Effects of Deindividuating Situational Cues and Aggressive Models On Subjective Deindividuation and Aggression

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Journal of Personality and Social Psychology A

1980, Vol. 39. No. 1, 104-113 ^

Effects of Deindividuating Situational Cues and Aggressive


Models on Subjective Deindividuation and Aggression
Steven Prentice-Dunn and Ronald W. Rogers
University of Alabama

This experiment demonstrated that a subjective state of deindividuation medi-


ates the effect df deindividuating situational cues on aggression displayed by
small groups (n — A) of coacting aggressors. The deindividuated state was
composed of two factors, Self-Awareness and. Altered Experiencing, both of
which had a causal influence on aggressive behavior. These data' are inter-
preted in terms of deindividuation theories which assume that certain' input
variables redure self-awareness andtforkcern about social evaluation and thereby
weaken the restraints against expressing .antisocial behavior. Also as predicted,
compared with a nb-model control condition, a high-aggressive model dis-
inhibited overt displays of aggression, whereas a low-aggressive model inhibited
aggression, among, both individuated and deindividuated group members. '

Myths may be interpreted as emotional Le Bon's (1896) analysis of the antisocial


jjrojections of our internal processes, trans- aspect of crowd behavior led Festinger, Pepi-
lated into visual images. They present exis- tone, and Newcomb (1952) to formulate and
tential dilemmas in a concise, essential form. investigate the seminal concept of "deindi-
Mythical figures have been created, for ex-, viduation." Recently, Zimbardo (1969) and
ample, to help us understand and reconcile Diener (1977) conceptualized deindividuation
the dilemma of our dualistic. nature, human as a process in which antecedent social condi-
and animal. Thus the Minotaur1, although half tions lessen self-awareness and reduce concern
with evaluation by others, thereby weakening
human, was half bull, a monster that had to
restraints against the expression of Undesira-
be appeased by devouring human sacrifices. ble behavior. More recently, Diener (1977,
The roots of such myths have a powerful hold 1979a) integrated Duval and Wicklund's
on the human imagination. Thus, 32 centuries •(1972) objective self-awareness theory with
after the Minotaur was slain, Le Bon. (1896) • deindividuation theory. This theory empha-
confronted our dualistic nature in his study of sizes that the antecedent social conditions
crowds, in which our humanitarian quality (I.e.,' Zimbardo's "input" variables) must
can be submerged by the bestial. Character- operate through reduced self-awareness to
ized in terms more palatable to the times, the produce aggression. Loss of self-awareness,
image of' the Minotaur became the "group dissolution of identity, and diminished, ca-
mind," part rational (humanistic) and part pacity for self-regulation increase the likeli-
primordial, destructive instincts (animalistic). hood that people will engage in antinormative
behavior.
The weak link in this tricomponential'
The preparation of this article was partially sup- model is the absence of evidence for the ex-
ported by Research Grants Committee of the Uni-
versity of Alabama Grant 1011 to the second author. istence of a deindividuated state and its
The authors gratefully acknowledge David Speed and role in mediating between input and output
Todd Peebles for their assistance in the collection of variables. Zimbardo (1969) inferred the in-
the data.
ternal state from stimulus conditions and
Requests for reprints should be sent to Ronald W.
Rogers, Department of Psychology, University of response variables. Diener's (e.g., 1976) re-
Alabama, University, Alabama 35486. search program, which has revealed evidence

Copyright 1980 by the American Psychological Association, Inc. OO22-3S14/8O/39O1-O104J0O.75

104
DEINDIVIDUATION AND AGGRESSION 105

of a deindividuated state, has consistently Therefore, the first purpose of the present
denied the subjective state a causal role in the experiment was to determine if deindividuat-
release of aggressive behavior. Nevertheless, ing situational cues produce an internal de-
the existence of an internal state of deindi- individuated state ajid if this state, in turn,
viduation and its role in mediating aggression mediates aggressive behavior among individu-
must be demonstrated to support deindividu- als in small groups. Several manipulations
ation theory and to rule out plausible rival were combined to maximize the likelihood of
interpretations such as an operant explana- producing the theoretically relevant state. The
tion that "Zimbardo's 'input variables' are a deindividuating input variables, all of which
summary list of discriminative cues which have been demonstrated to facilitate aggres-
release unrestrained behavior" (Diener, 1977, sion (see reviews by Diener, 1977; Dipboye,
p. 144). Thus, the important research task is 1977), were (a) ajjonymity to the experi-
to identify the natnre of, the internal state menter, (b) ^honymity to the victim, (c) no
and determine-if it can be linked to observable responsibility for harm doing, (d) dim light-
antecedent conditions, and measurable re- ing, and (e) noise-induced arousal. I t was
sponses. predicted that deindividuating situational cues
Three published studies have investigated would produce more aggression than individu-
whether deindividuating situational cues are ating cues. Furthermore, it was anticipated
accompanied by an internal state of deindi- that a factor analysis of p retrospective ques^ ;
viduation. Diener, Dineen, Endresen, Beaman, tionnaire designed to assess the hypothesized
and Fraser (1975) used a retrospective ques- subjective state would yield two factors simi-
tionnaire but found no evidence of a deindi- lar to those reported by Diener (1979b). To
viduated state or of any relationship between investigate the causal role of the deindividu-
the questionnaire data and aggression. These ated state, we used structural equation mod* •
authors attributed the, null findings to the els, especially path analysis, which are hjfciflg
fact that the input variables were not suf-
used with increasing frequency, by social psy-
ficient to produce the relevant state. In a
chologists to, investigate causal networks
subsequent study, Diener (1976) obtained
among nonexperimental data (e.g., Rogers &
evidence of a multifactor subjective state;
Mewborn, 1976; Simonton, 1977; Tesser &
however, the measures of the state were not
Paulhus, 1976). Path analysis is useful .for
correlated with the aggression data. Thus,
the observed deindividuated state did not evaluating a priori theory by determining-if
mediate, oc cause, the observed aggression. In the obtained correlations among the relevant
the third study, Diener (1979b) examined variables are consistent with the causal model;
the -effects of deindividuating cues on socially it can be valuable when used in conjunction
inappropriate but nonaggressive behaviors with an experimental design.. Based on-, the
(e.g., eating mud, listing friends' faults). A deindividuation theories of Zimbardo (1969)
factor analysis ol the self-report data yielded and Diener (1977, 1979a), it was assumed
a two-factor solution.. The first factor was that the internal state of deindividuation
interpreted as lack of, self-awareness, labeled would have a' causal influence on aggression.
Deindividuation, and the second factor was Thus, the path analysis addressed the ques-
labeled Altered Experiencing. Thus, some evi- tion, Does the deindividuation state mediate
dence for a deindividuated state is emerging, the effects, of the situational variables on ag-
but this' state has not yet been causally re- gressive behavior?
lated to aggression. As Diener (1976, 1977) Although deindividuation lowers the thresh-
noted, confirmation of the theory of deindi- old for expressing normally inhibited behavior,
viduation requires more than the demonstra-
it does not automatically produce aggressive
tion of a relationship between independent and
dependent variables; evidence of the media- behavior. The behavior that ensues depends,
tional role of a deindividuated state is of cru- in large measure, on other situational variables.
cial importance. Collective violence frequently occurs in situa-
tions in which aggressive models are readily
106 STEVEN PRENTICE-DUNN AND RONALD W. ROGERS

available: In summary, small groups Were exposed to


either deindividuating or individuating situa-
Without conscious thought of his action he darted tional cues. Based oji deindividuation theory,
into Ihe street and hurled the empty pop bottle in
his h«nd toward the last of the departing black-and-
it was predicted that the deindividuating
white cars. Striking the rear fender of Sgt. Rankin's cues would not only increase aggressive re-
car, it shattered, and it was as if in that shattering sponding but also would produce an internal
the thousand people lining the street found their state of deindividuation that mediates the
oyfa 'release . . . . Amidst the rending sounds of
tearing metal, splintering glass, cries of bewilder-
effect of the situational cues on aggression. In
ment and shouts of triumph, the Los Angeles up- addition, it was predicted that, compared with
rising had begun. (Conot, 1967, p. 29) a no-model control condition, subjects ex-
posed to a high-aggressive model would be-
On the other hand, potentially disastrous have more aggressively, whereas subjects ex-
crowd situations can be averted by a person posed to a low-aggressive model would act less
exhibiting calm, restrained actions. A aggressively. This prediction was based on the
Therefore, the second purpose of the present assumption that a high-aggressive model
experiment was to investigate the effects of makes aggressive acts salient and appropriate,
aggressive and nonaggressive modeling cues whereas a low-aggressive model signifies that
on aggression, especially ihe aggression dis- aggressive responses are not appropriate.
played by people experiencing the internal
state of deindividuation. According to social Method
learning theory (e.g., Bandura, 1977), models
"(a) make behavioral alternatives salient, (b) Design and Subjects
sijgntfy what is appropriate behavior in a situa- A 2 X 3 factorial design was used, with two
tiqn, and (c) provide information about the between-subjects experimental manipulations: (a)
consequences of an act. The aggression-disin- deindividuating cues versus individuating cues and
(b) high-aggressive model versus low-aggressive
hibiting effects of aggressive models and the model versus no model. The research participants
aggression-inhibiting effects of nonaggressive were 75 males enrolled in the introductory psychol-
models have been well established in dyads ogy courses who volunteered to earn extra credit. In
in which one individual harms another (see one replication, 3 subjects failed to follow instruc-
tions, and thus these data were deleted. The femain-
reviews by Bandura, 1973; Baron, 1977). ing 72 males were assigned randomly to each of the
However, only one published study has exam- experimental conditions, yielding 12 subjects per cell.
ined aggressive modeling cues in a group con-
text. Diener et al. (1975) found that sub-
Apparatus
jects exposed to a filmed aggressive model
behaved more aggressively than subjects who The shock apparatuses were modified Buss ag-
viewed a nonaggressive model. The present gression machines connected to a polygraph. Each of
study sought to extend this finding in several the four aggression machines had 10 push-button
switches that could be depressed to deliver "shocks"
respects: (a) A live model was substituted of progressively increasing intensity. Of course,
for the filmed model, (b) administration of shocks were not actually delivered. A Grason-Stadler
shock was substituted for Diener's procedures, noise generator (Model "No. '901A) was used to
produce white noise in the deindividuating cues
which allowed noninjurious assault, and (c) condition.
a no-model control condition was added to
determine if differences in aggressive behavior
were produced by the disinhibiting effects of Procedure
aggressive models or by the inhibiting effects
Subjects arrived and were tested in groups of
of nonaggressive models. Thus, the present four; three were naive participants, and one was our
study sought to determine if individuals in assistant. The study was explained as a combination
small groups would behave more aggressively of two experiments. The naive subjects had signed
up for an experiment entitled "Behavior Modifica-
when exposed to a live high-aggressive model tion." The other study was "Biofeedback." After
and less violently when exposed to a live low- explaining the concepts of behavior modification and
aggressive model. biofeedback, we indicated that the response of interest
f'l
DEINDIVIDUATION AND AGGRESSION 107
in both studies was heart rate. It was explained that could choose any intensity they wished on each trial.
another male student from the introductory psy- These instructions were designed to eliminate any
chology course had volunteered for the biofeedback potential altruistic motivation, and they made clear
study. (Actually, he was another assistant of the that use of the lowest possible intensity on every
experimenter's.) This student, who would arrive trial would fulfill the. requirements of tie experi-
later, would be attempting to maintain his heart menter. Use of any intensity greater than that of
rate at a predesignated high level. Whenever his heart Switch 1 would only result in additional pain to the
rate fell below the predetermined level, the subjects biofeedback subject.
in the behavior modification study would admin- The experimenter then left to bring the biofeed-
ister an electric shock. The purpose of having groups back subject from a waiting room to the experi-
of four behavior modifiers was explained as an mental room. By leaving the doors open, the subjects
attempt to establish a laboratory analog of a ward heard the instructions given by the experimenter to
at the local state hospital in which behavior modi- the biofeedback subject about his role. Thus, the
fiers actually worked in small groups. naive subjects would easily hear but not see their
We explained to all subjects that they received future victim.
their extra credit points for simply showing up and
that they could discontinue at any time. Each sub-
ject was asked if he had any questions about or
Experimental Manipulations
any objections to the use of electric shock. All ques-
The first manipulation attempted to differentiate
tions were answered, and no one declined to par-
maximally between deindividuating situational cue*
ticipate. In addition, written informed consent was
and individuating ones. In the deindividuating cues
obtained. Two mild sample shocks were admin-
condition, subjects were not addressed by name by
istered to the behavior modifiers (i.e., the naive
the experimenter. They were informed that the
subjects) via finger electrode. The shocks were frdm
shock intensities they used were of no interest to
Switches 4 (.3 mA) and 6 (.45 mA) on the aggres-
the experimenter and that he would not know which^
sion machine, and each lasted for 1 sec. These sam-
intensities they selected (anonymity to the experi-1'
ples were administered to convince the subjects that
mentcr). Subjects were further informed that th«y
the apparatus really worked and to give them some
would not meet or see the biofeedback subject
idea of the intensities they would be delivering.
(anonymity to the victim). The experimenter iwli-
The subjects were then taken to an adjoining cated that he assumed full responsibility. for. the
room, seated at the aggression machines with parti- biofeedback subject's well-being (no responsibility
tions that blocked observation of others' responses for harm doing). Finally, white noise (65 dB
(thus, responses were experimentally independent), fSPL]) was played in the dimly lit room under the
and given instructions for operating the shock appa- guise of eliminating any extraneous noise from the
ratus. Each time the biofeedback subject's heart hall or other experimental rooms (arousal). Thus, as
rate fell below the predetermined level, a signal light in Zimbardo's (1969) first experiment, the subjects'
would be illuminated on their panel. Then they identlfiability was minimized.
were to select and depress 1 of the 10 switches on
In the individuating cues condition, the subjects
their panels. It was explained that the higher the
wore name tags and were addressed by their first
level chosen, the stronger the shock administered.
names. As in Zimbardo's (1969) study, the "unique
The "shock" received by the biofeedback subject
reactions" of each subject were emphasized, and
was alleged to be the average of the intensities se-
the experimenter expressed his interest in the shock
lected by the four behavior modifiers. Although the
intensities used by the subjects. Subjects were in-
duration of the shock was not mentioned, it was
formed that they would meet the biofeedback sub-
implicit that each shock would last for the same
ject on completion of the study. It was emphasized
length of time as the sample shocks. Because of the
that the biofeedback subject's well-being was the
emphasis placed on the shock intensities, they were
responsibility of each individual behavior modifier.
the primary measure of aggression.
The room was well lit, and no white noise was
Each group was presented with 20 signal lights broadcast.
over the course of the experiment. The interval After delivering all instructions, the experimenter
between the appearance of any two signal lights was stated that an equipment check was needed before
initially chosen randomly, ranging from 45 to 75 proceeding with the actual studies. It was noted that
seconds. The intervals were then held constant across the check would also give the biofeedback subject
subsequent replications the opportunity to experience several shocks before
The final instruction given to the subjects was that the session actually began. One of the behavior
any of the 10 shock switches would be sufficient modification subjects (our assistant) was asked to
for the purposes of the experiment. It was explained deliver five sample shocks while the others watched
that the equipment had been designed with different him The signal light was illuminated five times at
shock intensities because we had not known how 10-sec intervals. In the high-aggressive model con-
strong the shocks would have to be to increase dition, our model pressed, in order, Switches 9, 10,
heart rate. We explained we had discovered that the 8, 9, and 10 for approximately 1.5 sec each. In the
different intensities all had equal effects on the bio- low-aggressive model condition, our assistant de-
feedback subject's heart rate, so the naive subjects pressed, in order, Switches 2, 1, 3, 2, and 1 for
108 I STEVEN PRENTICE-DUNN AND RONALD W. ROGERS

*ble 1 However, any event (e.g., time of day, biasing


n Shock Intensities as a Function of the conversation, other extraneous factors) that affected
10 Independent Variables individuals participating in the experiment at the
same time, thus rendering their responses dependent,
would appear as a significant departure from _the
Efperimlntal condition Mean shock intensities remaining replications within that treatment cett. 8ta
multivariate analyses of variance (one per experi-
Modeling condition mental cell) were computed among the replications
Low-aggressive model • 4.5 (groups) within each cell. This analysis yielded no
, N o model 5.4 significant effects (mean p > .25), confirming the
High-aggressive model 6.2 independence of the responses.

Situational cues
Results
Individuating cues 4.3 Aggression
"Deindividuating cues 6.4
A multivariate analysis of variance was per-
approximately .25 sec each. In the no-model condi-
formed on .the sums of the shock intensity and
tion, this equipment check phase was omitted. After duration scores. This analysis yielded a main
this manipulation, the actual session began and con- effect associated with the situational cues
tinued for a total of 20 trials. manipulation; Wilks's lambda (A) was .636,
F(2 V 116) = 14.76, p < .0001. T*he type-of-
Postexperimental Session model variable also had a main effect, A =
Following the last shock trial, subjects completed .780, F(4,116) = .3.84, p < .01. The inter-
a questionnaire containing 19 items designed to action effect was not significant (p > .5). The
assess (on 10-point Liken rating scales) an in- two effects significant in the multivariate
ternal state of deindividuation. A second question- analysis were then examined by univariate
naire assessed suspicions about the experiment. Six
subjects suspected that shocks were not actually de- analyses of variance. These univariate analy-
livered, and one -correctly. guessed that the model ses revealed a main effect for the situational
was our assistant. These subjects were deleted from cues variable on shock intensity, F ( l , 59) =
the data analyses. The data in this unequal n design 24.73, p < .0001, and on shock duration,
were analyzed with the complete least squares model
recommended by Overall, Spiegel, and Cohen (1975) F ( l , 5 9 ) = *,44, p < .01. An inspection Of
because this model meets the criterion of estimating the mean intensity scores presented in-Table
the same parameters as those .estimated in ah pr- t indicates that subjects in the deindividuat-
thogonal design. After each experimental session, ing cues condition used higher shock intensi-
each subject was thanked and given a full debriefing
that was based on Mill's (1976) recommendations. ties than those in the individuating cues con-
Finally, a questionnaire was given to each student' dition. A comparison of the duration scores
in a stamped envelope addressed to the Department indicated that slightly longer durations were
of Psychology Committee on Ethics. These anony- used in the individuating cues condition than
mous responses were returned by 65% of the sub-
jects. Of the respondents, 100% understood why the in the deindividuating cues condition (.2 sec).
deception had been necessary and did not resent it. Table 1 also shows that the model variable
Abo, everyone stated they would be willing to par- had the predicted effect on shock intensity,
ticipate in another, similar experiment. F(2,59) = 5.26, p < .01. An identical pat-
tern of results was obtained on the measure
Data Analysis . of shock duration, F(2, 59) = 3.49, p < .05.
A Duncan multiple-range test (p < .05) re-
Should the unit of analysis be the individual or
the group? We selected the individual for theo-
vealed that, compared with the no-model
retical and statistical reasons. Theoretically, we (baseline) condition, the low-aggressive model
conceptualized deindividuation as did Singer, Brush, suppressed aggression, whereas the high-aggres-
and Lublin (1%S) and Zimbardo (1969): Deindi- sive model intensified aggressive responding.
viduation is an intraindividuil phenomenon that
occurs within a group context. In addition, parti-
tions made it impossible for a subject to determine Internal State of Deindividuation
what shock intensities the others were selecting;
thus, the responses were experimentally independent, The 19 retrospective self-report items de-
permitting the individual to be the unit of analysis. signed to assess the subjective state of deindi-
DEINDIVlDCATION AND AGGRESSION 109

viduation proposed by Diener (1979b) and Table 2


Zimbardo (1969) were factor analyzed to Factor Loadings on the Rotated Factors
ascertain the existence of a state that might
be labeled deindividuation. A scree plot of item Factor loadfyg
figienvalues (Cattell, 1966), provided by a
Factor 1 (Altered Experience)
principal-axis factor analysis, indicated a two-
factor solution. Thus, two factors were ro- Thinking was somewhat altered .74 •
Emotions were different from
tated using the varirrtax method. The factor normal .63
loadings, which were greater than or equal to Felt aroused .72
.4 on only one of the factors, are shown in Responsibility was shared .44
Table 2. The factors were stable (reliabilities Time seemed to go quickly .69
were .85). (An analysis of only subjects in Thoughts were concentrated on
the moment .68
the deindividuation condition yielded the same Session was enjoyable .72 •
two factors.) The only items that did not load Willing to volunteer for similar
clearly on only one factor were the two^ items s\udy .65
Liked other group members .70
that assessed anonymity to the experimenter
Feeding of togetherness among
and anonymity to the victim, hoth of. which .- -group members .61
loaded positively (greater than .4) on the first
factor and,Begatively (greater than .4) on the Factor 2 (Self-Awareness)
second fartor. The first factor w^as composed Felt self-conscious .74
of variables that might be labeled Altered Ex- Heightened sense of individual
perience. It is highly simHar to Diener's identity .45
(1979b) factor, Altered Experiencing. The Felt inhibited .6.?
I had responsibility for harm doing .83
second factor might be labeled Self-Awareness Concerned with what experimenter
because the items loading on it assess self- thought of me • .81
consciousness and the ensuing inhibition and Concerned with what victim
thought of me .75
concern with evaluation by others. Diener's
Concerned with what other group
(1979b) second factor, which he called Dein- members thought of me .71
dividuation, contained some of- these same
items in addition to several that. loaded on Note. For Factor 1, reliability (coefficient alpha) was
our Altered Experience factor. Although the .854, and 26% of total variance was accounted for.
For Factor 2,'reliability (coefficient alpha) was .851,
Self-Awareness factor is highly similar to early and 23% of total variance was accounted for.
definitions of deindividuation (e.g., Festinger
et al., 1952; Singer et al., 1965), we reserve
the term deindividuation to refer to the entire scores on the Altered Experience factor (M —
subjective state. Our data suggest that the 63.9) than, the subjects in the individuating
subjective state of deindividuation Is com- cues condition {M = 52.8) but lower scores
posed of two factors: Altered Experience and on the Self-Awareness factor (Ms = 30.8 and
Self-Awareness. 41.2, respectively).
Now that the existence of *hese two factors
has been established, it might be asked Deindividuation and Aggression
whether the factors reliably discriminate be-
tween the individuating and deindividuating - Pearson product-moment correlations were
cues conditions. The sum of each subject's computed between the raw score sums on the
raw scores on the Altered Experience items Altered Experience and Self-Awareness fac-
and on the Self-Awareness items were sub- tors and shock intensity and duration. For
mitted to a multivariate analysis of variance. Altered Experience and shock intensity, r(63)
A significant difference was found between = .35, p < .01; for Self-Awareness and in-
the two situational cues conditions, A = .806, tensity, r(63) = —.31, p < .02. The correla-
F(2, 124) = 7.06, p < .002. Subjects in the tions between shock duration and each of the
deindividuating cues condition had higher two factors did not differ significantly from
zero.
110 STEVEN PRENTICE-DUNN AND RONALD W. ROGERS

facilitate antisocial behavior without arousing


a deindividuated state (Diener et al., 1975).
Thus the present study provided the first
demonstration that the internal state medi-
ates the effects of deindividuation cues %|J,
antisocial behavior.
The internal state of deindividuation had
two components, Altered Experience and Self-
Awareness, that were highly similar to Dien-
er's (1979b) findings. The presence of the
Figure 1. Structural model of deindividuation add Self-Awareness factor confirms deindividua-
aggression with path coefficients—standardized par' tion theory's (Diener, 1977; Dipboye, 1977;
tial regression coefficients, p < .01. (The double- Zimbardo, 1969) major assumption that a
headed curved arrow denotes a correlation coeffi- reduction in self-awareness and concern for
cient.)
social ^valuation is a crucial mediator in the
deindividuation process. The deindividuating
The correlation matrix among the two fac- cues lowered self-awareness and altered nor-
tors of the state of deindividuation and shock mal cognitive and affective processes. This
intensity was decomposed into components , deindividuated state may have weakened the
corresponding to paths. The significant path restraints against behaving aggressively,
coefficients (standardized partial regression co- which are normally maintained by internal
efficients) are diagrammed in Figure 1. The and external norms of social propriety.
model indicates that both Altered Experience These findings also demonstrate that Al-
and Self-Awareness had a direct effect on tered Experience, the second facet of the de-
shock intensity and that both had an indirect individuated state, had an independent effect
effect via the correlation between the two on aggression. That is, regardless of whether
factors. Thus, it is apparent that the ob- a crowd member loses self-consciousness, if
tained structural model conformed closely to deindividuating cues alter cognitive and af-
the a priori causal model. In addition, it is fective processes, violence may result. Hind-
possible to verify the validity of the structural sight suggests that the Altered Experience
model (cf. Blalock, 1971). First, when shock factor might have been predicted from an in-
intensity was regressed on Altered Experience spection of Zimbardo's (1969, Figure 1) list-
and Self-Awareness, the unstandardized partial ing of input variables. The Self-Awareness fac-
regression coefficients were significant. Thus, tor, or identifiability, would be expected to be
the model was not misspecified. Second, affected by the antecedent conditions of ano-
not only is the obtained causal net- nymity, diffused responsibility, and group
work interpretable, but it conforms to the presence. However, many of the remaining
theoretically predicted model. Finally, the input variables (e.g., altered temporal per-
structural model adequately accounts for the spective, arousal, physical involvement in the
zero-order correlations among the variables; act, novel situation, altered state of conscious-
thus, the structural model can be said to fit ness) were the experiences reported by our
the data. subjects that loaded on the Altered Experi-
ence factor. Finally, liking for the other group
Discussion members, which Festinger et al. (1952)
posited as a consequence of deindividuation,
The results of the present experiment was part of the Altered Experience factor.
clearly supported the hypothesis, derived
from deindividuation theory (Diener, 1977; Although we have emphasized the media-
Zimbardo, 1969), that deindividuating situa- tional role of the deindividuated state, we
tional cues produce an internal state of de- should not overlook the main effect of the situ-
individuation that mediates aggression. It has ational cues variable on aggression: Subjects
been shown that deindividuating cues can in the deindividuating cues condition used
DEINDIVIDUATION AND AGGRESSION 111

stronger shock intensifies than those in the The data implicating the deindividuated
individuating cues condition. Although the state as the causal mediator must be interpreted
theoretical independent variable of deindividu- with caution. The regression and path analy-
ation has been operationalized in many ways, ses cannot prove that the proposed theory of
Jflts finding corroborates several previous in- deindividuation is correct, but they did con-
vestigations (see reviews by Diener, 1977; firm that the proposed causal model was con-
Dipboye, 1977). sistent with the obtained correlational data.
One unexpected finding was that subjects in However, if one wishes to be more conserva-
the deindividuating cues condition adminis- tive about inferring causality, it is comforting
tered shocks for .2, sec less than those in the to recognize that the structural model (see
individuating cues condition. The experimental , Figure' \) is a simple regression model, and in-
instructions never mentioned the duration of terpretatUms can be restricted to that level.
the shocks but referred repeatedly to the dif- In addition, it might be argued that our a
fering shock intensities. (e.g., Subjects could priori model should have been a nonrecursive
use any intensities they wished, the shqck re- one (e.g., allowing aggression to affect the de-
ceived by the victim would be the average of individuation state). Nevertheless, the sequence
the intensities selected by the group mem- suggested by our data should be regarded as
bers, the sample shocks differed only in inten- only a first approximation to a complete de-
sity, etc.UfcThus, this finding is difficult to scription of the relationship between deindi-
interpret.TJne possible explanation is that in- viduation and aggression. Future research (es-
tensity and duration were used as compensa- pecially time-series designs with lagged mea-
tory response systems, which occurs in inter- surements of subjective deindividuation and
racial aggression (Wilson & Rogers, 1975). aggression) may reveal complex patterns of
That is, the deindividuated subjects may have chain regressions and feedback loops; how-
compensated for the stronger intensities they ever, neither current theories nor empirical
delivered by administering them for shorter findings suggest such a model. Also, that sub-
periods of time. This interpretatibn suggests jects knew they were .being observed in a
that intensity and duration would be corre- scientific study undoubtedly attenuated the
lated negatively. However, the correlation be- strength of the deindividuated state, a con-
tween the measures was + .23, nonsignificant, dition that, if averted, could make the inter-
thus ruling out this interpretation. (The cor- nal state and the ensuing aggression' even
relation within the individuating cues condi- more intense. Despite the limitations of trie
tion was + .08.) Furthermore, any interpreta- present methodology, it is the strongest dem-
tion of the duration data in terms of shorter onstration yet of the existence of an internal
durations indicating less motivation to harm state of deindividuation and its mediational
is not consistent with the deindividuated state, role in aggression.
in which the deindividuated subjects felt less Another limitation of the study is that one
inhibited, less self-conscious, less responsible, can only get out of a factor analysis what one
and less concerned about what others thought puts into it. If additional deindividuating cues
of them. Another interpretation is that, be- were to be used and if questionnaires con-
cause of the altered experiences and increased tained more diverse items, additional factors
arousal, the deindividuated subjects were more might emerge. One likely candidate is a posi-
active and energetic. Hence their physical acts tive affect factor, which has been postulated
were both brief and intense, much like vigorous by some theorists (e.g., Zimbardo, 1969).
hitting and stabbing. Whatever the interpre- However, our factors resembled Diener's
tation of this finding, its importance is greatly (1979b), despite the fact that we used dif-
diminished by the fact that, in contrast with ferent deindividuating cues and different anti-
the intensity data, shock durations did not social behaviors (socially deviant behavior
correlate with the Altered Experience or Self- versus shock administration). Taken together,
Awareness factors; that is, the duration data these data suggest that the internal state of
were not related to the deindividuated state. deindividuation is composed of at least two
112 STEVEN PRENTICE-DUNN AND RONALD W. ROGERS

stable components. Baron, R. A. Human aggreWon. New York: Plenum


1 Press, 1977.
The main effect of the modeling cues on ag-
Blalock, H. M. (Ed.). Causal models in the social
gression revealed that, as predicted, compared sciences. Chicago: Aldine-Atherton, 1971.
with the no-model baseline condition, (a) the Cattell, R. B. The scree Jest for the number
high-aggressive model disinhibited aggression factors. Multivariate Behavioral Research, 1966,
* t y . weakening the observers' restraints that 1, 245-276.
normally inhibit aggression, and (b) the low- Conot, R. Rivers of blood,.years *»/ darkness. New
York: Bantam Books, 1967.
aggresftive model inhibited aggression by Diener, E, Effects of prior destructive behavior,
strengthening the observers' restraints against anonymity, and group presence on deindividua-
'expressing aggression. These findings"attend tion and aggression. Joitrnal of Personality and
previous studies by ill Minn li <ljuj) M^it In i Social Psychology, 1976, 33, 497-507.
aggressive and nonaggressive mo^lSVaffect Diener, E. Dejndividuation: Causes and conse-
. quences. Social Behavior and Personality, 1977, S,
collective behavior in small groups..''.Group 14.1-155.
members who were exposed to the deindividu- Diener, E. Deindividuation: The absence o/ self-
ating cues and who experienced the subjective awareness and self-regulation in group members.
state were influenced by the social models. The In p. Paulus ( E d ) , The psychology of group in-
restraints against expressing aggression, which fluence. Hilkdale, N J . : Erlbaum, 1979. (a)
Diener, E. Deindividuation, i If mill, and dis-
had been weakened by the deindividuating
inbibitlon: Journal of Personality andoocial Psy-
cues, were further weakened by exposure to chology, 1979, 37, 1160-1171. (b)
a high-aggressive model, but these restraints Diener, E.,: Dineen, J., Endresen, K., Beaman, A. L.,
were strengthened by exposure to the low-ag- & Fraser, S. C. Effects of altered responsibility,
gressive model. Thus,,in naturalistic settings, cognitive set, and modeling on physical aggression
the modeling cues present in a crowd can af- and deindividuation.- Journal of Personality and
Social Psychology, 1975, 31, 328-337.
fect deindividuated individuals as well as indi- Dipboye, R. L. Alternative approaches to deindividu-
viduated ones by signifying what behaviors, ation. Psychological Bulletin, 1977, 84, 1057-1075.
are appropriate in the situation and what the Duval, S., & Wicklund, R. A. A theory of objective
consequences will be. The presence.©? a highly self-awareness. New York: Academic Press, 1972.
aggressive model in a.crowd of deindividuated Festinger, L., Pepftone, A., & Newcomb, T. Some
consequences of de-individuation in a group.
individuals may intensify physical: assaults . Journal of Abnormal and Social Psychology, 1952,
against hapless victims. On- the other hand, a 47, 382-389.
low-aggressive or nonaggressive model could Le Bon, G. The crowd. A study of the popular mind.
exert a restraining influence even on. dein- • London: Ernest Benn, 1896.
dividuated individuals, thus suggesting one Mills, J. A procedure for explaining experiments in-
volving deception. Personality and Social Psychol-
method, of controlling potentially violent mobs.
ogy Bulletin, 1976, 2, 3-13.
The Minotaur may still reside within us. Overall, J. E., Spiegel, D. K., & Cohen, J. Equiva-
Although deindividuating situatfonal cues and lence of .orthogonal and nonorthogonal analysis of
aggressive models may arouse the-bestial half, variance. Psychological Bulletin, 197S, 82, 182-186.
individuating cues and calm, restrained models Rogers, R. W., & Mewborn, C. R. Fear appeals and
may, if not slay the beast, at least summon, attitude change: Effects of a threat's noxiousness,
probability of occurrence, and the efficacy of
the compassionate humanity we possess. coping responses. Journal of Personality and So-
cial Psychology, 1976, 34, 54-61.
Simonton, D. K. Eminence, creativity, and geo-
graphic marginality: A recursive structural equa-
References tion model. Journal of Personality and Social
Psychology, 1977, 35, 805-816.
Bandura, A. Aggression: A social learning analysis. Singer, J. E., Brush, C, & Lublin, S. C. Some as-
Englewood Cliffs, N.J.: Prentice-Hall, 1973. pects of deindividuation: Identification and con-
Bandura, A. Social learning theory. Englewood formity. Journal of Experimental Social Psychol-
Cliffs, N.J.: Prentice-Hall, 1977 ogy, 1965, 1, 356-378.
DEINDIVIDUATION A N b AGGRESSION 113

Tester,
ester, A., & Paulhus, IIWL.
^ L . Toward a causal model
mode Zimbardo, P. G. The human choice: Individuation,
of love. Journal of Personality and Social Psy reason, and order versos" deindividuation, impulse,
chology, 19*6, 34, 1095-1105. and chaos.-In W. J. Arnold & D. Levine (Eds.),
iUon, L., & Rogers, R. W The fire this time: Ef Nebraska Symposium on Motivation (Vol. 17).
fects of race of target, insult, and potential re Lincoln: University of Nebraska Prew, 1969. -
taliation on black aggression. Journal of Person
ality and Social Psychology, 1975, 32, 85?-«64. Received August IS, 1979. f *

You might also like