Effects of Social Exclusion on Cognitive Processes: Anticipated Aloneness
Reduces Intelligent Thought
Roy F. Baumeister
Case Western Reserve University
Jean M. Twenge
San Diego State University
Christopher K. Nuss
University of Minnesota
Three studies examined the effects of randomly assigned messages of social exclusion. In all 3 studies,
significant and large decrements in intelligent thought (including IQ and Graduate Record Examination
test performance) were found among people told they were likely to end up alone in life. The decline in
cognitive performance was found in complex cognitive tasks such as effortful logic and reasoning; simple
information processing remained intact despite the social exclusion. The effects were specific to social
exclusion, as participants who received predictions of future nonsocial misfortunes (accidents and
injuries) performed well on the cognitive tests. The cognitive impairments appeared to involve reductions
in both speed (effort) and accuracy. The effect was not mediated by mood.
The two most important and powerful adaptations of the human
race are intelligence and social structure. Human intelligence,
based on a large and powerful brain, is far superior to what has
been observed in almost every other species. Human social orga-
nization is likewise more complex and flexible than what has been
observed elsewhere. Barchas (1986) has proposed that the small
social group is the single most important adaptation of human
beings, more important even than intelligence, although certainly
other theorists would argue that intelligence deserves priority. In
any case, it is clear that the human capacity for intelligent, com-
plex thought has enabled people to solve problems and master their
environment to an unprecedented extent, and human social groups
have likewise contributed immensely to cultural and technological
progress, safety and comfort, and individual well-being.
The present investigation was designed to investigate the pos-
sible link between social belongingness and intelligent thought.
Specifically, we explored the impact of social exclusion on cog-
nitive processes. We led participants to believe that they would be
alone in later life, and we measured the impact of this message on
cognitive functioning. Practical and ethical constraints prevented
us from studying long-term deprivation and potentially permanent
changes in intelligence, but we were able to study short-term
effects of social exclusion on mental performance. Competing
predictions could be generated. From a simple evolutionary per-
spective, one might assume that it would be highly adaptive for
intelligent functioning to increase among people who expect to be
excluded from social groups. There is, after all, some overlap (and
hence potential redundancy) in the benefits to be gained from
social group membership and intelligent thought. To survive,
people need to make effective decisions, avoid danger, resolve
problems, cope with misfortunes, and obtain life-sustaining re-
sources. Belonging to a group can help one accomplish these,
insofar as group members support and protect each other, share
resources, and pool information, and a group leadership structure
may free many individuals from the burden of making all their
own decisions. In contrast, people who do not belong to a group
must accomplish these ends on their own, so their need for sharply
intelligent functioning is seemingly increased. Put another way,
one has to be more intelligent to live by one’s own wits than
simply to get by with help from one’s friends (i.e., by relying on
and benefiting from the group). It therefore seems optimal for
intelligence to shift into high gear among individuals thrust out
from social groups.
A quite different theoretical approach, however, contends that
the powerfully intelligent human brain is not a substitute for social
group membership but rather a tool for facilitating it. The debate
over whether intelligent thought or group sociality is the more
important and beneficial human adaptation becomes moot, because
the two are intricately linked. In this view, social cognition rather
than technical problem solving may be the crucial raison d’etre of
human intelligence. Some theorists have emphasized that most or
all cognition is socially shared, including the fact that its medium
(language) and content (things and events) both refer inevitably to
the cultural and social world that is constructed and maintained by
groups of people (Resnick, Levine, & Teasley, 1991). Research on
work groups has emphasized, for example, that group members
construct, communicate, maintain, and pass along bodies of infor-
Roy F. Baumeister, Department of Psychology, Case Western Reserve
University; Jean M. Twenge, Department of Psychology, San Diego State
University; Christopher K. Nuss, Department of Psychology, University of
Correspondence concerning this article should be addressed to Roy F.
Baumeister, Department of Psychology, Case Western Reserve University,
Cleveland, Ohio 44106-7123, or to Jean M. Twenge, Department of
Psychology, San Diego State University, 5500 Campanile Drive, San
Diego, California 92182-4611. E-mail: email@example.com, jtwenge@
mail.sdsu.edu, or firstname.lastname@example.org
Journal of Personality and Social Psychology Copyright 2002 by the American Psychological Association, Inc.
2002, Vol. 83, No. 4, 817–827 0022-3514/02/$5.00 DOI: 10.1037//0022-35126.96.36.1997
mation about the nature and identity of the group, about the
members of the group, and about the work tasks (Levine &
Moreland, 1991). Even such a seemingly simple task as learning to
play baseball is not something that occurs inside one child’s head
but rather involves learning to participate in a shared understand-
ing of events, and that in turn requires ongoing and complex
communication between players, coaches, and others (Heath,
From this line of reasoning, one could predict that socially
isolated individuals should have relatively less use for intelligent
thought, whereas a fully functioning intelligence is needed to
navigate the complex social structure and dynamics of human
communities. If intelligence is designed to help the individual
satisfy the need to belong, then being excluded from social groups
might cause a reduction or impairment of cognitive functioning.
The impairment hypothesis could be elaborated to involve emo-
tional distress. Assuming that the need to belong is in fact a basic
and powerful motivation (Baumeister & Leary, 1995), people
would be likely to feel emotional distress when that urge is
thwarted. Research on anxiety has repeatedly found social exclu-
sion to be the single most common cause, and, in fact, other than
fear of bodily harm, social exclusion is almost the only established
major source of anxiety (see Baumeister & Tice, 1990, for review).
Researchers studying ostracism have found significant emotional
distress among ostracized individuals (Williams, Cheung, & Choi,
2000). Being excluded from social groups might set off anxiety or
other forms of emotional distress that could produce a short-term
impairment of cognitive functioning.
Despite the plausible hypothesis that emotional distress might
mediate a reduction in intelligent thought as a result of social
exclusion, the present investigation was actually stimulated by a
rising skepticism about the role of emotion after social exclusion.
In a series of studies, we found that social exclusion produced
substantial increases in aggressive behavior (Twenge, Baumeister,
Tice, & Stucke, 2001), decreases in prosocial behavior (Twenge,
Ciarocco, & Baumeister, 2001), and increases in self-defeating
behavior (Twenge, Catanese, & Baumeister, 2002). We had ini-
tially thought that being excluded from social groups would cause
severe emotional distress, which in turn would lead to antisocial
and self-defeating behavior and various other undesirable out-
comes. The behavioral effects were clear and consistent, but the
emotional processes failed to play the mediating role we had
hypothesized. To be sure, we did usually find that there were some
small increases in emotional distress among people who were
rejected or excluded by others, but these effects were weak and
inconsistent. Moreover, the emotional effects repeatedly failed to
mediate the behavioral consequences, as tested by standard statis-
tical tests of mediation (Baron & Kenny, 1986).
Hence, we had to begin entertaining the view that many effects
of social exclusion were mediated by something other than emo-
tional distress. Cognitive impairments were a prime candidate. For
example, there are many signs that socially excluded or deprived
people commit more crimes, such as in the disproportionately high
rate of offspring of single parents found in criminal prisons (Gott-
fredson & Hirschi, 1990). Research also finds that single men
commit more crimes than married men (Sampson & Laub, 1990)
and that rejected children are more aggressive than popular, ac-
cepted children (Newcomb, Bukowski, & Pattee, 1993). Yet stud-
ies of criminals have found better evidence of low intelligence than
of emotional distress, and psychologically inclined criminologists
tend to emphasize the importance of low intelligence (e.g., Gott-
fredson & Hirschi, 1990; National Research Council, 1993; Walsh,
Beyer, & Petee, 1987; Wilson & Herrnstein, 1985). Clearly, such
arguments are indirect and, moreover, highly ambiguous with
regard to the direction of causality. That is, even if there is a
significant link between low intelligence and antisocial behavior,
does the antisocial or socially isolated stance cause poor thinking,
or vice versa? Experimental tests (e.g., the present investigation)
are necessary to answer that question.
How are these cognitive decrements mediated? On conceptual
grounds, at least three major hypotheses could be put forward. The
first involves an increase in arousal, stemming perhaps from nas-
cent anxiety and uncertainty over the prospect of social exclusion.
According to the standard effects of arousal, simple tasks would be
facilitated, whereas complex ones would be impaired (Zajonc,
1965). Arousal would make people work faster, whereas accuracy
might suffer. Some signs of heightened arousal would also pre-
sumably be evident.
The second possibility is that people might find the prospect of
social exclusion threatening and respond by stifling their emo-
tional reaction. This effort would preoccupy the self-regulation
system, and previous work has suggested that one common set of
self-regulatory resources is used for many different kinds of self-
regulation as well as other controlled processes requiring the self’s
executive function (Baumeister, Bratslavsky, Muraven, & Tice,
1998; Muraven, Tice, Baumeister, 1998). If the self’s resources
were used in suppressing emotion, they would be less available for
controlling cognitive processes. Hence, the more automatic cog-
nitive processes would operate in a relatively unimpaired fashion,
whereas controlled ones would suffer. This theory predicts little or
no report of emotional distress.
The third possibility proposes that people ruminate about the
social exclusion, and this preoccupation with their inner thoughts
distracts them from processing incoming information. The reduced
attention to subsequent tasks would impair performance on nearly
all cognitive tasks or at least all that require attention.
In short, there are reasonable theoretical grounds for making
quite different predictions about the potential impact of social
exclusion on cognitive functioning. On an a priori basis, facilita-
tion and impairment are both plausible, and emotional distress
might or might not play a decisive mediating role. Impairment
could be mediated by heightened arousal, by the commitment of
the executive function to affect regulation, or by attentional load in
connection with rumination.
Our first experiment used a broad intelligence (IQ) test to
investigate the possibility that social exclusion would alter cogni-
tive functioning. We used the General Mental Abilities Test
(Janda, 1996; Janda, Fulk, Janda, & Wallace, 1995), which was
developed to provide a written, multiple-choice intelligence test
suitable for administration to groups. It includes measures of
verbal reasoning, mathematical ability, and spatial ability.
To manipulate social exclusion, we used a procedure developed
by Twenge, Baumeister, et al. (2001). Participants were given a
personality inventory and then provided with false feedback. In the
crucial future alone condition, they were told that the results of the
BAUMEISTER, TWENGE, AND NUSS
test enabled the researchers to predict with high likelihood that the
participants were the sort of people who would end up alone in life.
Because this feedback was randomly assigned and would therefore
be given to at least some people who presumably have rich
networks of friends, several steps were taken to enhance credibil-
ity. First, each participant was given accurate feedback about his or
her level of extraversion, on the basis of an accurate scoring of the
trait scale. Second, the experimenter added that the participant may
well have many friends and acquaintances at present, which is
common among young people in university settings. The experi-
menter went on to say, however, that once the person is past the
age at which people are constantly forming new friendships, the
current friends will tend to drift apart and not be replaced, so that
the individual will spend more and more time alone.
Two control groups were used. The first was the simple opposite
of the future alone manipulation. These future belonging partici-
pants were told that the personality test feedback revealed that they
were likely to spend the rest of their life surrounded by people who
care about them. The social network was predicted to include a
long, stable marriage and some lasting friendships.
The other control group was designed to provide unwelcome,
aversive feedback that would not involve the dimension of belong-
ingness. This misfortune control condition informed participants
that later in life they would become increasingly accident prone.
They would suffer broken bones and other mishaps requiring visits
to hospital emergency rooms. If the effects of the future alone
condition are due simply to receiving a bad, undesirable forecast
about one’s future, then the misfortune control group should yield
similar effects. In contrast, if the misfortune control and future
alone groups are substantially different, then one may infer with
greater confidence that the effects of the future alone manipulation
are specific to social exclusion.
Participants. The participants were 40 undergraduates participating to
fulfill a course requirement in introductory psychology. There were 19 men
and 21 women, 70% White and 30% racial minority. Average age was 19.1
Materials and procedure. The consent form said the study’s purpose
was to explore how personality relates to performance, but the experi-
menter did not verbally explain the purpose of the study. Participants first
completed a personality questionnaire (the Eysenck Personality Question-
naire; Eysenck & Eysenck, 1975). They were then randomly assigned to
hear one of three descriptions about their future life, ostensibly based on
their responses to the personality inventory. To gain credibility, the exper-
imenter first provided an accurate assessment of the subject’s extraversion
score, consisting of correct feedback about whether the score was high,
medium, or low on this scale. The experimenter used this as a segue into
reading a randomly assigned personality type description. One of three
descriptions was read. In the future belonging condition, the participant
was told, “You’re the type who has rewarding relationships throughout life.
You’re likely to have a long and stable marriage and have friendships that
will last into your later years. The odds are that you’ll always have friends
and people who care about you.”
In contrast, people in the future alone condition were told,
You’re the type who will end up alone later in life. You may have
friends and relationships now, but by your mid-20s most of these will
have drifted away. You may even marry or have several marriages,
but these are likely to be short-lived and not continue into your 30s.
Relationships don’t last, and when you’re past the age where people
are constantly forming new relationships, the odds are you’ll end up
being alone more and more.
Last, a misfortune control condition was included in which people were
You’re likely to be accident prone later in life—you might break an
arm or a leg a few times, or maybe be injured in car accidents. Even
if you haven’t been accident prone before, these things will show up
later in life, and the odds are you will have a lot of accidents.
This condition was intended to describe a negative outcome that was not
connected to social exclusion.
After hearing the randomly assigned description, the participants filled
out an item asking them to rate their current mood from 1 (negative)to7
(positive). The experimenter then told them that they would have 6 min to
complete as many items as they could on an intelligence test (the General
Mental Abilities Test, described previously; Janda, 1996; Janda, Fulk,
Janda, & Wallace, 1995). The experimenter collected the test after a timing
bell recorded the end of the 6 min. The tests were then scored against a key
(Janda, 1996) to determine the number of questions the participant got
right. Number of questions wrong and number of attempts were also
After completing the test, participants were fully debriefed. They were
told that the feedback about their extraversion score was true but that the
further feedback was a randomly assigned description. Particular care was
taken to ensure that participants in the future alone group understood that
the prediction was random and not true for them. No participant was
permitted to leave the laboratory until he or she expressed aloud that he or
she understood that the manipulations were assigned at random and thus
had no basis in fact. The experimenter also apologized for giving the false
feedback and explained the rationale for the deception. Participants were
also informed that the intelligence test was not really a good measurement
of IQ under the conditions, because they had received the false feedback
and were given only 6 min to work on a test that takes most people 20–30
min to complete.
IQ test performance. Intelligent performance was apparently
impaired by the diagnostic prediction of future social exclusion. As
Table 1 shows, participants in the future alone condition answered
significantly fewer questions correctly, as compared with partici-
pants in the future belonging and misfortune conditions, F(2,
37) ⫽ 5.44, p ⬍ .01. A post hoc Tukey’s honestly significant
difference (HSD) test showed that the future alone condition was
significantly different from the other two conditions at p ⬍ .05.
Thus, hearing that one was likely to be alone later in life affected
Performance on the General Mental Abilities Test
No. right 18.92 7.39 24.77 4.49 25.79 5.10 5.44**
No. attempts 27.23 8.23 33.54 4.88 31.64 5.29 3.46*
No. wrong 8.46 4.03 8.85 4.30 5.86 3.68 2.25
(right/attempts) 0.69 0.14 0.74 0.10 0.82 0.11 4.29*
* p ⬍ .05. ** p ⬍ .01.
SOCIAL EXCLUSION AND INTELLIGENT THOUGHT
performance on a timed cognitive test. When the future alone and
misfortune control groups were compared (the most rigorous test
of the effect), the effect size for performance on the IQ test was
d ⫽ .98, which is conventionally described as a large effect (i.e.,
anything higher than .80; Cohen, 1977). In other words, social
exclusion feedback produced a substantial decrement in intelligent
The decline in cognitive performance cannot be attributed to the
simple fact of hearing any sort of bad news. Participants in the
misfortune control condition, who heard that their future life would
be disrupted by frequent accidents, scored as well on the IQ test as
those in the future belonging group. Only the bad feedback that
pertained specifically to social exclusion produced a significant
drop in intelligence test scores.
There were no sex differences (main effects or interactions) for
any of these analyses or those following. Thus, we do not discuss
the issue further.
Attempts and accuracy. We recorded the number of test ques-
tions attempted by each participant, as a rough measure of effort.
This analysis again showed significant variation among the three
conditions, F(2, 37) ⫽ 3.46, p ⬍ .05. Participants in the future
alone condition attempted the fewest problems. Again, the deficit
was specific to feedback about social exclusion, insofar as partic-
ipants in the misfortune control condition attempted as many
problems (if not more) as did people in the future belonging
condition (a post hoc Tukey’s HSD test showed a p ⬍ .05 differ-
ence between the future alone and misfortune control groups).
Counting the number of errors yielded a different result, how-
ever. Participants in the misfortune control condition (M ⫽ 8.85;
see Table 1) gave as many wrong answers as did those in the future
alone condition (M ⫽ 8.46). In that sense, either kind of bad news
seemed capable of causing people to make mistakes (compared
with the lower number of errors made in the future belonging
condition, M ⫽ 5.86; however, this difference was not significant).
We also examined accuracy rates, which we computed by dividing
each participant’s number of correct responses by the number of
problems the person attempted. An analysis of variance (ANOVA)
on these scores again yielded significant variation among condi-
tions, F(2, 37) ⫽ 4.29, p ⬍ .03. A post hoc Tukey test showed that
the future alone and future belonging groups differed significantly
at p ⬍ .05; the misfortune control group was not significantly
different from the other two groups, lying somewhere in the
How did misfortune control participants manage to earn good
overall scores despite an elevated rate of mistakes? Their relatively
high number of attempts provides the explanation. They attempted
more problems than the future alone participants, which enabled
them to get almost as many correct as did people in the future
belonging condition while simultaneously getting as many wrong
as people in the future alone condition.
These results are relevant to understanding the effects of social
exclusion. The low scores achieved by participants in the future
alone condition apparently resulted from two separate impair-
ments. These participants made several more mistakes, and they
also attempted fewer problems than did people in the other con-
ditions (particularly the misfortune control condition). We said in
the introduction that the heightened arousal explanation for im-
pairments predicted an increase in speed and a decrease in accu-
racy. The misfortune control condition conformed to this pattern,
but the future alone condition did not. Indeed, the significant
reduction in speed among the future alone participants speaks
strongly against the hypothesis that the effects of social exclusion
were mediated by an increase in arousal.
Mood and emotion. We also wanted to assess whether mood
was driving the effects. An ANOVA on the one-item measure
revealed significant differences in mood among the three groups,
F(2, 37) ⫽ 4.32, p ⬍ .03. The future alone (M ⫽ 4.00, SD ⫽ 0.91)
and misfortune (M ⫽ 4.23, SD ⫽ 0.93) groups both reported
neutral mood on the 1–7 scale, whereas the future belonging group
reported a slightly more positive mood (M ⫽ 4.93, SD ⫽ 0.73).
This pattern casts doubt on any hypothesis that mood is a medi-
ating factor, because the misfortune control group resembled the
future alone group in mood but resembled the future belonging
group in IQ test scores.
Still, we conducted additional tests to investigate the possibility
of mediation by mood. In these analyses, we compare the future
alone condition with the misfortune control condition, as this
represents the most stringent test of the mood mediation hypoth-
esis. Consistent with the ANOVA reported earlier, we found a
significant correlation, r(24) ⫽ .45, p ⬍ .03, indicating poorer
performance by socially excluded individuals. Next, we computed
the correlation between mood and IQ test performance and found
that there was no relationship between mood and performance,
r(24) ⫽ .11, ns.
If mood mediates the path between social exclusion and IQ test
performance, then the correlation between exclusion (condition)
and IQ test performance should become nonsignificant when con-
trolled for mood (Baron & Kenny, 1986). However, this was not
the case. When controlled for mood, social exclusion (condition)
was still correlated with IQ test performance, r(23) ⫽ .44, p ⬍ .03.
(Note that the partial correlation reduces the degrees of freedom
by 1.) The mood mediation hypothesis also predicted that mood
would be significantly correlated with IQ test performance even
after exclusion (condition) was controlled. This was not the case;
the partial correlation of mood and IQ test performance, controlled
for exclusion, was not significant, r(23) ⫽ .06, ns. All of these
results are contrary to the hypothesis of mediation by mood.
Four conclusions were suggested by this first study. First, a
diagnostic forecast of future social exclusion caused a significant
drop in intelligent performance. Second, this was not due to
participants merely hearing any bad forecast, because a prediction
of being accident prone did not produce the same decrement.
Third, the decline in performance reflected both a higher rate of
errors and a reduced number of problems attempted (i.e., both
speed and accuracy deteriorated). The loss of accuracy may be a
result of participants hearing any sort of bad news, because it was
also found in the misfortune control group, but the decline in speed
(number of problems attempted) was unique to the socially ex-
cluded individuals. The decline in speed also contradicts the pre-
diction based on heightened arousal as mediator of the effects of
social exclusion. Fourth, bad mood did not mediate these effects.
The second experiment was designed to investigate the effects
of social exclusion on learning and memory. Experiment 1 shows
BAUMEISTER, TWENGE, AND NUSS
that social exclusion can impair subsequent performance on an
intelligence test. In principle, this could be caused either by im-
paired processing of new information (encoding) or by impaired
retrieval of stored information from memory. Experiment 2 seeks
to tease these apart. To accomplish this, we varied the sequence of
steps in the procedure. Obviously, the memory retrieval test has to
be administered after the encoding task, by definition. But by
altering the timing of the exclusion manipulation and the debrief-
ing, we were able to ensure that only the encoding or only the
retrieval task was performed while the participant was under the
influence of the social exclusion manipulation.
More precisely, one condition (recall affected) permitted par-
ticipants to encode the information first. Participants read passages
from the Reading Comprehension portion of the Graduate Record
Examination (GRE). After they finished reading, they received the
manipulation of social exclusion, followed by the test for recall of
the passages they had read, followed, finally, by debriefing. Thus,
participants in that condition only performed the memory recall
task while under the influence of the exclusion manipulation. In
the other condition (encoding affected), the social exclusion ma-
nipulation was administered prior to the encoding (reading) of the
passage. Then participants were debriefed to remove the effects of
the social exclusion manipulation, and recall was tested afterward.
In this condition, therefore, only the encoding was performed
while the participant was under the influence of the social exclu-
sion manipulation. This resulted in a 3 ⫻ 2 experimental design,
with three exclusion manipulation conditions and two encoding
versus recall affected conditions.
A second procedural change introduced in Experiment 2 was
that we removed the time limit for the test. Experts disagree as to
whether intelligent performance is best measured with timed or
untimed tests, and in any case it was plausible that the strict time
limit used in Experiment 1 could have had some differential effect
on performers in the different conditions. (Observed differences in
number of questions attempted lend credence to the notion that the
time limit had a differential impact on performance scores.) To
increase generality, therefore, we allowed participants in Experi-
ment 2 as much time as they wanted to complete the recall test.
We also followed the GRE’s own procedure of including both
easy and difficult sections. The easy passage was simple and short,
and the questions focused on recapitulating material from the
passage. The difficult passage was longer and more complicated,
and the questions required the person to be able to reason and think
(as in understanding implications) about the material.
We predicted that social exclusion would lead to a general
impairment of intelligent performance on all measures combined.
The different mediating theories would make different predictions
about the pattern of impairments. The arousal explanation would
predict improved performance on the easy task and impaired
performance on the difficult task. Furthermore, the effects of
arousal should be most pronounced on retrieval, but arousal could
affect encoding also if it leads participants to screen out some
relevant information. The self-regulation of affect theory predicts
that only the most controlled processes should show impairments,
and these should mainly involve the difficult recall questions
(which required some active thinking rather than just rote mem-
ory). Thus, the decrements should mainly be found on the recall
(not encoding) task and in the high difficulty condition. Last, the
rumination hypothesis predicts impairments on both the encoding
and the retrieval tasks, because the inner process of rumination
would take attention away from the job of encoding external
information and would also distract the participant from figuring
out what information needed to be retrieved to answer the memory
questions. Both the easy and the difficult versions should show
some decrements (though not necessarily to the same degree),
because the inner distraction would prevent information from
being encoded and processed.
The participants were 65 undergraduates participating to fulfill a course
requirement in introductory psychology. There were 41 men and 24
women, 72% White and 28% racial minority. Average age was 19.2 years.
The consent form told participants that they would complete a person-
ality measure and perform some judgment tasks. The exclusion manipula-
tion was identical to that used in Experiment 1: Participants completed the
Eysenck Personality Questionnaire and were randomly assigned to hear
that they would likely end up alone later in life (future alone), would suffer
frequent accidents (misfortune), or would have many stable, fulfilling
relationships throughout life (future belonging). They then completed a
one-item mood measure, rating their mood from 1 (negative)to7
Participants were given 3 min to read two passages taken from the
Verbal sections of a GRE. One passage was long and difficult, and the
other passage was short and comparatively easy. Participants then an-
swered 12 multiple-choice questions about the passages—7 about the
difficult passage, and 5 about the easy passage. Half of these questions
were taken from the actual GRE questions, and the other half were written
for this experiment.
There was no time limit for answering these questions; participants were
told they could take as long as they liked, and they were not timed. The
number of correct answers on each passage served as the main dependent
measure. Because all participants gave an answer to all questions on the
untimed test, it was not possible to analyze number of attempts, and error
rates were perfectly correlated with the number of correct answers.
We randomly assigned participants to one of two sequences of events. In
the recall affected condition, participants read the passage (encoding),
performed the short filler task of rating pictures of nature scenes, received
the exclusion feedback (exclusion), answered the questions (recall), and
were then told that the exclusion feedback was not true (debriefing). Thus,
in the recall affected condition, participants encoded the information under
normal conditions but retrieved the information under the influence of the
In the encoding affected condition, in contrast, participants received the
exclusion feedback first (exclusion), read the passages (encoding), per-
formed the filler task, were told the exclusion feedback was not true
(debriefing), and then answered the questions (recall). Thus, in the encod-
ing affected condition, participants encoded the information under condi-
tions of exclusion but retrieved the information under the normal condi-
tions, after having heard that the exclusion feedback was not true. The time
elapsed between encoding and recall was about the same in both orders of
presentation (administering the exclusion feedback took about as long as
debriefing the participant). This created three future outcome conditions
and two encoding versus recall conditions, resulting in a 3 ⫻ 2 design. We
also examined difficulty of passage (easy vs. difficult) as a within-subject
Cognitive performance. We analyzed the data separately for
the difficult as opposed to the easy questions, because the two tests
involved different passages and different numbers of questions,
SOCIAL EXCLUSION AND INTELLIGENT THOUGHT
thereby rendering direct comparison of scores on the two sets of
questions difficult to interpret. Performance on the difficult ques-
tions was analyzed with a 3 ⫻ 2 ANOVA. The interaction between
sequence (i.e., encoding affected vs. recall affected) and exclusion
feedback (future alone, future belonging, or misfortune control)
was significant, F(2, 59) ⫽ 3.21, p ⬍ .05. Table 2 shows the means
for this experiment. As inspection of Table 2 reveals, performance
quality was quite similar in all cells except for one: Future alone
participants in the recall affected condition performed significantly
worse on the difficult task, as compared with participants in all
other conditions. An ANOVA on the three cells in the recall
affected condition revealed significant variation, F(2, 33) ⫽ 4.91,
p ⬍ .01. For a comparison of the future alone and misfortune
control conditions, the effect size for the difficult questions was
d ⫽ 1.01, indicating a large difference in memory performance. No
significant differences were found within the encoding affected
condition. Thus, participants in the future alone condition showed
deficits in the recall affected condition but not in the encoding
Of particular interest was the difference between the recall
affected and the encoding affected conditions. Performance level
was essentially the same in these two procedures in the misfortune
control condition. Likewise, future belonging participants per-
formed about the same in both. But the future alone participants
performed significantly worse in the recall affected condition than
in the encoding affected condition, t(19) ⫽ 3.18, p ⬍ .005. Thus,
recall of information was impaired by social exclusion, whereas
encoding of information was not.
An ANOVA on the easy questions revealed no significant
effects. In particular, future alone participants performed just as
well as other participants on the easy questions—neither better nor
worse. We did conduct a repeated measures ANOVA within the
recall conditions, and it yielded a significant interaction between
social feedback condition and difficulty of task, F(2, 33) ⫽ 5.18,
p ⬍ .02. (Although direct comparison of answers to the easy
versus the difficult task may seem inappropriate, there was no
main effect for task difficulty, presumably because the greater
number of difficult questions partly offset their greater difficulty,
so that people got about the same number of easy and difficult
questions correct.) This confirmed that the only notable perfor-
mance decrement involved the future alone participants and the
difficult questions. These results suggest that social exclusion
impairs only the recall and use of complex information. This
suggests a decrement in controlled processing and executive
Mood. Again, we investigated the possibility that emotional
responses mediated the cognitive impairments. Mood reports did
not differ among the three social exclusion conditions, F(2,
62) ⫽ 1.73. The future alone (M ⫽ 4.22, SD ⫽ 1.00), misfortune
(M ⫽ 4.25, SD ⫽ 0.79), and future belonging (M ⫽ 4.64,
SD ⫽ 0.66) groups all reported moods near the neutral midpoint of
the scale. Mood also did not differ when examined for the recall
groups only, F(2, 33) ⬍ 1.00, ns, or for encoding groups only, F(2,
26) ⫽ 2.13, ns.A2⫻ 3 ANOVA showed no main effects and no
interactions for mood among the conditions.
Although the lack of significant differences in mood rendered a
mediation of the cognitive impairments unlikely, we conducted a
full mediation analysis for the recall groups’ performance on the
difficult passage. As in Experiment 1, we compared the future
alone and misfortune control groups. The simple bivariate corre-
lation between exclusion feedback (condition) and performance
was r(24) ⫽ .46, p ⬍ .03. Controlled for mood, this correlation
was unchanged, r(23) ⫽ .46, p ⬍ .03. Mood was not significantly
correlated with performance, r(24) ⫽ .05, and this was unchanged
when mood was controlled for exclusion (condition), r(23) ⫽
⫺.04. All of these results are directly contrary to the hypothesis of
mediation by mood.
The findings of Experiment 2 narrow the focus of the cognitive
impairments caused by social exclusion. Feedback that forecast a
lonely future led to significant and large impairments on a recall
test that was difficult. There was no corresponding impairment of
performance on an easy test. There was also no improvement on
the easy task, contrary to the view that heightened arousal mediates
the effects of social exclusion.
In contrast to the relatively poor recall, there was no sign of
impaired encoding among the socially excluded. When people read
the passage first and took the recall test under the influence of
social exclusion, they performed poorly. In contrast, if they read
the passage while under the influence of social exclusion—but
were debriefed, which thereby removed the effects of social ex-
clusion, before the recall test—they performed fine. In other
words, people seemed quite capable of processing information into
memory storage despite having received a message of social
exclusion, but social exclusion impaired their ability to retrieve
information from memory and use it to answer difficult, thought-
provoking questions. This pattern of results seems less consistent
with an explanation in terms of attention deficits (e.g., if the person
were distracted by rumination) than with an explanation in terms
of some deficit in executive function and, as a result, an impair-
ment specific to controlled processes.
As in Experiment 1, there was no sign that mood effects medi-
ated the cognitive impairments. Even the future alone participants
did not report any substantial increase in emotional distress, and
mood ratings were unrelated to cognitive performance.
Performance of Graduate Record Examination Reading
Easy passage Difficult passage
MSDF MSD F
Future alone 4.25 1.42 2.75 1.60
Misfortune 4.67 0.89 4.58 2.02
Future belonging 3.83 1.11 4.75 1.54
Future alone 3.78 1.20 4.78 1.20
Misfortune 4.70 0.48 4.60 2.17
Future belonging 3.50 1.43 4.40 2.01
Note. Scores indicate number correct; higher score indicates better per-
formance. The easy passage contained five questions; the difficult passage
contained seven questions.
dfs ⫽ 2, 33.
dfs ⫽ 2, 26.
† p ⫽ .06. ** p ⬍ .01.
BAUMEISTER, TWENGE, AND NUSS
Experiment 3 was designed to help differentiate between two
ways of interpreting the findings of Experiment 2. One way of
interpreting the results is that social exclusion feedback impairs
recall rather than encoding. The other line of interpretation is that
it affects complex cognitive tasks that require active thinking,
whereas simple and basic information processing remains unaf-
fected. The ambiguity arose because the difficult questions on the
reading comprehension test tended to require some reasoning
rather than simple rote memory of information.
Experiment 3 therefore presented participants with one of two
cognitive tasks. Half the participants were given questions from
the Analytical section of the GRE. This test is explicitly designed
to measure reasoning ability, so it seemed a good way to test for
impairments of logic and reasoning without the confound of mem-
ory retrieval deficits. In contrast, the remaining participants were
given a recall test involving nonsense syllables, which is a standard
method used by cognitive psychologists to study memory without
any confounding effects of reasoning or even meaningful thought.
If the main effect of social exclusion is on recall, then excluded
participants should show deficits on the nonsense syllable task but
not on the analytical reasoning task. In contrast, if the main impact
is on logic and reasoning, then the decrements should be found on
the analytical task and not on the nonsense syllable task. In terms
of our mediation theories, the rumination and distraction theory
predicts decrements on both tasks because of attentional load,
whereas the executive function theory predicts that the decrements
should mainly be found in the reasoning task. Reasoning requires
the self to make an active effort to extrapolate from facts by
drawing conclusions, considering alternatives, and distinguishing
among possible implications.
An additional purpose of Experiment 3 is to try a different
method of measuring emotion and mood. The first two experi-
ments failed to find much in the way of emotional distress, but it
is conceivable that the simple, one-item measure of emotion was
somehow insensitive or unsuited to capture the emotional impact
of the social feedback manipulation. Experiment 3 uses the Posi-
tive and Negative Affect Scale (PANAS; Watson, Clark, & Tel-
legen, 1988), which is a standard, respected, and frequently used
measure of emotion.
The participants were 82 undergraduates participating to fulfill a course
requirement in introductory psychology. There were 48 men and 34
women, 71% White and 29% racial minority. Average age was 18.7 years.
The consent form said the study’s purpose was to explore how person-
ality relates to performance, but the experimenter did not verbally explain
the purpose of the study. The exclusion feedback manipulation was iden-
tical to that used in Experiments 1 and 2: Participants completed the
Eysenck Personality Questionnaire and were randomly assigned to hear
that they would likely end up alone later in life (future alone), would suffer
frequent accidents (misfortune), or would have fulfilling relationships
(future belonging). They then completed the PANAS mood measure,
which yields separate scores for positive mood and negative mood. In this
sample, the positive mood scale had an internal reliability of
⫽ .85, and
the negative mood scale had an internal reliability of
Participants were then assigned to one of two tasks: solving GRE
Analytical section problems or memorizing and retrieving nonsense sylla-
bles. (Thus, this was a between-subjects, 3 ⫻ 2 design). Participants in the
GRE Analytical section condition heard the social exclusion feedback,
completed the PANAS, and were given 12 min to answer 12 analytical
reasoning questions (5 questions on one logic problem, 3 reasoning ques-
tions, and 4 questions on another logic problem. These questions are rather
long, and thus we do not reproduce examples here.) The dependent variable
was the number of questions answered correctly. Number wrong, number
of attempts, and success rate (number right divided by attempts) were also
In the nonsense syllables task, participants were given 60 s to memorize
a list of 15 nonsense syllables that varied in how easy they were to
pronounce. All of the syllables consisted of three letters (e.g., FUM, TYJ,
KOY, WEV, JEF, PIH). Participants then worked on math problems for
90 s as a filler. They then received the social exclusion feedback and
completed the PANAS measure. They were then given 90 s to recall as
many syllables as they could. The dependent variable was the number of
syllables recalled correctly. Number wrong, number of attempts, and
success rate (number right divided by attempts) were also measured.
Cognitive performance. The results of this study suggest that
social exclusion impairs reasoning but not simple recall. A one-
way ANOVA on the Analytical GRE test scores revealed signif-
icant variation among the three conditions, F(2, 44) ⫽ 5.43, p ⬍
.01. As shown in Table 3, participants in the future alone condition
answered significantly fewer of the GRE Analytical questions
correctly. For a comparison of the future alone and misfortune
control conditions, the effect size for this task was d ⫽ 1.13,
indicating a large difference.
In contrast, an ANOVA on the nonsense syllable scores failed to
find any differences between the conditions, F(2, 32) ⫽ 0.39, ns.
As Table 3 shows, there was not even any slight trend toward
inferior performance among the socially excluded participants.
Clearly, their ability to recall the nonsense syllables was unim-
paired. Nor can this be attributed to a ceiling effect due to the
nonsense syllable task being too easy (comparable to the no-
difference finding in the easy questions in Experiment 3), because,
on average, participants could only correctly recall between 5
and 6 of the 15 stimulus syllables.
Performance on Graduate Record Examination (GRE)
Analytical Section and Nonsense Syllables Recall
FMSD M SDMSD
No. right 4.67 2.13 6.81 1.68 6.63 2.16 5.43**
No. attempts 6.94 2.05 10.31 2.09 8.51 2.34 9.46***
No. wrong 2.27 2.22 3.50 2.00 1.88 2.00 2.67††
(right/attempts) 0.69 0.27 0.68 0.17 0.80 0.20 1.45
No. right 5.75 1.54 5.67 1.61 5.18 1.83 0.39
No. attempts 6.67 1.30 6.92 1.24 6.09 2.26 0.75
No. wrong 0.92 0.67 1.25 1.14 0.91 1.14 0.45
(right/attempts) 0.86 0.12 0.82 0.19 0.86 0.17 0.30
dfs ⫽ 2, 44.
dfs ⫽ 2, 32.
†† p ⫽ .08. ** p ⬍ .01. *** p ⬍ .001.
SOCIAL EXCLUSION AND INTELLIGENT THOUGHT
To confirm the difference between the two tasks, we conducted
a3⫻ 2 ANOVA. The interaction between task and exclusion was
significant, F(2, 76) ⫽ 3.65, p ⬍ .05. Because some might object
that the number of nonsense syllables recalled was not directly
comparable to the number of logic questions answered correctly,
we standardized scores within task type and repeated the analysis.
The result was nearly the same, with the interaction between task
type and social feedback significant, F(2, 76) ⫽ 3.20, p ⬍ .05.
Thus, future alone participants were hindered in their reasoning
ability but not in their simple recall ability.
Attempts and accuracy rate. We conducted a series of addi-
tional analyses to investigate possible differences in accuracy rate,
speed, and effort. The results of these one-way ANOVAs (within
task type) are summarized in Table 3.
The only significant difference to emerge from these analyses
involved the number of questions attempted on the Analytical test.
Future alone participants attempted the fewest, whereas misfortune
control participants attempted the most. This result was compara-
ble to what we found in Experiment 1 (and, once again, it directly
contradicts the view that increased arousal mediates the effects of
social exclusion). There was a marginal trend toward more errors
by the misfortune control participants. (If this is treated as a
replication of Experiment 1, a one-tailed test is appropriate, and
the difference was significant at p ⬍ .05.) The pattern of results for
the success rate (which we computed by dividing the number right
by the number attempted) was the same as in Experiment 1, but the
result was not significant, possibly because of a floor effect created
by the relatively small number of questions. In any case, it appears
that misfortune control participants resembled and surpassed the
future belonging participants in how hard they tried, but they
resembled the future alone participants in their success rate.
On the nonsense syllable task, no differences were found on any
of the measures. Given the nature of the task, it may be more
difficult to measure attempts than on other tasks. In any case, these
results are consistent with the view that social exclusion feedback
has no effect on people’s ability to store and remember informa-
tion. This pattern is difficult to reconcile with the hypothesis that
excluded people were actively ruminating about the message of
exclusion and so were less able to attend to the information coming
in from their environment.
Mood and emotion. The PANAS permitted a more detailed
and thorough assessment of mood than was possible in Experi-
ments 1 and 2. Across both task conditions, ANOVA on PANAS
scores revealed significant variation in negative affect as a func-
tion of social feedback, F(2, 79) ⫽ 3.33, p ⬍ .05. Future alone
participants reported the most negative mood (M ⫽ 15.48,
SD ⫽ 6.57), followed by misfortune (M ⫽ 14.07, SD ⫽ 3.50) and
future belonging (M ⫽ 12.33, SD ⫽ 2.34) participants. The dif-
ference between the future alone and future belonging group was
d ⫽ 0.71, a moderate to large effect size. A 2 ⫻ 3 ANOVA showed
no interaction and no main effect for task; the main effect for
exclusion condition was significant, F(2, 79) ⫽ 3.16, p ⬍ .05.
There were no differences in positive mood, F(2, 79) ⫽ 0.65. A
2 ⫻ 3 ANOVA showed no main effects and no interactions for
positive mood. Thus, social exclusion increased negative mood but
did not affect positive mood.
The possibility of mediation by mood was only relevant to
participants who worked on the GRE Analytical task (because
there were no differences in cognitive performance on the other,
nonsense syllable task). Contrary to the mood mediation hypoth-
esis, when we restricted the ANOVA to participants who per-
formed the Analytical task, the analysis failed to show any signif-
icant variation in negative mood as a function of exclusion
feedback, F(2, 44) ⫽ 2.13, ns (M ⫽ 15.27, SD ⫽ 6.60 for future
alone; M ⫽ 13.63, SD ⫽ 3.34 for misfortune; and M ⫽ 12.00,
SD ⫽ 2.25 for future belonging). There were also no differences in
positive mood, F(2, 44) ⫽ 0.84, ns.
Although the mood differences among participants who per-
formed the Analytical task failed to reach significance, we went
ahead with a mediation analysis. Similar to the previous analyses,
we examined the future alone and misfortune control conditions
only. The simple bivariate correlation between social exclusion
and performance on the GRE Analytical problems was r(29) ⫽
.50, p ⬍ .004. Controlled for negative mood, the correlation
between exclusion and performance was r(28) ⫽ .50, p ⬍ .005.
Controlled for positive mood, it was again unchanged from the
bivariate, r(28) ⫽ .48, p ⬍ .008. The bivariate correlation between
performance and negative mood was r(29) ⫽⫺.05, ns. When
controlled for condition, the correlation was r(28) ⫽ .03, ns. The
bivariate correlation between performance and positive mood was
r(29) ⫽⫺.32, ns; after condition was controlled for, this was
r(28) ⫽⫺.26, ns. In short, these data satisfied none of the
requirements for a conclusion that mood mediated between con-
dition and performance.
To thoroughly investigate this issue, we also examined specific
facets of negative mood as measured by the PANAS. As a main
effect, there were no significant differences among conditions in
anxiety (measured by the two items “jittery” and “nervous”), F(2,
44) ⫽ 0.83, ns. There were nearly significant differences in two
fear items (“scared” and “afraid”), F(2, 44) ⫽ 2.91, p ⫽ .06, and
significant differences in two general distress items (“upset” and
“distressed”), F(2, 44) ⫽ 5.91, p ⬍ .006. However, none of these
specific facets mediated the effect of exclusion on analytical
problem performance. They did not correlate with performance
when controlled for exclusion condition, and the correlation be-
tween exclusion and performance was not changed significantly
when controlled for any of the facets.
Experiment 3 sheds additional light on the nature of the cogni-
tive impairments caused by social exclusion. The results indicate
that a diagnostic forecast of future social exclusion can cause a
significant reduction in logic and reasoning ability, but this fore-
cast did not affect the ability to recall simple information. More
precisely, people who were told they would end up alone in life
performed worse than other participants on questions drawn from
the GRE Analytical test. They were able to recall nonsense sylla-
bles just as effectively as people in other conditions, however.
These findings point to an impairment in controlled processes
rather than a broad attentional decrement, insofar as rejected
people were fully able to encode and retrieve information as long
as they did not have to engage in active reasoning. Put another
way, the results are most consistent with the view that social
exclusion leads to a deficit in controlled processes and executive
function, possibly caused by the need to devote the self’s regula-
tory resources to stifling emotional distress. They do not fit the
view that recall in general is disrupted.
BAUMEISTER, TWENGE, AND NUSS
The effects were specific to social exclusion rather than being
broadly the result of hearing bad news. Participants in the misfor-
tune control condition, who were told that their future life would
contain many painful accidents and injuries, got just as many items
correct as did people who received the welcome feedback that their
future life would be full of rich, satisfying relationships. The
anticipation of future social isolation was thus more damaging than
the forecast of future harm and misfortune.
The findings of Experiment 3 also echo those of Experiment 1
with regard to effort and accuracy. Participants in the misfortune
control group were not totally unaffected by the forecast of future
bad times, but they responded by increasing their efforts to do well
on the task. They attempted more problems. The trend toward
making more errors was also replicated (although it only reached
significance by a one-tailed test). These participants’ error rate was
identical to that of the future alone participants and lower than that
of the future belonging group, although the error rate differences
failed to reach significance. Still, these results are broadly consis-
tent with the view that the anticipation of social exclusion impairs
performance in two ways, both by reducing effort and by increas-
ing errors. Other kinds of bad news apparently only produce one of
those, namely the increase in errors, which can be effectively
offset by an increase in effort.
Experiment 3 used a more thorough and standard measure of
emotion than Experiments 1 and 2 used. Despite the improved
measurement, however, there was still no sign that mood mediated
the cognitive impairments. In fact, the data satisfy none of the
requirements for a mediation analysis. There is some indication
that social exclusion produced a moderate increase in negative
affect, but it failed to predict performance. Positive affect was
unaffected by the exclusion manipulation.
The results of this investigation can be summarized as follows.
In all three studies, people exhibited significant cognitive decre-
ments after they were told that they were likely to end up alone in
life. Thus, the prospect of social exclusion reduced people’s ca-
pacity for intelligent thought. Moreover, the decrements in intel-
ligent performance qualified as large effects every time.
The drop in intelligent performance appears to be specific to
social exclusion. All three studies contained a control group that
forecast future misfortune in the physical domain, such as acci-
dents and injuries. Thus, there were two conditions (future alone
and misfortune control) that conveyed bad news to participants.
Moreover, in the one study that found significant differences in
mood (Experiment 2), the misfortune control group yielded a mean
mood that was closer to that of the future alone group than to that
of the future belonging group. But these bad moods in the mis-
fortune group were not accompanied by comparable decrements in
cognitive performance. The misfortune control group consistently
scored as high as the future belonging group on the measures of
intelligent performance. Only the message of social exclusion led
to significant reductions in cognitive performance.
The forecast of social exclusion caused people to attempt fewer
problems as well as make more errors on those they did attempt.
Speed and accuracy are not necessarily linked in that way, and, in
fact, the traditional assumption and finding is that speed and
accuracy are generally associated in the opposite manner, so that
one improves at the other’s expense, such as when performers
trade off speed for accuracy. The simultaneous decline in both
speed and accuracy is reminiscent of serious, wide-ranging cog-
nitive impairments (e.g., from head injuries) rather than a strategic
Not all cognitive functions were uniformly impaired by our
manipulation of social exclusion. People who expected to end up
alone showed the biggest decrements mainly in reasoning and
thinking. Simple information processing seemed unaffected. Spe-
cifically, we found significant and large declines in performance
on an intelligence test (Experiment 1), on challenging questions
that involved recalling a complex passage and answering questions
on the basis of that information (Experiment 2), and on a test of
logic and reasoning (Experiment 3). In contrast, social exclusion
produced no decrement in people’s responses to relatively simple
and straightforward questions about a simple reading passage
(Experiment 2), nor on people’s ability to perform a rote memory
task involving goal-directed retrieval of nonsense syllables (Ex-
periment 3). Experiment 2 also found that people who read a
complex passage after receiving the social exclusion feedback
were able to recall it without any noticeable decrement, provided
that we had removed the message of social exclusion by debriefing
them before they took the recall test. This shows that they were
able to encode the information effectively into memory storage
even though they had just received the message of social exclu-
sion. After all, no matter how good one’s memory is, one cannot
recall something one has never learned.
These tasks differ in their reliance on executive function. Direct
retrieval of information is generally regarded as an automatic
process, at least once the goal of retrieving the information has
been activated (Logan, 1989). Thus, simply recalling exactly what
one has been told does not invoke executive function. On both the
nonsense syllables task and the easy recall task, direct retrieval was
all that was required, and rejected people performed fine. In
contrast, automaticity is undermined on tasks that require effort
and control (Bargh, 1994; Kahneman, 1973), such as engaging in
logical reasoning or extrapolating from information in memory,
and on these tasks, the socially excluded people showed significant
The fact that socially excluded people performed well on many
tests helps rule out some alternative interpretations. For example,
some might suggest that the social exclusion feedback made par-
ticipants angry at the experimenter or reduced their willingness to
complete the tasks in the experiment. (To be sure, such a reaction
might have been expected to yield clearer evidence of antipathy on
the mood measures, and at least negative moods would have been
more strongly correlated with cognitive performance than we
found, so these interpretations are inconsistent with some find-
ings.) But it is hard to imagine that these supposedly disgruntled
participants would have encoded the information in the complex
reading passage so effectively or matched the high achieving
groups on the nonsense syllable task. Likewise, the debriefing in
Experiment 2 involved telling participants that the experimenter
had deceived them, and if they were already disgruntled, that
would have likely made them even more so. They should therefore
have performed especially poorly on the recall test that came after
the debriefing. Instead, we found that they performed especially
well. In a similar vein, if social exclusion feedback conveyed some
implicit experimenter demand characteristic that induced partici-
SOCIAL EXCLUSION AND INTELLIGENT THOUGHT
pants to cooperate with the experimenter by performing poorly,
then this feedback should have yielded more uniform decrements
in performance, including on the nonsense syllables task, but it did
not. In short, it does not appear that the results can easily be
explained on the basis of demand characteristics or disgruntled
We had three theories about how social exclusion might impair
cognitive performance. The theory based on heightened arousal
was contradicted by multiple findings. Arousal should have pro-
duced an increase in speed of performance, but, instead, the
socially excluded people showed a decrease in speed in both
studies in which we measured it (Experiments 1 and 3; Experi-
ment 2 used untimed tests). Also, arousal should have facilitated
performance on the simple tasks, but there was no sign of that
either. Last, the reports failed to show much sign of emotional
distress, and, indeed, socially excluded participants tended to re-
port neutral moods. The slight increases in distress in Experiment 3
did not mediate performance. Hence, we reject the arousal
Another possible explanation involves attention deficits caused
by ruminating about the message of exclusion. Several findings
appear inconsistent with this explanation. Socially excluded people
were quite capable of encoding and recalling nonsense syllables
(Experiment 3) and of encoding meaningful information into
memory (Experiment 2). If their attention had been reduced by
inner distractions, they would have shown decrements in their
intake of new information, but they did not. Hence, the attention
deficit explanation also seems unlikely.
Instead, the results seem most consistent with the view that
social exclusion specifically impairs controlled processes, such as
by monopolizing some of the resources of the self’s executive
function. Impairments were found on tasks that required active
thinking, such as reasoning and logic, whereas the relatively au-
tomatic (less efficient, less controlled) tasks were unaffected. Re-
cent work has suggested that the self’s executive function uses a
common resource for diverse acts of volition and self-regulation,
including controlling feelings, controlling thoughts, resisting
temptation, making choices, and responding actively instead of
passively (Baumeister et al., 1998; Muraven et al., 1998). Al-
though we have no direct evidence that socially excluded partici-
pants were engaged in self-regulation, the surprising lack of emo-
tion they reported in response to what must have come as upsetting
news is at least consistent with the view that they suppressed their
negative feelings. In Experiment 1, for example, the mean affec-
tive state in the future exclusion condition was precisely on the
neutral point of the scale, despite the fact that participants had been
told that much of their future was likely to be spent all alone in the
world. This effort at affect regulation (if that is indeed what it was)
would have used up some of the self’s executive resources, ren-
dering it difficult to think and reason actively but allowing the
simpler and more automatic cognitive processes to proceed.
In sum, we think we can best explain the pattern of cognitive
decrements by proposing that social exclusion constitutes a threat-
ening, aversive event but that people strive to suppress their
emotional distress, and the resulting drain on their executive func-
tion impairs their controlled processes. Although this explanation
fits what we found, it is desirable for future research to provide
some direct evidence that people respond to social exclusion with
an active effort to suppress emotion. We recognize the method-
ological difficulty of obtaining such evidence, however.
The view that social exclusion leads to a breakdown in self-
regulation can potentially explain far more than the present results.
In other studies, we have found that socially excluded people
behave in more aggressive ways (Twenge, Baumeister, Tice, &
Stucke, 2001), whereas prosocial behavior is reduced (Twenge,
Ciarocco, & Baumeister, 2001). It is crucial that these studies
likewise failed to find much in the way of strong emotional distress
or mediation by mood, even though they included studies that
manipulated direct, current social rejection rather than the mere
forecast of future exclusion like the present three studies. The
aggression and prosocial behavior findings had a paradoxical
aspect, because one might think that social exclusion should mo-
tivate people to try harder to gain acceptance, such as by reducing
aggression and increasing helpful, generous, and cooperative be-
havior, whereas the opposite appears to be more correct. Yet a
breakdown in self-regulation would potentially explain both pat-
terns of findings, insofar as prosocial behavior often requires one
to put the collective ahead of self-interest, and aggressive impulses
are often held back by inner restraints.
One important limitation is that the present studies investigate
how social exclusion affects intelligent thought about matters
having nothing to do with social belongingness per se. Other work
has found that social exclusion can have specific and positive
effects on cognitive processes pertaining to belongingness and the
self (e.g., Gardner, Pickett, & Brewer, 2000; Leary, 1990; Leary,
Tambor, Terdal, & Downs, 1995; Williams, 1997; Williams &
Another limitation of the present work is that the future social
relationships were generally presented in a positive light. The
choice was essentially between being embedded in a network of
satisfying, positive relationships or being alone. We did not in-
clude a control group in which people expected a future that would
involve being firmly embedded in a network of conflict-filled,
hostile, or antagonistic relationships. Such a prediction might be
difficult to make plausible in cultures such as that of the United
States, in which most people can leave and replace aversive or
abusive relationships, but it might be particularly relevant to other
cultures in which people are locked into firm social groups regard-
less of personal preference. There is some evidence that the need
to belong is not satisfied by hostile, conflictual, or antagonistic
relationships (see Baumeister & Leary, 1995). For example, Vi-
nokur and van Ryn (1993) found that the benefits of social support
were limited to positive, pleasant interactions, and, indeed, social
relationships characterized by conflict, criticism, and undermining
were harmful to mental health. Because our theorizing began with
the need to belong, we assumed that aversive relationships would
be similar to if not worse than being alone, but the data from the
present experiments did not test this.
We began by suggesting that human survival in the evolutionary
past could be facilitated either by intelligent problem solving or by
affiliation with a social group—and that it seemed adaptive for the
failure of one strategy to result in a facilitation of the other. This
line of reasoning predicts that people who failed to secure mem-
bership in social groups should respond with an increase in the
capacity for intelligent thought.
Our three studies consistently show the opposite effect, how-
ever. Substantial decrements in intelligent thought were repeatedly
BAUMEISTER, TWENGE, AND NUSS
found among people who received messages of social exclusion.
Rather than compensating for each other, intelligent thought and
social inclusion seem to have a positive, direct relationship, if
anything. Our results are more consistent with the view that
intelligence evolved as a means to support and facilitate social
relations rather than to compensate for the absence of their advan-
tages (although our data fall far short of permitting such a sweep-
ing conclusion). Our findings could even be taken to suggest that
people responded as if being excluded from social groups removed
the need for intelligent thought. Although some may regard social
cognition as a small aspect or special case of human thought, the
present results offer at least a hint that cognition in general may
have an important social basis.
Barchas, P. (1986). A sociophysiological orientation to small groups. In E.
Lawler (Ed.), Advances in group processes (Vol. 3, pp. 209–246).
Greenwich, CT: JAI Press.
Bargh, J. A. (1994). The four horsemen of automaticity: Awareness,
intention, efficiency, and control in social cognition. In R. S. Wyer, Jr.,
& T. K. Srull (Eds.), Handbook of social cognition (pp. 1–40). Hillsdale,
Baron, R., & 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.
Baumeister, R. F., Bratslavsky, E., Muraven, M., & Tice, D. M. (1998).
Ego depletion: Is the active self a limited resource? Journal of Person-
ality and Social Psychology, 74, 1252–1265.
Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for
interpersonal attachments as a fundamental human motivation. Psycho-
logical Bulletin, 117, 497–529.
Baumeister, R. F., & Tice, D. M. (1990). Anxiety and social exclusion.
Journal of Social and Clinical Psychology, 9, 165–195.
Cohen, J. (1977). Statistical power analysis for the behavioral sciences.
New York: Academic Press.
Eysenck, H. J., & Eysenck, S. B. G. (1975). Manual of the Eysenck
Personality Questionnaire. San Diego, CA: EdITS.
Gardner, W. L., Pickett, C. L., & Brewer, M. B. (2000). Social exclusion
and selective memory: How the need to belong influences memory for
social events. Personality and Social Psychology Bulletin, 26, 486–496.
Gottfredson, M. R., & Hirschi, T. (1990). A general theory of crime.
Stanford, CA: Stanford University Press.
Heath, S. B. (1991). “It’s about winning!” The language of knowledge in
baseball. In L. Resnick, J. Levine, & S. Teasley (Eds.), Perspectives on
socially shared cognition (pp. 101–124). Washington, DC: American
Janda, L. (1996). The psychologists’ book of self-tests. New York: Berkley.
Janda, L. H., Fulk, J., Janda, M., & Wallace, J. (1995). The development of
a test of General Mental Abilities. Unpublished manuscript, Old Domin-
Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ:
Leary, M. R. (1990). Responses to social exclusion: Social anxiety, jeal-
ousy, loneliness, depression, and low self-esteem. Journal of Social and
Clinical Psychology, 9, 221–229.
Leary, M. R., Tambor, E. S., Terdal, S. K., & Downs, D. L. (1995).
Self-esteem as an interpersonal monitor: The sociometer hypothesis.
Journal of Personality and Social Psychology, 68, 518–530.
Levine, J. M., & Moreland, R. L. (1991). Culture and socialization in work
groups. In L. Resnick, J. Levine, & S. Teasley (Eds.), Perspectives on
socially shared cognition (pp. 257–279). Washington, DC: American
Logan, G. D. (1989). Automaticity and cognitive control. In J. S. Uleman
& J. A. Bargh (Eds.), Unintended thought (pp. 52–74). New York:
Muraven, M., Tice, D. M., & Baumeister, R. F. (1998). Self-control as
limited resource: Regulatory depletion patterns. Journal of Personality
and Social Psychology, 74, 774–789.
National Research Council. (1993). Understanding and preventing vio-
lence. Washington, DC: National Academy Press.
Newcomb, A. F., Bukowski, W. M., & Pattee, L. (1993). Children’s peer
relations: A meta-analytic review of popular, rejected, neglected, con-
troversial, and average sociometric status. Psychological Bulletin, 113,
Resnick, L. B., Levine, J. M., & Teasley, S. D. (1991). Perspectives on
socially shared cognition. Washington, DC: American Psychological
Sampson, R. J., & Laub, J. H. (1993). Crime in the making: Pathways and
turning points through life. Cambridge, MA: Harvard University Press.
Twenge, J. M., Baumeister, R. F., Tice, D. M., & Stucke, T. S. (2001). If
you can’t join them, beat them: Effects of social exclusion on aggressive
behavior. Journal of Personality and Social Psychology, 81, 1058–1069.
Twenge, J. M., Catanese, K. R., & Baumeister, R. F. (2002). Social
exclusion causes self-defeating behavior. Journal of Personality and
Social Psychology, 83, 606–615.
Twenge, J. M., Ciarocco, N. J., & Baumeister, R. F. (2001). Help! I need
somebody: Effects of social exclusion on prosocial behavior. Manuscript
submitted for publication.
Vinokur, A. D., & van Ryn, M. (1993). Social support and undermining in
close relationships: Their independent effects on mental health of un-
employed persons. Journal of Personality and Social Psychology, 65,
Walsh, A., Beyer, J. A., & Petee, T. A. (1987). Violent delinquency: An
examination of psychopathic typologies. Journal of Genetic Psychology,
Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and vali-
dation of brief measures of positive and negative affect: The PANAS
scales. Journal of Personality and Social Psychology, 54, 1063–1070.
Williams, K. D. (1997). Social ostracism. In R. M. Kowalski (Ed.),
Aversive interpersonal behaviors (pp. 133–170). New York: Plenum.
Williams, K. D., Cheung, C. K. T., & Choi, W. (2000). CyberOstracism:
Effects of being ignored over the Internet. Journal of Personality and
Social Psychology, 79, 748–762.
Williams, K. D., & Sommer, K. L. (1997). Social ostracism by one’s
coworkers: Does rejection lead to loafing or compensation? Personality
and Social Psychology Bulletin, 23, 693–706.
Wilson, J. O., & Herrnstein, J. Q. (1985). Crime and human nature. New
York: Simon & Schuster.
Zajonc, R. B. (1965). Social facilitation. Science, 149, 269–274.
Received June 7, 2001
Revision received March 18, 2002
Accepted April 1, 2002 䡲
SOCIAL EXCLUSION AND INTELLIGENT THOUGHT