Personality and Social Psychology Bulletin

http://psp.sagepub.com

Social Influence in Computer-Mediated Communication: The Effects of Anonymity on Group Behavior

Tom Postmes, Russell Spears, Khaled Sakhel and Daphne de Groot
Pers Soc Psychol Bull 2001; 27; 1243
DOI: 10.1177/01461672012710001

The online version of this article can be found at:

http://psp.sagepub.com/cgi/content/abstract/27/10/1243

Published by:

http://www.sagepublications.com

On behalf of:

Society for Personality and Social Psychology, Inc.

Additional services and information for Personality and Social Psychology Bulletin can be found at:

Email Alerts: http://psp.sagepub.com/cgi/alerts

Subscriptions: http://psp.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Citations http://psp.sagepub.com/cgi/content/refs/27/10/1243

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
Postmes et al. / COMPUTER-MEDIATED COMMUNICATION

Social Influence in Computer-Mediated

Communication: The Effects of
Anonymity on Group Behavior

Tom Postmes
Russell Spears
Khaled Sakhel
Daphne de Groot
University of Amsterdam

Two studies examined hypotheses derived from a Social Identity in the interaction between group members who are
model of Deindividuation Effects (SIDE) as applied to social anonymous.
influence in computer-mediated communication (CMC) in Although most group theories do not explicitly rule
groups. This model predicts that anonymity can increase social out the possibility of social influence when group mem-
influence if a common group identity is salient. In a first study, bers are unknown to each other or do not physically
group members were primed with a certain type of social behavior interact, many of these theories would expect social
(efficiency vs. prosocial norms). Consistent with the model, influence to increase as a function of the intensity of
anonymous groups displayed prime-consistent behavior in their social contact. For example, the theory of “normative
task solutions, whereas identifiable groups did not. This sug- influence” (Deutsch & Gerard, 1955) argues that group
gests that the primed norm took root in anonymous groups to a influence depends on social pressure from others and
greater extent than in identifiable groups. A second study repli- that this pressure can best be exercised when group
cated this effect and showed that nonprimed group members con- members are identifiable, under surveillance, and thus
formed to the behavior of primed members, but only when anony- accountable to the group for their responses. Social
mous, suggesting that the primed norm was socially transmitted impact theory (Latané, 1981) extends the theory of nor-
within the group. Implications for social influence in small mative influence and suggests that social influence will
groups are discussed. increase with the “immediacy” of its members (their
proximity in space or time). Short, Williams, and Chris-
tie (1976) refer to a closely related concept of “social
This article is concerned with processes of social influ- presence.” Identifiability, as opposed to anonymity,
would be expected to enhance immediacy and social
ence in groups communicating by means of computers.
A common feature of communication via e-mail and the presence and thus facilitate social influence. Similarly,
Internet is the relative anonymity of contact with others, theories of social influence based on interpersonal
especially in initial interactions. In two studies, we inves- attraction (Lott & Lott, 1965) or interdependence
tigate the effect of visual anonymity on social influence
in computer-mediated communication (CMC). In the Authors’ Note: The research of Tom Postmes is supported by a fellow-
process, we address basic issues of general concern to ship of the Royal Netherlands Academy of Arts and Sciences. The au-
social psychology and examine the effects of this increas- thors wish to thank Jolanda Jetten, Richard Moreland, Diederik Stapel,
ingly popular communication medium. Deriving predic- Garold Stasser, and two anonymous reviewers for their comments on
earlier drafts and Chick Judd for his methodological advice. Corre-
tions from the Social Identity model of Deindividuation
spondence concerning this article should be addressed to the first au-
Effects (SIDE) (Reicher, Spears, & Postmes, 1995), we thor, University of Exeter, School of Psychology, EX4 4QG, Exeter,
try to show in the first study that anonymity can enhance United Kingdom; e-mail: t.postmes@exeter.ac.uk.
the influence of a primed norm. The second study in- PSPB, Vol. 27 No. 10, October 2001 1243-1254
vestigates evidence for the transmission of this norm © 2001 by the Society for Personality and Social Psychology, Inc.

1243

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

1244 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

(Lewin, 1948/1997) of group members suggest that face- ence. We now address a framework that attempts further
to-face interaction should strengthen the interpersonal specification of the effects of anonymity within the
bonds that transmit social influence, whereas isolation group.
and anonymity should weaken them. In summary, in all The SIDE model. The SIDE model (Reicher, Spears, &
of these cases, the anonymity of group members might Postmes, 1995) builds on self-categorization theory and
be expected, either explicitly or implicitly, to weaken tries to extend it to provide a more detailed analysis of
social influence by and within the group. the effects of situational factors such as anonymity on
However, consideration of other theories of the social influence processes. The SIDE model acknowl-
group in social psychology suggests that anonymity edges that under certain conditions identifiability can
might not necessarily undermine social influence. increase accountability and influence behavior. How-
Allport (1985) proposed that social psychology was not ever, this generally concerns compliance to the norms of
just concerned with the effect of the “real presence” of powerful outgroups to avoid behavior that would other-
others on behavior but also with the effects of their wise be sanctioned, rather than referring to true influ-
“imagined or implied presence.” Classic research on the ence in the sense of private acceptance (Reicher et al.,
influence of “reference groups” also showed that we can 1995; Turner, 1991). More generally, in terms of the
be influenced by social groups or categories that might power of the ingroup to exert true influence, the SIDE
not necessarily be present in situ and to which we might model proposes that when a social identity is already
not even belong (French & Raven, 1959; Hyman, 1942). salient (i.e., when people define themselves as members
More recently, self-categorization theory (Turner, of a group rather than as individuals), visual anonymity
1987, 1991) has proposed a group-based theory of social can actually enhance group salience and its related effects
influence (originally called “referent informational (group identification). Following self-categorization
influence”) that does not relate influence to the theory, this should then lead to enhanced social influ-
identifiability or surveillance of group members. Rather, ence in line with group norms (Turner, 1987, 1991).
social influence is an internal and willing process stem- It is argued that visual anonymity obscures personal
ming from a social definition of the self and social valida- features and interpersonal differences and thereby
tion of one’s views in relation to the group (Turner, 1987, diminishes the relative importance of interpersonal con-
1991). In these terms, social influence is in the first cerns in favor of a focus on the known or emergent char-
instance cognitively mediated by one’s self-categorization acteristics of the group as a whole. Provided a common
as a group member, rather than by processes involving identity is available, anonymity thus increases the
social contact per se. However, this does not mean that salience of group identity and group identification,
social influence is just a “cognitive” matter; the strength thereby enhancing the group’s influence. In sum,
of group self-categorization is closely bound up with the according to the SIDE model, anonymity should accen-
affective and emotional significance attached to this social tuate the effects of the salient social identity and the
self-definition, as measured by constructs such as group dominant normative response associated with it
identification. Self-categorization theory indicates that (Reicher et al., 1995). One of the main aims of the pres-
social influence should generally increase as a function ent research is to investigate whether anonymity can
of group identification or of variables that increase it. actually enhance group influence in CMC, as predicted
Although the indicators of group identification can be by the SIDE model, or whether identifiability enhances
individually measured, it is not conceptualized as a fixed influence as suggested by the classical theories of group
individual difference variable: It should increase with influence referred to earlier.
factors that facilitate a shift in self-definition from the It is important to distinguish the SIDE model from
personal to the group level of self-categorization deindividuation theory (e.g., Diener, 1980; Zimbardo,
(Turner, 1987). 1969). “Classical” deindividuation theory (e.g.,
Although self-categorization theory does not sug- Zimbardo, 1969) is most relevant here because, like the
gest that anonymity in the group should weaken social SIDE model, it is concerned with the effects of anonym-
influence, it does not explicitly state that it should ity within the group. It proposes that immersion and ano-
strengthen it, either. In fact, it has been argued by some nymity in the group can lead to reduced self-awareness
self-categorization researchers that the very copresence (a state of deindividuation), which results in antinor-
of others can render the group unit salient and thus mative behavior. With regard to social influence, de-
increase group identification and social influence individuation caused by anonymity would therefore
(Abrams, Wetherell, Cochrane, Hogg, & Turner, 1990). imply reduced self-awareness and reduced influence of
As elaborated further below, this article examines the social norms and standards. The contemporary version
opposite; namely, that anonymity can enhance identifi- of deindividuation theory (Prentice-Dunn & Rogers,
cation with the group and thereby enhance social influ- 1989) has moved away from anonymity as the basis for a

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

 Postmes et al. / COMPUTER-MEDIATED COMMUNICATION 1245

deindividuated state and considers deindividuation to enhance normative behavior has been investigated in
be the product of reduced private self-awareness intragroup (e.g., Spears, Lea, & Lee, 1990) and inter-
brought about by arousal coupled with immersion in the group settings (e.g., Postmes, Spears, & Lea, 1999), as
group. However, in common with the classical theory well as indirectly in a meta-analysis (Postmes & Spears,
based on anonymity, this theory also views the 1998). However, none of these studies have investigated
deindividuated state as one of reduced self-regulation the actual social influence process. Moreover, none of
resulting in unresponsiveness to social norms and stan- these studies provides a direct experimental manipula-
dards, leading to antinormative behavior (see Postmes & tion of the central theoretical variable of group norm.
Spears, 1998). With these shortcomings in mind, we set out providing a
Unlike the SIDE model, deindividuation theories more direct test of the SIDE model.
would therefore seem to predict that anonymity either For many groups, the norms may be obvious and
fosters unresponsiveness to social identities and their explicit to group members. This is less often the case for
associated norms, or behavior that is generally antinorma- ephemeral groups or collectives such as crowds that
tive, or both. The SIDE model was developed as an meet on an ad hoc basis (Reicher, 1987). The lack of
attempt to provide an alternative (normative) account of clearly specified norms also might characterize many
the diverse “deindividuation” effects found in the litera- computerized groups, because the formation of these
ture. A recent meta-analysis of the deindividuation litera- groups is not always embedded in a predefined social
ture indicates that there is far stronger support for this structure. When norms are unknown, groups have to
SIDE analysis, which proposes that anonymity in the rely on the inductive construction of social norms
group enhances conformity to group norms, than for (Postmes, Spears, & Cihangir, 2001; Postmes, Spears, &
either classical or contemporary deindividuation the- Lea, 2000). By this inductive process, we mean that norms
ory’s predictions (Postmes & Spears, 1998). The present are inferred from the common behavior or predominant
studies are not intend ed as d ir ect tes ts o f attributes of typical members (Reicher, 1987; Turner,
deindividuation theories versus the SIDE model; how- 1982), somewhat similar to Sherif’s (1935) classic study of
ever, they may help to shed some light on the norm formation using the autokinetic effect. In practical
deindividuation effects investigated in this literature. terms, this feature of computerized groups provides a
CMC and deindividuation. If we now turn to the litera- context in which we can attempt to manipulate norms
ture on social influence in CMC we find that theory and without having to rely on preexisting group attitudes.
research has tended to endorse the interdependence/ Manipulations of group norms are not easy to achieve
surveillance conceptualization of the group. The growth in small groups. Mostly, social norms are manipulated by
of CMC and other forms of mediated communication providing group members with (false) feedback con-
has raised researchers’ concerns with the relative ano- cerning the predominant mode of behavior in their
nymity of CMC and more recently of the Internet as a group (e.g., Spears et al., 1990). A problem with this kind
whole because it would decrease the “social presence” of feedback manipulation is that demand characteristics
(Short et al., 1976) or reduce “social cues” (Kiesler, may be too apparent and may even evoke reactance or
Siegel, & McGuire, 1984; Rutter, 1984) conveyed in contrast effects (Postmes et al., in press). A more desir-
interaction. Consistent with deindividuation theory, this able manipulation of group norms would therefore be
reduction of social information is seen as producing neg- less obtrusive. If the manipulated behavior were already
ative outcomes: It reduces self-awareness, decreases evident in the in vivo group, members could infer a
attraction to the group, stimulates antinormative behav- group norm. One way of achieving this is to use confed-
ior, and decreases social influence (cf. McLeod, Baron, erates. However, using confederates poses a threat to the
Marti, & Yoon, 1997). CMC also tends to be viewed as an ecological validity of group processes. A more elegant
impersonal medium that, in the terminology of Bales approach is to activate behavioral responses in partici-
(1950), facilitates task-oriented exchanges as opposed to pants by means of preconscious priming manipulations.
socioemotional interaction (Daft & Lengel, 1984). The effects of priming on judgment tasks (Srull &
Against this general trend, theorists have more Wyer, 1979) and on behavioral measures (Bargh, Chen, &
recently proposed that CMC groups are ultimately no Burrows, 1996) are well-known. For example, Bargh et al.
less prone to the operation of group-based normative (1996) primed participants with the trait “politeness”
principles than are face-to-face groups (e.g., McKenna & and observed that participants were subsequently much
Bargh, 1998; Spears & Lea, 1992). As we have seen, the more reluctant to interrupt another person than those
SIDE model proposes that certain features of CMC, such primed with “rudeness.” Although it is not entirely clear
as relative anonymity, can actually intensify social influ- what process is responsible for the effect of primes on
ence (see Postmes, Spears, & Lea, 1998, for a review). behavior, it seems that priming behavioral dimensions
With regard to CMC, the prediction that anonymity can can trigger similar behavior. Although effects of such

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

1246 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

primes have not yet been demonstrated in the context of one session. In the identifiable condition, a digitized pic-
small group behavior, if behavior is triggered by a prime ture was taken of participants. When all members of the
in members of the group, a behavioral group norm two groups that were run simultaneously were present,
could be induced by those same members. In the fol- instructions were given to commence the scrambled-
lowing study, we prime contrasting norms to assess sentence task. After 10 minutes, the computer signaled
whether group-based CMC is sensitive to variation in the with a beep that participants should stop the scrambled-
content of the norms induced and whether this is accen- sentence test immediately and proceed to the next task.
tuated under conditions of anonymity rather than This was a computer task designed to enhance group
identifiability. cohesiveness and was necessary to ensure that the group
developed a common group identity and that this iden-
STUDY 1 tity was salient throughout the study (see Spears et al.,
1990). Participants were allocated to groups identified as
The purpose of Study 1 is to demonstrate that ano- Group A or Group B. It was clear to participants that the
nymity can increase the influence of group norms in dis- two groups were present in the laboratory simulta-
cussion via CMC. This is achieved by manipulating ano- neously, although there was no direct interaction
nymity and the group norm. Anonymity is manipulated between groups. An intergroup context enhances the
by either individually identifying participants to each salience of group identity (Spears et al., 1990; Turner,
other during interaction by means of pictures of each 1987). In the task, participants were required to estimate
group member displayed on the computer screens. The the number of black squares on the screen amid a num-
manipulation of group norms is achieved indirectly by ber of white and gray squares. Participants had to maxi-
priming participants in a prior, seemingly unrelated, mize the score of their group, and they got feedback
task with instances of two contrasting social norms about group choices but not about individual scores.
(stressing either efficiency or prosocial behavior) by The feedback was manipulated such that groups con-
means of a scrambled-sentences test (Srull & Wyer, 1979; verged in the final judgments. Previous research indi-
see below). We assume that participants will induce a cated that this task enhances self-categorization in terms
group norm from the behavior in the group (cf. Sherif, of the group identity (Doosje, Spears, & Koomen, 1995).
1935). Implicit in this test was the competition for the highest
Predictions based on the SIDE model are that the group score with the other group, further designed to
manipulated behaviors will assume normative qualities, ensure a degree of identification with the group.
particularly in anonymous groups. It is assumed that Participants then proceeded to the next task, a group
these norms will embed themselves in the group over the discussion, and were instructed to discuss solutions to a
course of the interaction. The expression of an efficiency- dilemma for 15 minutes via a computer-conferencing
oriented or prosocial norm is expected to be reflected by system. Groups were not required to reach any decision
the content of the interaction. Moreover, efficiency-ori- or consensus. Groups discussed the following scenario:
ented groups are predicted to prefer more business-like
“rational” alternatives, whereas prosocially primed Problems have arisen in a medium-size hospital.
groups prefer more socially considerate choices. Once Through an increase in the number of patients, the hos-
again, these effects are expected to be strongest for pital has problems providing adequate care to all
anonymous groups. patients in time. The management of the hospital wants
to solve the problem by increasing the efficiency in the
Method organization. . . . The doctors and nurses want to secure
patient welfare and prevent patients’ being treated as
Participants. Seventy-five undergraduates, 41 women
numbers. . . . Discuss possible solutions to this problem
and 34 men, participated for course credits. Participants with your group.
were randomly divided into 21 groups of three (9 groups)
or four (12 groups). One person did not fill in the ques-
The scenario was kept ambiguous so that there was no
tionnaires as instructed and was omitted from the analy-
one “correct” solution. A prosocial approach to the solu-
sis of the questionnaire data. Two groups were run in
tion emphasized the importance of patients’ feelings,
each session.
personal contact, quality of treatment, and a patient-
Procedure. After entering the laboratory, participants centered approach. An efficiency-oriented approach to
were placed in a cubicle isolated from the other partici- solutions stressed the importance of the hospital’s task,
pants, where they remained during the whole study. Par- efficiency, costs, speed of work, and a rational approach
ticipants therefore were generally unaware who else par- in general. These suggested approaches were later used
ticipated in the study. Participants were told that the in counting elements of the discussion, as described
study combined a number of unrelated experiments in below. After 15 minutes of discussion, participants filled

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

 Postmes et al. / COMPUTER-MEDIATED COMMUNICATION 1247

out a questionnaire and were debriefed about the pur- measured with three questions (a = .85); for example,
pose of the study. “At this moment I identify with group [A or B].” Checks
were made of participants’ private self-awareness using a
Apparatus. Groups communicated via a synchronous
two-item scale suggested by Matheson and Zanna
computer-conferencing system on Macintosh comput-
(1990); for example, “I was aware of the way my mind
ers. The system allowed users to send messages within
worked” (a = .78). The reason to include this scale was to
one group by entering text into a small window at the
check whether the anonymity manipulation would
bottom of their screen. After pressing the return button,
decrease self-awareness, as suggested by deindividuation
the message was sent to all other members of their group
theory (Diener, 1980).
within 0.4 seconds. Each message appeared on the
screens of the group in a large scrollable discussion win- An open-ended question asked participants to indi-
dow, in order of sending. Each message was identified cate what, in their own opinion, would be the best solu-
with its sender’s user name, consisting of a letter, A or B, tion to the problem. The answers to this question were
and a number, 1 through 4. The letter signified the coded by two independent raters, blind to treatment con-
group in which participants were placed and the digit ditions, on a 5-point scale from 1 (a completely efficiency-
was an individual identifier; thus, the computer software oriented solution) (e.g., efficiency increases, management
provided a tool to have group discussions that resemble, improvements) to 5 (a completely prosocial solution) (e.g.,
as it were, transcripts of a face-to-face discussion. focused on patient care). The raters achieved good reli-
ability (intraclass r = .87) and resolved the inconsisten-
Manipulations. Norms were manipulated with a cies in their judgments. The final question asked partici-
scrambled-sentence task to prime certain kinds of behav- pants to indicate whether they thought some parts of the
ior. Scrambled sentences were presented as being experiment were related to each other and, if so, why.
selected from conversations of fellow students to ensure The content of the discussions was analyzed using a
relevance of the prime to the group setting. The tasks computer program counting the occurrences of words
consisted of 37 scrambled sentences, 7 of which were in the text. The number of words and sentences were
neutrally phrased. The remaining 30 sentences con- counted. Counts were made of the number of prosocial
tained prosocial traits or verbs (e.g., warm, social, senti- relevant words (feeling, personal, treatment, patient,
mental; to help, support, and sympathize) or efficiency- and social) and the number of efficiency-oriented words
oriented traits or verbs (e.g., efficient, rational; to solve, (task, efficient, costs, quick, and rational). These words
work, and measure). These words were embedded in were drawn from the scenario text. The number of self-
sentences that would describe prosocial or efficiency-ori- references were counted as an additional self-awareness
ented behaviors or people when placed in correct order. measure (Davis & Brock, 1975).
For example, the stimulus sentence, “The efficiently
work is in office the remarkably done” is correctly Results
unscrambled by forming the sentence, “The work in the
Questionnaire data. None of the participants saw a rela-
office is done remarkably efficiently.” An example of a
tion between the scrambled-sentence test and the discus-
prosocial priming sentence would be, “Group members
sion, which confirms that the prime was unobtrusive.
support each other all the way.” Participants were given
Questionnaire data were analyzed at the group level in a
10 minutes to unscramble as many sentences as possible.
2 (anonymity) ´ 2 (prime) between-groups analysis of
On average, participants unscrambled 23 sentences.
variance. The manipulation check showed that the ano-
The reason to impose a time constraint was that it has
nymity manipulation worked well. Groups in the anony-
been suggested that unfinished tasks have a stronger
mous condition indicated that their group was more
priming effect than tasks that are completed (Martin,
anonymous to them (M = 4.04, SD = 1.00) compared with
1986). The anonymity manipulation was achieved by
identifiable groups (M = 3.19, SD = 0.69), F(1, 17) = 4.87,
presenting identifiable participants with pictures of the
p < .05. (Because analyses conducted at the group level
group members (including themselves) at the top of
decrease the power of these tests, significant effects
their screen during discussion. Each picture was identi-
reveal substantial effect sizes; in this case, h2 = .22, which
fied with the user name. In the anonymous condition, no
corresponds to an r of .47.) The main effect of prime and
pictures were shown.
the interaction were not significant.
Dependent variables. The questionnaire consisted of 7- The individual solutions were rated by coders such
point rating scales (1 = not at all, 7 = very much). An ano- that higher numbers indicate more perceived efficiency
nymity check on the effectiveness of the anonymity content in the solutions. A 2 ´ 2 ANOVA showed no reli-
manipulation consisted of a two-item scale (a = .71); for able main effects, and the predicted interaction was signif-
example, “The people we interacted with were person- icant, F(1, 17) = 11.93, p < .01. As can be seen in Figure 1,
ally identifiable to me.” Social identification was anonymous groups favored solutions that were consis-

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

1248 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

5.00
dicted direction: Anonymous groups used more effi-
Prosocial Prime ciency-oriented words (M = 5.16, SD = 2.80) compared
Efficiency-oriented Solutions -->

Efficiency Prime
with identifiable groups (M = 3.12, SD = 0.92), F(1, 17) =
4.00 3.70 3.75
2.73, p = .12. In prosocial-primed groups, anonymous
2.98
groups produced more prosocial words (M = 3.20, SD =
3.00
2.51
1.65) compared with identifiable groups (M = 1.37, SD =
1.12), F(1, 17) = 4.70, p < .05.
2.00
Discussion
1.00
This study set out to demonstrate that anonymity in a
Anonymous Identifiable
group can promote normative behavior. Results indicate
Figure 1 Solutions proposed after discussion, Study 1. that anonymous group members conform to the stan-
dards set by a prime, whereas identifiable group mem-
bers do not. This effect occurs foremost in the solutions
tent with the prime they had received. Anonymous to the problem: These were consistent with the prime
prosocial primed groups suggested more prosocial solu- only in anonymous groups. This finding is corroborated
tions (M = 2.51, SD = 0.28) and efficiency-primed groups by the analysis of the use of words during the discussion.
favored more efficiency-oriented solutions (M = 3.70, SD = The fact that the solutions were given after discussion
0.89). Simple effects indicated this was a significant dif- and in private suggests that public conformity to the per-
ference, F(1, 17) = 9.59, p < .01. Identifiable groups did ceived norm is not a likely explanation for these find-
not differ significantly. In the identifiable condition, ings.3 Public conformity is an even less likely explanation
there was a tendency for prosocial-primed groups to for these effects because conformity is generally stronger
favor more efficiency-oriented solutions (M = 3.75, SD = in groups whose members are identifiable.
0.71) compared with efficiency-primed groups (M = 2.98, The finding of increased compliance in anonymous
SD = 0.65), F(1, 17) = 3.08, p < .10. groups is at odds with the suggestions made by classical
The identification scale did not show the predicted deindividuation theory (e.g., Zimbardo, 1969). The
main effect of anonymity, F(1, 17) = 2.65, p = .12, ns, present results show no evidence that anonymity has any
although the means differed in the predicted direction impact on participants’ self-awareness, as is predicted by
(Manonymous = 3.69, SD = .58; Midentifiable = 3.27, SD = 0.68). deindividuation theory. It seems unlikely, therefore, that
No other effects were reliable. Also, no significant effects anonymity in these computer-mediated groups induced
were found on the private self-awareness scale. Most a state of psychological deindividuation. Although
important, anonymity did not reliably affect private self- deindividuation theorists have noted the mixed effects
awareness, F(1, 17) = 0.13, ns. produced by anonymity in the group (e.g., Diener,
1980), a suitable explanation for this variability was
Discussion content. Counts of units of text typed during never found. The SIDE model does offer an alternative
the discussion were converted to group averages. No sig- explanation, suggesting that anonymity may obscure
nificant differences were found across conditions for the individual inputs and thereby enhance the salience of
number of words, sentences, or characters typed by the the collective and of its norms (Postmes & Spears, 1998).
groups.1 No reliable differences were found for the num- Indeed, the result that social influence was stronger in
ber of self-references, either: Similar to the private-self anonymous groups would seem to support this view.
awareness scale, there was no effect of anonymity on the However, identification with the group was not reliably
number of self-references, F(1, 17) = 0.63, ns. enhanced by anonymity in the group. This failure could
The number of prosocial and efficiency-oriented be due to a lack of power: Analyses are generally less pow-
words used during discussion is displayed in Figure 2. A erful at the group level than at the individual level. In
MANOVA across these two measures indicated that no addition, the failure to find reliable identification effects
main effects were reliable, but the Anonymity ´ Priming may be due to suboptimal operationalization. Questions
interaction was significant, F(2, 16) = 7.73, p < .01.2 referred to identification “at this moment.” This phras-
Inspection of the univariate effects for efficiency-oriented ing is problematic to the extent that what is relevant for
and prosocial words separately indicated that both inter- the conformity to group norms is identification with the
actions were significant: F(1, 17) = 7.83, p < .02, for effi- group during the discussion. The follow-up study exam-
ciency-oriented words (Figure 2a) and F(1, 17) = 4.90, p < ines this explanation.
.05, for prosocial words (Figure 2b). The simple effect Study 1 suggests that greater behavioral conformity to
for efficiency-oriented groups showed a trend in the pre- the primed norm can be found in the anonymous group,

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

 Postmes et al. / COMPUTER-MEDIATED COMMUNICATION 1249

7.00

5.97
6.00 Proscial Prime
Efficiency Prime
5.16
5.00
Average Number of Words

4.00
3.23 3.12 3.20 3.15
3.00
2.33

2.00
1.37

1.00

0.00
Anonymous Identifiable Anonymous Identifiable

A: Efficiency Oriented Words B: Prosocial Words

Figure 2 Number of efficiency-oriented and prosocial words by condition corrected for the total number of words per group, Study 1.

but there is no conclusive evidence that active social would be supported if no systematic differences exist
influence causes the Anonymity ´ Priming interaction. between neutrally primed subgroups and efficiency-ori-
The design of this priming study does not allow us to rule ented subgroups. Rather, anonymous groups as a whole
out that influence is a purely intraindividual process are predicted to display more efficiency-oriented behav-
derived from the primed norm and that this process ior compared with identifiable groups. In addition, neu-
does not involve communication between group mem- trally primed subgroups should become more efficiency
bers. Study 2 is designed to establish that the effect of oriented over time in the anonymous condition but not
primed behavior under conditions of anonymity is in the identifiable condition.
indeed caused by social influence in the sense that it Another question of interest is whether participants
involves the social transmission of (mutual) influence perceive their group to be more or less efficiency ori-
between individuals. This would allow for more definite ented in the predicted direction. Although such con-
inference of social influence as responsible for the scious awareness need not necessarily exist for confor-
increase in primed behavior in anonymous groups. mity to a group norm, the SIDE model’s predictions
The method chosen to investigate this question was to would be supported if participants could identify the
prime only two out of four group members with an effi- group atmosphere correctly. Moreover, if participants
ciency prime and the two other group members with a generalize the efficiency orientedness of group mem-
neutral prime. No prosocial primes were used in Study 2 bers to a characteristic feature of the group in other con-
because we had already established the generality of the texts (i.e., if they perceive a group norm), this would be a
effect across priming norms in Study 1. The follow-up’s direct indication that norm formation has taken place
main purpose was to investigate whether neutrally (Postmes et al., in press). Thus, Study 2 includes mea-
primed participants would, over time, come to display sures of the group atmosphere and group norm. Study 2
similar behavior to participants who were primed with effi- also controls for the potential effect of taking pictures of
ciency. Thus, groups consisted of an efficiency-primed identifiable participants by taking pictures of anony-
and a neutrally primed subgroup. If the alternative inter- mous participants as well. This is to ensure that the factor
pretation suggested above is true, then the efficiency anonymity is limited to the discussion phase, when it is
prime would have an effect on the efficiency-primed sub- hypothesized to have its impact, and rules out the possi-
group only, not on the neutrally primed subgroups. bility that taking photos could itself account for differ-
However, if social transmission of the norm occurs ences between conditions. Study 2 reexamines the pro-
within groups, then neutral subgroups and primed sub- posed mediating role of group identification by asking
groups should, over time, conform in the anonymous participants for their retrospective identification with
condition. Hence, our explanation of effects in Study 1 the group during the discussion.

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

1250 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

STUDY 2 figure that could range from zero to the number of words
in a contribution, in practice either zero or one). The word
Method count was then regressed on the statement number, which
Participants and procedure. After dropping two groups renders a score ranging from –1 (all efficiency-related
from the analysis who did not discuss the topic, 64 partic- words at the beginning of the conversation) to 1 (all effi-
ipants, 45 women and 19 men, who were divided across ciency words at the end of the conversation). These stan-
16 groups of 4, participated in return for course credit. dardized regression coefficients were then averaged
The procedure was similar to the previous study, with the across the five efficiency words for each individual par-
exception that digitized pictures were taken of all partici- ticipant and subsequently averaged per subgroup.
pants. Participants entered the cubicle, performed the Results
scrambled-sentence task (an efficiency or neutral
prime), performed the ostensibly unrelated group task, Results were analyzed at the group level. Because each
and discussed the topic via CMC. After discussion, partic- group consisted of two subgroups (one efficiency
ipants filled out a questionnaire and were debriefed primed and one neutrally primed), we had to solve the
about the purpose of the study. problem of dependence between measures. Analyses
focused on group means and on subgroup difference
Manipulations. As in Study 1, a scrambled-sentence scores. Group means were analyzed to test for the main
test was used as a prime. In each group, two out of four effect of the anonymity manipulation. We shall refer to
members were given efficiency primes (the efficiency these as the group scores in the Results section. The sub-
subgroup), and two members were given neutral primes group difference score is calculated by subtracting the
(the neutral subgroup). The efficiency prime was identi- neutral subgroup mean from the efficiency subgroup
cal to Study 1. The neutral prime resembled the effi- mean and is an elegant way of correcting for the inter-
ciency prime in number of words and complexity but did dependence among subgroups. If efficiency subgroups
not contain prosocial or efficiency-oriented stimuli. The have a higher score on a measure than neutral sub-
anonymity manipulation was identical to Study 1. groups, this results in a positive subgroup difference
Dependent variables. The questionnaire consisted of 9- score. These scores are used to establish the effect of the
point rating scales (1 = not at all, 9 = very much). The ano- priming manipulation and of the Anonymity ´ Prime
nymity check was identical to that used in Study 1. Identi- interaction. A main effect of the priming manipulation
fication was measured with the same three questions as exists if the subgroup difference score is greater or
in Study 1, but these questions were rephrased to ask for smaller than zero across conditions. There is an inter-
the degree of retrospective identification during the dis- action between anonymity and priming manipulations if
cussion: “I identified with the group during the discus- the subgroup difference is different for anonymous than
sion” (a = .86). Participants were asked to indicate their for identifiable groups. t tests were conducted to com-
opinion of the degree of efficiency orientedness of the pare group scores and subgroup difference scores.
group atmosphere with two items (a = .77): “Our group Questionnaire data. The manipulation check of the
discussed rationally” and “I think the people in this anonymity manipulation had the predicted result.
group were sensitive to each other” (recoded). In addi- Group scores in the anonymous condition indicated
tion, a measure was added of the perceived group norm group members were more anonymous to each other (M =
with two items (a = .69): “I have the impression this 5.38, SD = 0.86) compared to identifiable groups (M = 6.34,
group is generally business-like” and “The people in this SD = 0.82), t(14) = 2.41, p < .05.4
group usually make rational choices.” As in Study 1, an As predicted, the group atmosphere was rated to be
open-ended question asked participants to indicate what more efficiency oriented in the anonymous condition
would be the best solution to the problem, and (group score M = 5.94, SD = 0.98) compared with the
responses were coded (intraclass r = .79). The content of identifiable condition (M = 5.10, SD = 0.87), t(14) = 1.79,
the discussions was analyzed similar to Study 1. In addi- p < .05. Of importance, the subgroup difference scores
tion, we computed the increase (or decrease) of the were not significantly different from zero or from each
number of efficiency words over time. This was other (Midentifiable = –0.25, SD = 0.91, Manonymous = –0.04, SD =
operationalized as a standardized regression coefficient 0.95; negative scores indicate neutral subgroups per-
that was computed for each individual’s use of a certain ceived a more efficiency-oriented atmosphere). This
efficiency word during the discussion. For each contri- indicates that efficiency subgroups and neutral sub-
bution to the discussion (ranging from one to the total groups did not differ among each other in their percep-
number of contributions made by this individual), we tions of the group norm. Thus, for both the efficiency
counted the number of times that this word occurred (a and neutrally primed subgroups, anonymous partici-

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

 Postmes et al. / COMPUTER-MEDIATED COMMUNICATION 1251

pants perceived their groups to be more efficiency ori-

ented than identifiable groups. The same pattern was
obtained for the group norm measure. Group scores (.56*) .22
Anonymity Efficiency-oriented
indicated that groups in the anonymous condition per- Solutions
ceived a more efficiency-oriented group norm (M = 5.34,
.58* .57*
SD = 0.93) compared with identifiable groups (M = 4.53,
SD = 0.79) t(14) = 1.88, p < .05. Subgroup differences
indicated that the priming main effect and the interac-
tion were not significant. Thus, both atmosphere and Identification

the group norm were as predicted.

With respect to the coders’ ratings of individual solu-
tions, the predicted pattern was found. Solutions in the
Figure 3 Path analysis of the predicted mediation, Study 2.
anonymous condition were more efficiency oriented (M = *p < .05.
3.43, SD = 0.72) than solutions in the identifiable condi-
tion (M = 2.56, SD = 0.66), t(14) = 2.50, p < .05. No signifi-
cant effects were found for the priming manipulation or
for the interaction. Also consistent with the prediction 0.65, ns. Thus, both with regard to average number of
was group identification. In anonymous groups, identifi- efficiency words used and with regard to the changes
cation was higher (group M = 5.05, SD = 0.85) compared over time, predictions were confirmed.
with identifiable groups (M = 4.11, SD = 0.53), t(14) = Mediation. To test the predictions of the SIDE model
2.65, p < .05. The subgroup difference scores showed no with respect to the mediating role of group identifica-
significant priming main effect or interaction effect. tion, we further examined the causal links between the
Discussion content. With regard to the counts of the num- input variable (anonymity contrast coded such that 1
ber of sentences, words, and characters, no significant dif- indicates anonymity and –1 indicates identifiability), the
ferences were found. The number of self-references did process variable (social identification), and the outcome
not differ between anonymous (M = 1.41, SD = 1.04) and variable (the coders’ ratings of individual solutions to
identifiable groups (M = 1.64, SD = 0.76), t(14) = 0.51, ns. the dilemma). In a series of regression analyses, a path
With regard to the number of efficiency-oriented words analysis—displayed in Figure 3—was conducted to estab-
used during discussion, and the developments over lish mediation of identification (Baron & Kenny, 1986).
time, predictions also were confirmed. Anonymous Initial regressions established that anonymity has an
groups used more efficiency-oriented words (M = 6.53, impact on the solutions (b = .56, p < .05), that anonymity
SD = 3.13) than did identifiable groups (M = 3.87, SD = has a significant impact on identification (b = .58, p <
2.00), t(14) = 2.30, p < .05. No significant differences .05), and that identification and solutions are reliably
were found between subgroups: Neither the priming related (b = .70, p < .01). A fourth regression analysis
main effect nor the interaction effect was significant. showed that when solutions are the dependent measure,
Regarding the developments over time, reliable effects and anonymity and identification are entered as predic-
were obtained for anonymity, t(14) = 3.77, p < .01, and for tors, only the effect of identification is reliable (b = .57,
the interaction between anonymity and priming condi- p < .05), whereas the effect of anonymity is not (b = .22,
tion, t(14) = 2.46, p < .05. The development in time of p = .35, ns). Moreover, the percentage of variance
the number of efficiency-oriented words differed reli- explained by the path model with only anonymity as a
ably from zero in one condition: The neutrally primed predictor of solutions (31%) improves reliably when
subgroups in the anonymous condition showed an identification is added as a predictor, r2 = 53%, Dr2 = .22,
increase of efficiency words over time (M = 0.20, SD = Fchange(1, 13) = 6.05, p < .05. If, subsequently, anonymity is
0.12), t(7) = 4.81, p < .01. The changes over time in this dropped from the model as a predictor, the drop in vari-
subgroup condition were reliably greater than the ance explained is nonsignificant, r2 = 49%, Dr2 = –.03,
changes over time in the other half of the group, the effi- Fchange(1, 15) = 0.93, ns. This suggests that identification
ciency primed subgroups in the anonymous condition mediates the effect of anonymity on norm-consistent
(M = 0.06, SD = 0.12), t(7) = 3.32, p < .01. In contrast, the solutions.
subgroups in the identifiable condition did not differ
Discussion
reliably from each other (or from zero), the changes
over time in the neutrally primed subgroups (M = –0.05, Results of this study replicate and extend the findings
SD = 0.09) being similar to changes over time in the effi- of Study 1 for efficiency-oriented groups. The hypothesis
ciency primed subgroups (M = 0.00, SD = 0.12), t(7) = that anonymity in the group can facilitate social influ-

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

1252 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

ence was supported. Anonymous groups conformed to This suggests that visual anonymity increases group
the efficiency-oriented behavior that was primed in half members’ identification with the group during discus-
of the group. Neutrally primed subgroups displayed nor- sion, which in turn increases the social influence within
mative behavior as well as efficiency-primed subgroups. the group.
This was evidenced by the proposed solutions after dis- These findings extend and develop the empirical
cussion and by the discussion content. Moreover, neu- base for the assumptions of the SIDE model, particularly
trally primed subgroups accommodated to the primed in the context of CMC. Earlier research suggested that
subgroup when they were anonymous, such that they anonymity could enhance social influence of a group’s
used more efficiency words over time. In addition to norm when group identity is salient (Postmes & Spears,
these behavioral measures, and compared with identifi- 1998; Spears et al., 1990). The current studies provide
able participants, anonymous participants perceived a direct support for this suggestion, in particular because
more efficiency-oriented atmosphere and inferred the the group norms were manipulated experimentally in a
existence of an efficiency-oriented social norm. Thus, subtle manner between conditions. Thus, we avoid possi-
the results suggest that priming led to normative behav- ble demand characteristics, allowing us to place greater
ior in groups whose members were anonymous to each faith in the causal inferences that can be drawn about the
other but not in groups whose members were effects of social norms in combination with anonymity.
identifiable. Moreover, the present study provides insight into the
The fact that there was no difference between efficiency- proposed mediating mechanisms in terms of social influ-
primed and neutrally primed group members on any of ence and enhanced identification with the group that
these measures (except for the changes over time) con- take place under conditions of anonymity. Nonetheless,
firms the assumption that the influence is genuinely the correlational nature of the mediation effects sug-
social and not an intraindividual process of picking up gests that further investigation of the process is required.
on previously primed behavior. Thus, the hypothesis was Further research also should attempt to provide a
supported that normative processes may shape behavior more refined analysis of anonymity, both in terms of the
in anonymous groups, despite the less direct (i.e., visi- dimension and degrees of anonymity and in terms of
ble) social contact of group members with each other. anonymity of others to oneself versus identifiability of
Finally, Study 2 shows some support for the proposed oneself to others. As we indicated in the introduction,
processes underlying these findings, identification the SIDE model proposes that the perceived anonymity
mediating the effect. It should be noted, however, that of group members to oneself is the critical factor pre-
due to the correlational nature of the mediational analy- dicted to affect group salience, identification, and
ses conducted, these process inferences can never be as ingroup influence. Identifiability to others, however, can
strong as when the processes are manipulated directly. influence strategic behavior, especially where one feels
In addition, it is possible that identification mediates the accountable to a group (Reicher et al., 1995). This route
effect because it is highly correlated with other variables to behavioral influence cannot explain the social influ-
that are the “true” mediator (group salience would be ence effects in the present study because it predicts
one likely candidate). Nonetheless, the path analysis fits greater “influence” under conditions of identifiability.
the predictions well and suggests that anonymity However, research separating out these aspects of ano-
increased social identification, which in turn leads to nymity is an important next step.
normative behavior. These findings cannot be reconciled easily with classi-
cal deindividuation theory, or with theoretical
GENERAL DISCUSSION approaches to CMC, which are based on deindividuation
Two studies confirm predictions derived from the theory. As indicated earlier, decreased self-awareness is
SIDE model. Anonymous groups displayed primed the primary mechanism proposed to be responsible for
behavior, whereas identifiable groups did not. That this decreasing attention to social norms and standards
finding is due to social influence is strengthened by the according to both theories. However, no evidence was
finding that in anonymous groups a neutrally primed found that anonymity reduces self-awareness and, more-
subgroup conforms to the efficiency-oriented behavior over, anonymous groups evidenced more social influ-
primed in fellow group members. Thus, visually anony- ence, not less. Nevertheless, it is important to note that
mous groups appear to be more conducive to social the type of social influence documented in this study
influence in line with a primed group norm than identi- does not correspond to the type of social norms impli-
fiable groups. The underlying process proposed by the cated in deindividuation theory or those that have been
SIDE model received some support in Study 2. Results central to CMC theorizing. As outlined by Diener
show that the effect of visual anonymity on normative (1980), deindividuation theory postulates transgression
behavior is mediated by identification with the group. of general societal norms as a result of anonymity. These

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

 Postmes et al. / COMPUTER-MEDIATED COMMUNICATION 1253

studies examined local group norms that do not neces- 2. Although the overall number of efficiency-oriented words
appeared larger than the number of socioemotional words, not too
sarily correspond to those broader societal norms. The much value can be attached to this finding because the normal occur-
present results therefore imply that our understanding rence of these words is probably not equal.
of anonymity’s effects in groups and collectives (whether 3. Codings of the solutions proposed during the interaction (i.e., in
public) show the exact same pattern of results as the private responses
in real life or on the Internet) might benefit from distin- afterward. These are not reported because of the redundancy of the
guishing local and societal norms and making more effects.
apparent when each will exert its influence (Postmes & 4. Due to the directional nature of the predictions, one-tailed tests
are reported.
Spears, 1998; Reicher et al., 1995). In sum, the results are 5. It is important to note that our procedures were designed to pro-
more consistent with predictions of the SIDE model vide a salient group identity. According to the Social Identity model of
than with predictions of deindividuation theory or Deindividuation Effects (SIDE) model, this is required for such a pro-
cess to occur. Not making the common group identity salient may lead
approaches derived from it and applied to CMC. to different anonymity effects (Postmes, Spears, & Lea, 1998; Spears,
We can also make a fundamental point with regard to Lea, & Lee, 1990; Spears & Lea, 1992).
social influence on the basis of this research. The impli-
cation of the findings is that factors such as copresence, REFERENCES
interdependence, and identifiability are unlikely to be
Abrams, D., Wetherell, M. S., Cochrane, S., Hogg, M. A., & Turner, J. C.
the only or even the primary factors that determine the (1990). Knowing what you think by knowing who you are: Self-
influence of the group on its members (cf. Deutsch & categorization and the nature of norm formation, conformity and
Gerard, 1955; Latané, 1981). This is because most influ- group polarization. British Journal of Social Psychology, 29, 97-119.
Allport, G. W. (1985). The historical background of social psychol-
ence was found in those groups where these factors ogy. In G. Lindzey & E. Aronson (Eds.), The handbook of social psy-
could not have played a role for the simple reason that chology (Vol. 1, pp. 1-46). Reading, MA: Addison-Wesley.
members were anonymous. In contrast, groups whose Bales, R. F. (1950). Interaction process analysis: A method for the study of
small groups. Reading, MA: Addison-Wesley.
members were identifiable displayed less influence. The Bargh, J. A., Chen, M., & Burrows, L. (1996). The automaticity of
results suggest that social influence at the group level social behavior: Direct effects of trait concept and stereotype acti-
originates not merely from the interpersonal relations vation on action. Journal of Personality and Social Psychology, 71, 230-
244.
within the group but that social influence is also Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator vari-
grounded in the relation of group members to the group able distinction in social psychological research: Conceptual, stra-
as a whole, such that greater identification with the tegic and statistical considerations. Journal of Personality and Social
Psychology, 51, 1173-1182.
group fosters greater social influence.5 Our results there- Daft, R. L., & Lengel, R. H. (1984). Information richness: A new
fore support a “cognitive” redefinition of the group approach to managerial behavior and organizational design.
(Turner, 1982). Research in Organizational Behavior, 6, 191-233.
Davis, D., & Brock, T. C. (1975). Use of first-person pronouns as a
However, our results also make clear that the group function of increased objective self-awareness and performance
is not just a mere cognitive concept in that social inter- feedback. Journal of Experimental Social Psychology, 11, 381-388.
action is required for social influence to occur. Interac- Deutsch, M., & Gerard, H. B. (1955). A study of normative and infor-
mational social influences upon individual judgement. Journal of
tion and communication between individuals serves to Personality and Social Psychology, 51, 629-636.
establish and strengthen a local group norm (Haslam, Diener, E. (1980). Deindividuation: The absence of self-awareness
1997). In this sense, the group is both cognitive and and self-regulation in group members. In P. B. Paulus (Ed.), Psy-
chology of group influence (pp. 209-242). Hillsdale, NJ: Lawrence
social at the same time. We argue that social norms can Erlbaum.
be induced from social interaction and that identifica- Doosje, B., Spears, R., & Koomen, W. (1995). When bad isn’t all bad:
tion with the group is a prerequisite for such norm con- Strategic use of sample information in generalization and stereo-
typing. Journal of Experimental Social Psychology, 69, 642-655.
struction to occur. Furthermore, assuming some level of French, J.R.P., & Raven, B. (1959). The bases of social power. In
identification with the group, anonymity can further D. Cartwright (Ed.), Studies in social power (pp. 150-167). Ann
enhance the salience of the group, and the power of its Arbor, MI: Institute for Social Research.
Haslam, A. (1997). Stereotyping and social influence: The foundations
norms, by making sure that individuals do not get in its of stereotype consensus. In R. Spears, P. J. Oakes, N. Ellemers, &
way (by either deflecting from its norms or by undermin- S. A. Haslam (Eds.), The social psychology of stereotyping and group life
ing the group level of categorization). Paradoxically, (pp. 119-143). Oxford, UK: Basil Blackwell.
Hyman, H. (1942). The social psychology of status. Archives of Psychol-
reducing the presence of the individuals within the ogy, 38 (Whole No. 269), 94.
group may actually serve to accentuate the presence of Kiesler, S., Siegel, J., & McGuire, T. W. (1984). Social psychological
the group within the individual. aspects of computer-mediated communication. American Psycholo-
gist, 39, 1123-1134.
Latané, B. (1981). The psychology of social impact. American Psycholo-
NOTES gist, 36, 343-356.
Lewin, K. (1997). Resolving social conflicts: Selected papers on group
1. The counts of units of text reported are corrected for the number dynamics (G. Weiss Lewin, Ed.). Washington, DC: American Psy-
of words typed per group to minimize variance. Reported effects are chological Association. (Original work published 1948)
similar for both corrected and uncorrected counts. The correction Lott, A. J., & Lott, B. E. (1965). Group cohesiveness as interpersonal
consisted of division of counts by the number of words and multiplica- attraction: A review of relationships with antecedent and conse-
tion by 120, the average number of words overall. quent variables. Psychological Bulletin, 64, 259-309.

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009

1254 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN

Martin, L. L. (1986). Set/reset: Use and disuse of concepts in impres- (Eds.), European review of social psychology, Vol. 6 (pp. 161-198).
sion formation. Journal of Personality and Social Psychology, 51, 493- Chichester, UK: Wiley.
504. Rutter, D. R. (1984). Looking and seeing: The role of visual communication
Matheson, K., & Zanna, M. P. (1990). Computer-mediated communi- in social interaction. Chichester, UK: Wiley.
cations: The focus is on me [Special issue: Computing: Social and Sherif, M. (1935). A study of some social factors in perception.
policy issues]. Social Science Computer Review, 8, 1-12. Archives of Psychology, 187, 1-60.
McKenna, K.Y.A., & Bargh, J. A. (1998). Coming out in the age of the Short, J. A., Williams, E., & Christie, B. (1976). The social psychology of
Internet: Identity “demarginalization” through virtual group par- telecommunications. Chichester, UK: Wiley.
ticipation. Journal of Personality and Social Psychology, 75, 681-694. Spears, R., & Lea, M. (1992). Social influence and the influence of the
McLeod, P. L., Baron, R. S., Marti, M. W., & Yoon, K. (1997). The eyes “social” in computer-mediated communication. In M. Lea (Ed.),
have it: Minority influence in face-to-face and computer-mediated Contexts of computer-mediated communication (pp. 30-65). Hemel
group discussion. Journal of Applied Psychology, 82, 706-718. Hempstead, UK: Harvester Wheatsheaf.
Postmes, T., & Spears, R. (1998). Deindividuation and anti-normative Spears, R., Lea, M., & Lee, S. (1990). De-individuation and group
behavior: A meta-analysis. Psychological Bulletin, 123, 238-259. polarization in computer-mediated communication. British Jour-
Postmes, T., Spears, R., & Lea, M. (1998). Breaching or building social nal of Social Psychology, 29, 121-134.
boundaries? SIDE-effects of computer-mediated communication. Srull, T. K., & Wyer, R. S., Jr. (1979). The role of category accessibility
Communication Research, 25, 689-715. in the interpretation of information about persons: Some deter-
Postmes, T., Spears, R., & Lea, M. (1999). Inter-group differentiation in minants and implications. Journal of Personality and Social Psychol-
computer-mediated communication: Effects of depersonalization. Manu- ogy, 37, 1660-1672.
script submitted for publication. Turner, J. C. (1982). Towards a cognitive redefinition of the group. In
Postmes, T., Spears, R., & Lea, M. (2000). The formation of group H. Tajfel (Ed.), Social identity and intergroup relations. Cambridge,
norms in computer-mediated communication. Human Communi- UK: Cambridge University Press.
cation Research, 26, 341-371. Turner, J. C. (1987). The analysis of social influence. In J. C. Turner,
Postmes, T., Spears, R., & Cihangir (2001). Quality of decision mak- M. A. Hogg, P. J. Oakes, S. D. Reicher, & M. S. Wetherell (Eds.),
ing and group norms. Journal of Personality and Social Psychology, 80, Rediscovering the social group: A self-categorization theory. Oxford, UK:
918-930. Basil Blackwell.
Prentice-Dunn, S., & Rogers, R. W. (1989). Deindividuation and the Turner, J. C. (1991). Social influence. Milton Keynes, UK: Open Univer-
self-regulation of behavior. In P. B. Paulus (Ed.), The psychology of sity Press.
group influence (2nd ed., pp. 86-109). Hillsdale, NJ: Lawrence Zimbardo, P. G. (1969). The human choice: Individuation, reason,
Erlbaum. and order vs. deindividuation, impulse and chaos. In W. J. Arnold &
Reicher, S. D. (1987). Crowd behaviour as social action. In J. C. Turner, D. Levine (Eds.), Nebraska symposium on motivation (pp. 237-307).
M. A. Hogg, P. J. Oakes, S. D. Reicher, & M. S. Wetherell (Eds.),
Lincoln: University of Nebraska Press.
Rediscovering the social group: A self-categorization theory (pp. 171-
202). Oxford, UK: Basil Blackwell.
Reicher, S. D., Spears, R., & Postmes, T. (1995). A social identity model Received June 2, 1999
of deindividuation phenomena. In W. Stroebe & M. Hewstone Revision accepted September 26, 2000

Downloaded from http://psp.sagepub.com at NORTHWESTERN UNIV LIBRARY on March 27, 2009