
or have at least experienced a close call or two, they are likely to
empathize with your misfortune and judge you charitably as a
result—more charitably than you expect.
But now imagine that you have been a participant in one of
Milgram’s (1974) obedience experiments, and, like the modal
participant, you went along with the experimenter’s command to
administer what appeared to be a near-lethal dose of electricity to
another human being. You would probably be ashamed if others
learned of your actions, and, again, you would probably be worried
about being judged harshly—this time as a weak or heartless
individual with deficient moral scruples. Here, however, your fears
are likely to be confirmed. Observers would almost certainly never
have been in such a situation themselves and would have a hard
time accurately imagining the pressures acting on you. In fact,
when imagining what it would be like to be in your shoes, almost
everyone would conclude that they would have acted differently
(Bierbrauer, 1976; Epley & Dunning, 2000; Sherman, 1980). Ob-
servers are thus unlikely to look on you with much empathy, and
you would in fact be seen as weak or uncaring—just as you feared.
Thus, observers are more likely to pass harsh judgments when
they have difficulty adopting an embarrassed actor’s perspective
than when they can easily imagine being in the actor’s position.
Yet we propose that this difference is a feature of the human
condition that people often fail to consider when anticipating how
they will be seen by others. The net result is a tendency to
1
Empathy is a multidimensional concept involving both affective and
cognitive components (Davis, 1983), including concern or sympathy (e.g.,
Batson et al., 1991; Cialdini, Brown, Lewis, Luce, & Neuberg, 1997),
personal distress (e.g., Mikulincer et al., 2001), and perspective taking
(e.g., Chartrand & Bargh, 1999). Our use of the term throughout this article
is limited to this last (cognitive) component—the capacity or inclination to
adopt another person’s perspective—and should not be confused with other
aspects of the term.
2
We have shown elsewhere that people tend to overestimate how much
they will be noticed and how extremely they will be judged for both
positive and negative behaviors (Gilovich et al., 2000; Savitsky et al.,
2001; Van Boven et al., 1999). However, a number of psychological
processes collude to make these effects more reliable and pervasive for
negative behaviors than for positive behaviors (see Savitsky et al., 2001).
As a result, in this article we focus exclusively on people’s estimates of
how harshly they will be judged after potentially embarrassing failures or
mishaps. From an applied perspective, furthermore, this type of error
probably carries the greatest consequence and is thus more pressing to
understand.
301
EMPATHY NEGLECT

overestimate how harshly one will be judged by observers who can
adopt one’s perspective.
We conducted four experiments to test our thesis. Three of these
shared a common design in which participants were placed in an
embarrassing situation—struggling to answer difficult word puz-
zles, performing poorly on a test of general knowledge, or singing
the “Star Spangled Banner”—while being watched by observers.
Some observers were in a position to empathize with the actor’s
plight; others were not. We expected observers who could empa-
thize with the actor to form more charitable impressions of the
actor than observers who could not. But because actors were likely
to be focused inordinately on their own embarrassing blunder, we
expected them to be insensitive to the difference in observers’
orientations and therefore to anticipate being judged equally
harshly by both. The fourth study explored whether this insensi-
tivity is indeed the product of a tendency to overlook empathy by
investigating whether people become better calibrated when ex-
plicitly asked to consider another’s empathic orientation.
We believe a general failure to anticipate others’ empathic
orientation can help make sense of actors’ tendency to overesti-
mate how harshly they are judged by others, even if those others
are prone to the correspondence bias.
Study 1
In early November 1999, presidential candidate George W.
Bush sat down for a television interview and was subjected to a
pop quiz. To gauge his knowledge of foreign policy, the inter-
viewer asked Bush to name the leaders of four international “hot
spots”: Chechnya, India, Pakistan, and Taiwan. Bush went 1 for 4,
answering correctly only that Lee Teng-hai was the current leader
of Taiwan.
Hoping to minimize the damage caused by her boss’s lackluster
performance, Bush’s campaign manager argued that the pop quiz
was picky and revealed little about Bush’s command of foreign
policy. This political spin, however, may have been unnecessary.
Although political pundits chastised Bush for his poor perfor-
mance (e.g., Dowd, 1999; “A Pop Quiz,” 1999), his popularity
among rank-and-file voters suffered no downturn after the quiz
(Moore, 1999). After all, most Americans likely found that they,
too, were stumped by the questions and were reluctant to judge
Bush harshly as a result. Still, we suspect that such inferential
charity was lost on Candidate Bush, who, caught up in his embar-
rassing failure, likely believed that others would now think even
less of his intellect than they already did.
We designed something of a reenactment of Bush’s pop quiz to
test our hypothesis. But instead of questioning participants about
foreign relations, we questioned them about the relations between
words. The items were difficult, so we expected participants to
perform poorly. Each participant’s poor performance was viewed
by an observer stationed behind a one-way mirror in an adjacent
room. In one condition, the observer was given the answers to the
test items beforehand; in the other condition, he or she was not.
We made three predictions. First, we expected observers who
had been given the answers to the puzzles (i.e., informed observ-
ers) to experience a curse of knowledge (Camerer, Loewenstein, &
Weber, 1989) and judge the questions to be easier than those who
did not have the answers (i.e., naive observers). Research has
shown that once a person is given some information—told, for
example, that “the goose hangs high” means that times are good,
not bad—they are cursed with this knowledge and can have a
difficult time imagining the perspective of someone who is not
privy to the same information (Keysar & Bly, 1995). Second, we
expected this curse of knowledge to lead informed observers to
have difficulty empathizing with the actor’s plight and cause them
to judge the actor’s intellectual ability harshly. In contrast, we
expected naive observers to recognize more readily that they
would also have been unable to answer the items, making them,
like those who witnessed Bush’s quiz, unlikely to render harsh
judgments of the actor. Finally, we predicted that actors would fail
to recognize the extent to which observers’ empathic orientations
would influence their impressions and thus would anticipate being
judged harshly by both informed and naive observers.
Method
Participants. Fifty Cornell University undergraduates participated in
same-sex pairs in exchange for extra credit in their psychology or human
development courses.
Procedure. Participants were informed that the experiment involved
“integrative orientation ability” and people’s perceptions of that ability.
They were told that integrative orientation consists of “the ability to see
connections between various stimuli” and that it is related to one’s intel-
ligence and creativity. Participants were then told that they would be
assigned to one of two roles in the experiment, either the solver or the
observer. The solver would be given a test of integrative orientation ability
by the experimenter and asked to respond verbally to each item. The
observer would simply watch the solver take the test through a one-way
mirror from an adjacent, soundproof room. Although observers could not
hear the questions or the solver’s responses, participants were told that the
observer would have a copy of the test and would learn of the solver’s
performance on each question—an incorrect answer would be signaled by
a sharp buzzing tone, and a correct answer by a chime.
The test of integrative orientation consisted of 10 word triads (e.g.,
shopping, washing, picture). For each triad, the solver was to think of a
fourth word associated with all three (e.g., window; cf. Brown, 1990;
McFarlin & Blascovich, 1984). Other items included skunk, kings, boiled;
chamber, staff, box; and jump, kill, bliss.
3
Participants were given a sample
question and answer to clarify the test format. Because we were interested
in people’s reactions after an embarrassing performance, participants were
given a difficult version of the task. Those confronted with this version in
previous research have generally answered only 2 or 3 of the 10 items
correctly (Beauregard & Dunning, 1998; Kruger, 1999).
Observers were given one of two copies of the integrative orientation
test. In one condition, informed observers were told, in the presence of the
solver, that their copy of the test contained both the test items and the
answers. In the other condition, naive observers were told, also in the
presence of the solver, that their copy of the test contained only the test
items, not the answers. Observers in both conditions were instructed to
record the solver’s performance during the testing phase.
During the test itself, solvers were told whether their answers were
correct but were unable to hear the tones presented to observers. Solvers
did not receive the answers to the test until the experiment was completed.
Dependent measures. Following the test, observers rated the solver on
seven dimensions related to intellectual ability: integrative orientation,
general knowledge, creativity, intelligence, test-taking ability, the ability to
see connections between various stimuli, and the ability to think clearly.
3
Cabbage, music, and joy, respectively.
302
EPLEY, SAVITSKY, AND GILOVICH

All ratings were made in comparison with the average Cornell student on
scales ranging from 0 (much worse than average) to 100 (much better than
average), with 50 labeled average. Solvers, in contrast, anticipated how
they would be rated by the observer on these seven dimensions.
Next, observers estimated how many items they would have answered
correctly had they been the solver and also the average number of items
they believed were answered correctly by all solvers throughout the ex-
periment. Solvers also estimated the number of items observers would have
gotten right had they been the solver and the average number of items
answered correctly by all solvers. They also estimated the number of items
observers would say they would have answered correctly had they been the
solver. Finally, observers and solvers indicated the actual number of items
the solver answered correctly.
Results
Gender did not influence the results of this or any other exper-
iment reported in this article and is not discussed further.
Solver’s performance. As expected, solvers generally per-
formed poorly, answering an average of 1.3 of the 10 items
correctly. Solvers believed they had solved fewer items than the
observer would have solved if he or she had been the solver
(M ⫽ 3.68) and fewer than what they believed to be the average
performance of all solvers (M ⫽ 3.92), paired ts(24) ⫽ 7.02
and 6.87, respectively, ps ⬍ .001. Solvers’ performance did not
vary by the observer’s condition, t(23) ⬍ 1, and was correctly
recalled in all experimental sessions by both solvers and observers.
Actual and anticipated ratings. We expected informed observ-
ers to experience a curse of knowledge and find the items easier
than would naive observers. As anticipated, informed observers
indicated that they would have answered more items correctly if
they had been the solver (M ⫽ 3.75) than did naive observers
(M ⫽ 1.84), t(23) ⫽ 2.46, p ⬍ .05, and also tended to believe that
the average solver would answer more items correctly (M ⫽ 3.17)
than did naive observers (M ⫽ 2.04), t(23) ⫽ 1.66, p ⫽ .11.
These results suggest that informed observers found it difficult
to imagine just how hard the questions would have seemed had
they not known the answers—a tendency that would make it
harder for them to empathize with the solver’s difficulties. As a
result, we expected informed observers to form a more negative
impression of the solver’s intellectual ability than would naive
observers. Recall that both actual and anticipated ratings were
made along seven dimensions related to intellectual ability. These
dimensions were highly intercorrelated, so they were collapsed
into a composite index for both anticipated and actual evaluations
(both
␣
s ⬎ .87). As can be seen in Figure 1, an analysis of this
composite index revealed that informed observers did indeed eval-
uate the solver’s intellectual ability more harshly than did naive
observers, t(23) ⫽ 2.20, p ⬍ .05.
Nevertheless, we expected this difference between informed and
naive observers to be lost on the solvers themselves, who, wrapped
up in their own embarrassing failure, would overlook the observ-
ers’ empathic orientation and anticipate being judged equally
harshly by both informed and naive observers. To evaluate this last
prediction, we first examined solvers’ estimates of the number of
items they thought observers would say they would be able to
solve correctly. As expected, solvers anticipated that informed
(M ⫽ 4.75) and naive observers (M ⫽ 5.31) would expect to
perform equally well had they been in the solvers’ position,
t(23) ⬍ 1. Thus, solvers failed to appreciate the difference in
observers’ estimates that resulted from the curse of knowledge.
Notice, moreover, that these estimates are substantially higher than
the 1.3 questions solvers actually answered correctly, paired
t(24) ⫽ 10.38, p ⬍ .0001. Clearly, solvers thought they had done
poorly and were expecting little empathy from either observer,
whether they possessed the answers or not. And indeed, as can be
seen in Figure 1, solvers expected to be judged equally harshly by
informed and naive observers alike, t(23) ⫽ 1.04, ns.
To test the statistical significance of this overall pattern, we
submitted anticipated and actual impressions of the solvers to a 2
(impression: anticipated vs. actual) ⫻ 2 (observer: naive vs. in-
formed) mixed-model analysis of variance (ANOVA). This anal-
ysis revealed a main effect for impression, suggesting that solvers
generally overestimated how harshly they would be judged by
observers, F(1, 23) ⫽ 31.33, p ⬍ .001. This effect was qualified,
however, by the predicted interaction, F(1, 23) ⫽ 5.87, p ⬍ .05.
Naive observers tended to judge the solver more charitably than
did informed observers, a difference that was lost on the solvers.
Although solvers generally thought they would be judged more
harshly than they actually were, this effect was particularly pro-
nounced when the observer could readily adopt the solver’s
perspective.
Mediational analysis. We contend that observers’ judgments
of the solver were mediated in part by their ability to empathize
with the solver’s difficulties. Informed observers judged the solver
more harshly, we believe, because the answers seemed obvious to
them (“Why can’t the solver get any of these right?”). Naive
observers, on the other hand, did not have the answers and thus
were in the same befuddled state as the solvers themselves (“I
can’t answer any of these either”). If so, the difference between
informed and naive observers’ ratings should be less pronounced
when the observers’ ability to empathize is statistically controlled.
To test this hypothesis, we derived an index of empathic orien-
tation by subtracting the number of questions observers anticipated
they would have answered correctly from the number the solver
they viewed actually answered correctly. We reasoned that an
observer who thought he or she would have answered, say, six
more questions correctly than the solver actually did would expe-
Figure 1. Anticipated and actual impressions of the solver, Study 1.
303
EMPATHY NEGLECT

rience greater difficulty adopting the solver’s perspective than
would an observer who thought he or she would have answered
only one more item correctly.
As expected, this index revealed significantly less empathy
among informed observers than among naive observers (
␤
⫽ .46),
t(24) ⫽ 2.47, p ⬍ .05. Informed observers predicted that they
would have answered 2.25 more questions correctly than the
solvers actually did; naive observers predicted that they would
have answered only 0.69 more questions correctly. In addition, this
empathy index was related to observers’ ratings of the solvers
(
␤
⫽⫺.49), t(24) ⫽⫺2.67, p ⬍ .05. Those who were better able
to empathize with the solver’s difficulties formed more charitable
impressions than did those who were less able. Of key importance,
moreover, the empathy index remained marginally significant
when both the empathy index and the observer’s condition were
entered into the regression (
␤
⫽⫺.37), t(24) ⫽⫺1.85, p ⫽ .08,
whereas the effect of the observers’ condition became nonsignif-
icant (
␤
⫽ .24), t(24) ⫽ 1.21, p ⫽ .24. This change in the
predictive power of the observer’s condition was marginally sig-
nificant by a Sobel test (z ⫽ 1.84, p ⫽ .07), indicating that
observers’ judgments were partially mediated by their ability to
adopt the solver’s perspective. Those who could imagine being in
the solver’s shoes formed more charitable impressions than did
those who could not.
Discussion
These data support the contention that one reason people over-
estimate how harshly their failures will be judged by others is that
they neglect to consider the extent to which others can empathize
with the difficulties they face and moderate their harsh inferences
accordingly. Observers in this experiment who could easily adopt
the solver’s perspective formed more charitable impressions than
did those who were less able to do so. This difference, however,
was lost on the solvers, who expected to be judged equally harshly
by both.
But perhaps the observers’ empathic orientation in this study
was lost on the solvers because our manipulation of empathy was
too subtle. Well-read social psychologists might understand the
curse of knowledge and anticipate its ramifications, but the every-
day, intuitive psychologist may not. In addition, our manipulation
of the observers’ empathic orientation was between subjects,
which may have reduced its salience. Although we made every
effort to make the observers’ perspective obvious to solvers, it may
be that such a manipulation would have more impact in a within-
subject design. Study 2 uses exactly this design.
Study 2
This study was patterned after Ross et al.’s (1977) “quiz show”
experiment, in which some participants (questioners) were asked
to create 10 difficult general-knowledge questions that were then
posed to other participants (contestants). Because questioners in
that study were able to create questions that capitalized on their
own idiosyncratic knowledge, contestants typically answered few
questions correctly. We expected contestants in our rendition to
perform poorly as well.
Unlike Ross et al. (1977), our key interest was how contestants
would anticipate being rated by each of two observers who dif-
fered in their likely empathic orientation. One, the inside observer,
sat beside the contestant throughout the procedure. The inside
observer was thus privy to a broad range of information, including
the contestant’s demeanor, the contestant’s (generally incorrect)
responses, and, most important, the difficulty of the questions. We
expected this information to make it easy for the inside observer to
empathize with the contestant’s failure—after all, the inside ob-
server was in fundamentally the same position as the contestant
and was unlikely to know the answers either.
The other, outside observer, was escorted to an adjacent, sound-
proof room from which he or she watched the procedure through
a one-way mirror. The contestant’s performance was signaled to
the outside observer, but he or she was given no additional infor-
mation on which to base an impression. The outside observer was
thus unaware of the difficulty of the questions and of the contes-
tant’s particular responses. As a result, outside observers had no
firm basis for knowing whether they would have had similar
difficulty answering the questions had they been in the contestant’s
shoes. They were, in essence, unaware of the particular challenges
confronting the contestant and could base an assessment only on
the number of correct (and incorrect) answers.
Because of this differential ability to adopt the contestant’s
perspective, we predicted that outside observers would form more
negative impressions of the contestant’s intellectual ability than
would inside observers. However, despite the within-subject de-
sign of this study, we again expected this difference to be lost on
the contestants, who would anticipate being judged equally harshly
by both.
Method
Participants. Forty-eight Cornell University undergraduates partici-
pated for extra credit in their psychology or human development courses.
They were run in same-sex groups of 4 that had been screened beforehand
to ensure that all were strangers.
Procedure. Participants were told that the purpose of the experiment
was to investigate the processes by which people form impressions of one
another’s intellectual abilities. They were further informed that they would
be playing a quiz game in which one person would be randomly assigned
to the role of questioner, another to the role of contestant, a third to the role
of inside observer, and the last to the role of outside observer. Participants
drew slips of paper to learn their assignment.
All participants were then given verbal descriptions of their assigned
roles in the presence of the entire group. Questioners learned that they were
to generate 10 difficult but not impossible questions. They were told to
avoid both easy (e.g., “Who founded Cornell University?”) and unfair
questions (e.g., “How many pet turtles do I have?”) and to focus instead on
areas in which they had particular knowledge or expertise (e.g., history,
sports, literature). They were led to a separate cubicle and were given 15
min to generate their questions, each of which was to have only one correct
answer.
Contestants learned that their task was to answer these questions to the
best of their ability. Inside observers learned that they would simply watch
the quiz game, acting as something like a studio audience, whereas outside
observers learned that they would watch the quiz game from an adjacent,
soundproof room. Outside observers were told that although they would be
unable to hear any of the questions or answers, they would be informed of
304
EPLEY, SAVITSKY, AND GILOVICH

the contestant’s performance by means of a signaling device (the same
used in Study 1).
4
Prior to the quiz game, the outside observer was taken to an adjacent,
soundproof room, seated in front of a one-way mirror, and given a quick
demonstration of the signaling device. The inside observer, in contrast, was
seated directly beside the contestant, across the table from the questioner.
The questioner then posed each question to the contestant, waited for a
response, supplied the correct answer if necessary, and signaled the outside
observer. Questioners and contestants were instructed to follow this script
carefully and not to engage in extraneous conversation. After all 10 items
had been asked, the experimenter announced—once to those in the room
and once to the outside observer—the number and percentage of correct
responses provided by the contestant.
Dependent measures. Following the quiz game, all participants rated
the contestant and questioner on five dimensions related to intellectual
ability: level of general knowledge, test-taking ability, memory for isolated
facts, ability to answer general-knowledge questions, and level of intelli-
gence. All ratings were made in comparison with the average Cornell
student on scales ranging from 0 (much worse than the average Cornell
student) to 100 (much better than the average Cornell student), with 50
labeled average. After indicating their actual impressions, contestants were
asked to anticipate how they would be rated by the other 3 participants on
the same five dimensions.
Results
Because the five ratings were highly correlated, they were
collapsed into a composite index for both anticipated and actual
judgments (all
␣
s ⬎ .88).
Contestant’s performance and the correspondence bias. Con-
testants generally performed poorly, answering an average of 2.8
of the 10 questions correctly. Consistent with previous research on
the correspondence bias (Ross et al., 1977), participants generally
rated the contestants’ intellectual ability more poorly than the
questioners’, F(1, 11) ⫽ 23.84, p ⬍ .01. This main effect was
qualified, however, by the participants’ role, F(3, 33) ⫽ 8.84, p ⬍
.01. As can be seen in Table 1, outside observers, who were least
able to adopt the contestant’s perspective, exhibited the strongest
correspondence bias, rating the contestant much more harshly than
they rated the questioner, paired t(11) ⫽ 4.74, p ⬍ .001. Contes-
tants and inside observers also rated the contestant more harshly
than the questioner, paired ts(11) ⫽ 3.05 and 3.83, respectively,
ps ⬍ .05, but rated the contestant far more charitably than did
outside observers, paired ts(11) ⫽ 4.51 and 4.60, respectively,
ps ⬍ .01. Also in line with previous research, questioners them-
selves did not commit the correspondence bias, paired t(11) ⬍ 1.
Contestants: Actual and anticipated impressions. We ex-
pected contestants to be somewhat embarrassed by their inability
to answer the questions asked of them and to anticipate being
judged harshly by the observers as a result. Indeed, contestants
expected to be rated significantly below average (i.e., below the
scale midpoint of 50, which was labeled average) by those who
witnessed their performance (M ⫽ 40.64), t(11) ⫽ 2.63, p ⬍ .05,
and thought they would be rated more harshly by others than they
rated themselves (M ⫽ 53.80), paired t(11) ⫽ 3.67, p ⬍ .01.
As expected, these fears of harsh judgment were generally
exaggerated. A 2 (impressions: anticipated vs. actual) ⫻ 3 (rater:
questioner, inside observer, outside observer) repeated measures
ANOVA revealed a main effect for impressions, indicating that
contestants generally overestimated how harshly they would be
judged by others in the group, F(1, 11) ⫽ 5.51, p ⬍ .05. This effect
was qualified, however, by the predicted interaction with the
participant’s role, F(2, 22) ⫽ 4.73, p ⬍ .05. As can be seen in
Figure 2, questioners and inside observers, who could understand
the challenges confronting the contestants, formed relatively char-
itable impressions of the contestants’ intellectual ability
(Ms ⫽ 52.00 and 51.33, respectively). Outside observers, on the
other hand, who were unaware of contestants’ difficult situation,
formed significantly more negative impressions than did either
questioners or inside observers (M ⫽ 38.00). A planned contrast
comparing the first two ratings with the latter was significant,
t(11) ⫽ 4.87, p ⬍ .001.
Despite this large difference in actual impressions, contestants
made no allowance for the observers’ empathic orientations and
anticipated being judged equally harshly by all, F(2, 22) ⬍ 1.
Thus, contestants correctly anticipated the negative evaluations on
the part of the outside observer, paired t(11) ⬍ 1, but overlooked
the extent to which an empathic orientation would produce more
4
To be consistent with the procedures described in Ross et al. (1977),
we asked contestants and both observers to generate their own lists of
questions while they waited for the questioner to complete his or hers. They
(unlike the questioners) were told to generate relatively easy questions that
could be answered by approximately 90% of high school students.
Figure 2. Anticipated and actual impressions of the contestant, Study 2.
Table 1
Participants’ Ratings of the Contestant and Questioner
Ratings by
Ratings of
DifferenceContestant Questioner
Contestant 53.83 64.08 10.25*
Questioner 52.00 50.00 ⫺2.00
Inside observer 51.33 58.66 7.33**
Outside observer 38.00 56.38 18.38***
* p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.
305
EMPATHY NEGLECT

charitable impressions in the minds of questioners and inside
observers, paired ts(11) ⫽ 2.60 and 2.72, respectively, ps ⬍ .05.
Discussion
These results provide further evidence that one reason people
overestimate how harshly they will be judged following an em-
barrassing episode is that they fail to consider others’ empathic
orientation. Contestants who performed poorly on a test of general
knowledge thought they would be rated harshly by all observers
regardless of the observer’s ability to understand their difficult
situation. This failure occurred even though the difference in
observers’ empathic orientation was highlighted by the use of a
within-subject design.
We believe the results of Studies 1 and 2 help to reconcile the
apparent contradiction between people’s tendency to overestimate
how harshly they will be judged and observers’ tendency to take
behavior at face value and make overly dispositional inferences.
Although people are indeed inclined to jump from acts to dispo-
sitions, they are more likely to reign in or adjust those dispositional
inferences when they can easily adopt an actor’s perspective and
empathize with his or her predicament. In embarrassing moments,
this often leads observers to form relatively charitable impressions
of embarrassed actors—a tendency that seems to be lost on the
actors themselves. As a result, actors tend to expect more negative
inferences from empathic observers than they actually receive.
Given the results of Studies 1 and 2, there can be little doubt that
those who commit a blunder fail to anticipate charitable judgments
from those who can empathize with them. But why? We maintain
that individuals who have committed a blunder tend to overlook
when others are likely to empathize with them. This may occur for
a variety of reasons, the most well documented of which is that
people in embarrassing moments tend to be inordinately focused
on themselves and their own phenomenology (Savitsky et al.,
2001). Note, however, that there is an alternative interpretation.
People may be fully aware of when others will empathize with
their plight but fail to understand what effect an empathic orien-
tation has on another’s judgments. Contestants in Study 2, for
example, may have understood that inside observers would adopt
their perspective more easily than would outside observers. They
may not have understood, however, the implications of this dif-
ference for how observers were likely to evaluate them.
One way to determine whether people overlook others’ em-
pathic orientation or whether they instead misunderstand the link
between empathy and charity is to examine a situation in which the
situational factors that induce empathy are so obvious that actors
could hardly fail to consider them when making their judgments.
Under these circumstances, those who commit a blunder may be
able to escape, at least partially, their egocentric perspective and
realize that empathic observers will judge them relatively charita-
bly. But if the error lies in a failure to understand the link between
empathy and charity, then efforts to render differences in observ-
ers’ empathic orientations salient would be unlikely to affect how
actors expect to be judged. We designed Study 3 to investigate this
issue.
Study 3
Participants in this experiment were asked to perform a task we
thought would be mildly embarrassing for almost anyone—singing
the “Star Spangled Banner” a cappella in front of a small audience
(Apsler, 1975; R. S. Miller, 1987). To make the task especially
challenging, we asked participants to chew a wad of gum while
singing. Singers were then evaluated by two observers, one of
whom was in the room with them and was thus aware of the
gum-chewing constraint and another who heard only an audio
recording of their performance and knew nothing of the gum. We
expected observers who were aware of the gum to be better able to
empathize with the singer’s predicament and thus form more
charitable impressions than would observers who were not aware
of the gum-chewing constraint.
Furthermore, we expected this difference between observers—
between those who knew about the gum and those who did not—to
be quite prominent in the singers’ minds. After all, chewing gum
was a defining feature of this embarrassing event. We contend that
the actors in Studies 1 and 2 failed to distinguish between the
judgments of empathic and unempathic observers because the
situational manipulation that induced empathy in those studies
escaped their attention. If this is true, then making that manipula-
tion too salient to be overlooked should give rise to the insight,
among singers, that the two observers are likely to differ in their
appraisals of the performance. If, on the other hand, participants’
failure to draw the appropriate distinction in Studies 1 and 2
stemmed from a misunderstanding of the link between empathy
and charity, then singers should expect to be evaluated equally
harshly by both observers.
Method
Participants. Sixty Cornell University undergraduates participated for
extra credit in their psychology or human development courses.
Procedure. Participants were recruited in pairs and told that the ex-
periment was an investigation of people’s ability to perform two tasks at
once—in this case, singing while chewing gum. Participants were informed
that they would be randomly assigned to one of two roles. One of them, the
singer, would be asked to perform his or her best rendition of the “Star
Spangled Banner” while chewing an unwieldy wad of bubble gum. The
other, the inside observer, would simply watch the singer’s performance.
Once roles were assigned, singers were situated behind a music stand at
one end of the laboratory, and observers were seated at a table at the other
end. Singers were then presented with an assortment of different flavors of
bubble gum and asked to select several pieces to chew during their
performance. Singers were required to chew approximately 1.5 square in.
(14.44 cm
2
) of gum, two pieces of one popular brand or five pieces of
another. Singers were told that their rate of chewing during the song would
be monitored and that they would be asked to sing a second time if they
failed to chew continuously (no participant had to sing twice). Singers were
then given a copy of the lyrics to the first verse of the “Star Spangled
Banner” and a small hand-held microphone attached to a video camera.
The experimenter activated the video camera and signaled the singer to
begin when ready.
The audio portion of the videotape was played for a 3rd participant
randomly selected from the same participant pool in a separate experimen-
306
EPLEY, SAVITSKY, AND GILOVICH

tal session.
5
This participant, the outside observer, was given little infor-
mation about the experiment—simply that another Cornell student had
been asked to sing the “Star Spangled Banner.” Outside observers neither
met the singer nor knew his or her identity.
Dependent measures. Following the (actual or recorded) performance,
observers were asked to rate the singer’s ability on three dimensions:
general singing ability, vocal pitch (the ability to sing in tune), and vocal
clarity (the ability to sing clearly). These ratings were made in comparison
with the average Cornell student on scales ranging from 0 (much worse
than average) to 100 (much better than average), with 50 labeled average.
Singers were told that their performance would be evaluated both by the
observer who had watched their performance and by another observer who
would hear an audio tape of their performance without being told anything
else about the experiment. Singers then anticipated how they would be
rated by each observer on the same three scales.
Results and Discussion
Because the three ratings were highly correlated, they were
collapsed into a composite index for both anticipated and actual
judgments (all
␣
s ⬎ .75).
The “Star Spangled Banner” is a difficult tune to carry under the
best of circumstances, and a mouth full of gum does not make it
any easier. Our informal observations during the experiment more
than confirmed our expectations that most singers would perform
poorly. Nevertheless, in line with our previous experiments, we
expected inside observers, who were aware of the singers’ con-
straint, to find it easier to empathize with their difficulties and
judge them more charitably than would outside observers, who
were not aware of the constraint.
Of key interest was whether singers would anticipate this dif-
ference. Because the gum was so salient to both singers and inside
observers and because its hindrance to good singing was so obvi-
ous to both, we expected singers to anticipate being judged more
charitably by inside observers, who knew about the gum-chewing
constraint, than by outside observers, who did not.
To test these hypotheses, we submitted participants’ anticipated
and actual evaluations to a 2 (rater: inside vs. outside ob-
server) ⫻ 2 (impression: anticipated vs. actual) repeated measures
ANOVA. As expected, both main effects in this analysis were
significant, Fs(1, 19) ⫽ 18.88 and 35.04 for rater and impression,
respectively, ps ⬍ .001. These effects were qualified, however, by
a significant interaction, F(1, 19) ⫽ 7.43, p ⬍ .05. As can be seen
in Figure 3, inside observers, who could better understand the
singers’ predicament, formed significantly more charitable impres-
sions of the singer’s ability than did outside observers, paired
t(19) ⫽ 5.53, p ⬍ .0001. In addition, singers showed some appre-
ciation of this difference, correctly anticipating that they would be
judged more charitably by inside than outside observers, paired
t(19) ⫽ 3.16, p ⬍ .005.
These data indicate that people understand the link between
empathy and charity and that when the situational variables that
distinguish an empathic observer from an unempathic one are
particularly pronounced, people will adjust their expectations ac-
cordingly. The problem appears to be that, in many cases, such as
in Studies 1 and 2, the variables that induce such an orientation are
difficult to identify because they are generally background features
of the situation.
Note, however, that the significant interaction indicates that
although singers understood that empathic observers would be
more charitable than would unempathic observers, they still un-
derestimated the impact of the observer’s empathic orientation.
This finding suggests that increasing people’s attention to an
observer’s empathic orientation may only be a partial remedy for
inaccurate expectations. There may be other mechanisms that
hinder people’s ability to fully appreciate when others are likely to
empathize and when they are not. We return to this issue in the
General Discussion.
Study 4
Study 3 casts doubt on the possibility that people overlook
empathy because they misunderstand its psychological impact.
When the cause of their poor performance was too salient to be
missed, participants were confident that observers would not miss
it, and this resulted in more accurate predictions than those ob-
served in Studies 1 and 2. But note that we neither measured nor
manipulated the extent to which participants thought that observers
were likely to empathize with them, so an important part of the
evidentiary support for our thesis is missing. Study 4 was designed
to provide that evidence. We reasoned that explicitly asking par-
ticipants to evaluate an observer’s likely empathic orientation
would lead them to notice an important variable they would
otherwise have overlooked. This, in turn, should lead actors to
expect a difference in the judgments rendered by empathic and
unempathic observers—expectations we did not observe from the
actual actors in Studies 1 and 2.
More specifically, we asked participants to imagine being in one
of the preceding three experiments. Participants in the control
condition simply estimated how they would be judged on the same
dimensions used in the original experiments—just as actors in
those experiments had done. Participants in the empathy condition,
in contrast, first rated how much they thought observers would be
able to empathize with them and then estimated how they would be
judged. Because participants in the original experiments over-
5
This session was run as soon after the original as possible, often
within 5 min from the end of the original session and no longer than 3 days
afterward.
Figure 3. Anticipated and actual impressions of the singer, Study 3.
307
EMPATHY NEGLECT

looked the impact of the observers’ empathic orientations in Stud-
ies 1 and 2 but not Study 3, we expected that participants led to
consider empathy in this experiment would anticipate being judged
more charitably by the empathic observer than by the unempathic
observer in Studies 1 and 2 but that this manipulation would have
little or no impact in Study 3. After all, the problem with the actual
singers in Study 3 was not that they failed to understand an
observer’s likely empathic orientation but that they underestimated
its impact.
This analysis makes it clear why explicit measures of antici-
pated empathy could not have been collected in Studies 1–3. We
contend that in the normal course of events, an observers’ em-
pathic orientation is likely to be overlooked by those caught in the
throes of an embarrassing moment. As we have shown elsewhere
(Savitsky et al., 2001), those in an embarrassing moment tend to be
inordinately focused on themselves and their own phenomenology
when anticipating how they appear to others (see also Kenny &
Depaulo, 1993; Gilovich, Savitsky, & Medvec, 1998; Gilovich et
al., 2000). But what this egocentrism would otherwise render
unavailable becomes quite noticeable when it is the object of an
explicit query. If people naturally overestimate how harshly they
will be judged because they neglect to consider the extent to which
an observer can empathize with them, then leading people to
consider empathy should produce more charitable and more accu-
rate anticipated judgments. After all, our thesis is not that people
cannot see the implications of empathic and unempathic observers
but that such distinctions are generally overlooked or neglected in
embarrassing moments.
Method
Participants. One hundred fifty Cornell University undergraduates
participated for extra credit in their psychology or human development
courses. Fifty participants imagined participating as actors in each of
Studies 1, 2, and 3.
Procedure. Participants were given detailed descriptions of the proce-
dures of one of the three previous experiments and then asked to imagine
that they had participated as an actor in that experiment—that they had
been a solver in Study 1, a contestant in Study 2, or a singer in Study 3. For
Studies 1 and 2, participants were asked to imagine that their performance
was at a level typical of participants in the original studies (i.e., one item
correct in Study 1, two items correct in Study 2). For Study 3, participants
were simply asked to imagine, as best they could, how their performance
would look and sound.
After reading the description and engaging in the pertinent act of
imagination, participants were asked to estimate how they would have been
judged by observers. These judgments were made in a manner akin to the
judgments made in the original experiments. Thus, half of the participants
who imagined having participated in Study 1 predicted how they would
have been judged by an observer who possessed the correct answers to the
word puzzles, and half predicted how they would have been judged by an
observer who did not possess the answers (i.e., a between-subjects manip-
ulation, as in the original study). In contrast, participants who imagined
having participated in Study 2 or Study 3 predicted how they would have
been rated by each of two observers—one who heard the difficult ques-
tions and one who did not for those who imagined participating in Study 2,
and one who knew about the chewing gum and one who did not for those
who imagined participating in Study 3 (i.e., a within-subject manipulation,
as in the original studies).
Before anticipating how they would be evaluated by each of the observ-
ers, participants in the empathy condition rated the extent to which the
observer in question would be able to “empathize with the difficult situa-
tion you were in...orimagine being ‘in your shoes.’” These ratings were
made on 11-point scales ranging from 1 (not at all)to11(a great deal).
Participants in the control condition did not make this rating before
anticipating how they would be judged.
Results
We predicted that participants in the control condition would
overlook the impact of observers’ empathic orientations in Stud-
ies 1 and 2 and thus anticipate being judged harshly by both
observers, just like actors in the original experiments. In contrast,
we predicted that participants in the empathy condition who ex-
plicitly considered the extent to which an observer could adopt
their perspective would incorporate this consideration into their
judgments and anticipate more charitable ratings from observers
who could adopt their perspective. We predicted that this manip-
ulation would not influence judgments in Study 3, however, be-
cause that experiment was explicitly designed to highlight the
observer’s empathic orientation and thus would be taken into
account even by those not explicitly asked to consider it.
Table 2 presents the anticipated judgments of participants in the
empathy and control conditions for each of the three experiments
(averaged across the specific items into a composite index for each
experiment, all
␣
s ⬎ .70). Because participants were not randomly
assigned to the scenarios, we submitted these data to three separate
2 ⫻ 2 ANOVAs, one for each experiment, with one factor for
whether participants considered empathy and another for the type
of observer (empathic vs. unempathic). These analyses revealed a
significant interaction among participants who imagined them-
selves participating in Study 1, F(3, 46) ⫽ 9.31, p ⬍ .01, and
Study 2, F(1, 48) ⫽ 5.68, p ⬍ .05. Participants who initially
evaluated the observers’ empathy (i.e., those in the empathy con-
dition) anticipated being judged more charitably by naive than by
informed observers in Study 1, t(19) ⫽ 2.44, p ⬍ .05, and also
more charitably by inside than by outside observers in Study 2,
paired t(24) ⫽ 6.51, p ⬍ .001. In contrast, participants in the
control condition did not make the same distinction between
Table 2
Anticipated Ratings by Participants Who Imagined Being an
Actor in Study 1, Study 2, or Study 3 After Explicitly
Evaluating the Observer’s Ability to Empathize or
Without Doing So
Simulated study
Condition
Empathy evaluated Control
Study 1: Integrative orientation
Informed observer 34.50 45.83
Naive observer 48.57 37.14
a
Study 2: The quiz bowl
Outside observer 34.24 35.20
Inside observer 50.16 38.80
a
Study 3: Singing
Outside observer 23.97 25.17
Inside observer 43.97 41.95
Note. Means that differ significantly (p ⬍ .05) within rows are marked
by subscript.
308
EPLEY, SAVITSKY, AND GILOVICH

empathic and unempathic observers, just like actors in the original
experiments. If anything, these participants anticipated a differ-
ence in the opposite direction from the actual effect for Study 1
(replicating the same pattern seen among actors in the original
experiment as well), t(27) ⫽ 1.81, p ⫽ .08, and no difference for
Study 2 (t ⬍ 1).
This occurred, it appears, because participants who explicitly
considered empathy anticipated that naive observers (M ⫽ 8.00)
would adopt their perspective more easily than would informed
observers (M ⫽ 4.78) in Study 1, t(19) ⫽ 2.76, p ⬍ .05, and that
inside observers (M ⫽ 7.56) would be better able to adopt their
perspective than would outside observers (M ⫽ 4.06) in Study 2,
paired t(24) ⫽ 8.16, p ⬍ .001. These empathy ratings correlated
strongly with anticipated judgments in Study 1, r(21) ⫽ .72, p ⬍
.01, and the difference between the observers’ ability to empathize
was strongly correlated with the difference in anticipated judg-
ments in Study 2, r(25) ⫽ .53, p ⬍ .01.
Those who imagined participating in Study 3 anticipated being
judged more charitably by the inside observer than by the outside
observer, F(1, 48) ⫽ 117.05, p ⬍ .001, regardless of whether they
explicitly considered empathy (F for interaction ⬍ 1). Participants
in the empathy condition felt that the inside observer (M ⫽ 9.00)
would find it easier to adopt their perspective than would the
outside observer (M ⫽ 2.73), paired t(25) ⫽ 16.56. The lack of a
significant interaction suggests that these considerations were al-
ready on the minds of participants in the control condition.
Because observers’ likely empathic orientation in Study 1 was
manipulated between participants, we were able to determine in
this study whether perceptions of the observer’s empathic orien-
tation mediated participants’ anticipated judgments. To do this we
first predicted, using linear regression, participants’ anticipated
empathy from whether or not the observers were said to have
possessed the answers to the items (i.e., from experimental con-
dition). We then predicted participants’ anticipated judgments
from their anticipated empathy. Finally, we predicted participants’
anticipated judgments simultaneously from both anticipated em-
pathy and experimental condition (Baron & Kenny, 1986).
Whether or not observers had the answers significantly pre-
dicted their perceived ability to empathize (
␤
⫽ .53), t(20) ⫽ 2.76,
p ⬍ .05, and perceived empathy in turn predicted participants’
anticipated judgments (
␤
⫽ .72), t(24) ⫽ 4.54, p ⬍ .001. Most
important, when participants’ anticipated judgments were pre-
dicted from both condition and anticipated empathy, there was a
significant effect of anticipated empathy (
␤
⫽ .64), t(24) ⫽ 3.39,
p ⬍ .01, but no effect of experimental condition (
␤
⫽ .14,
t ⬍ 1.00). This change in the predictive power of the experimental
condition was significant by a Sobel test (z ⫽ 2.40, p ⬍ .05),
indicating that participants’ anticipated judgments in the empathy
condition were mediated by their perceptions of the observer’s
ability to empathize with their embarrassing failure.
Discussion
Taken together, these data demonstrate more conclusively that
people in the midst of an embarrassing moment tend to overlook
the impact of others’ empathic orientation and are consistent with
our previous findings that this may occur because people are
excessively focused on their own embarrassing blunder or mishap
(Savitsky et al., 2001). Participants who imagined being actors in
Studies 1 and 2, just like those in the original experiments, failed
to anticipate more charitable judgments from observers who could
empathize with them than from observers who could not. This
oversight was overcome by simply asking participants to consider
the extent to which an observer would be able to empathize with
them before anticipating how they would be judged. The fact that
participants were perfectly able to identify charitable and unchar-
itable judges when asked to consider empathy also shows that
people possess the knowledge to make more calibrated judgments
but that this knowledge might be overlooked in the midst of an
embarrassing moment.
General Discussion
One of the first things a student of social psychology typically
learns is that people (Westerners at least) underestimate the impact
of situational variables on others’ behavior and draw stronger
dispositional inferences on the basis of observed behavior than is
warranted (e.g., Gilbert & Malone, 1995). Confusion might set in
when this same student also learns that people tend to overestimate
the extent to which others are likely to draw strong dispositional
inferences about them (Savitsky et al., 2001; Van Boven et al.,
1999). If people are so inclined to take behavior at face value,
where is there room for exaggeration?
This article has attempted to head off any potential confusion by
examining the critically important role played by the observer’s
ability to empathize or imagine being in the actor’s position.
Consistent with past research, we found that observers who could
empathize with the actor’s predicament were more charitable in
their assessments than were those who were not in a position to
empathize. We also found, however, that actors typically failed to
anticipate the significance of this difference in observers’ empathic
perspectives. This failure led actors to assume that empathic ob-
servers would form stronger dispositional inferences about them
than they actually did. In four experiments, we demonstrated how
this tendency to overlook empathy leads people to overestimate
how harshly they are judged for their failures and mishaps.
In Studies 1 and 2, observers in a position to adopt an actor’s
perspective formed more charitable impressions of the actor’s
intellectual ability than did observers who were not in such a
position. In both experiments, however, the actors failed to antic-
ipate this difference, expecting to be judged equally harshly by
both. This failure occurred even when both empathic and unem-
pathic observers were included in the same experimental session
(Study 2). Studies 3 and 4 rule out the possibility that people are
simply unaware of the link between empathy and judgmental
charity by demonstrating that actors do anticipate a difference
between empathic and unempathic observers when the distinction
between the two is made particularly obvious or when they are
asked explicitly to consider another’s empathic orientation. Of
course, because the situational factors that induce empathy in
everyday life tend to be more subtle than those we used in Study 3
and because individuals rarely pause to assess explicitly the extent
to which others empathize with them, the results of Studies 1
and 2, in which actors failed to consider observers’ empathy
altogether, are probably more typical.
309
EMPATHY NEGLECT

Readers might wonder, though, whether the results of these
experiments were artifactually produced by observers’ reluctance
to admit to passing harsh judgments on an embarrassed actor.
Inside observers in Study 3, for example, may have thought the
actor they observed had a perfectly hideous voice but were reluc-
tant to admit being such a harsh critic. But this alternative inter-
pretation runs afoul of two findings. First, we have demonstrated
elsewhere that people are fully willing to admit to judging others
harshly when they have little information other than an embarrass-
ing blunder on which to base an impression or are unaware of the
prevailing situational constraints (Savitsky et al., 2001). Second
and most important, observers in the current studies who were
unable to adopt an actor’s perspective were also perfectly willing
to express their harsh judgments.
Readers might also worry that self-presentational concerns were
at work in actors’ predictions of how they would be judged, with
a sense of false modesty making the actors reluctant to report that
they expected observers to rate them charitably. But this is a
similarly unsatisfactory account of our findings. Note that the
singers in Study 3 and all participants in the empathy condition of
Study 4 anticipated different evaluations on the part of empathic
and unempathic observers—differences that would not have
emerged if actors were simply withholding their true expectations
out of false modesty.
It thus seems clear that an important reason why people often
overestimate how harshly they will be judged for a public failure
or embarrassing mishap is that they fail to recognize when others
will empathize with their predicament and moderate their judg-
ments accordingly. But note that the results of Study 3 suggest that
even when an observer’s empathic orientation is blatantly obvious,
people still fail to adjust sufficiently for another’s empathic char-
ity. We believe this insufficient adjustment is produced by at least
three additional barriers that hinder a full appreciation of another’s
empathic orientation.
The first is one we alluded to earlier—that people who commit
an embarrassing blunder tend to focus excessively on the blunder
itself when anticipating how they will be judged and neglect to
consider other factors that may moderate observers’ impressions
(Savitsky et al., 2001). Embarrassing blunders, after all, do not
take place in isolation but instead are part of an ongoing social
context. Musicians who miss one key note in a concert still hit
countless others, and a speaker who blows an opening joke often
has 49 minutes to recover. To the extent that actors focus too much
attention on one embarrassing event and fail to consider the
context surrounding that event, they will overestimate the impact
of their focal blunders. People may therefore direct too little
attention to others’ empathic orientation because their attention is
too focused elsewhere.
A second reason that people often anticipate excessively harsh
judgments on the part of others, even when they recognize that the
observer can see things from their perspective, is that people tend
to hold overly cynical beliefs about others. People tend to believe
that others will behave less morally than they will themselves
(Alicke, 1985; Allison, Messick, & Goethals, 1989; Epley &
Dunning, 2000; Goethals, 1986), that others will claim more
responsibility for joint tasks than they actually deserve (Kruger &
Gilovich, 1999), and, more generally, that others are narrowly
motivated to maximize their material and psychological self-
interest (D. T. Miller, 2000). This naive cynicism (Kruger &
Gilovich, 1999) is also reflected in the tendency for people to
believe that others are motivated to try to think highly of them-
selves by disparaging others (Savitsky et al., 2001).
This dark view of others’ motivations may contribute to exag-
gerated fears of harsh evaluations by leading people to believe that
others are inclined to think that they would do better—that ob-
servers are likely to believe that they, unlike the actor, would act
morally, resist temptation, or exercise greater skill and ability. One
ancillary finding from Study 1 is consistent with this possibility.
Recall that observers in that experiment were asked to anticipate
how many items they would have answered correctly if they had
been the solver. Solvers, in contrast, were asked to anticipate how
many items observers would believe they would have answered
correctly if they had been the solver. We found that solvers
expected both empathic and unempathic observers to anticipate
answering more items correctly than those observers themselves
did, F(1, 23) ⫽ 22.81, p ⬍ .01. Solvers may have anticipated little
judgmental charity from the observers, then, because they exag-
gerated the observers’ anticipated performance.
A third psychological barrier that may hamper people’s ability
to appreciate others’ empathy rests on beliefs not about others’
willingness to adopt their perspective but rather about others’
ability to do so. Actors may fail to anticipate the empathy that
comes from shared understanding because they are unaware of
how often others have experienced blunders and setbacks similar
to their own. Forgetting an acquaintance’s name, for example, may
seem like one’s own personal affliction while one is struggling to
cover it up, but such a misstep is probably not as rare as one might
think. Because people try to cover up their embarrassing actions,
the average person is likely to falsely assume that he or she is more
prone than the average person to suffer from such mishaps (D. T.
Miller & Prentice, 1994; Nisbett & Ross, 1980).
To examine this contention, we asked 20 Cornell students to
consider a list of 14 embarrassing mishaps and to indicate whether
they had committed each of them within the last 5 years. They
were also asked (in counterbalanced order with the previous ques-
tion) to estimate the percentage of their peers who would indicate
having committed each of the blunders. If people think their own
missteps are relatively unique, then the percentage who indicate
that they have suffered each blunder should be higher than the
average estimated percentage of others who have committed it. As
can be seen in Table 3, participants did indeed think they com-
mitted the embarrassing events (M ⫽ 76.1%) more than their peers
had (M ⫽ 63.9%), paired t(19) ⫽ 3.04, p ⬍ .01. This pattern
emerged on 11 of the 14 events (p ⬍ .01 by binomial test). Of
these 11, 9 were significant at the .01 level.
This tendency to overestimate the uniqueness of one’s personal
failings may be fairly common. After all, because people are
almost always aware of their own behavior, their own missteps,
blunders, and failures are hard to miss. Not so with the shortcom-
ings of others, who may slip while one is not looking, forget
someone else’s name, or bungle a talk that one happens to miss.
Because people are necessarily present at all of their embarrassing
moments but not others’, they may conclude that they are uniquely
prone to embarrassing mishaps.
Although we have been most concerned in this analysis with
those determinants of empathic orientation that people fail to
310
EPLEY, SAVITSKY, AND GILOVICH

recognize, we suspect there are also many determinants they do
anticipate. Doubtless people expect to be judged especially char-
itably by friends, relatives, and acquaintances, and doubtless these
expectations are, on the whole, confirmed. The tale we are telling,
then, is a familiar one—of people failing to take account of
transient situational variables (in this case, contextual determinants
of empathy) while also giving considerable weight to less tran-
sient, person-centered variables (in this case, the observer’s moti-
vation to empathize or not because of a preexisting relationship
with the actor). Thus, recent research on the spotlight effect is not,
after all, at variance with the more established literature on the
fundamental attribution error. As this analysis makes clear, the two
literatures fit together in several respects like hand in glove.
References
A pop quiz for Mr. Bush. (1999, November 6). The New York Times, p.
A16.
Alicke, M. D. (1985). Global self-evaluation as determined by the desir-
ability and controllability of trait adjectives. Journal of Personality and
Social Psychology, 49, 1621–1630.
Allison, S. T., Messick, D. M., & Goethals, G. R. (1989). On being better
but not smarter than others: The Muhammad Ali effect. Social Cogni-
tion, 7, 275–295.
Apsler, R. (1975). Effects of embarrassment on behavior towards others.
Journal of Personality and Social Psychology, 32, 145–153.
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal of Personality and Social Psychol-
ogy, 51, 1173–1182.
Batson, C. D., Batson, J. G., Slingsby, J. K., Harrell, K. L., Peekna, H. M.,
& Todd, R. M. (1991). Empathic joy and the empathy–altruism hypoth-
esis. Journal of Personality and Social Psychology, 61, 413–426.
Beauregard, K., & Dunning, D. (1998). Turning up the contrast: Self-
enhancement motives prompt egocentric contrast effects in social judg-
ments. Journal of Personality and Social Psychology, 74, 606–621.
Bierbrauer, G. (1976). Why did he do it? Attribution of obedience and the
phenomenon of dispositional bias. European Journal of Social Psychol-
ogy, 9, 67–84.
Brown, J. (1990). Evaluating one’s abilities: Shortcuts and stumbling
blocks on the road to self-knowledge. Journal of Experimental Social
Psychology, 26, 149–167.
Camerer, C., Loewenstein, G., & Weber, M. (1989). The curse of knowl-
edge in economic settings: An experimental analysis. Journal of Polit-
ical Economy, 97, 1232–1254.
Chartrand, T. L., & Bargh, J. A. (1999). The chameleon effect: The
perception–behavior link and social interaction. Journal of Personality
and Social Psychology, 76, 893–910.
Cialdini, R. B., Brown, S. L., Lewis, B. P., Luce, C., & Neuberg, S. J.
(1997). Reinterpreting the empathy–altruism relationship: When one
into one equals oneness. Journal of Personality and Social Psychol-
ogy, 73, 481–494.
Clark, J. V., & Arkowitz, H. (1975). Social anxiety and self-evaluation of
interpersonal performance. Psychological Reports, 36, 211–221.
Cohen, D., Nisbett, R. E., Bowdle, B. F., & Schwartz, N. (1996). Insult,
aggression, and the southern culture of honor: An “experimental eth-
nography. ” Journal of Personality and Social Psychology, 70, 945–960.
Cole, S. W., Kemeney, M. E., & Taylor, S. E. (1997). Social identity and
physical health: Accelerated HIV progression in rejection-sensitive gay
men. Journal of Personality and Social Psychology, 72, 320–335.
Daly, M., & Wilson, M. (1988). Homicide. Hawthorne, NY: Aldine de
Gruyter.
Davis, M. H. (1983). Measuring individual differences in empathy: Evi-
dence for a multidimensional approach. Journal of Personality and
Social Psychology, 44, 113–126.
Dowd, M. (1999, November 7). Name that general! New York Times, p.
A15.
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.
Fenigstein, A., & Vanable, P. A. (1992). Paranoia and self-consciousness.
Journal of Personality and Social Psychology, 62, 129–138.
Gilbert, D. T. (1989). Thinking lightly about others: Automatic compo-
nents of the social inference process. In J. Uleman & J. A. Bargh (Eds.),
Unintended thought (pp. 189–211). New York: Guilford Press.
Table 3
Percentage of Participants Who Have Committed Various Blunders and Their Estimates of the
Percentage of Their Peers Who Have Done So
Behavior
% committed
by self
Average %
committed by peers Difference
Tripped over a crack in the sidewalk 100 77 23***
Left bathroom with zipper down 70 65 5
Spilled water on pants (so it looked like they
had wet themselves) 80 63 17***
Tripped over shoelaces 55 57 ⫺2
Worn clothes with unnoticed but obvious stain 85 67 18***
Committed an embarrassing slip of the tongue 85 82 3
Forgotten name of acquaintance 95 79 16***
Said something embarrassing in class 75 58 17**
Forgotten what to say in middle of a sentence 100 73 27***
Spilled drink at party 90 70 20***
Dropped tray in cafeteria 35 39 ⫺4
Accidentally triggered security system 20 24 ⫺4
Had coughing or sneezing fit in class 75 53 22**
Had a bad hair day 100 92 8**
** p ⬍ .01. *** p ⬍ .001.
311
EMPATHY NEGLECT

Gilbert, D. T., & Jones, E. E. (1986). Perceiver induced constraint: Inter-
pretations of self-generated reality. Journal of Personality and Social
Psychology, 50, 269–280.
Gilbert, D. T., & Malone, P. S. (1995). The correspondence bias. Psycho-
logical Bulletin, 117, 21–38.
Gilovich, T., Medvec, V. H., & Savitsky, K. (2000). The spotlight effect in
social judgment: An egocentric bias in estimates of the salience of one’s
own actions and appearance. Journal of Personality and Social Psychol-
ogy, 78, 211–222.
Gilovich, T., & Savitsky, K. (1999). The spotlight effect and the illusion of
transparency: Egocentric assessments of how we’re seen by others.
Current Directions in Psychological Science, 8, 165–168.
Gilovich, T., Savitsky, K., & Medvec, V. (1998). The illusion of transpar-
ency: Biased assessments of others’ ability to read one’s emotional
states. Journal of Personality and Social Psychology, 75, 332–346.
Goethals, G. R. (1986). Fabricating and ignoring social reality: Self-
serving estimates of consensus. In J. M. Olson, C. P. Herman, & M. P.
Zanna (Eds.), Social comparison and relative deprivation: The Ontario
symposium (Vol. 4, pp. 135–157). Hillsdale, NJ: Erlbaum.
Goffman, E. (1959). The presentation of self in everyday life. New York:
Doubleday.
Griffin, D. W., & Ross, L. (1991). Subjective construal, social inference,
and human misunderstanding. In L. Berkowitz (Ed.), Advances in ex-
perimental social psychology (Vol. 24, pp. 319–359). San Diego, CA:
Academic Press.
Jones, E. E., & Harris, V. A. (1967). The attribution of attitudes. Journal
of Experimental and Social Psychology, 3, 1–24.
Kenny, D. A., & DePaulo, B. M. (1993). Do people know how others view
them? An empirical and theoretical account. Psychological Bulletin,
114, 145–161.
Keysar, B., & Bly, B. (1995). Intuitions of the transparency of idioms: Can
one keep a secret by spilling the beans? Journal of Memory and Lan-
guage, 34, 89–109.
Kruger, J. (1999). Lake Wobegon be gone! The “below-average effect” and
the egocentric nature of comparative ability judgments. Journal of
Personality and Social Psychology, 77, 221–232.
Kruger, J., & Gilovich, T. (1999). “Naı¨ve cynicism” in everyday theories
of responsibility assessment: On biased assumptions of bias. Journal of
Personality and Social Psychology, 76, 743–753.
McFarlin, D. B., & Blascovich, J. (1984). Associates Test (RAT) as an
alternative to illusory performance feedback: A methodological note.
Basic & Applied Social Psychology, 5, 223–229.
Mikulincer, M., Gillath, O., Halevy, V., Avihou, N., Avidan, S., & Eshkoli,
N. (2001). Attachment theory and reactions to others’ needs: Evidence
that activation of the sense of attachment security promotes empathic
responses. Journal of Personality and Social Psychology, 81, 1205–
1224.
Milgram, S. (1974). Obedience to authority: An experimental view. New
York: Harper Torchbooks.
Miller, D. T. (2000). The norm of self-interest. American Psychologist, 54,
1053–1060.
Miller, D. T., & Prentice, D. A. (1994). Collective errors and errors about
the collective. Personality and Social Psychology Bulletin, 20, 541–550.
Miller, R. S. (1987). Empathic embarrassment: Situational and personal
determinants of reactions to the embarrassment of another. Journal of
Personality and Social Psychology, 53, 1061–1069.
Moore, D. W. (1999). Bradley stronger candidate than Gore against Bush
or McCain in general election. Gallup Poll Monthly, 411, 11–15.
Nisbett, R., & Ross, L. (1980). Human inference: Strategies and short-
comings of social judgment. Englewood Cliffs, NJ: Prentice-Hall.
Regan, D., & Totten, J. (1975). Empathy and attribution: Turning observers
into actors. Journal of Personality and Social Psychology, 32, 850–856.
Ross, L. (1977). The intuitive psychologist and his shortcomings: Distor-
tions in the attribution process. In L. Berkowitz (Ed.), Advances in
experimental social psychology (Vol. 10, pp. 174–221). New York:
Academic Press.
Ross, L., Amabile, T. M., & Steinmetz, J. L. (1977). Social roles, social
control, and biases in social-perception processes. Journal of Personality
and Social Psychology, 35, 485–494.
Ross, L., & Ward, A. (1996). Naı¨ve realism in everyday life: Implications
for social conflict and misunderstanding. In T. Brown, E. Reed, & E.
Turiel (Eds.), Values and knowledge (pp. 103–135). Hillsdale, NJ:
Erlbaum.
Savitsky, K., Epley, N., & Gilovich, T. (2001). Is it as bad as we think?
Overestimating the impact of our failures, shortcomings, and mishaps.
Journal of Personality and Social Psychology, 81, 44–56.
Schkade, D. A., & Kahneman, D. (1998). Does living in California make
people happy? A focusing illusion in judgments of life satisfaction.
Psychological Science, 9, 340–346.
Shafer, D. (1974). Suicide in childhood and early adolescence. Journal of
Child Psychology and Psychiatry, 45, 406–451.
Shafer, D. (1988). The epidemiology of teen suicide: An examination of
risk factors. Journal of Clinical Psychiatry, 49, 36–41.
Sherman, S. J. (1980). On the self-erasing nature of errors of prediction.
Journal of Personality and Social Psychology, 39, 211–221.
Storms, M. D. (1973). Videotape and the attribution process: Reversing
actors’ and observers’ points of view. Journal of Personality and Social
Psychology, 27, 165–175.
Van Boven, L. D., Kamada, A., & Gilovich, T. (1999). The perceiver as
perceived: Everyday intuitions about the correspondence bias. Journal of
Personality and Social Psychology, 77, 1188–1199.
Wilson, T. D., Wheatley, T., Meyers, J., Gilbert, D. T., & Axsom, D.
(2000). Focalism: A source of durability bias in affective forecasting.
Journal of Personality and Social Psychology, 78, 821–836.
Zimbardo, P. G. (1990). Shyness: What it is, what to do about it (Rev. ed.).
Reading, MA: Addison-Wesley.
Received September 2, 2000
Revision received January 28, 2002
Accepted January 28, 2002 䡲
312
EPLEY, SAVITSKY, AND GILOVICH