Elaborating the police perspective: The role of perceptions and experience in the explanation of crowd conflict

European Journal of Social Psychology

Eur. J. Soc. Psychol. 39, 991–1001 (2009)

Published online 17 February 2009 in Wiley InterScience

(www.interscience.wiley.com) DOI: 10.1002/ejsp.603

*E

C

Research article

Elaborating the police perspective: The role of perceptions and experience inthe explanation of crowd conflict

Correspondence to: Gabriele Prati, Dipartimen-mail: [email protected]

opyright # 2009 John Wiley & Son

GABRIELE PRATI* AND LUCA PIETRANTONIDepartment of Education, University of Bologna, Italy

Abstract

Classical theories of crowd behaviour view crowd conflict as deriving from the pathology of the crowd itself. Recent

developments in crowd psychology as the elaborated social identity model (ESIM) conceptualize crowd behaviour as a

dynamic intergroup process between demonstrators and police. The present study assessed exposure to crowd conflict,

adherence to classical views of crowd behaviour, public order policing methods and attributions of responsibility for crowd

conflict among 352 Italian police officers. Results showed that exposure to crowd conflict was related to adherence to classical

views of crowd, which, in turn, was related to ‘bad practices’ of public order policing and to system-justificatory

attributions. Overall, these results offer support and extend the police perspective within the ESIM model. Practical

implications for public order policing strategies and training are also discussed. Copyright# 2009 JohnWiley& Sons, Ltd.

Classical theories of crowd behaviour conceptualized conflict as deriving from the pathology of the crowd itself (e.g.

Allport, 1924; Le Bon, 1895). Recent development of classic models of crowd behaviour into de-individuation theory

(Prentice-Dunn & Rogers, 1982, 1989; Postmes & Spears, 1998) assumed that crowd members lose control of antisocial

impulses as result of two processes: (a) Loss of private self-consciousness due to heightened levels of group cohesiveness

and arousal; (b) loss of public self-consciousness, due to anonymity, diffused responsibility and a lack of accountability.

Stott and Reicher (1998) criticized this model of crowd behaviour and they put forward a different model based on the

social identity theory (SIT). The elaborated social identity model (ESIM; Drury & Reicher, 2000; Reicher, 1996, 2001;

Stott & Reicher, 1998) took into account, in explaining crowd behaviours, the context in the form of actions of the other

group, relationship between categories in context, historical situatedness and intergroup relations. The ESIM of crowd

behaviour is derived from the social identity tradition and is based on principles from SIT (Tajfel & Turner, 1979) and self-

categorization theory (SCT: Turner, 1987). The ESIM suggested that crowd members share a common social identity

defined on the basis of the relationship between categories in context and tend to act accordingly. Furthermore the ESIM

conceptualized social identity as an ongoing process depending on intergroup relation: Crowd members’ norms and

identities that specify normative behaviours can change over time as a function of interaction with an out-group. On the

other hand, intergroup interaction can modify social identity and normative behaviours. According to Drury, Stott, and

Farsides (2003) crowd conflict can arise where there is (1) a discrepancy between theway in which identity and behaviours

are defined by crowd members themselves and by out-group members (for example the use of indiscriminate coercive

force could be perceived legitimated by police and illegitimate by demonstrators); (2) empowerment or confidence in

group ability to impose its understanding of the situation on the out-group members (e.g. demonstrators could feel the

power to collectively resist police action). Thus the dynamics of legitimacy and power between groups are crucial in

understanding conflicts. However, when it comes to study crowd conflict the focus was likely to ignore police perspective

due to various difficulties (see Drury & Stott, 2001, for a full discussion of these issues). More recently, studies

to di Scienze dell’Educazione, Universita di Bologna, via Filippo Re, 6-40126 Bologna, Italy.

s, Ltd.

Received 27 May 2008

Accepted 16 November 2008

992 Gabriele Prati and Luca Pietrantoni

investigating crowd dynamics have begun to take an interest in the police perspective (Cronin & Reicher, 2006; Drury

et al., 2003; Reicher, 1996; Stott & Drury, 2000; Stott & Reicher, 1998).

DEVELOPING THE POLICE PERSPECTIVE

According to Reicher (1996) research in crowd behaviour should include police perspective in order to gain a better

understanding of its dynamics. Stott and Reicher’s (1998) interview study identified three main themes related to the police

theories of the crowd. The first theme concerned the composition of crowds. On the one hand, police officers recognized

that crowds are typically heterogeneous and generally representative of society in general. On the other hand, for these

officers crowd was divided into minority and majority. The minority is powerful and violent while the majority is

characterized as being peaceful and composed by ordinary people that in a crowd context become incapable of judgment

and hence particularly prone to manipulation. According to Stott and Reicher (1998) the description of minority and

majority resembled the two classic theories of crowd action. The explanation of minority members’ behaviour as a

function of individual pathology mirrors Allport’s approach. In the case of the majority, the explanation in terms of crowd

context leading ordinary people to become pathological mirrors Le Bon’s ideas.

The dynamic of a powerful violent minority and a susceptible majority leaded to the second theme concerning the

homogeneous threat posed by crowds. Once people get together in the crowd, crowd members are all potentially

dangerous. Given this sense of danger about crowd conflict, in general, police officers express concern for their own safety.

The last theme is about public order policing. Fear and danger experienced by police officers have three major

implications. First, if crowd members were perceived to be all potentially dangerous, then in case of conflict it was

recommended strict control and quick intervention against crowds. Second, if crowd danger was perceived homogeneous,

then there were tactical reasons for treating a crowd as a single unit. Finally, police officers tended to deny any possible

responsibility for crowd conflict and to attribute it to outside forces. In other words, the violent and irrational nature of the

crowd was a sufficient explanation for conflict and there was no reason to address the role of the police tactics (e.g. use of

coercion, undifferentiated intervention) in the dynamic.

Drury et al.’s (2003) survey study involving 80 riot-trained officers from different United Kingdom forces explored the

validity and generalizability of the results about police perceptions and practices. The authors used a questionnaire

containing a series of items based on those main themes identified in Stott and Reicher’s (1998) interview study. Results

showed that police officers view football and demonstration crowds as heterogeneous and, at the same time, comprising

powerful minorities and labile majorities. Consequently, police officers recommended strict control and quick intervention

to prevent crowd conflict and did not see their own practices as responsible for the escalation of conflict. However police

officers did not endorse or reject the view that crowd members are all potentially dangerous and there are tactical reasons

for treating a crowd as a single unit. Finally, results showed the relation between perceptions and practices. More

specifically, the perception of a homogeneous threat partially mediated the relation of perception of dichotomous

composition to coercive policing perception of dichotomous composition led to the through the partial mediation of the

perception of a homogeneous threat. Moreover, the necessity of coercive intervention mediated the relation between the

perception of a homogeneous threat and the attributions of responsibility for crowd conflict to participants. Tactical

reasons for treating a crowd as a single unit and coercive policing predicted denial of responsibility for conflict. Police

perceptions of crowds and practices may sometimes become a self-fulfilling prophecy in that they may lead previously

distinct groups to unite in common hostility to the police. Research on demonstration riots (Reicher, 1996; Stott & Drury,

2000) confirmed this process found in qualitative and quantitative studies.

These studies on police perspective were developed among British police officers. We still do not know if this

perspective is unique to British police officers or it is a pan-European issue. One of the aims of this study is to make cross-

cultural validation of research in this domain.

According to Reicher (2001), the ideological view of crowd serves three functions: Denial of protester’s voice because

it has no sense, denial of responsibility of society and legitimacy of repression. At this point a question arises: Why police

officers endorse this view of crowd dynamics? One reason is that crowds indeed contain minorities inclined to instigate

widespread violent conflict (Stott, Hutchison, &Drury, 2001; Stott, Adang, Livingstone, & Schreiber, 2007; Stott &Drury,

1999, 2000). However, this fact does not justify the view of the crowd as irrational and prone to the influence of violent

minorities. A more plausible reason may be that ideology of the crowd is very common in popular culture (e.g.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 39, 991–1001 (2009)

DOI: 10.1002/ejsp

Exposure to crowd conflict and ESIM 993

McClelland, 1989) and thus in police officers. Another reason concerns police training based on ideological (e.g. Le

Bonian) models of the crowd (Cerrah, 1998).

In addition to these hypotheses, we argue that endorsing the view of crowds as suggestible and comprising violent

minorities could be considered a cynical attitude towards crowd members due to stress of crowd conflict. In other words,

the exposure to crowd conflict constitutes a ‘critical incident’ (Paton, 1997), in that it is a personal experience involving

actual or threatened death or serious injury. Research findings highlight that duty-related stressors among police officers

are related to cynical and suspicious attitude towards people (Robinson, Sigman, &Wilson, 1997). Moreover, exposure to

violence is a very negative job stressor and could contribute to depersonalization,1 a dimension of the burnout syndrome

(Maslach, Schaufeli, & Leiter, 2001) defined as a negative, insensitive and cynical attitude towards public or recipients.

Indeed depersonalization correlated strongly with more cynical attitudes towards the public (Stearns & Moore, 1993) and

predicted more positive attitude towards the use of violence and actual violent behaviour during a police action (Kop,

Euwema, & Schaufeli, 1999). In this regard, greater experience of crowd conflict should be linked to a higher level of

adherence to an ideological view of crowd resulting from stress related cynical attitude. Although Drury et al. (2003) did

not identify differences between those who had low and those who had high experience of both policing and participating

crowds, we argue that involvement in actual crowd conflict might be related to differences in adherence to ideological view

of crowd. Whereas personal experience of crowd events such as concert or sport events may be really pleasant,

involvement in crowd conflict is a potentially traumatic event. In this study, we aimed at verifying the effect of experience

of crowd conflict, while Drury et al. simply investigated experience of crowds per se.

In this study, we attempted to explore the validity and generalizability of the results from Stott and Reicher’s (1998) and

Drury et al.’s (2003) researches as well as test for the role of experience of crowd conflict. In this research, we involved a

sample of police officers that was larger and from a different country (Italy). As in Drury et al.’s (2003) study, we made two

main hypotheses. First, we predicted that police officers would endorse the view that (1) crowd composition is

heterogeneous and dichotomous; (2) crowd members pose a homogeneous danger; (3) in case of conflict strict control and

quick intervention are needed; (4) there are tactical reasons for treating the crowds as a unit; (5) police tactics are not

responsible for the development of crowd conflict. Second, we expected the following mediational processes (see

Figure 1): (1) Perception of a homogeneous threat mediates the effect of perceiving a dichotomous composition on

coercive policing; (2) the relation of homogeneous threat to denial of police responsibility is mediated by coercive policing

and tactical reasons for treating crowd as one; (3) perception of a dichotomous composition might be expected to predict

relatively more self- or system-justificatory attributions, indirectly by perceptions of homogeneous threat, endorsement of

coercive policing and tactical reasons for treating crowd as one.

In addition to Drury et al.’s (2003) study, it was expected that (1) given that the homogeneity of crowd danger is to be

related to homogeneity of police response and that perception of dichotomous crowd composition is to be related to

homogeneous threat of crowds (Stott & Reicher, 1998), perception of a homogeneous threat mediates the effect of

perceiving a dichotomous composition on tactical reasons for treating crowd as one; (2) police officers’ experience of

crowd conflict might affect perceptions of crowd events and crowd members in a dichotomous way (e.g. crowds as

suggestible and comprising violent minorities) and thus endorse the view that crowds pose a homogeneous threat. More

specifically, perception of dichotomous crowd composition mediates the relation of exposure to crowd conflict to

perception of a homogeneous threat.

METHODS

Participants

A total of 353 Italian police officers participated in the study. None refused to take part. We did not take any measures of

age or rank in order to ensure anonymity of the sample as a whole. All participants were men. Length of service ranged

from 0 to 31 years (M¼ 6.95, SD¼ 7.42).

1Here the term depersonalization has a different meaning from the one used in self-categorization theory describing a state of diminished awareness ofinterpersonal differences and increased attention to one’s social, as opposed to personal, self-identity.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 39, 991–1001 (2009)

DOI: 10.1002/ejsp

Figure 1. The path model (N¼ 352). Note: Exposure to crowd conflicts is dummy coded: non-exposed to crowd conflict¼ 0 andexposed to crowd conflict¼ 1; * p< .05, **p< .01, ***p< .001.

994 Gabriele Prati and Luca Pietrantoni

Materials and Measures

The instrument is a questionnaire containing 20 items based on the main themes emerged in Stott and Reicher’s (1998)

interview study and operationalized in Drury et al.’s (2003) survey. Items are worded in Italian language.

Participants rated all these items on a seven-point Likert-type scale ranging from�3 (strongly disagree) toþ3 (strongly

agree). Finally, we asked police officers to report the times they were involved in sport and demonstrating crowd conflicts

as part of their duty.

Procedure

The present study employed a survey methodology, using a questionnaire. The questionnaire was administered

immediately prior to taking part in official training course as part of police education. Participants were asked to fill out the

questionnaire as a way of introducing them to the topic of public order policing.

RESULTS

Missing Data Analysis and Imputation

Subjects who had filled out less than 50% of the questionnaire were excluded from the analyses (N¼ 1). After this deletion

of cases, there were no variables with more than 5% missing values. Listwise deletion of cases would have resulted in

significant sample loss in our main multivariate analyses. To avoid any bias, this could have introduced in our results, we

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 39, 991–1001 (2009)

DOI: 10.1002/ejsp

Exposure to crowd conflict and ESIM 995

imputed missing data on these variables (Schafer & Graham, 2002). Using the expectation-maximization (EM) method,

we conducted Little’s MCAR test to examine whether the missing data were random. The x2 was significant suggesting

our missing data were not completely at random (x2 (x2¼ 1127.52) df¼ 918, p< .001) (i.e. probability of missing data on

a particular variable can depend on other observed variables, but not on the variable itself when controlling for the other

observed variables). According to Schafer and Graham (2002) we imputed missing values with the EM methods using

NORM v. 2.03 software (Schafer, 2000) in order to produce a dataset without missing data.

Endorsement of Accounts of the Crowd and Exposure to Crowd Conflict

We examined participants’ endorsement of each of the general themes related to police perceptions of crowd behaviour

that were identified in Stott and Reicher’s (1998) interview study and operationalized in Drury et al.’s (2003) survey. In

order to test police endorsements of the statements in the questionnaire, we focused on significance levels of difference of

the means from the midpoint of zero and 95% confidence intervals (CIs). Police officers endorsed the statements: ‘People

of all sorts can be found among demonstration (M¼ 2.33; t (351)¼ 34.73, p< 0.001) and sport crowds (M¼ 1.40; t

(351)¼ 16.02, p< 0.001)’ and that, at the same time, crowds consist of a powerful antagonistic minority and a susceptible

majority: ‘Crowd is divided into a violent minority and a suggestible majority’ (M¼ 1.43; t (351)¼ 17.30, p< 0.001),

‘Professional agitators are skilled at inciting violent behaviour among previously peaceful members of demonstrating/

football crowds’ (M¼ 1.61; t (351)¼ 20.67, p< 0.001) and ‘Crowd conflict is due to a manipulative minority and a

mindless majority’’ (M¼ 1.13; t (351)¼ 13.37, p< 0.001).

Participants did not clearly endorse the view that crowds pose a homogeneous threat: They slightly agreed with the

statement ‘In a crowd, people tend to act irrationally’ (M¼ 0.51; t (351)¼ 5.80, p< 0.001) and ‘Once violence starts, the

crowd tends to sustain it’ (M¼ 0.53; t (351)¼ 6.16, p< 0.001) but they slightly disagreewith item 7 (‘Once disorders start,

even the most ordinary law-abiding citizens tend to become violent’M¼�0.22; t (351)¼�2.18, p< 0.05), and item 10

(‘It does not matter how much heterogeneous crowd composition might be, in case of crowd conflict everybody joins

together’ M¼�0.20; t (351)¼�2.05, p< 0.05). The confidence intervals of items 8 (‘It only takes one to commit

violence and the rest of the crowd is doing it’ M¼�0.02; t (351)¼�0.23, p> .05) and 15 (‘When going into situations

involving big crowds, I am worried about my safety’ M¼ 0.01; t (351)¼ 0.06, p> 0.05) included zero and were not

significant.

Police endorsed the statements concerning tactical reasons for treating crowd members as one in item 11 (‘By the time

the police take any serious action against violent members of a crowd, most genuinely peaceful crowd members will have

retreated to a place of safety’ M¼ 1.02; t (351)¼ 10.29, p< 0.001), 12 (‘In case of crowd conflict, there is difficulty of

identifying individual crowd members: Thus it is impossible to treat crowd members differently’M¼ 1.78; t (351)¼ 9.66,

p< 0.001), 13 (‘It is impossible to identify the perpetrators of violence among crowd members’M¼ 0.49; t (351)¼ 4.75,

p< 0.001) and 14 (‘It’s inevitable that police intervention in crowd conflict involves innocent people’ M¼ 1.24; t

(351)¼ 13.35, p< 0.001).

The officers in the sample agreed with coercive policingmethods stated in items 16 (‘Crowds must be strictly controlled

in order to prevent widespread violence erupting’), 17 (‘Police should suppress every potential threat to public order’

M¼ 1.35; t (351)¼ 14.46, p< 0.001), 18 (‘If there is the possibility of crowd conflict, then police should use coercive

force against the crowd’ M¼ 1.07; t (351)¼ 10.76, p< 0.001) and 19 (‘Should crowd members deviate in any way from

what had been agreed beforehand with the police that would constitute a breakdown of control and would constitute

incipient disorder’ M¼ 1.29; t (351)¼ 17.41, p< 0.001). Finally, participants did not see police tactics themselves as

having any responsibility for the initiation or development of public disorder as stated in item 20 (‘When violence

involving crowds occurs, the police are not responsible for either the initiation or any escalation of such violence’

M¼ 0.93; t (351)¼ 8.97, p< 0.001).

A total of 119 (34%) police officers were not exposed to sport and/or demonstrations crowd conflicts as part of their

duty. Police officers were exposed to a mean of 3.92 (SD¼ 6.80, 95%CIs¼ 3.21–3.64) sport crowd conflicts and to a mean

of 2.06 (SD¼ 4.42, 95%CIs¼ 1.60–2.53) demonstration crowd conflicts. The distributions of these measures were highly

positive skewed (8.80 and 14.55, respectively) in that the frequent scores were clustered at the lower end of the scale.

Given that normality assumptions were severely violated for these items we used the dummy version of exposure to crowd

conflict (non-exposed vs. exposed) for subsequent analysis.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 39, 991–1001 (2009)

DOI: 10.1002/ejsp

Table 1. Factor pattern matrix using principal axis factor and direct quartimin rotations

Variable 1: HT 2: CP 3: HC 4: TR 5: DC 6 h2

Item 1 �.072 �.080 .854 �.044 .065 �.050 .712Item 2 .006 .060 .478 .010 .013 .054 .253Item 3 �.004 �.097 .082 �.032 .531 �.037 .309Item 4 .106 �.037 .003 .053 .567 .183 .394Item 5 .083 .120 �.005 �.034 .576 �.110 .393Item 6 .429 .107 .062 �.017 .121 �.194 .296Item 7 .612 �.023 .022 �.016 .015 �.134 .383Item 8 .615 .039 �.035 �.010 .109 .172 .504Item 9 .684 �.039 �.030 �.045 .027 .202 .550Item 10 .723 .017 �.058 .005 �.029 �.019 .511Item 11 �.035 .069 .010 �.002 .040 .430 .204Item 12 .060 �.044 .033 �.661 �.112 .136 .477Item 13 .014 �.003 .029 �.663 �.091 �.184 .440Item 14 �.023 .056 .054 �.512 .058 .009 .291Item 15 �.021 �.005 �.095 �.243 .140 .042 .096Item 16 .100 .366 .012 �.018 �.008 .013 .177Item 17 �.103 .545 .040 .043 .006 .098 .314Item 18 .010 .863 �.045 �.042 �.028 �.089 .718Item 19 .158 .052 .064 �.251 .128 .080 .194Item 20 .132 .089 .071 �.061 �.122 .207 .118EVs 3.107 1.271 1.068 .888 .576 .423 7.333% of variance 18.4 9.2 8.0 7.1 6.1 5.4 54.3

Note: Coefficients greater than j.35j are in bold face and retained for that factor. Percentage variance is postrotation. The eigenvalue of the seventh,unretained factor was .969. h2¼ communality coefficient. HT¼ homogeneous threat of crowds; CP¼ coercive policing; HC¼ heterogeneouscomposition of crowd; TR¼ tactical reasons for treating crowd as one; DC¼ dichotomous composition of crowd.

996 Gabriele Prati and Luca Pietrantoni

Exploratory Factor Analysis of the Statements

The items were factor analysed using principal axis factoring followed by quartimin rotation. In order to identify the

number of factors, we employed parallel analysis (Horn, 1965; O’Connor, 2000) that indicated a six-factor solution. A

total of 54.3% variance was explained by exploratory factor analysis. The variance explained by each factor of the rotated

six-factor solutions was, respectively, 18.4, 9.2, 8.0, 7.2, 6.1 and 5.4%. Table 1 shows the full factor pattern matrix and

communalities. Absolute factor loadings greater than .35 were considered salient. Five of six dimensions were labelled:

Homogeneous threat of crowds, coercive policing, heterogeneous composition of crowd, tactical reasons for treating

crowd as one and dichotomous composition of crowd. The sixth factor was unretained because it consisted of only one

item loading greater than .35. Items 11, 15 and 19 were dropped because of their low factor loadings on the first five factors.

We retained item 20 related to police responsibility for conflict for subsequent analysis because, according to the theoretical

model, we expected low factor loading on the other factors. In order to investigate how stable this factor analytic solution was,

the analysis was repeated without the items that were dropped. We found support for a five-factor solution with the same

factors and similar factor loadings. Again item 20 showed low factor loadings (<.15) on all factors.

Path Model of the Relations Between Measures

We used structural equation modelling to test our proposed model of the role of police perceptions in mediating the

relationship between exposure to crowd conflicts and crowd policing strategy. The hypothesized path model (see Figure 1)

showed relatively good fit to the data, Satorra–Bentler Scaled x2 (100, N¼ 352)¼ 135.6054, p¼ .01, CFI¼ .96,

TLI¼ .95, RMSEA¼ .032, 90%CI for RMSEA¼ .016–.045. This model and their parameters are summarized in

Figure 1. Post hoc power analysis for covariance structure modelling indicated that our sample size was adequately large

for power of .80 (MacCallum, Browne, & Sugawara, 1996). As predicted, there was a significant pathway from exposure

to crowd conflict to perception of dichotomous crowd. Police perception of dichotomous crowd was related to perception

of homogeneous threat. In turn, perception of homogeneous threat predicted tactical reasons for treating crowd as one and

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 39, 991–1001 (2009)

DOI: 10.1002/ejsp

Exposure to crowd conflict and ESIM 997

coercive policing. Finally, there were significant pathways from tactical reasons for treating crowd as one and coercive

policing to denial of police responsibility for crowd conflict.

Mediation Analyses

Holmbeck (1997) and Preacher and Hayes (2004) made a distinction between indirect and mediated effects. Mediation is a

special, more restrictive, type of indirect effect that occurs when there is a significant total effect. In such case, the mediator

does significantly account for the total effect. Moreover, in the case of partial mediation the mediator accounts for a

significant portion of the total effect but a significant direct effect also remains, whereas in the case of complete mediation

the mediator fully accounts for the significant total effect (James, Mulaik, & Brett, 2006).

On the basis of the theoretical model and the axiom of parsimony (James et al., 2006), the base model for tests of

mediation is complete mediation. According to Mathieu and Taylor’s (2006) decision tree for evidence supporting

different intervening effects, complete mediation condition requires first a significant total effect. Failing that indirect

effect condition is considered. As illustrated in the upper triangle of Table 2, there were significant correlations (i.e. total

effect) between: (1) Exposure to crowd conflict and perception of a homogeneous threat, (2) perception of dichotomous

composition and tactical reasons for treating crowd as one; (3) perception of dichotomous composition and coercive

policing; (4) perception of homogeneous threat and denial of police responsibility for crowd conflict. The correlation

between perception of dichotomous composition and denial of police responsibility for crowd conflict was not significant,

thus an alternative hypothesis of an indirect effect was considered.

In testing for the significance of the indirect effect, MacKinnon, Fairchild, and Fritz (2007) recommended resampling

methods (5000 bootstrapped resamples). Power analysis for bias-corrected bootstrap tests of mediation revealed that our

sample size is adequate to achieve at least .80 statistical power to detect an effect (Fritz & MacKinnon, 2007). An indirect

effect is said to occur if the 95%CI for estimates does not contain zero. As evidenced in Table 3 all the indirect effects were

Table 2. Intercorrelation between measures (n¼ 352)

1 2 3 4 5 6

1. ECy — .16�� .12� .05 .04 .012. DC — .47��� .18��� .15�� .053. HT — .38�� .32�� .11�

4. TR — .12� .18���

5. CP — .19���

6. DR

Note: yEC is dummy coded: Non-exposed to crowd conflict¼ 0 and exposed to crowd conflict¼ 1; EC¼ exposure to crowd conflicts; DC¼ dichotomouscomposition of crowds; HT¼ homogeneous threat of crowds; TR¼ tactical reasons for treating crowd as one; CP¼ coercive policing; DR¼ denial ofresponsibility for crowd conflict.�p< .05; ��p< .01; ���p< .001.

Table 3. Indirect effects and estimation of the direct effect model

Predictor variable Mediator Dependent

Indirect effects Direct effect model Direct effect

Estimatey SE BC 95%CI DSBx2(1)z Estimatey

EC DC HT .228�� .10 (.06, .45) .532 .07DC HT TR .294��� .09 (.15, .53) .256 �.08DC HT CP .140��� .06 (.06, .31) 5.720� �.21�

DC HT, TR and CP DR .137�� .06 (.05, .29) .299 .10HT TR and CP DR .171�� .06 (.07, .29) 1.207 .19

Note: yEstimates are unstandardized. zSBx2¼Satorra–Bentler scaled statistic x2 difference value with one degrees of freedom using the no direct effectmodel as a base (see Figure 1); BC 95%IC¼Bias-corrected bootstrap confidence interval; EC¼ exposure to crowd conflicts; DC¼ dichotomouscomposition of crowds; HT¼ homogeneous threat of crowds; TR¼ tactical reasons for treating crowd as one; CP¼ coercive policing; DR¼ denial ofresponsibility for crowd conflict.�p< .05; ��p< .01; ���p< .001.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 39, 991–1001 (2009)

DOI: 10.1002/ejsp

998 Gabriele Prati and Luca Pietrantoni

significant. Moreover, complete mediation depends on the non-significance of direct effect. Therefore, the model with a

direct effect should not show a significant improvement over the model without direct path depicted in Figure 1. Therefore,

we calculated and tested the significance of the difference between the goodness-of-fit x2 values for the two models. In the

event of significant improvement of model with a direct effect, partial mediation is said to occur. Results are showed in

Table 3: (1) Perception of dichotomous crowd composition completely mediated the relation of exposure to crowd conflict

to perception of a homogeneous threat; (2) perception of a homogeneous threat completely mediated the effect of

perception of dichotomous composition on tactical reasons for treating crowd as one; (3) perception of a homogeneous

threat was a suppressor variable on the effect of perception of dichotomous composition on coercive policing. According

to Shrout and Bolger (2002), suppression occurs when the indirect effect has the opposite sign of the direct effect: In this

case, it is considered more appropriate to interpret the direct effect (adjusted for the suppressor) of perception of

dichotomous composition on coercive policing than the total effect; (4) there was an indirect effect of perception of

dichotomous composition to denial of police responsibility for crowd conflict via homogeneous threat, tactical reasons for

treating crowd and coercive policing; (5) the relation of homogeneous threat to denial of responsibility was completely

mediated by coercive policing and tactical reasons for treating crowd.

DISCUSSION

The present study addressed the police perspective in the explanation of crowd dynamics. Broadly, our results support and

extend the analysis of Stott and Reicher (1998) and Drury et al. (2003). One of the major contributions that this paper

makes is the cross-cultural validation of the link between officer’s theoretical model of the crowd and their adherence to

reactive use of indiscriminate force.

Police officers saw that both sport and demonstration crowds have a heterogeneous composition. In spite of the fact that

crowds, especially the demonstration ones, were perceived as composed by a variety of people, police officers made a

distinction between a majority and a minority. The minority is intentionally violent and capable of manipulating the

mindless susceptible majority. In sum, police officers perceived crowds’ composition as heterogeneous and, at the same

time, constructed a dichotomy between minority and majority especially when it comes to explain crowd conflict.

In situations of conflict, officers did not clearly agree—nor did they disagree—with the perception of crowd members

to be generally homogeneous in composition in terms of the danger that they represent to public order and to their safety.

This result is in line with Drury et al.’s (2003) survey study but is out of line with Stott and Reicher’s (1998) interview

study. If we consider the words used in the items, we can see that they tended to agree with items related to crowd

behaviour but disagree with items related to individual behaviour (in a crowd context). In other words, police officers

endorsed the view that the persons tend to be irrational in crowd context but not necessarily violent. However, when it

comes to talk about crowd behaviour, officers endorsed the view that it tends to be violent. It is likely that police officers

felt too simplistic the items related to individual behaviour, but not the items related to crowd behaviour. Ideology and

social representation of crowd dynamics might have played a role in adherence to perception of homogeneous threat of

crowd. The contradictory nature of common sense discourse is evident when the crowd is conceptualized as inherently

violent but the individual behaviour in a crowd context is not conceptualized as violent.

Police officers clearly endorsed the view that crowd should be treated as one and must be dealt with harshly methods

and that their own practices are not responsible for the escalation of conflict. The only difference with Drury et al.’s (2003)

study is that in their study there was not a clear endorsement of tactical reasons for treating the crowd as one. It seems that

British police officers, in comparison to the Italian colleagues, recognize more the need for differentiation in crowd

policing. This is probably due to a different training program and organizational culture.

We may speculate about what zero represented on the response scales used. It is possible that subjects may have felt

genuinely ambiguous about an item or that they may have veered towards the midpoint because they wanted to indicate

that their answer was ‘it depends’. In this case, the midpoint could reflect conditionality, which is different from neutrality

of opinion. However the midpoint was never the most frequently occurring value in the set of scores (mode). Although, we

cannot reject the possibility that some police officers that may have felt ambiguity or reflected on conditionality, we can

say that these cases were, at most, a small minority of the sample. Future studies should elucidate whether these processes

are responsible.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 39, 991–1001 (2009)

DOI: 10.1002/ejsp

Exposure to crowd conflict and ESIM 999

Taken as a whole, our path model confirmed the results of Drury et al. (2003). We found out the same relation between

perceptions of crowd and policing methods and, in turn, to denial of responsibility. Mediation analyses confirmed what the

authors stated: Endorsement of indiscriminate and harsh actions was expected to demand more system-justification (denial

of responsibility) than endorsement of mere perception. More specifically, the perception of dichotomous composition of

crowd was directly related to perception of homogeneous threat as result of the social influence of a violent minority on a

susceptible majority. The relation between the view of the crowd as inherently threatening and the denial of police

responsibility for crowd conflict was not direct but was completely explained by the endorsement of coercive policing

methods.

However, our results differed from Drury et al.’s (2003) study in two main points. First, homogeneous crowd was not a

mediator but a suppressor of the relationship between perceptions of dichotomous composition and endorsement of

coercive policing methods. More specifically, police officers, who endorsed the view that the composition of the crowd is

dichotomous, tended to perceive homogeneous threat of the crowd and do not support coercive policing methods. Given

that the relation of perception of homogeneous crowd and endorsement of coercive policing methods was positive, a

suppression effect was present within this mediation model in that the direct and mediated effects of perception of

dichotomous composition of the crowd on endorsement of coercive policing methods had opposite signs. It is likely that

police officers who recognized the inherent dichotomous nature of the crowd tended also to recognize that harshly policing

methods might involve also the innocent majority and thus do not recommend them. However, given that police officers

tended to view the majority as susceptible and prone to manipulation in case of conflict, all crowd members could be

considered dangerous, and thus they recommend harsh policing methods. Second, the endorsement of tactical reasons to

treat crowd as one completely explained the relation between perception of homogeneous threat and denial of

responsibility for crowd conflict. This result confirmed the hypothesis that if police officers tend to endorse the view that

crowd is a homogeneous threat, thus it should be easier for them to make a good case for tactical reason for treating crowd

members as a single body. Moreover, mediation analysis showed that the tendency to treat crowd members as one was

related to the theories that are held about crowds.

In additions to Drury et al.’s (2003) study, our results showed a significant indirect effect between perception of the

crowd’s dichotomous composition and denial of responsibility. These variables were indirectly related through significant

relationships with a linking mechanism though the nature of the relation is distal. As in Drury et al.’s study, there was not a

significant direct relationship between perception of the crowd’s dichotomous composition and denial of responsibility.

More importantly, the results of this study showed that there was a relation between exposure to crowd conflict and

endorsement of the dichotomous view of the crowd. Moreover, this kind of perception of the crowd explained the relation

between exposure to crowd conflict and perception of homogeneous threat of crowd. It seems likely that the exposure to

life-threating events might be related to a stricter adherence to ideological view of crowd as result of stress related cynical

attitude. The literature showed a relation between duty-related stressors and cynical attitude (Kop et al., 1999; Robinson

et al., 1997; Stearns & Moore, 1993). According to the theory proposed by Niederhoffer (1967), police officers come into

the occupation with idealistic aspirations (e.g. serving the community), but quickly come to realize the hard realities of the

world and of police work (in this case, involvement in violent crowd events). Over time, police officers then become

increasingly stressed and may lose a sense of purpose. Cynicism may be a way to cope with what is perceived to be an

unfriendly and insecure world, providing a convenient explanation for constant disillusionment and a way of acting out

anger and resentment. Cynicism as a coping method might be related to stress resulting from involvement in crowd conflict

in that it poses a threat to the physical integrity of self or others (Violanti & Marshall, 1983). In public-order policing,

cynical reactions could take the form of adherence to an ideology that views crowd conflict as deriving from the pathology

of the crowd itself or the form of approval of coercive and undifferentiated interventions. This happens because highly

stressful experiences violate a person’s expectations or well-established schematized knowledge of the self and the world

and can result in changes in one’s beliefs and personal assumptions as a response (Janoff-Bulman, 1989). However, even if

stress is a particularly important issue that is ignored in this domain, the stress hypothesis does not intend to explain

entirely the adherence to an ideological view of crowd dynamics. Cultural and organizational factors contribute as well to

the development of police perceptions and practices.

Talking about practical implications, this study stresses the importance of public order policing strategy. According to

Stott and Reicher (1998), the adherence to classical view of crowds (e.g. Le Bon’s or Allport’s tradition) could become a

self-fulfilling prophecy if it constitutes the basis of policing strategies. Research evidences showed the effectiveness of a

model of ‘good practice’ derived from empirical research (e.g. Adang & Stott, 2004; Cronin & Reicher, 2006; Stott et al.,

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 39, 991–1001 (2009)

DOI: 10.1002/ejsp

1000 Gabriele Prati and Luca Pietrantoni

2001, 2007). Reicher, Stott, Cronin, and Adang (2004) developed new guidelines for public order policing based on four

specific principles of crowd policing: Education about the social identities of the various groups in the crowd, facilitation

of crowd aims, proper communication with crowd members and differentiation of crowd members. Recently, Stott and

Pearson (2007) added the concepts of dynamic risk perception and graded policing response to these principles. Moreover,

the present study supports Drury et al.’s (2003) demand for a police training based on the ESIM model of the crowd and

against ideological models of the crowd. The impact of exposure to crowd conflict has practical implications because it

underlines the need for monitoring stress-related reactions. Moreover, given that exposure to crowd conflict results in

greater adherence to classical view of the crowd, we suggest the implementation of group sessions for police officers after

crowd conflict where participants can discuss their experience and where a crowd psychologist stimulate them to

recognize the complexity of conflict dynamics.

To conclude, the results of this paper evidence the role of police perceptions and practices concerning crowd dynamics

in the maintenance of public order. Although the design of the study is cross-sectional, and thus we cannot draw causal

inference, the results of the study show that exposure to crowd conflict and adherence to ideological view of crowd are

related to ‘bad practices’ of public order policing and, in turn, to system-justificatory attributions. Broadly, this study

support and extend the police perspective within the ESIM model and stresses the importance of implementing specific

public order strategies and training programmes derived from recent empirical studies on the field of crowd psychology.

ACKNOWLEDGEMENTS

The authors thank Lorenzo Luini and Fabrizio Marini for their collaboration in this research project and Zlatina Kostova

for her comments. They also thank who kindly volunteered to participate in the study. Finally, the authors thank the three

anonymous reviewers for their very insightful comments.

REFERENCES

Adang, O. M. J., & Stott, C. J. (2004). Preparing for Euro 2004: Policing international football matches in Portugal. Unpublished reportfor the Portuguese Public Security Police.

Allport, F. (1924). Social psychology. Boston: Houghton Mifflin.Cronin, P., & Reicher, S. (2006). A study of the factors that influence how senior officers police crowd events: On SIDE outside thelaboratory. British Journal of Social Psychology, 45, 175–196.

Drury, J., & Reicher, S. D. (2000). Collective action and psychological change: The emergence of new social identities. British Journalof Social Psychology, 39, 579–604.

Drury, J., & Stott, C. (2001). ‘‘Bias’’ as a research strategy in participant observation: The case of intergroup conflict. Field Methods, 13,47–67.

Drury, J., Stott, C. J., & Farsides, T. (2003). The role of police perceptions and practices in the development of ‘public disorder’. Journalof Applied Social Psychology, 33, 1480–1500.

Fritz, M. S., & MacKinnon, D. P. (2007). Required sample size to detect the mediated effect. Psychological Science, 18, 233–239.Holmbeck, G. N. (1997). Toward terminological, conceptual, and statistical clarity in the study of mediators and moderators: Examplesfrom the child-clinical and pediatric psychology literatures. Journal of Consulting and Clinical Psychology, 65, 599–610.

Horn, J. L. (1965). A rationale and test for the number of factors in factor analysis. Psychometrika, 30, 179–185.James, L. R., Mulaik, S. A., & Brett, J. M. (2006). A tale of two methods. Organizational Research Methods, 9, 233–244.Janoff-Bulman, R. (1989). Assumptive worlds and the stress of traumatic events: Applications of the schema construct. SocialCognition, 7, 113–136.

Kop, N., Euwema, M., & Schaufeli, W. (1999). Burnout, job stress and violent behaviour among Dutch police officers.Work & Stress,13, 326–340.

Le Bon, G. (1895 trans. 1947). The crowd: A study of the popular mind. London: Ernest Benn.MacCallum, R. C., Browne, M.W., & Sugawara, H.M. (1996). Power analysis and determination of sample size for covariance structuremodeling. Psychological Methods, 1, 130–149.

MacKinnon, D. P., Fairchild, A. J., & Fritz, M. S. (2007). Mediation analysis. Annual Review of Psychology, 58, 593–614.Maslach, C., Schaufeli, W. B., & Leiter, M. P. (2001). Job burnout. Annual Review of Psychology, 52, 397–422.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 39, 991–1001 (2009)

DOI: 10.1002/ejsp

Exposure to crowd conflict and ESIM 1001

Mathieu, J. E., & Taylor, S. R. (2006). Clarifying conditions and decision points for mediational type inferences in organizationalbehavior. Journal of Organizational Behavior, 27, 1031–1056.

McClelland, J. S. (1989). The crowd and the mob: From Plato to Canetti. London, UK: Unwin Hyman.Niederhoffer, A. (1967). Behind the shield: The police in urban society. Garden City, NY: Anchor Press.O’Connor, B. P. (2000). SPSS and SAS programs for determining the number of components using parallel analysis and Velicer’s MAPtest. Behavior Research Methods, Instrumentation, and Computers, 32, 396–402.

Paton, D. (1997). Dealing with traumatic incidents in the workplace (3rd ed.). Queensland, Australia: Gull.Postmes, T., & Spears, R. (1998). Deindividuation and antinormative behavior: A meta-analysis. Psychological Bulletin, 123, 238–259.Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects in simple mediation models. BehaviorResearch Methods, Instruments, & Computers, 36, 717–731.

Prentice-Dunn, S., & Rogers, R. W. (1982). Effects of public and private self-awareness on deindividuation and aggression. Journal ofPersonality and Social Psychology, 43, 503–513.

Prentice-Dunn, S., & Rogers, R. W. (1989). Deindividuation and the self-regulation of behavior. In P. B. Paulus (Ed.), The psychology ofgroup influence (2nd ed., pp. 86–109). Hillsdale, NJ: Erlbaum.

Reicher, S. (1996). ‘‘The battle of Westminster’’: Developing the social identity model of crowd behaviour in order to explain theinitiation and development of collective conflict. European Journal of Social Psychology, 26, 115–134.

Reicher, S. (2001). The psychology of crowd dynamics. In M. Hogg, & R. S. Tindale (Eds.), The Blackwell handbook of socialpsychology: Group processes (pp. 182–208). Oxford: Blackwell.

Reicher, S., Stott, C., Cronin, P., & Adang, O. (2004). An integrated approach to crowd psychology and public order policing. Policing:An International Journal of Police Strategies and Management, 27, 558–572.

Robinson, H. M., Sigman, M. R., & Wilson, J. P. (1997). Duty-related stressors and PTSD symptoms in suburban police officers.Psychological Reports, 81, 835–845.

Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art. Psychological Methods, 7, 147–177.Schafer, J. L. (2000). NORM: Multiple imputation of incomplete multivariate data under a normal model. version 2.03, Software forWindows 95/98/NT, available from http://www.stat.psu.edu/�jls/misoftwa.html

Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and nonexperimental studies: new procedures and recommendations.Psychological Methods, 7, 422–445.

Stearns, G. M., & Moore, R. J. (1993). The physical and psychological correlates of job burnout in the royal Canadian mounted police.Canadian Journal of Criminology, 35, 127–148.

Stott, C., & Drury, J. (1999). The intergroup dynamics of empowerment: A social identity model. In P. Bagguley, & J. Hearns (Eds.),Transforming politics: Power and resistance (pp. 32–45). New York, NY: Macmillan.

Stott, C., & Drury, J. (2000). Crowds, context, and identity: Dynamic categorization processes in the ‘‘poll tax riot’’. Human Relations,53, 247–273.

Stott, C., & Pearson, G. (2007). Football ‘hooliganism’: Policing and the war on the ‘English disease’. London: Pennant Books.Stott, C., & Reicher, S. D. (1998). Crowd action as intergroup process: Introducing the police perspective. European Journal of SocialPsychology, 28, 509–529.

Stott, C., Adang, O., Livingstone, A., & Schreiber, M. (2007). Variability in the collective behaviour of England fans at Euro 2004:Intergroup relations, identity content and social change. European Journal of social psychology, 37, 75–100.

Stott, C., Hutchison, P., & Drury, J. (2001). ‘‘Hooligans’’ abroad? Intergroup dynamics, social identity, and participation in collective‘‘disorder’’ at the 1998 World Cup finals. British Journal of Social Psychology, 40, 359–384.

Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. Austin, & S. Worchel (Eds.), The socialpsychology of intergroup relations. Montery, CA: Brooks Cole.

Turner, J. C. (1987). Introducing the problem: Individual and group. In J. C. Turner, M. A. Hogg, P. J. Oakes, S. D. Reicher, & M. S.Wetherell (Eds.), Rediscovering the social group: A self-categorization theory. Oxford: Basil Blackwell.

Violanti, J. M., & Marshall, J. R. (1983). The police stress process. Journal of Police Science, and Administration, 11, 389–394.

Copyright # 2009 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 39, 991–1001 (2009)

DOI: 10.1002/ejsp