How low can you go ? Ostracism by a computer is sufficient to lower self-reported levels of belonging, control, self-esteem, and meaningful existence

ARTICLE IN PRESS
Journal of
Experimental
Social Psychology
Journal of Experimental
Experimen
Social Psychology
Psycholog xxx (2004) xxx–xxx
www.elsevier.com/locate/jesp
How low can you go? Ostracism by a computer is su?cient to
lower self-reported levels of belonging, control, self-esteem,
and meaningful existenceq
Lisa Zadro,a,* Kipling D. Williams,b,* and Rick Richardsona
a School of Psychology, Faculty of Science, University of New South Wales, UNSW Sydney 2052, Australia
b Department of Psychology, Macquarie University, Sydney, NSW 2109, Australia
Received 20 June 2003; revised 22 October 2003
Abstract
Previous research has demonstrated self-reports of lower levels of four fundamental needs as a result of short periods of face-to-
face ostracism, as well as short periods of Internet ostracism (Cyberball), even when the ostracizing others are unseen, unknown, and
not-to-be met. In an attempt to reduce the ostracism experience to a level that would no longer be aversive, we (in Study 1) con-
vinced participants that they were playing Cyberball against a computer, yet still found comparable negative impact compared to
when the participants thought they were being ostracized by real others. In Study 2, we took this a step further, and additionally
manipulated whether the participants were told the computer or humans were scripted (or told) what to do in the game. Once again,
even after removing all remnants of sinister attributions, ostracism was similarly aversive. We interpret these results as strong
evidence for a very primitive and automatic adaptive sensitivity to even the slightest hint of social exclusion.
Ó 2004 Elsevier Inc. All rights reserved.
Ostracism—the act of being excluded and ignored
self-esteem, and meaningful existence. These studies
(Williams, 2001)—is ubiquitous. Reviewing the etho-
consisted largely of a laboratory-based ball-tossing
logical, anthropological, and social psychological liter-
paradigm, in which participants partook in a sponta-
ature reveals that ostracism is used by many species, by
neous ball-toss game with two other confederate par-
children and adults, in primitive tribes and modern in-
ticipants. When the other two individuals began tossing
dustrialized societies, as a formal method of reprimand
the ball just between themselves, ostracized participants
among nations, institutions, and organizations and as
slumped in their chairs, looking despondent, after only
an informal emergent reaction on playgrounds and
4 min. Studies of long-term ostracism report incidences
hallways, with large groups, small groups, and dyads.
of attempted suicide and depression (Williams & Zadro,
Ostracism is also powerful. Studies have shown that
2001), and even mass-shootings (Leary, Kowalski,
people subjected to ostracism for a short period of time
Smith, & Phillips, 2003).
report worsened mood, anger, and lower levels of four
Recently, Williams, Cheung, and Choi (2000) re-
state measures of needs proposed by Williams (1997,
ported that individuals who played a virtual ball-toss
2001) to be threatened by ostracism: belonging, control,
game on the computer, ostensibly with others who were
logged on to the website, also reported worsened mood
q
and lower need levels. Additionally, if given the op-
This research was funded by an Australian Research Council
portunity to make spatial judgments with a new group
Grant to the second author, and comprised part of the ?rst authorÕs
doctoral dissertation. We would like to extend our thanks to Bibb
of individuals, ostracized participants were more likely
Latan
e for comments that initially triggered our interest in pursuing
to conform to this groupÕs unanimously incorrect an-
this line of inquiry, and Keith Lim and Trevor Case for their technical
swers.
assistance.
*
One goal of the Williams et al. (2000) research was to
Corresponding authors.
?rst establish a baseline condition in which ostracism
E-mail addresses: l.zadro@unsw.edu.au (L. Zadro), kip@psy.mq.
edu.au (K.D. Williams).
would have no e?ects, and then to add on necessary
0022-1031/$ - see front matter Ó 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.jesp.2003.11.006

---

ARTICLE IN PRESS
2
L. Zadro et al. / Journal of Experimental Social Psychology xxx (2004) xxx–xxx
factors until ostracism had a negative impact on the four
visualization, and that to assist them in practicing their
needs. The authors were surprised to ?nd out that what
skills at mental visualization they would be playing an
they thought would be the baseline turned out to be
Internet ball-toss game on the computer. They were told
su?ciently adverse. One goal of the present studies,
that performance in the game was unimportant, and
therefore, was to create conditions even less meaningful
instead, the game was merely a means for them to en-
than those in the Williams et al. (2000) studies, in order
gage their mental visualization skills. They were asked to
to determine the necessary and su?cient conditions for
visualize the situation, themselves, and the other players.
ostracism to have an aversive impact. To the extent that
The game was accessed via the Internet (a downloadable
such a baseline can be established, then we can gain a
version of this game is available at: http://www.psy.m-
clearer idea what aspects of ostracism are essential.
q.edu.au/sta?/kip/Announce/cyberball). The game de-
picts three ball-tossers, the middle one representing the
participant. The game is animated and shows the icon
Study 1
throwing a ball to one of the other two. When the ball
was tossed to the participants, they were instructed to
In Study 1, participants were either ignored or in-
click on one of the other two icons to indicate their in-
cluded during Cyberball—a cyber analogue of a ball-
tended recipient, and the ball would move toward that
tossing game (Williams et al., 2000, 2002)—by two
icon. The game was set for 40 total throws (the game
other players whose identity was manipulated. Targets
lasted approximately 6 min). Once the instructions were
were told that they were playing Cyberball with either
read, the participant clicked the ‘‘Next’’ link and the
two computer-generated players or two human players
program randomly assigned them to one of the four
prior to the start of the game. If the identity of the
conditions. At the end of the game, the website in-
source is an important component in determining the
structed participants to inform the experimenter that
aversiveness of ostracism, then targets who are ostra-
they had ?nished, and they were then instructed to ?ll
cized by two human players should report lower levels
out a post-experiment questionnaire.
of primary needs than targets who are ostracized by
Inclusion/ostracism manipulation. If assigned to the
two computer-generated players or targets who are
inclusion condition, participants received the ball for
included in the game. If the identity of the source is not
roughly one-third of the total throws. If assigned to the
important—rather, the very act of ostracism is aversive
ostracism condition, participants received the ball twice
enough to induce deleterious psychological e?ects—
at the beginning of the game, and for the remaining
then targets who are ostracized by humans or com-
time, never received the ball again.
puters should both report lower levels of mood and of
Sources manipulation. Half the participants were told
the four primary needs when compared to targets who
that they were playing with two other individuals who
are included in the game.
were stationed in similar laboratories at two other uni-
versities in Sydney. This cover story was augmented by
Method
staged phone calls to the other experimenters making
sure that their participants were ready to go. The other
Participants and design
half were told that they were playing the game with a
Eighty ?rst-year undergraduates enrolled in intro-
computer.
ductory psychology at the University of New South
Dependent measures. The questionnaire contained
Wales were randomly assigned to a 2 (inclusionary sta-
several manipulation checks for inclusion/ostracism:
tus: ostracism vs. inclusion) Â 2 (attributed source:
‘‘What percent of the throws were thrown to you?,’’ ‘‘To
computer-generated players or human players) between-
what extent were you included by the other participants
S design. Participants volunteered to take part in the
during the game?,’’ and a 9-point bipolar scale (‘‘ac-
experiment in return for course credit. Eighteen partic-
cepted/rejected’’). The questionnaire also contained a
ipants were excluded because of technical di?culties
number of questions that asked participants to assess
with the computers and the Internet connection, thus
their levels of four needs that they felt during the game.
the ?nal experiment consisted of 62 participants (20
These needs were: belonging (‘‘I felt poorly accepted by
males, 42 female, M age ¼ 19.9, SD ¼ 2:7).
the other participants,’’ ‘‘I felt as though I had made a
‘‘connection’’ or bonded with one or more of the par-
Procedure
ticipants during the Cyberball game,’’ ‘‘I felt like an
One participant per session arrived at the laboratory,
outsider during the Cyberball game’’), control (‘‘I felt
and was seated in front of a computer.1 Participants
that I was able to throw the ball as often as I wanted
were told that the study involved the e?ects of mental
during the game,’’ ‘‘I felt somewhat frustrated during
the Cyberball game,’’ ‘‘I felt in control during the Cy-
1
berball game’’), self-esteem (‘‘During the Cyberball
In both studies, cardiovascular measures were taken periodically,
but only the self-report data are reported in the present paper.
game, I felt good about myself,’’ ‘‘I felt that the other

---

ARTICLE IN PRESS
L. Zadro et al. / Journal of Experimental Social Psychology xxx (2004) xxx–xxx
3
Table 1
Means and standard deviations (in parentheses) of variables in Study 1 (all scales 1 ¼ not at all to 9 ¼ very much so, unless otherwise stated)
Source
Human
Computer
Inclusion
Ostracism
Inclusion
Ostracism
ðn ¼ 18Þ
ðn ¼ 15Þ
ðn ¼ 17Þ
ðn ¼ 12Þ
Fundamental needsa
Belonging (a ¼ :74)
6.4 (1.5)
3.4 (2.1)
6.5 (1.7)
3.7 (2.3)
F ð1; 57Þ ¼ 34:99, p < :0005b
Control (a ¼ :72)
6.4 (1.7)
3.2 (1.6)
5.8 (1.7)
3.8 (1.9)
F ð1; 57Þ ¼ 32:5, p < :0005
Self-esteem (a ¼ :70)
7.1 (1.2)
5.6 (2.1)
6.9 (1.1)
5.5 (2.1)
F ð1; 57Þ ¼ 11:1, p ¼ :002
Meaningful existence (a ¼ :66)
6.8 (1.4)
3.8 (1.8)
6.5 (1.5)
3.7 (1.7)
F ð1; 58Þ ¼ 50:7, p < :0005
Moodc
6.5 (1.2)
6.4 (1.4)
6.5 (1.1)
6.5 (1.2)
Ancillary variables
I enjoyed playing the Cyberball game
4.9 (2.4)
2.8 (2.0)
4.5 (2.2)
2.9 (2.5)
F ð1; 58Þ ¼ 10:2, p ¼ :002
I felt angry during the Cyberball game
1.8 (1.6)
2.1 (1.5)
1.2 (.39)
3.3 (2.5)
F ð1; 58Þ ¼ 8:2, p ¼ :006
Manipulation checks
To what extent were you included by the participants during the game?
7.1 (1.6)
2.7 (2.0)
6.1 (1.8)
2.8 (2.1)
F ð1; 58Þ ¼ 67:3, p < :0005
What percentage of throws do you think your received during the
37.1 (13.9)
8.3 (4.8)
40.1 (17.8)
9.3 (7.5)
Cyberball game?
F ð1; 58Þ ¼ 83:9, p < :0005
Rejected–acceptedd
6.8 (1.5)
4.0 (2.1)
6.2 (1.9)
4.2 (2.2)
F ð1; 57Þ ¼ 24:3, p < :0005
a Each fundamental need score represents an average of three questions.
b All F values refer to signi?cant ostracism vs. inclusion main e?ects.
c Total mood score was an average of four 9-point bipolar questions.
d This was a 9-point scale with rejected–accepted as anchors.
participants failed to perceive me as a worthy and like-
Results
able person,’’ ‘‘I felt somewhat inadequate during the
Cyberball game’’), and meaningful existence (‘‘I felt that
Manipulation checks. There were three manipulation
my performance [e.g., catching the ball, deciding whom
checks assessing inclusionary status. As shown in Table
to throw the ball to] had some e?ect on the direction of
1, participants in the ostracism condition reported that
the game,’’ ‘‘I felt non-existent during the Cyberball
they felt signi?cantly less included and more rejected
game,’’ ‘‘I felt as though my existence was meaningless
than participants in the inclusion condition (smallest F
during the Cyberball game’’). Mood was assessed using
was for rejection, F ð1; 57Þ ¼ 24:3, p < :0001). Partici-
four bipolar questions (bad/good, happy/sad, tense/re-
pants in the ostracism condition also reported that they
laxed, and aroused/not aroused). The questionnaire also
received the ball less often during the game than par-
contained two ancillary variables (‘‘I felt angry during
ticipants in the inclusion condition, F ð1; 58Þ ¼ 83:9,
the Cyberball game’’ and ‘‘I enjoyed playing the Cy-
p < :0001.2 This suggests that participants correctly
berball game’’). Unless otherwise stated, all questions
perceived whether they were included or ostracized
were rated on 9-point scales (where 1 ¼ not at all, and
during the game.
9 ¼ very much so).
To assess the source manipulation, a verbal manip-
After the participants indicated that they had ?nished
ulation check was carried out at the end of the study
the questionnaire, the experimenter asked participants
prior to debrie?ng. All but two participants correctly
about their thoughts/feelings during the study, and
identi?ed whether they played the game with computer
performed a verbal manipulation check by asking par-
or human players. The two aberrant participants (both
ticipants whether they had played the game with uni-
in the human players condition) reported having played
versity students or computer players. They were then
the game with two computers rather than two humans.
thoroughly debriefed about the aims of the study,
thanked, and given course credit for participating in the
2 Degrees of freedom vary slightly in the analyses here and in Study
study.
2 because of occasional missing data.

---

ARTICLE IN PRESS
4
L. Zadro et al. / Journal of Experimental Social Psychology xxx (2004) xxx–xxx
The game had malfunctioned during their participation
F ð1; 58Þ ¼ 8:2, p ¼ :006. There was no main e?ect for
(a factor that may have led them to realize that they
whether they played with humans or computers, but
could not have been playing with two humans). Con-
there was a signi?cant interaction, F ð1; 58Þ ¼ 4:7,
sequently, their self-report data were not included in the
p ¼ :034, such that anger was higher only when they
analysis.
were ostracized by the computer.
Self-reported levels of needs. The items assessing the
Ostracized participants reported enjoying the game
four needs (state measures of belonging, control, self-
less compared with included participants, F ð1; 58Þ ¼
esteem, and meaningful existence) were reverse scored
10:2, p ¼ :002. There was no signi?cant main e?ect of
where necessary and the internal consistency of the items
source identity, or interaction between inclusionary
assessing each need were examined. CronbachÕs alpha
status and source identity, for enjoyment of the game,
coe?cients for each need were: belonging ¼ 74; con-
both F s < 1, ns.
trol ¼ 72;
self-esteem ¼ 70;
and
meaningful
exis-
tence ¼ 66. The coe?cients suggested a reasonable level
Discussion
of internal consistency for each need; thus the average
for the items assessing each need were used in the
In Study 1, we found that compared to included
analysis.
participants, ostracized participants reported lower lev-
The main ?ndings of this study, that ostracized par-
els on the state measures of the four needs. Moreover,
ticipants reported lower levels of the needs independent
ostracized participants also reported feeling angrier and
of the source, can be seen in Fig. 1 (which depicts a
enjoyed the game less than included participants. Con-
composite score for the four conditions tested in this
trary to our expectations, mood reports were not af-
study). Statistical analysis of each need (see Table 1 for
fected by ostracism. It should be noted, however, that
all descriptive statistics) revealed that ostracized partic-
the literature to date is surprisingly inconsistent on the
ipants reported lower levels on each of the four needs
e?ects of social exclusion on mood. That is, while some
measured, smallest F ð1; 57Þ ¼ 11:1, p ¼ :002, for self-
have reported worsened mood following ostracism
esteem. There were no signi?cant main e?ects for source
(Williams, 2001) or rejection (Leary, Koch, & Hechen-
identity on self-reported needs, nor did the inclusionary
bleiker, 2001), others have failed to see an e?ect on
status interact with the source manipulation, all
mood following social exclusion (Twenge, Baumeister,
F s < 1:4, ns.
Tice, & Stucke, 2001; Twenge, Catanese, & Baumeister,
Mood. The mood items were reverse scored where
2002).
necessary (so that a higher score ¼ more positive mood)
Although the ostracism manipulation adversely af-
and an average mood score was calculated. There were
fected participants in accordance with previous ostra-
no signi?cant main e?ects or interactions for mood, all
cism research, whether participants were ostracized by
F s < 1:0, ns.
humans or the computer had no e?ect on their self-re-
Ancillary variables. Ostracized participants reported
ported need levels, or enjoyment of the game. Oddly
feeling angrier than included participants (see Table 1),
enough, the only impact that the human/computer ma-
nipulation had was the unexpected interaction showing
that being ostracized by a computer made participants
angrier than being ostracized by humans. Based on
comments made during the interviews it would seem
that this result was due to a violation of the basic as-
sumption that the computer is a tool to serve humans
and hence should not deliberately act to distress or
alienate them. For example, one participant (who hap-
pened to be a computer programmer) stated that he felt
incredibly angry and frustrated during the game because
‘‘the computer is supposed to serve me. ItÕs not supposed
to reject me.’’
Why would individuals report lower levels of be-
longing, control, self-esteem, and meaningful existence
when they have experienced ostracism in a 6-min game
of virtual ball toss, regardless of whether they were
playing with humans or a computer? We believe that
initial reactions to ostracism are, in BrewerÕs (2003)
terms, ‘‘deep,’’ rather than ‘‘high.’’ That is, it is our
Fig. 1. Mean self-reported levels of combined needs of belonging,
position that ostracism has such adaptive signi?cance
control, self-esteem, and meaningful existence, as a function of inclu-
sion or ostracism, and by either human or computer in Study 1.
for humans that there is essentially an early warning

---

ARTICLE IN PRESS
L. Zadro et al. / Journal of Experimental Social Psychology xxx (2004) xxx–xxx
5
system that is quick to perceive social exclusion and feel
players or two computer-generated players. In addition,
the associated pain, so that the individual is motivated
half of the participants were informed that the players
to make changes that will eliminate the ostracism.
(whether human or computers) were playing Cyberball
One potential problem with Study 1 is that partici-
according to a script given to them by the experimenter.
pants were not asked to indicate on the post-experi-
This script instructed the players to whom they were to
mental questionnaire whether they believed they were
throw the ball every time it was their turn to play.
playing the game with humans or computers. Although
Study 2 also included self-report manipulation
we assessed their perceptions in a post-experimental
checks to supplement the verbal manipulation checks
interview, it is possible that experimenter bias may have
used in Study 1, and some additional questions were
in?uenced our perceptions of their answers. Thus, al-
asked
to
assess
other aspects of
the Cyberball
though we think this is rather implausible, it is possible
experience.
that participants simply did not attend to our instruc-
tion about who they were playing the ball game with,
Method
and that is why we found no main e?ects or interactions
with the human/computer manipulation on the self-re-
Participants and design
ported needs. That we found one interaction on the
Seventy-seven undergraduates (30 males, 41 female,
measure of anger (in an unexpected direction, we might
M age ¼ 19:6 years, SD ¼ 1:9) enrolled in introductory
add), argues against this criticism that participants were
psychology at the University of New South Wales were
not attentive to the human/computer manipulation;
randomly assigned to a 2 (inclusionary status: inclusion
however, it is possible that this interaction was spurious.
vs. ostracism) Â 2 (source identity: computer generated
Therefore, in Study 2, we added explicit manipulation
vs. university students) Â 2 (attribution of choice:
checks to the post-experimental questionnaire assessing
scripted vs. unscripted) between-S design.3 Participants
the participantsÕ understanding of with whom they were
volunteered to take part in the experiment in exchange
playing. In Study 2, we also extended the present ex-
for course credit.
perimental design by adding a third factor: perceived
choice of the source. It may well be that to feel threats to
Procedure
the four needs an individual must attribute some sort of
The experiment was conducted on eight versions (one
sinister intent on the part of the sources (Kramer, 1994).
per condition) of an Internet website, which were iden-
What if we told them that the players with whom they
tical to those used in Study 1 except for modi?cations to
were playing were following a script that had been
the cover pages to accommodate the attribution of
provided to them, and that they had no choice but to
choice
manipulation.
The
ostracism/inclusion
and
throw the ball to the individual that the script indicated?
source identity manipulations were identical to those
If the needs that are threatened by ostracism require
used in Study 1.
some higher level interpretation of the ostracism event,
Attribution of choice manipulation. In the scripted
then we would expect this manipulation to eliminate the
conditions, the cover page instructed participants that
negative impact when participants viewed the ostracism
the other players (whether computer generated or hu-
behavior of the sources as outside their control. If,
man) would be playing the game according to a script,
however, the fundamental needs are a?ected without
and hence their actions were not spontaneous. In the
cognitive intervention, then manipulations aimed at re-
unscripted condition, participants were instructed that
ducing sinister attributions may have no impact at all.
the game was spontaneous and the players were free to
throw the ball to whomever they chose (in the case of the
computer-generated players, this spontaneous action
Study 2
was explained by saying that the players would be
throwing the ball randomly). In all conditions, partici-
In Study 2, we examined whether providing an ex-
pants were reminded that they were free to throw the
plicit and external reason for ostracism reduced its
ball to whomever they chose.
negative impact. If ostracized individuals know that the
Dependent measures. The questionnaire was essen-
reason they are not being thrown the ball has nothing to
tially the same as that used in Study 1, with the addition
do with them personally, but rather the other partici-
of manipulation check for source identity (‘‘Did you
pants (be they human or computer) are simply following
play the Cyberball game with: two students from Mac-
a script, will they still report reduced levels of the four
quarie and Sydney University or two computer-gener-
needs? If so, we believe this suggests that it is the per-
ated players?’’) and attribution of choice (‘‘Was the
ception of oneÕs own ostracism, not oneÕs understanding
sequence of throws by Player 1 and Player 2 scripted/
of it, that is immediately threatening.
Thus, participants in Study 2 were either included or
3 Initially, 120 participants were recruited, but a computer virus
ostracized from the Cyberball game by two human
prevented 43 participants from any form of participation.

---

ARTICLE IN PRESS
6
L. Zadro et al. / Journal of Experimental Social Psychology xxx (2004) xxx–xxx
pre-programmed or spontaneous?’’). One additional
Results
question was added to examine whether the manipula-
tions led participants to feel emotionally hurt (‘‘My
Manipulation checks. As shown in Table 2 (where
feelings were hurt during the game’’).
means and standard deviations can be found for all
After the participants completed the game, they were
dependent variables), our manipulations were perceived
directed to complete the post-study questionnaire.
as intended. Participants in the ostracism condition re-
Debrie?ng. As in the previous experiment, the ex-
ported that they felt less included than participants in
perimenter asked participants to state whether they had
the inclusion condition, and more rejected, (smallest F
played the game with two human players or two com-
was for rejection, F ð1; 69Þ ¼ 25:6, p < :0005). They also
puters, and whether or not the game had been scripted
reported receiving the ball less often during the game
or unscripted. The experimenter then fully debriefed
than included participants, F ð1; 66Þ ¼ 117:8, p < :0001
participants, ensuring that they were aware that they
The source identity manipulation was also successful in
were randomly assigned to conditions. Participants in
that only 2% of participants incorrectly identi?ed the
the ostracism condition were carefully debriefed about
identity of the players, and 4% of participants incor-
all aspects of the game, and were given extra informa-
rectly identi?ed the attribution of choice manipulation.
tion about the nature of ostracism. After answering any
Because there was no computer malfunction that could
remaining questions, participants were then thanked
easily account for these incorrect reports, all partici-
and dismissed.
pants were retained in the analyses.
Table 2
Means and standard deviations (in parenthesis) of variables in Study 2 (all scales 1 ¼ not at all to 9 ¼ very much so, unless otherwise stated)
Source
Human
Computer
Inclusion
Ostracism
Inclusion
Ostracism
Scripted
Unscripted Scripted
Unscripted Scripted
Unscripted Scripted
Unscripted
ðn ¼ 12Þ
ðn ¼ 8Þ
ðn ¼ 7Þ
ðn ¼ 11Þ
ðn ¼ 9Þ
ðn ¼ 10Þ
ðn ¼ 9Þ
ðn ¼ 11Þ
Fundamental needsa
Belonging
5.8 (1.5)
6.3 (2.0)
3.6 (1.9)
2.8 (1.2)
5.8 (1.6)
6.4 (1.4)
3.0 (1.4)
2.7 (1.3)
F ð1; 69Þ ¼ 74:8, p < :0005b
Control
5.8 (1.6)
6.8 (2.2)
2.7 (.90)
3.2 (1.5)
4.7 (.98)
5.5 (2.3)
2.9 (.96)
3.2 (1.6)
F ð1; 69Þ ¼ 55:0, p < :0005
Self-esteem
6.9 (1.0)
7.6 (1.3)
6.1 (1.7)
5.1 (1.9)
6.3 (2.1)
7.7 (1.6)
4.5 (1.7)
5.4 (1.4)
F ð1; 69Þ ¼ 25:2, p < :0005
Meaningful existence
6.1 (1.6)
7.6 (1.1)
2.8 (1.4)
3.6 (2.1)
5.7 (1.4)
6.2 (1.3)
3.7 (1.8)
3.7 (1.3)
F ð1; 69Þ ¼ 68:0, p < :0005
Moodc
6.8 (1.4)
6.7 (1.4)
6.1 (1.3)
6.0 (1.3)
7.0 (.80)
6.6 (1.8)
5.4 (1.7)
6.4 (.88)
F ð1; 69Þ ¼ 6:2, p ¼ :015
Ancillary variables
I felt angry during the Cyberball game
1.8 (1.8)
2.0 (2.1)
2.1 (1.4)
2.8 (1.8)
2.2 (1.5)
1.0 (.00)
4.0 (2.2)
4.0 (2.2)
F ð1; 69Þ ¼ 13:2, p ¼ :001
I enjoyed playing the Cyberball game
4.6 (2.3)
6.6 (1.8)
3.3 (1.5)
3.0 (1.7)
5.1 (1.8)
5.2 (2.2)
3.3 (2.2)
3.5 (1.8)
F ð1; 69Þ ¼ 22:2, p < :0005
My feelings were hurt during the
2.2 (2.4)
1.1 (.35)
1.1 (.38)
3.2 (2.2)
2.1 (1.8)
1.2 (.42)
4.3 (2.7)
3.0 (2.2)
Cyberball game
F ð1; 69Þ ¼ 8:3, p ¼ :005
Manipulation checks
To what extent were you included
6.2 (1.5)
6.9 (2.0)
2.4 (.53)
2.6 (1.5)
5.3 (1.7)
6.0 (1.9)
2.2 (.44)
2.6 (.69)
by the participants during the game?
F ð1; 69Þ ¼ 122:7, p < :0005
What percentage of throws did you
35.9 (14.2)
45.3 (15.0)
11.7 (8.5)
13.9 (8.2)
43.9 (7.7)
45.8 (13.9)
15.7 (9.1)
13.1 (4.6)
receive during the Cyberball game?
F ð1; 69Þ ¼ 117:8, p < :0005
Rejected–acceptedd
6.3 (1.8)
6.9 (2.2)
4.4 (2.1)
4.2 (2.0)
6.3 (1.9)
5.7 (2.6)
3.4 (2.1)
3.5 (1.9)
F ð1; 69Þ ¼ 25:6, p < :0005
a Each fundamental need score represents an average of three questions.
b All F values refer to signi?cant ostracism vs. inclusion main e?ects.
c Total mood score was an average of four 9-point bipolar questions.
d This was a 9-point scale with rejected–accepted as anchors.

---

ARTICLE IN PRESS
L. Zadro et al. / Journal of Experimental Social Psychology xxx (2004) xxx–xxx
7
Self-reported level of needs. As in Study 1, the items
if playing a computer or humans, but as in Study 1,
assessing each need were reverse scored where necessary
ostracized participants were angrier at being ostracized
and the internal consistency of the items were assessed.
by a computer than humans, F ð1; 34Þ ¼ 5:4, p ¼ :026.
CronbachÕs alpha coe?cients for each need were as
No other two-way interactions were signi?cant.
follows: belonging ¼ .71; control ¼ .80; self-esteem ¼ .76;
There was a signi?cant three-way interaction for hurt
and meaningful existence ¼ .69. The coe?cients sug-
feelings, F ð1; 69Þ ¼ 4:0, p ¼ :049, which suggested that
gested a reasonable level of internal consistency for each
when interacting with humans, participants reported
need; thus the average for the items assessing each need
higher levels of hurt feelings only when they were os-
were used in the analysis.
tracized by two players who had free choice as to whom
As in the previous study, there were signi?cant main
they could throw the ball. Participants who played the
e?ects for inclusionary status on the four needs such that
computer, however, reported more hurt feelings simply
participants who were ostracized reported lower levels
when they were ostracized, regardless of whether or not
of belonging, control, self-esteem, and meaningful exis-
the computer game had been scripted.
tence than participants who were included in the game,
(smallest
F
was
for
self-esteem,
F ð1; 69Þ ¼ 25:2,
p < :0005).
Discussion
There were no signi?cant main e?ects for source
identity on the primary needs (largest F was for control,
The results of Study 2 largely replicated the ?ndings
F ð1; 69Þ ¼ 2:2, p ¼ :147).
of Study 1. Once again, we found that ostracism resulted
Attribution of choice did not a?ect self-reported need
in lower self-reported levels of four needs. However, in
levels for belonging, control, or self-esteem, but there
Study 2 we also found that ostracism resulted in less
was a marginally signi?cant main e?ect for meaningful
positive mood than did inclusion. Thus, the inconsistent
existence, such that participants who believed the game
nature of mood e?ects following social exclusion that
was unscripted (free choice) reported higher levels of
exists in the broader literature was mirrored in our two
meaningful existence than participants who believed the
studies. We regard this inconsistency as evidence of the
players were scripted, F ð1; 69Þ ¼ 3:9, p ¼ :052.
less than robust e?ect of social exclusion on mood. In
There was also a marginally signi?cant interaction
any case, as Twenge et al. (2001, 2002) and Williams et
between inclusionary status and source identity for
al. (2000) have also demonstrated, regardless of whether
meaningful existence, F ð1; 69Þ ¼ 3:95, p ¼ :051. Follow
mood is a?ected by social exclusion, it does not mediate
up analyses revealed that the e?ects of inclusion and os-
need levels nor subsequent behaviors.
tracism on self-reported meaningful existence produced
More importantly for the present investigation, os-
more extreme di?erences when the sources were human
tracism by computers was just as unpleasant as ostra-
than when the sources were computers. No other two-way
cism by humans, and furthermore, it did not matter
interactions for the remaining needs were signi?cant, nor
whether the human or computer players were perceived
were there any signi?cant three-way interactions.
to have a choice as to whom they threw the ball. Thus,
Mood. The mood items were reverse scored where
once again, it appears that ostracism, per se, is felt im-
necessary (so that a higher score equalled more positive
mediately as a negative and depleting experience. Par-
mood) and an average mood score was calculated.
ticipantsÕ initial reactions to a short exposure to
Participants in the ostracism condition reported feeling
ostracism were not a?ected by two factors that would
more negative during the game than participants in the
generally be regarded as rendering the ostracism expe-
inclusion condition, F ð1; 69Þ ¼ 6:2, p ¼ :015. There were
rience meaningless: being ignored and excluded by a
no signi?cant main e?ects for source identity or attri-
computer, and knowing that the players were told (or
bution of choice, nor were there any signi?cant two or
programmed) not to throw the ball to them. Instead,
three way interactions, all Fs < 1:5, ns.
within minutes, feelings of belonging, control, self-es-
Ancillary variables. There were several e?ects of in-
teem, and meaningful existence are reduced, simply be-
clusionary status on the ancillary variables. Speci?cally,
cause the participants were not thrown a ball while
participants who were ostracized reported feeling an-
playing a relatively meaningless game that had no win-
grier, more hurt, and that they enjoyed the game less
ners or losers, with people who they do not know and
than participants who were included in the game,
will not meet.4
(smallest F was for anger, F ð1; 69Þ ¼ 8:3, p ¼ :005).
There were no signi?cant main e?ects for source
identity or attribution of choice on the ancillary variables.
4 As one reviewer suggested, it is possible that participants viewed
There was a signi?cant two-way interaction between
the scripted conditions as an attempt to ostracize them by the
experiment. Although we do not have any measures to explicitly
inclusionary status and source identity for anger,
address this possibility, this suggestion would lead to an expectation of
F ð1; 69Þ ¼ 5:0, p ¼ :028. Follow up analyses revealed
stronger e?ects under the scripted conditions (for human or computer
that participants who were included were no more angry
sources), which we did not ?nd.

---

ARTICLE IN PRESS
8
L. Zadro et al. / Journal of Experimental Social Psychology xxx (2004) xxx–xxx
General discussion
social signal that it produces widespread intrapsychic
reactions that serve to mobilize coping responses (see
The ?ndings of both studies lead us to conclude that
also, Panksepp, 2003).
ostracism is such an important warning signal that in-
dividuals are pre-cognitively attuned to its employment
on them. For primates, and many other species (see
Williams, 2001), ostracism means death. For humans, it
References
surely signals the potential for hard times, possibly loss
of contact with important others, loss of resources, and
Brewer, M. (2003). Implicit and explicit processes in social judgments
in some cases, death. Hence, it appears that even the
and decisions: An integration. In J. P. Forgas, K. D. Williams, &
slightest hint of ostracism, in the present case by a
W. von Hippel (Eds.), Responding to the social world: Implicit and
computer, is enough to trip o? emotional reactions that
explicit processes in social judgments and decisions. New York:
will activate coping strategies to increase oneÕs sub-
Cambridge University Press.
Eisenberger, N. A., Lieberman, M. D., & Williams, K. D. (2003). Does
sequent inclusion.
rejection hurt? An fMRI study of social exclusion. Science, 302,
Our ?nding that individuals can react to computers in
290–292.
a similar fashion to how they react to humans is also
Kramer, R. M. (1994). The sinister attribution error: Paranoid
consistent with NassÕs work on human–computer in-
cognition and collective distrust in groups and organizations.
teractions (e.g., Reeves & Nass, 1996). In a variety of
Motivation and Emotion, 18, 199–229.
Leary, M. R., Koch, E. J., & Hechenbleiker, N. R. (2001). Emotional
clever studies, Nass and his colleagues have shown that
responses to interpersonal rejection. In M. R. Leary (Ed.),
humans will reciprocate self-disclosures, be more stra-
Interpersonal rejection (pp. 145–166). New York: Oxford Univer-
tegically polite, and form in-group alliances with com-
sity Press.
puters, just as they have been shown to do with other
Leary, M., Kowalski, R. M., Smith, L., & Phillips, S. (2003). Teasing,
people. Whether this means that we anthropomorphize
rejection, and violence: Case studies of the school shootings.
Aggressive Behavior, 29, 202–214.
computers, enact mindless scripts to guide our social
Panksepp, J. (2003). Feeling the pain of social loss. Science, 302, 237–
behaviors, or that we have such a deep-rooted need to
239.
belong that it can be ful?lled by computers (as exem-
Reeves, B., & Nass, C. (1996). The media equation: How people treat
pli?ed in Steven SpielbergÕs AI where the mother grew to
computers, television, and new media like real people and places.
love a Cyborg son), is the subject matter for future in-
New York: Cambridge University Press.
Twenge, J. M., Baumeister, R. F., Tice, D. M., & Stucke, T. S. (2001).
vestigations.
If you canÕt join them, beat them: E?ects of social exclusion on
We remain motivated to ?nd conditions that involve
aggressive behavior. Journal of Personality and Social Psychology,
ignoring and exclusion but that are so minimal as to not
81, 1058–1069.
in?ict emotional damage. Based on the present ?ndings,
Twenge, J. M., Catanese, K. R., & Baumeister, R. F. (2002). Social
it appears that this search for the necessary and su?-
exclusion causes self-defeating behavior. Journal of Personality and
Social Psychology, 83, 606–615.
cient conditions for ostracism, at least when measured
Williams, K. D. (1997). Social ostracism. In R. M. Kowalski (Ed.),
during or soon after the ostracism, may be di?cult or
Aversive interpersonal behaviors (pp. 133–170). New York: Plenum.
impossible. It appears that ostracism is such a powerful
Williams, K. D. (2001). Ostracism: The power of silence. New York:
signal that even being ignored by a computer can acti-
Guilford Press.
vate strong reactions. Indeed, Eisenberger, Lieberman,
Williams, K. D., Cheung, C. K. T., & Choi, W. (2000). CyberOstra-
cism: E?ects of being ignored over the Internet. Journal of
and Williams (2003) showed activation of the anterior
Personality and Social Psychology, 79, 748–762.
cingulate cortex—the brain region that is also activated
Williams, K. D., Govan, C. L., Croker, V., Tynan, D., Cruickshank,
when individuals endure physical pain—as a result of
M., & Lam, A. (2002). Investigations into di?erences between
Cyberball ostracism, even when the participants were
social and cyber ostracism. Group Dynamics: Theory, Research, &
told their particular computer was not yet linked to the
Practice, 6, 65–77.
Williams, K. D., & Zadro, L. (2001). Ostracism: On being ignored,
other two participantsÕ computers. We are also exam-
excluded and rejected. In M. R. Leary (Ed.), Interpersonal rejection
ining the impact of these simple acts of ostracism on
(pp. 21–53). New York: Oxford University Press.
participantsÕ cardiovascular responses (Zadro, Barker,
Zadro, L., Barker, G. L., Richardson, R., & Williams, K. D. (2001).
Richardson, & Williams, 2001; Zadro, Walker, Wil-
Physiological impact of ostracism: Challenge and threat. Presented
liams, & Richardson, 2000). The amassed evidence
at the 73rd Annual Meeting of the Midwestern Psychological
Association. Chicago, Illinois.
suggests that at all levels of measurement, humans de-
Zadro, L., Walker, P., Williams, K., & Richardson, R. (2000). The
tect and su?er from the most minimal cues of ostracism,
psychophysiological e?ects of being ostracized. Presented at the 10th
supporting our view that ostracism is such a powerful
World Congress of Psychophysiology, Sydney.

---

Document Outline

• How low can you go? Ostracism by a computer is sufficient to lower self-reported levels of belonging, control, self-esteem, and meaningful existence
  • Study 1
    • Method
      • Participants and design
      • Procedure
    • Results
    • Discussion
  • Study 2
    • Method
      • Participants and design
      • Procedure
    • Results
  • Discussion
  • General discussion
  • References

---