2316914

Journal of Community & Applied Social Psychology, Vol. 1,249-267 (1991)
Police Intervention in Riots: the Role of
Accountability and Group Norms. A Field Experiment
MARCELINE B. R. KROON, DIK VAN KREVELD AND
JACOB M. RABBIE
Department of Social and Organizational Psychology, University of Utrecht, Heidelberglaan I ,
3584 CS Utrecht, The Netherlands
ABSTRACT
This study explores the role of perceived accountability in riot police action. The basic hypothesis is that accountability, provided that non-violent norms are made salient, will lessen the
chances of escalation of a conflict between police and demonstrators. Four platoons participated in a field experiment conducted at the Dutch Riot Police Academy, in which they
played a riot simulation. In a 2 x 2 design, the effects of accountability and Department
on attitudes and norms concerning the action of ‘police’ and ‘demonstrators’ were tested.
Measurements were made before and after the simulation. Internal analysis, involving redefined
accountability categories, provided support for our hypothesis. That is, perceived accountability proved to be related to a heightened public self-awareness, a less extremely positive evaluation of fellow group members, and less intergroup differentiation. The limitations of an
internal analysis are discussed, as well as the importance of the nature of organizational
norms and of the accountability forum in predicting the effect of accountability.
Key words: Accountability, crowd behaviour, intergroup conflict, group norms,
de-individuation.
For almost a century now, scientists from various disciplines have been trying to
explain the violence displayed by crowds involved in a riot. One line of theorizing,
which has its roots in late nineteenth-century pioneering work on crowd behaviour
in France and Italy (see Van Ginneken, 1989, for an overview) states that members
of a crowd become ‘submerged’ in the mass; that is, are not readily identifiable
as individuals, lose their feelings of uniqueness and awareness of their personal standards. Festinger, Pepitone and Newcomb (1952) named this phenomenon ‘de-individuation’. According to classic de-individuation research, in this psychological state
the ‘learned concern for social evaluation and the self-observingaspect of conscience’
(Zimbardo, 1970, p. 253) are minimized, which causes the release of normally restrained behaviour (Festinger et al., 1953; Zimbardo, 1970).
In more recent theories of de-individuation it is argued that situational factors
can prevent group members from becoming self-aware, by directing their attention
away from their inner feelings and personal norms. Lack of identifiability within
a group, focusing on the group as a whole instead of on its individual members,
0885-6249191 JO40249-19$09.50
0 1991 by John Wiley & Sons, Ltd.
Received 30 M a y 1991
Accepted22 August 1991
250
M. B. R. Kroon
activities centred on outside events, etc., are such factors. According to Diener (1980),
a lack of self-awareness, together with conditions which prevent the development
of self-awareness, such as conformity pressures and the presence of role models,
will cause de-individuation. Under such de-individuating circumstances, individual
behaviour will be guided by random situational cues (Diener, 1980; Prentice-Dunn
and Rogers, 1983).
Prentice-Dunn and Rogers (1983) have specified this theory of de-individuation
in their ‘Differential Self-awareness Theory’, by distinguishing between ‘private’ and
‘public’ self-awareness. The former concept refers to attention focused inward, while
the latter refers to concern for the opinion of others. In their view, de-individuation
as an internal psychological state can only result from a lowered privute self-awareness. A lowered public self-awareness is not supposed to cause the psychological
state of de-individuation, although it does presumably facilitate the occurrence of
antisocial behaviour, because negative sanctions are not expected.
On a conceptual level the distinction between public and private self-awareness
is important, since it shows that two different mechanisms can account for the same
aggressive behaviour. In the field experiment to be reported here, however, we will
concentrate on public self-awareness, which, according to Prentice-Dunn and Rogers,
is triggered by ‘accountability cues’, such as identifiability and perceived responsibility.
In order to affect public self-awareness, accountability should be anticipated by
members of a crowd or group. Our basic hypothesis is that anticipated accountability,
by heightening public self-awareness, can inhibit crowd or group aggression. However, in view of other recent theories of crowd behaviour which stress the importance
of norms, this basic hypothesis will be specified in the following paragraphs. Although
Prentice-Dunn and Rogers do not mention this explicitly, the concept of public
self-awareness suggests that norms may play an important role in determining crowd
behaviour. That is, public self-awareness makes people take into account the norms
used by the external authorities or other fora who demand an account of their behaviour.
The theories of de-individuation mentioned so far have in common that they regard
de-individuated crowd behaviour, though not necessarily as negative, as ‘impulsive,
irrational, and de-regulated’. The ’emergent norm theory’, advanced by Turner and
Killian (1972), presents an opposite point of view. This theory states that the activities
of a set of prominent individuals, or ‘keynoters’, lead to the emergence of social
norms. That is, these keynoters function as it were like role models, their behaviour
being viewed by other crowd members as the right way to behave. In the course
of a crowd event, behaviour becomes guided by these group norms. Contrary to
de-individuation theory, emergent norm theory predicts that ‘the control of the crowd
is greatest among persons who are known to one another rather than among anonymous persons’ (p. 25).
Norms can develop in different ways. The behaviour of others can become a norm,
as in the case of the keynoters, but also the law, religion, organizational climate,
etc. are possible sources of norms. Here we define norms as expectations about
behaviour. Violation of these expectations can lead to negative sanctions, whereas
behaviour that matches the expectations may be positively sanctioned.
The assumption that de-individuated behaviour is not subject to norms remains
questionable, since research testing de-individuation against emergent norm theory
Riot police, accountability andgroup norms
251
has produced contradictory results (Rabbie and Visser, 1984; Mann, Newton and
Innes, 1982). Theoretically, the argument made by Diener (1980) and Prentice-Dunn
and Rogers (1983, 1989) that behaviour of de-individuated group members is not
normative in the sense that they find it morally correct, but instead is influenced
by situational cues such as role models, does not seem to be fully convincing. Although
these authors have empirically established that de-individuation lowers the sense
ofpersonal standards, it does not follow from their data that de-individuated persons
become insensitive to group normals as well. Moreover, the conceptual difference
between their ‘role models’ and the ‘keynoters’ of emergent norm theory is not made
clear by these researchers.
But emergent norm theory also has a fundamental problem. That is, it cannot
account for the specific content of emerging group norms, and therefore is unable
to predict the kind of behaviour a crowd will display (Rabbie, 1990). Emergent
norm theory can give only post-hoe explanations of the proceedings during a
crowd event and of its particular form. Therefore in the next section a third line
of recent theorizing, using the intergroup perspective, will be presented. This
approach offers a more comprehensive explanation o f crowd behaviour. Like
emergent norm theory, it also criticizes the fundamental notion of the irrationality
of de-individuated crowd behaviour (Rabbie and Visser, 1984; Reicher, 1984a,b;
Reicher and Potter, 1985).
The crucial point Reicher makes is that, in a state of de-individuation, the loss
of private self-awareness implies only the loss of a sense ofpersonal,not social identity
(cf. Tajfel, 1978; Turner, 1982). Indeed, by weakening the sense of personal identity,
in a group setting, de-individuation can increase the salience of social identity (i.e.
the knowledge of being a member of a specific social group or category, and
the importance attached to that membership). Thus, de-individuation resulting
from a focus on the group instead of on the individual member will make the
members behave in more or less the same way, based on the ingroup norms. In
other words, ‘crowd behaviour can be seen to have a clear social form’ (Reicher,
1984a, p. 342).
This approach makes it possible to account for the specific content of crowd
behaviour (e.g. when Germany beat England in the Football World Championship
1990, in subsequent riots in Britain mainly German targets were attacked (NRCHandelsblad, 5 July 1990)). Moreover, it can explain the clear limits of participation
in crowd action (e.g. why the riot police do not psychologically mix with the crowd)
(Reicher, 1984b) and the experimental findings that the nature of group behaviour
under de-individuating conditions depends on the norms made salient in the situation
(Johnson and Downing, 1979; Spears, Lea and Lee, 1990). Finally, it signifies that,
in order to inhibit crowd aggression, the anticipation of accountability alone
will not be enough to cause non-violent behaviour. In order to be effective, it is
essential that accountability will be accompanied by specific behavioural norms.
Only when people know what kind of behaviour is desired by the accountability
forum, can the anticipation of accountability stimulate them to display this specific
behaviour.
Different groups figuring in a particular situation may have contradictory norms,
as for example a crowd of hooligans versus riot police accompanying them to a
football match. This brings up the point that, in predicting the effect of accountability,
it is important to specify the locus of anonymity. Anonymity from fellow group
252
M. B. R. Kroon
members may have a largely different effect on crowd behaviour than anonymity
from outside agents (Mann et al., 1982; Rabbie and Visser, 1984), since the ingroup
norms may differ from the norms used by outsiders. In crowds, people seem to
be anonymous with respect to outsiders rather than to their fellow crowd members,
(Rabbie, 1990; Reicher, 1984b). Identifiability within a group of hooligans may stimulate group members to act aggressively, because under the prevailing ingroup norms
it heightens their status. Anonymity from the police, by lessening the concern for
the negative consequences of aggressive behaviour, may strengthen such an aggressive
tendency.
In an aggressive intergroup conflict, a certain balance exists between the display
of aggression and the negative sanctions that may follow. Each group that figures
in the conflict may expect certain costs and benefits of aggressive conduct; this applies
as much to hooligans, demonstrators, etc., as to the riot police. Also for riot police
officers, depending on their group norms, the relative anonymity from their superiors
or external agencies such as the press may encourage tough action.
Concluding, we see that in riots attention focused on outside events rather than
on personal feelings and norms, a feeling of belonging to a group rather than individuality, etc., may function to lower private self-awareness or personal identity, enhance
social identity and cause de-individuation. This in itself will lead to behaviour governed by the ingroup norms and modelled on the behaviour of fellow group members.
Moreover, diffusion of responsibility and anonymity from the outside world lower
public self-awareness and the concern for sanctions that could follow the display
of antisocial behaviour. In such crowd situations, where the incompatibility of group
interests will cause mutual antagonism between competing groups (Brown, 1988),
intergroup differentiation, discrimination of outgroups, and eventually disinhibited
aggression, are likely to occur.
The purpose of the present experiment is to test the effects of anticipated individual
accountability on the attitudes of members of the riot police involved in a simulated
intergroup conflict. We argue that antisocial behaviour and escalation of the intergroup conflict can be countered when group members are made individually accountable for their actions andwhen the party that asks them to account for their behaviour
sets an appropriate (anti-violent) norm. This party is called the forum of accountability, and will be represented in this study by the police authorities. This study focuses
on individual accountability, but as a shorthand this will further be referred to simply
as accountability.
Accountability, by increasing public self-awareness, can serve a dual goal. On
the one hand, accountability makes group members focus on the authorities, and
take their points of view into account. Accountability may reinforce the conformity
to the ‘peaceful’ norm, known to be set by the accountability forum (i.e. the authorities). It has been found in studies on aggression that anticipated accountability
decreases aggressive behaviour, especially when a non-aggressive norm is made salient
(Diener et al., 1975; Prentice-Dunn and Rogers, 1982; Rabbie and Goldenbeld, 1988)
and increases public self-awareness (Diener et al., 1975; Prentice-Dunn and Rogers,
1982). Therefore, we predict that public self-awareness, as measured by comments
such as ‘I was concerned with what other group memberdthe experimenter thought
of me’ (Rogers and Prentice-Dunn, 1981) will be highlighted when accountability
is expected. Besides, accountability will heighten conformity to the norm introduced
by the authorities. In view of this norm, which advocates moderation of violence
Riot police, accountability andgroup norms
253
against the other party, accountability will reduce intergroup differentiation and
hostility.
On the other hand, individual accountability will induce group members to focus
on their own behaviour rather than on the group. Being individually accountable,
people should account for their own actions, regardless of the performance of the
rest of the group. They have to rely entirely on themselves concerning the impression
they make on the accountability forum. Consequently, accountability will have an
individuating effect. Thus, we hypothesize that accountability may lessen the identification with and cohesion of the ingroup, which is involved in the intergroup conflict.
In this experiment we chose the behaviour of the riot police during a simulated
riot which was part of the training of riot police officers, to test our predictions
concerning the effects of accountability. In these 'riots', one group of police officers
played the role of riot police, and the other group played the role of demonstrators,
as will be explained more elaborately in the Method section. The party that plays
the role of the police is probably the easiest to make accountable for their actions
beforehand, since they get instructions on how to handle the situation, which may
be accompanied by the setting of appropriate behavioural norms and the announcement that excessive use of violence will not be tolerated. The role of the other party
demands a more provocative attitude, after all, they are assigned to create a situation
in which riot police action is warranted.
Summarizing,we advance the following hypotheses:
(1) Perceived accountability will increase public self-awareness, i.e. it will heighten
the concern with the opinion of the accountability forum.
(2) Perceived accountability will increase the conformity to the norms set by the accountability forum, which advocate moderation of the use of violence.
(3) In view of the anti-violent norm, accountability will lessen intergroup diflerentiation,
and will reduce mutual hostility.
(4) Individual accountability will lessen ingroup cohesion and ingroup identijication.
METHOD
Subjects
A field experiment was conducted at the Dutch Riot Police Academy. This Academy
consists of two Departments, located at different sites. From each Department, two
platoons participated in the study.'
'Of the 145 participants, 110 were men and 35 were women, their ages varying
between 21 and 42 years with a mean of 25.6. They had on the average 4.3 years
of service, with a range of 1-19.
Participants came from various local Dutch Municipal Police Forces. They were
assigned to one of the two Departments according to its geographical distance to
' In all, we recruited 167 participants from four different platoons; 145 of these belonged to the Municipal
Police, and 20 belonged to the State Police. It turned out that these two groups responded quite differently
on all dependent measures. However, due to the relatively small number of State Police participants,
it was not feasible to treat the kind of police force as an extra independent variable. Therefore, we
decided to delete the data of State Police participants.
254
M . B. R.Kroon
their home town. Apart from geographical considerations, respondents were randomly assigned to platoons by the police authorities.
Design
We used a 2 x 2 factorial design with accountability and Department as factors.
Although the focus of the experiment is on effects of accountability, the fact that
the Riot Police Academy has two Department made it necessary to investigate possible Department effects. The time when measures were made constituted a withinsubject factor. Respondents filled in a questionnaire at the beginning (pre-test) as
well as at the end of the training (post-test). These questionnaires were not identical,
i.e. not all items were retained in every questionnaire. This will be explained further
in the Dependent measures section. In Table 1 the design is presented schematically.
Table 1 .
Design
Accountability
Department A
Department B
No accountability
Pre-test
Post-test
Pre-test
Post-test
platoon 1
platoon 3
platoon 1
platoon 3
platoon 4
platoon 2
platoon 4
platoon 2
Accountability munipulation
The manipulation was introduced twice, in order to enhance its impact. It was administered at the beginning of the training (first week) and again just before the riot
simulation (third week). It consisted of the announcement made by the trainers,
and repeated by the experimenter, that participants would be held individually
accountable for their actions, and that use of violence should be avoided as much
as possible. It was stressed that after the action there would be a team meeting
in which the trainers would evaluate the behaviour of all group members, using
as a standard that use of violence should be moderate and suited to situational
demands.
We deliberately asked the trainers to hold participants accountable, because
trainers, unlike for instance platoon o r section commanders-although they were
not really part of the ingroupseemed to be seen by participants as more experienced
colleagues, representing an attractive reference group. Trainers consistently stressed
the ‘we’ feeling and behaved in a cordial and fraternal way, e.g. allowing themselves,
despite differences in rank, to be called by their first name. As members of an attractive
reference group, trainers seem to have a right to define ingroup norms. Other authorities, however, such as the platoon and section commanders, were rather seen as
inexperienced outgroup members, who do not really know what it is like to face
a crowd. Therefore, accountability announced by the section commanders would
at best stimulate compliance to a powerful outgroup. Conversely, if accountability
were introduced by the trainers, it would induce acceptance of norms set by those
well acquainted with realities of the riot situation.
Because we intended to test the effect of accountability onpoEice action, the trainers
Riot police, accountability and group norms
255
as well as the experimenter specified that respondents would be held accountable
only when playing the role of riot police and not when playing demonstrators. However, circumstances did not allow us to announce accountability only for those who
immediately afterwards would play the role of riot police. The announcement of
accountability could be made only at the beginning of the whole simulation, when
one section would be playing riot police, and the others playing demonstrators.
In the middle of the simulation, when the two sections changed roles, there was
neither time nor place to announce accountability again. Therefore, accountability
was announced at the beginning of the simulation to both sections together, irrespective of the role they would be playing first. This means that the accountability manipulation may have had a stronger impact on participants who played the police role
first than on those initially playing the demonstrator role. Thus, overall effects of
accountability may have been weakened.
Procedure
The experimentwas conducted during the 4-week training period of riot police officers
which took place at both Departments at the same time. Each Department trains
one platoon per training period. So, every time, two platoons are being trained
simultaneously. Since we needed four platoons to complete the design, we conducted
the experiment in two subsequent training periods.
In the first period we administered the accountability manipulation in Department
A, while Department B served as a control. In the second period (about a month
later) this procedure was reversed. The members of the two platoons that participated
in the second period in which the experiment was conducted, did not know the
experiment had been conducted before. The platoons that participated consisted
of two sections, who worked separately most of the time. In this way two different
groups were formed within platoons.
The study was introduced as an investigation of group development and impression
formation. Participants were assured that their responses would be treated confidentially. They responded anonymously, i.e. under a code which precluded tracing their
names.
In the first week of the training we administered the accountability manipulation
for the first time. After that, respondents filled in a pre-test (see Dependent measures
below). The experimentalphase took place about 2 weeks later, when a riot simulation
exercise was carried out. Before the action started, the accountability manipulation
was administered for the second time. In this simulation the two sections of a platoon
acted against each other. It consisted of two parts: in the first part one section
played the role of police and the other section acted as demonstrators; after that
there was a short break to change roles (and clothes), and then the second part
was enacted.
Some days later, participants completed a questionnaire of post-measures (see
Dependent measures). We did not debrief the respondents immediately, in order
to prevent them informing future respondents about the aim of the experiment.
Instead, we gave them the opportunity to obtain a written report with full information
about the experiment upon completion of the study. Only one respondent used this
opportunity. A popular version of the research report (Kroon and Richelle, 1991)
has also appeared in one of the journals of the Dutch police, which is read by
most police officers.
256
M. B. R. Kroon
Dependent measures
We made measurements at two different points in time. These repeated measurements
were made at the beginning and end of the training period. Questionnaires with
seven-point, or in a few items five-point, Likert-type response scales were used.
The pre-test consisted of clusters of items measuring group cohesiveness, social
identification, evaluation of the own group and the other group, norms concerning
the use of violence, and a manipulation check.
The post-test contained all items included in the pre-test. Apart from this, some
additional items concerning public self-awareness, the evaluation of the action of
the own group and the other group (on the dimensions :control of violence, legitimacy
of the use of violence and competitiveness), and manipulation checks were included.
Furthermore, some items measured participants’ own norms concerning the use of
violence, and others measured the norms they thought their trainers used.
Since the two questionnaires were not identical-that is, their items only partially
overlapped-factor analysis was not feasible. Therefore, we selected all items that
figured in both pre- and post-test, and clustered these into different groups according
to their theoretical meaning. These groups were then subjected to reliability analysis,
and items were excluded from the groups if necessary. This procedure yielded four
identical clusters for the pre- and post-test: aggressiveness of the own section and
of the other section, and cohesion of the own section and of the other section (see
Appendix). Furthermore, eight additional clusters for the post-test only were found,
six of them concerning the evaluation of the own section and the other section,
respectively, on the dimensions :control of violence, appropriateness of the action,
and competitiveness. The remaining two clusters represented public self-awareness
and trainers’ norms concerning the use of violence, as perceived by trainees (see
Appendix). In further analyses these clusters were used. Items were analysed separately only when they did not fit into any cluster.
RESULTS
As is often the case in field studies, the realities of the research context make strict
control difficult. In this Riot Police Academy, circumstances did not allow us to
carry out the research design exactly the way we planned it.
This section will start with a discussion of the unforeseen variables that arose,
and the way we dealt with these and changed our plan of analysis. Secondly, we
will present the results.
The unforeseen
Unfortunately, in Department B, circumstances prevented us from executing the
experimental manipulation adequately. In general, things proceeded less smoothly
in this Department. This was probably due to the fact that a fundamental reorganization was going on there at the time of the experiment. The reorganization particularly
affected trainers, whose cooperation was vital for the execution of the manipulation.
Under the circumstances, trainers in Department B were less prepared to satisfy
the demands of the experiment than trainers in Department A. Consequently, in
Department B we were obliged to ask section commanders instead of trainers to
carry out the accountability manipulation. As already mentioned in the Method
Riot police, accountability andgroup norms
257
section, section commanders, unlike trainers, are seen as inexperienced outgroup
members. Therefore, this was a crucial modification of the manipulation of accountability. Not surprisingly, then, as we will see next, the two versions of the accountability
manipulation each had quite a different impact.
Table 2. Awareness of accountability
Department A
Department B
All
Accountability No accountability Accountability No accountability
n =31
21.4%
n = 14
n=5
9.7%
3.4%
n=13
9.0%
n=63
43.5%
Not aware
n=7
4.8%
n=21
18.6%
n=29
20.0%
n=19
13.10%
n=82
56.5%
All
n = 38
26.2%
n= 41
28.3%
n=34
23.4%
n = 32
22.1%
Aware
Although the manipulation of accountability had been effective in both Department
A (chi-square(,) 16.22, p < 0.001) and Department B (chi-square(,) 4.35, p < 0.05),
the effect in each Department was in the opposite direction. In Department A, as
expected, in the accountability condition most respondents reported taking account
that they would be evaluated after the action, while in the control condition most
respondents reportedly did not take the evaluation into account. In Department
B the reverse pattern of results was obtained (see Table 2).
Apart from the fact that the manipulation was carried out differently and inadequately in Department B, and apparently had the opposite effect there, we also have
some evidence for a general reactance effect in this Department. Not only were
the trainers less willing to cooperate, but also the non-response rate was considerably
higher in Department B than in Department A for both questionnaires. For each
respondent we computed a response rate score for each questionnaire separately
by adding up his or her missing values. Thus, two response rate scores were obtained.
A MANOVA with Department and accountability as factors yielded significant
(3,160)= 6.47, p < 0.001;
differences between Departments on both scores. (Fmultivariate
Funivariate(1,162) = 7.01, p < 0.01; Funivariate(1,162)
= 11.29, p < 0.01). There were no effects
of accountability, nor were there any interaction effects.
Not only was accountability announced by ‘outgroup’ section commanders instead
of ‘ingroup’ trainers in Department B, but the norms prevailing in this Department
were also less congruent with the experimental manipulation, which advocated moderation of the use of violence, than those in Department A. In Department B, respondents more strongly opposed the idea that the real task of the riot police was to
prevent disturbances by posing a threat instead of acting (F1,14,1= 4.03, p < 0.05).
They were also less prepared to negotiate with the other party if possible
(F(1,141)= 5.90, p < 0.05). Finally, when asked about the norms of their trainers, trai(4,70) = 3.77, p < 0.01). In
nees in each Department responded differently (Fmultivariate
Department B, trainees to a larger degree thought their trainers expected them to
let the other party feel who was strongest (F,,73)= 13.81, p < 0.001) and to retaliate
(F(1,71)= 3.54, p < 0.07). Thus, in Department B a slightly less ‘peaceful’ norm con-
258
M. B. R.Kroon
cerning the use of violence appears to have governed the behaviour of trainees than
in Department A. Consequently, the accountability manipulation may have created
a relatively new situation in Department B, while in Department A it meant only
reinforcing current practice. In Department B a form of reactance, expressed by
the declining of norms introduced by the authorities, and reluctance to cooperate
with the experiment, may have arisen because accountability was announced by
the ‘wrong’ persons, advocating an uncommon norm. Furthermore, since trainers
in Department B overtly showed their reluctance to cooperate with the experiment,
the chances that norms introduced by the authorities would become accepted as
ingroup norms by trainees, became very slight indeed. Therefore, caution is warranted: possible effects of accountability in Department B cannot be interpreted
in the same way as in Department A.
Considering these differences in organizational climate between Departments, an
analysis using the original experimental design does not seem feasible. The unforeseen
circumstances made us change our plan of analysis.
To begin with, we examined the effect of accountability in Department A separately,
because there the manipulation was executed properly. In this Department there
was hardly any difference between the accountability and the no-accountability condition on the dependent variables of interest, which may be accounted for by several
factors. In the first place, accountability was only one of the many factors that
affected participants in this natural context. Secondly, the norms that favoured control instead of contestation that were introduced in the accountability condition
closely resembled the real norms of this Department. Thus, our control condition,
in which neither accountability nor norms were introduced experimentally, may in
fact not have differed very much from the accountability condition. Thirdly, the
impact of accountability may have been weakened by the fact that, during the riot
simulation, accountability was announced only once, irrespective of the role respondents would be going to play first. A stronger effect might have been obtained if
it had been possible in the experimental condition to make all groups of respondents
accountable directly before they played the police role. Finally, the reduced sample
size may have played its part.
Because of the problems with our experimental manipulation mentioned above,
we tried to gain additional insight in the impact of accountability in a different
way. We hypothesized that although the manipulation of accountability may not
have worked well enough, it might still be possible that in both Departments some
people feel highly accountable and others feel hardly accountable. This led us to
perform an ‘internal’ analysis, using the data of both Departments, in which we
used awareness of accountability instead of manipulated accountability as a factor.
In this analysis we examined the relationship between awareness of accountability
and the attitudes and norms of respondents during the action.
The primary goal of this analysis is to generate hypotheses, rather than to test
them.
The internal analysis: awareness of accountubility
Awareness of accountability was measured on a two-point scale, by asking respondents whether they took into account the evaluation which would be held after
the action.
We divided respondents into two groups according to their responses on this
Riot police, accountability andgroup norms
259
item. This resulted in the design presented below (Table 3). The same repeated measurements are included here as in the experimental design, presented in Table 1. As
Table 3 shows, there are unequal cell frequencies. In Department A the number
of respondents who feel accountable is somewhat larger than the number of those
who do not (45 against 34). In Department B the difference between the number
of respondents who do and who do not feel accountable is in the opposite direction
and much larger (18 against 48). In order to avoid contamination of accountability
and Department effects, we chose the regression approach of MANOVA, with awareness of accountability and Department as factors. In this way, an effect is adjusted
for all other effects in the model.
Table 3. Design of the internal analysis
~
High accountable
Low accountable
Dept. A
Dept. B
Dept. A
Dept. B
All
n=45
n=18
n = 34
n=48
n = 145
When pre-test scores of specific items were supposed to influence post-test scores,
they were used as covariates. This will be indicated at the results concerned.
Awareness of accountability
Public self-awareness. It was expected that perceived accountability will increase
public self-awareness (hypothesis I). On two measures of public self-awareness a
multivariate significant effect of awareness of accountability is found (F(2,70) = 3.61;
p < 0.04), with a univariate significant effect on the extent to which participants
report to take their fellow group members’ views into account during the action
(F(1,71) = 7.16; p < 0.01). As can be seen in Figure I, the more participants feel
accountable, the more they are concerned with the opinion of their fellow group
members. Thus, in accordance with hypothesis 1, awareness of accountability is
related to a heightened public self-awareness.
Use of violence. As expected, awareness of accountability leads to less acceptance
of the use of violence during confrontation exercises. Acceptance of violence was
measured by asking subjects’ opinions about the statements: ‘An exercise should
preferably be carried out without violence’ and ‘The real task of the riot police
is to threaten, not to act’. We analysed the post-test scores as well as the post-test-pretest difference scores of these two items together (multivariate: F(4,138) = 3.93;
p < 0.01). The former item yielded a marginally significant univariate effect of awareness of accountability (F(1,141) = 3 . 6 8 ; ~< 0.06). Figure 2 shows that when awareness
of accountability was high there tended to be more agreement with the statement
that confrontation exercises should preferably be carried out without the use of
violence, than when awareness of accountability was low.
However, as can be seen in Figure 2, unexpectedly, participants who feel accountable tend to accept more violence in actual practice than in training. Conversely,
participants who do not feel accountable consider as much violence acceptable in
practice as those who do feel accountable, but even more in training settings (multivar-
260
M. B. R.Kroon
~
high
low
awareness of a c c o u n t a b i l i t y
measures a r e on a scale from l = l o w t o 5 = h i g h
Figure 1. Public self-awareness
iate: F(3,69) = 3.10; p < 0.04; univariate: F(1,71) = 4.78; p < 0.04). Thus, awareness
of accountability is only in training settings related to a more moderate attitude
towards the use of violence. This may be due to the fact that the non-violent norm
which was stressed before the action started, and was reinforced by accountability,
exclusively dealt with behaviour in a training setting. Besides, the introduction of
accountability, in which it was stressed that violence should be used only when
the situation demanded, may have made respondents more realistically appraise the
necessity of the use of force in different settings. That is, the chances of being confronted with really violent demonstrators are probably higher in practice than in
training. Consequently, real-life confrontations may on the average require a firmer
approach than confrontation exercises. It should be noted that acceptance of violence
in practice was moderate, never high. Anyhow, these findings indicate that it may
be worthwhile to examine the effect of anticipated accountability in combination
with a non-violent norm not only in training settings, but in actual practice as well.
Group cohesion. In hypothesis 4 we predicted that accountability would lessen
group cohesion. All five measures concerning group cohesion were analysed by a
2 x 2 (awareness of accountability x Department) MANCOVA, controlling for
initial group cohesion. In this way, possible differences in group cohesion existing
at the beginning of the study were adjusted for. The results are displayed in Table
4,showing a multivariate significant effect of awareness of accountability. Subsequent
univariate tests indicated that, when group members expect to be held accountable,
as predicted, they perceive their fellow group members as less sympathetic than
when they do not feel accountable (see Table 4). However, contrary to our prediction,
final reported attraction to the ingroup is the same in both conditions. Also, there
are no differences in reported ingroup identification.
Riot police, accountability and group norms
26 1
LEGEND
W2 use
of violence in training
use of violence against civilians
high low
awareness of accountability
measures a r e on a scale f r o m l=low t o 5=high
Figure 2. Use of violence
Table 4. Group cohesion
Effect: covariate: initial
Effect: awareness of
cohesion
accountability
Multivariate F Univariate F Multivariate F Univariate F
(5,136)
(1,140)
(5,136)
(1,140)
Dependent measures
General attraction to ingroup
Willingness to cooperate in
future
Appropriateness own
group’s action
Difference between
appropriateness other and
own group’s action
Likeableness of fellow group
members
* p < 0.05; * * p < 0.01;
5.35*
10.19**
ns.
n.s.
ns.
2.81*
n.s.
ns.
1.61**
6.12***
12.88**
13.28***
***p < 0.001.
Thus, perceived accountability has no effect on ingroup identification, and there
is some ambiguity in the results on ingroup cohesion. Taking a closer look at these
results it appears that only participants’ judgements of their fellow group members,
and not their evaluations of their group as a whole, are affected by the feeling of
accountability. Besides estimates of likeableness, also estimates of the appropriateness
and aggressiveness of the action of fellow group members are less extreme when
respondents feel accountable than when they do not feel accountable (41,140) = 7.67;
M. B. R. Kroon
262
< 0.01; F(l,I41) = 14.59; p < 0.001). As can be seen in Figure 3, when participants
feel accountable, they judge their fellow group members to be less extremely likeable
and unaggressive, and their own group’s action to be less extremely appropriate.
Thus, though not directly lowering general feelings of ingroup cohesion, perceived
accountability positively induces a moderation of the evaluation of one’s fellow group
members.
p
LEGEND
“T
EEZl
attractiveness
appropriateness
EGI
aggressiveness
high
low
a w a r e n e s s of a c c o u n t a b i l i t y
scores
range f r o m l = l o w
t o 7=high, adjusted
f o r initial cohesion
Figure 3. Group cohesion and aggression
Intergroup differentiation. In accordance with hypothesis 3, when participants
feel accountable, their judgements of their own and the other group are in general
more similar than when they do not feel accountable (see Figure 4). In the first
place, this applies to judgements of control of violence and of aggressiveness as
measured on the post-test (Fmu,tivariate(4,138)
= 5.66; p < 0.001; univariate: (F(1,141)
= 4.56;
p < 0.04; F(1,141) = 14.63,p < 0.001). Secondly, less differentiation is found for estimates of the appropriateness of the own group’s and the other group’s action. Controlling for initial group cohesion, a MANCOVA on measures of estimated cohesion
and appropriateness of the own group’s and the other group’s action resulted in
a multivariate significant effect of awareness of accountability (F(5,,3h)
= 6.12;
p < 0.001; see Table 4). Respondents who felt accountable made no difference between
the appropriateness of their own or the other party’s action (MdiR
= 0.02), while
those who did not feel accountable thought the action of their own group was more
appropriate than the action of the other group (MdiR
= -0.83 F(,,,40)
= 12.88;
p < 0.001; see Figure 4). Concluding, as predicted, perceived accountability is related
to a reduced intergroup differentiation.
Riot police, accountability and group norms
263
LEGEND
BZi controlling
-1.5
of violence
I
J
high
a w a r e n e s s of
low
accountability
measures are o n a scale from -6
to 6
Figure 4. Intergroup differentiation
DISCUSSION
Differences in organizational climate between the two departments participating in
this study precluded a straightforward experimental test of the hypotheses. This
led us to explore the role of awareness of accountability in an internal analysis.
Since awareness of accountability categories were achievedpost hoe, such an analysis
yields only correlational information, and strictly speaking does not permit causal
inferences to be made. However, although the data do not allow for a rigorous
test of the hypotheses, they do provide a rich source for hypothesis generation.
The tests suggest that awareness of accountability is related to several aspects
of the way group members perceive an intergroup conflict. As predicted, participants
who feel accountable have a heightened public self-awareness, evaluate their own
group more moderately and engage in less intergroup differentiation.
The results on group cohesion, however, are contradictory. In accordance with
our hypothesis that perceived accoqntability will lower group cohesion, participants
who feel accountable report to like their fellow group members a little less as compared to the other participants. This fits the pattern that participants’ judgements
of their fellow group members are generally less extremely positive when they feel
accountable. Concerning the evaluation of fellow group members, perceived accountability appears to reduce a positive ingroup bias. However, attraction to the group
as a whole and ingroup identification do not suffer when participants feel accountable.
Thus, we cannot unconditionally maintain that perceived accountability will lower
group cohesion and identification. An explanation may be that part of the meaning
of accountability in this situation is that participants feel accountable to their group,
i.e. that social identity is salient. Especially when trainers, who are seen as more
experienced ingroup members, ask for an account of behaviour, this explanation
264
M. B. R.Kroon
seems to be valid. Besides, accountability may have been felt by participants as
a common fate, which served to confirm the feeling of unity within groups. Therefore,
attraction to the group in general and ingroup identification may not have been
weakened by perceived accountability.
Taken together, these results suggest that awareness of accountability can play
a role in reducing the likelihood of escalation of conflicts between police and demonstrators. We cannot reach this conclusion without restraint, however, because we
are dealing with correlational evidence, and for additional reasons as follow.
To begin with, the exact meaning of ‘awareness of accountability’ is not entirely
clear. It can be determined by different factors, such as personality, the culture
of the organization, and of course by our manipulation of accountability. Our data
suggest that the latter two factors, either alone or in combination, have been important. That is, many more people expressed awareness of accountability in Department
A than in Department B, where the manipulation was executed differently.
Secondly, which role do the norms set by the authorities play in causing awareness
of accountability? What difference will it make whether or not these norms match
the existing culture of the organization and/or the norms of trainees? On the one
hand, one can presume that congruency between authority’s and trainee’s norms
will result in more awareness of accountability. That is, having to render account
gives the opportunity to display behaviour guided by norms that are also favoured
by the authorities, and thus possibly will be positively sanctioned. On the other
hand, it seems equally probable that incongruency will heighten awareness of accountability, because of the cognitive processing that new and uncommon messages tend
to induce (cf. Moscovici, 1980 and Nemeth, 1986, on the increased cognitive processing brought about by deviant minority points of view).
In this experiment, congruency between the norms introduced by the authorities
and norms already existing within the organization (as existed in Department A)
seems to have heightened the awareness of accountability, while incongruency (as
existed in Department B) rather seems to have caused a form of reactance. This
raises a third problem. Since the norms introduced by the authorities fit in better
with the organizational climate of Department A than of Department B, as explained
earlier, it is not quite clear how one should interpret ‘awareness of accountability’
for respondents from Department B. However, since no interactions between awareness of accountability and Department are found, this need not be very problematic.
Apparently, when respondents in Department B reported being aware of accountability, they also knew the non-violent norms the authorities propagated.
Finally, we face the question whether a heightened sense of accountability caused
mere strategic responses, or meant acceptance of norms promoting moderation of
violence, as it were adopting these as ingroup norms. This points to the importance
of specifying whether the forum of accountability be an ingroup or an outgroup.
Accountability to the ingroup should stimulate behaviour according to ingroup
norms. Conversely, accountability to an outgroup may induce actual acceptance
of outgroup norms only when the norms of the two groups are matched. When
ingroup and outgroup norms do not match, behaviour according to outgroup norms
will result only if the outgroup is more powerful. In the latter case, one should
merely expect short-term compliance to the outgroup norms. These norms will not
be accepted by the ingroup, and consequently they will not influence future behaviour.
In the present study the reactance we found in Department B may have been
Riot police, accountability and group norms
265
caused by the ‘wrong’norms being advocated by the ‘wrong’people. As mentioned,
in this Department the non-violent norms were introduced by section commanders,
who were regarded as an outgroup. Although formally more powerful than the riot
police officers, section commanders possessed limited actual power. That is, in the
heat of the action, for example, it was not exceptional that orders of section commanders were ignored, simply by pretending the communication devices did not
work. Thus, it is not very surprising that in Department B the non-violent norms
advocated by section commanders generally were rejected. In Department A these
norms were, as intended, announced by respected ‘ingroup’ trainers. Therefore, we
may expect participants in this Department not only complied with these norms,
but also adopted them as their own.
In sum, perceived accountability may lower the chances of escalation of an intergroup conflict, provided that non-violent norms are salient. In order to achieve
internalization of such norms, rather than temporary compliance or even reactance,
it is important that those who ask for an account of behaviour constitute an attractive
and credible reference group. Otherwise, they will rather be seen as priers, who
do not understand what riot police action is about, than as counsellors, whom one
had better listen to.
ACKNOWLEDGEMENTS
The author wishes to thank Paul Richelle for his assistance in conducting the experiment, and Dr Russell Spears and an anonymous reviewer for their insightful comments on earlier drafts of this manuscript.
REFERENCES
Brown, R. (1988). Group Processes: Dynamics Within and Between Groups, Blackwell, Oxford.
Diener, E. (1980). ‘Deindividuation: the absence of self-awareness and self-regulation in group
members’. In: Paulus, P.B. (Ed.), The Psychology of Group Znjluence, Erlbaum, Hillsdale,
NJ.
Festinger, L., Pepitone, A. and Newcomb, T. (1952). ‘Some consequences of deindividuation
in a group’, Journal of Abnormal and Social Psychology, 47,382-389.
Ginneken, J. van (1989). Crowds, Psychology and Politics 1871-1899. Dissertation, University
of Amsterdam.
Johnson, R.D. and Downing, L.L. (1979). ‘Deindividuation and valence cues: effects on prosocia1 and antisocial behavior’, Journal of Personality and Social Psychology, 37, 1632-1 538.
Kroon, M.B.R. and Richelle, P. (1991). ‘ME-er moet zich bewust zijn van doe1 optreden
(Riot police officers should be aware of aim of action)’, Algemeen Polititeblad, 8, 11-13
(in Dutch).
Mann, L., Newton, J.W. and Innes, J.M. (1982). ‘A test between deindividuation and emergent
norm theories of crowd aggression’, Journal of Personality and Social Psychology, 42, 260272.
Moscovici, S. (1980). ‘Toward a theory of conversion behavior’. In: Berkowitz, L. (Ed.),
Advances in Experimental Social Psychology, Academic Press, New York.
Nemeth, Ch.J. (1986). ‘Differential contributions ofmajority and minority influence’, Psychological Review, 93,23-32.
Prentice-Dunn, S. and Rogers, R.W. (1982). ‘Effects of public and private self-awareness
on deindividuation and aggression’, Journal of Personality and Social Psychology, 43,503513.
266
M. B. R. Kroon
Prentice-Dunn, S. and Rogers, R.W. (1983). ‘Deindividuation in aggression’. In: Green, R.G.
and Donnerstein, E.I. (Eds), Aggression. Theoretical and Empirical Reviews, Vol. 2, Academic Press, New York.
Prentice-Dunn, S. and Rogers, R.W. (1989). ‘Deindividuation and the self-regulation of behavior’. In: Paulus, P.B. (Ed.), Psychology of Group Influence, 2nd edn, Erlbaum, Hillsdale,
NJ.
Rabbie, J.M. (1990). ‘Agressieveconflicten tussen individuen en groepen: een sociaal-psychologische benadering (Aggressive conflicts between individuals and groups: a social-psychological approach). In: Van Hooff, J.A.R.A.M., Benthem van de Berg, J. and Rabbie, J.M.
(Eds), Oorlog. Multidisciplinaire Benaderingen ( War. Multidisciplinary Approaches), Stubeg
(In Dutch).
Rabbie, J.M. and Goldenbeld, Ch. (1988). ‘The effects of modelling and accountability on
aggressive intergroup behaviour’. Paper presented at the World Congress of the International
Society for Research on Aggression, July 1988, Swansea, UK.
Rabbie, J.M. and Visser, L. (1984). ‘Deindividuatie en de ontwikkeling van normen in groepen
toeschouwers, een veldexperiment (De-individuation and the development of norms in an
audience)’. In: Jaspars, J. and Van der Vlist, R. (Eds), Sociale Psychologie, IV, Samenleving
en Individuen (Social Psychology, vol. IV, Society and Individuals), Van Loghum Slaterus,
Deventer (In Dutch).
Reicher, S.D. (1984a). ‘Social influence in the crowd: attitudinal and behavioral effects of
de-individuation in conditions of high and low group salience’, British Journal of Social
Psychology, 23, 341-350.
Reicher, S.D. (1984b). ‘The St. Pauls’ riot: an explanation of the limits of crowd action in
terms of a social identity model’, European Journal ofsocial Psychology, 14, 1-21.
Reicher, S. and Potter, J. (1 985). ‘Psychologicaltheory as intergroup perspective: a comparative
analysis of “scientific” and “lay” accounts of crowd events’, Human Relations, 38, 167-1 89.
Rogers, R.W. and Prentice-Dunn, S. (1 981). ‘Deindividuation and anger-mediated interracial
aggression: unmasking regressive racism’, Journal of Personakty and Social Psychology,
41.63-73.
Spears, R., Lea, M. and Lee, S. (1991). ‘De-individuation and group polarization in computermediated communication’, British Journal of Social Psychology, 29, 121-1 34.
Tajfel, H. (Ed.) (1978). Diferentiation Between Social Groups: Studies in the Social Psychology
of Intergroup Relations, Academic Press, London.
Turner, J. (1982). ‘Towards a cognitive redefinition of the social group’. In: Tajfel, H. (Ed.),
Social Identity and Intergroup Relations, Cambridge University Press, Cambridge.
Turner, R.H. and Killian, L.M. (1972). Collective Behavior, Prentice Hall, Englewood Cliffs,
NJ.
Zimbardo, P.G. (1970). ‘The human choice: individuation, reason and order versus deindividuation, impulse and chaos’. In: Arnold, W.J. and Levine, D. (Eds), Nebraska Symposium
on Motivation, Vol. 17, University of Nebraska Press, Lincoln, NB.
Riot police, accountability and group norms
267
APPENDIX: RELIABILITY OF THE PRE- AND POST-TEST FACTORS
Evaluation own group Evaluation other group
Alpha
Alpha
Pre-test
Post-test
Pre-test
Post-test
Aggression factor
aggressive
active
hostile
undisciplined
irritable
harmful
0.72
0.80
0.85
0.78
Attraction factor
nice to work with
friendly
fraternal
unanimous
honest
0.62
0.78
0.69
0.79
Post-test
Alpha
Evaluation action,
own group
Evaluation action,
other group
Control of violence
not excitable
collected
well-considered
moderate
avoid escalation
negotiate if possible
do not be tough
control the situation
0.66
0.80
Appropriateness
good
responsible
0.66
0.59
Competitiveness
competitive
display strength
retaliate
0.72
0.7 1
Public self-awareness
concerned with evaluation by the trainer
concerned with opinion of team members
concerned with opinion of other party
0.59
Trainers’ norms concerning the use of violence
display strength
retaliate
0.72