BYSTANDER APATHY

The British Journal of Criminology, 2022, 62, 18–36 DOI: 10.1093/bjc/azab042 Advance access publication 15 June 2021 Article

Circles of Peace. A Video Analysis of Situational Group Formation and

Collective Third-Party Intervention in Violent Incidents

Don Weenink , Raheel Dhattiwala and David van der Duin* *Don Weenink, Department of Sociology, University of Amsterdam, Roeterseilandcampus Nieuwe

Achtergracht 166, 1018 WV Amsterdam, the Netherlands; [email protected]; Raheel Dhattiwala and David van der Duin, Department of Sociology, University of Amsterdam, Amsterdam, the Netherlands.

We seek to understand collective third-party intervention in violent incidents. Based on analyses of 131 video clips captured on mobile phones, we argue that bodily alignment between bystand- ers creates situational groups that encourage collective de-escalatory action. Qualitative analysis reveals that third parties generate situational groups when they create circular formations, provid- ing greater opportunities to notice each other’s monitoring of the situation while separating those focused on the incident from others just passing by. Statistical analysis shows that the formation of situational groups facilitates the collectivization of de-escalatory action. In contrast to popular but outdated ideas about the dangerous influence of assemblies on individual behaviour (allegedly leading to disorder or apathy), our findings emphasize the self-regulatory, defusing capacity of gath- erings.

Key Words: bystander, bystander effect, third-party intervention, violence, visual criminology, video analysis

I N T RO D U CT I O N Support for the ‘bystander effect’—the idea that larger the number of people present, less likely are they to intervene when others are in need—has weakened over time. Third parties are less passive than previously thought.1 In conducting a video analysis of street violence, our original aim was to identify the conditions that facilitate intervention. After viewing several videos, we observed how frequently bystanders engaged in collective de-escalatory action. This was sur- prising because violence presents an uncertain environment to be caught in, perhaps more so for the bystander. What is it, then, about being part of a collective that facilitates intervention in

1 The highly publicized absence of intervention in the Kitty Genovese murder, which went on to kindle the bystander effect, has since been found to be exaggerated (e.g. Manning et al. 2007).

© The Author(s) 2021. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact [email protected]

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

Circles of Peace • 19

an uncertain environment? The central objective of this paper became, therefore, to identify the processes that precede and facilitate collective intervention.

Following an inductive approach based on visual analysis of 131 incidents of physical violence and altercations that did not turn violent, we developed the notion of situational groups—the physical co-presence of people aware that they are together in the same situation—to explain collective third-party intervention. Our focus is on non-verbal, bodily forms of de-escalatory ac- tion.

Our paper relates to ‘the new visibility’ (Thompson 2005; Goldsmith 2010): the increased opportunities for civilians, law enforcement officers and authorities to monitor behaviour, re- sulting from the ubiquitous presence of recording devices and mass dissemination of video footage. Activists and ordinary civilians collect and compile phone recordings of police–ci- vilian encounters to evidence police misdemeanour ( Jones and Raymond 2012); the police employ CCTV footage to trace suspects; and in court, various forms of video material are used as (counter-) evidence (see Goodwin 1994 on the use of video recordings in the Rodney King trial). The world-wide protests against racism and excessive police violence sparked by bystander footage depicting the killing of George Floyd by police officers in Minneapolis in May 2020 are exemplary. On the other hand, older ideas about the passivity of bystanders now reappear in the form of complaints about people eagerly filming an incident rather than try- ing to de-escalate it. More fundamentally, phone filming might dehumanize the people they record into objects of entertainment, instead of generating empathy (cf. Baumann 1993: 177–8, quoted by Carrabine 2012: 467). Indeed, we often observed multiple recorders commenting on the entertainment value of the spectacle. However, our findings also clearly show that collective de-escalatory action is prevalent in the age of phone recording.

By studying third-party intervention as a collective effort, we advance earlier social psycho- logical studies on bystander behaviour that seek to explain (in)action at the level of individuals (see for a review Fischer et al. 2011). Our data allow analysing de-escalatory action as it happens in real-life situations rather than in laboratory settings. While ours is not the first study of by- stander behaviour based on video data (Levine et al. 2011; Bloch et al. 2018; Liebst et al. 2018, 2019; Philpot et al. 2019), our notion of situational group formation is original for it identifies the process that facilitates collective de-escalatory action.

F RO M T H E BY STA N D E R E F F ECT TO BY STA N D E R S I T UAT I O N A L   G RO U P S

Latane and Darley (1968) paved the way for the prolific social-psychological research tradition on the bystander effect: the finding that individuals are less likely to help people in need when more others are present as each tends to feel less responsible. However, most experiments that have found supportive evidence for the bystander effect did not concern (the threat of) actual violence (Fischer et al. 2006: 268; Levine and Manning 2013: 239). In the few earlier experi- mental studies that simulated the threat of violence (e.g. with actors simulating a near rape in a parking lot close to research participants) or in which incidents were convincingly enacted (e.g. a confederate’s screaming and moaning expressing great pain), the bystander effect was either not found or found to be reversed (Darley et al. 1973; Harari et al. 1985). In their meta-analysis of bystander research, Fischer et al. (2011: 527, 534) found that when communication between bystanders is real (rather than experimentally manipulated), when their presence is physical (rather than implied) and when they are unsuspecting individuals (rather than instructed con- federates), the presence of additional bystanders leads to more, not less, individual helping be- haviour in emergency situations (see also Greitemeyer and Mügge 2015).

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

20 • The British Journal of Criminology, 2022, Vol. 62, No. 1

Convincing evidence of helping behaviour in real-life violent incidents is provided by Levine et al.’s (2011) pioneering study. Based on 42 incidents captured on CCTV footage, they found that third parties were more likely to engage in de-escalatory rather than escalatory action, and did so more often as group size increased. The study also showed that de-escalatory behaviour is a collective effort (Levine et al. 2011: 409).

We propose that the process of bodily alignment involved in situational group formation is more likely to result in collective intervention. Here, we follow a clue provided by one of the founding fathers of the bystander effect paradigm. In 1973, Darley and colleagues found that the ‘helping rate was […] affected not simply by the presence of other bystanders, but by their physical orientation vis-à-vis each other; groups in a facing orientation were more likely to re- spond than non-facing groups’ (Darley et al. 1973: 396–8). People facing one another could monitor startle responses—‘an orientation toward the noise, a facial expression of concern, etc’.—among group members (Darley et al. 1973: 396).

Early sociological work on how individuals assemble in public space noted that a ‘point of common interest’ tends to generate a circular arrangement of individuals, a ring formation, be- cause it allows for the most efficient organization of individuals around such common point of interest (Milgram and Toch 1969: 518; quoted in McPhail 1994: 42–3). Following McPhail (1994: 48–9), collective orientations develop because the ‘extraordinary is noteworthy’. Specifically, he notes that arguments, fights and arrests can lead to shifts in orientation, which ‘are frequently followed by converging locomotion that results in the formation of arcs and rings around the actors’ (McPhail 1994: 49).

Bloch et al. (2018: 1057–8) studied the appearance of what they termed ‘caring collectives’— groups of bystanders who engage in coordinated caring and consoling behaviour towards vic- tims, ‘typically in the aftermath of the violent conflict’. Bloch et al. (2018: 1058) also note that the development of such collectives involves a specific spatial form: a crescent shape, a ring forma- tion around the victim, which provides, as Bloch et al. (2018: 1063) aptly quote Goffman (1961: 18), ‘an eye-to-eye ecological huddle that maximizes each participant’s opportunity to perceive the other participants’ monitoring of him’. Situational group formation, therefore, concerns the development of a shared sense of being together in the same situation, which can be attained by circular formations in which individuals align themselves around a common focus of attention.

One further step is needed to explain why situational group formation encourages people to engage in collective de-escalatory action. Durkheim (1995/1912: 231–2; see also Collins 2004: 48) noted that assemblies in which bodily aligned participants have developed a shared focus of attention generate feelings of group membership. Now if collective intervention works through feelings of group membership, we expect that third parties intervene in synchrony, matching their de-escalatory action in time (Mogan et al. 2017: 15).

Alternatively, it can be argued that the relationship between situational group formation and collective intervention works through the calculus of individual bystanders (see Felson and Tedeschi 1993): third parties may be more likely to intervene if they see that others who are watching could assist or sound the alarm if they get in danger, or they may be prompted to inter- vene as they recognize that others are inclined to do the same. If intervention follows individual calculus, bystanders are more likely to wait and take into account what the other bystanders are doing while they consider taking de-escalatory action. Below, we will consider the validity of the ‘groupness’ versus the ‘calculus’ explanations.

M ET H O D O LO G Y: DATA , S A M P L I N G, CO D I N G A N D A N A LY S I S Data and sampling

In contrast to prior studies on de-escalatory behaviour based on CCTV footage (Levine et al. 2011; Liebst et al. 2018; Philpot et al. 2019), we used videos captured on mobile phones. CCTV

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

Circles of Peace • 21

footage rarely has sound, while both verbal and nonverbal utterances can be useful to under- stand the unfolding interaction (Lindegaard and Bernasco 2018).

We obtained our data from freely available video uploading websites such as YouTube, Liveleaks and WorldStarHiphop. Search terms with the English keywords ‘brawl’, ‘fight’, ‘street fight’ and ‘assault’ yielded prolific data. The lion’s share of the clips was from English-speaking countries (mainly the United States and the United Kingdom, two from Australia and one from Canada), simply due to our English search terms but probably also reflecting the high num- ber of video uploads from these countries. The remaining clips were from Brazil, Colombia, Cuba, Dominican Republic, France, Germany, Japan, India, the Netherlands, the Philippines and Spain.

We used the following criteria to compile our data corpus. First, we only included clips in which at least two third parties were visible for the duration of the clip. Second, we added ex- treme cases to increase variation concerning the severity of the violence. On the one hand, we included incidents that involved just one push, slap or punch as well as altercations which did not end in violence, using the keywords ‘argument’ and ‘altercation’. On the other hand, we added clips in which antagonists displayed and/or used weapons (apart from guns), using key- words ‘knife fight’ and ‘machete fight’. Prior studies indicate that intervention is related to ‘emer- gency awareness’—how clear it is that help is required and the degree to which third parties perceive the distress of victims (Fischer et al. 2006: 276; 2011: 533). Therefore, we need to take the severity of the violence into account. Third, we excluded fights that seemed to be arranged on the basis of any of the following criteria: the fight was a schoolyard or backyard fight between young people and the bystanders, having about the same age, were yelling the names of the antagonists and/or the title of the clip indicated the names of the two antagonists (e.g. ‘Mike vs. Eric’); the defeated party visibly acknowledged his opponent’s victory; the antagonists were wearing protective clothing; or a referee was present.

A final note on gender: the overwhelming majority of incidents was between males (see Table 1). We purposively excluded females-only fights to restrict variation and also because they often seemed to be arranged (back yard fights with referees).

Because the use of phone recorded footage is uncommon in criminology and the social sciences more broadly, our data warrant further discussion. One concern about using phone- recorded footage uploaded onto websites is that we often know little about the rationale for their recording and uploading. There may thus be hidden selection biases in the material available on- line. Another concern is whether the recording itself influenced the behaviour of third parties. Experimental studies by Van Bommel et al. (2014) show that noticing the presence of CCTV cameras or visibility via webcam makes third parties more likely to intervene. Nevertheless, it may not be the presence of cameras per se but the imagined audience as well as the kind of be- haviour thought to be (un)acceptable for this audience that affects individual behaviour (Briggs et al. 2016: 6). In our data, the recorders conveyed ambiguous messages. Some communicated excitement rather than concern. Only in a few cases, the act of recording was meant to deter ant- agonists from turning to violence (‘Don’t do it! They are taping it!’ and ‘I have it all on camera’). Several videos showed more than one person recording the altercation. Whatever the moral message, the recording itself probably contributed to establishing a shared focus of attention. In other words, our data may be positively biased towards the formation of situational groups. Third, one might think that our data are skewed towards atypical, more noteworthy incidents of street violence that attract circular formations around them. However, in the majority of our cases, we did not observe situational group formation, and we did not consider these cases less noteworthy than the ones in which bystanders created circular arrangements around the inci- dent. Fourth, our video material is drawn from a population unknown to us. However, given the variation in de-escalatory behaviour, situational group formation as well as in the severity

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

22 • The British Journal of Criminology, 2022, Vol. 62, No. 1

of the violence (see Table 1), we trust that our data allow to explore the theoretical connection between situational group formation and collective third-party intervention.

Coding and analysis Our study involved repeated sessions of watching and discussing the clips for over two years. The coding scheme we developed provided indicators used in the statistical analyses. The quali- tative analysis was based on transcripts of three contrasting video clips. We will first describe the numerical indicators and then discuss the transcripts.

For each clip, we coded the presence of circular formations when at least three people were standing facing one another, their faces oriented towards the incident and their backs away from it. We only noted this form of bodily alignment if it appeared before anyone, if at all, intervened. To enhance the precision and transparency of our coding, we timestamped our codes (in min- utes and seconds) and noted which persons were part of these formations. We coded the for- mation of lines prior to bodily de-escalation when at least two people stood next to each other with their faces oriented towards the antagonists. People sitting in a line (e.g. in a bus or train) were not coded as such, unless they got up from their seats to form a line as they stood together.

We coded and timestamped bodily de-escalatory behaviour when we observed third parties taking one of the following actions: walking towards the antagonists with open-handed ges- tures; waving their arms; making contact by touching or patting the antagonists; restraining or moving an antagonist away from the altercation; pulling people apart and putting one’s body in between the antagonists (these codes are largely derived from Levine et al. 2011: 408). We also noted who performed these actions and when.

We coded the following indicators that we suspected would affect the likelihood of situational groups forming and/or for bystanders engaging in bodily de-escalatory action: First, we noted how often third parties signalled their attentiveness to the incident prior to bodily de-escalatory behav- iour (see, e.g., the storylines of clips 18 and 27 below). We counted vocal responses (escalatory or de-escalatory verbal utterances and exclamations of excitement or aversion) and pointing ges- tures in the direction of the antagonists.2 Second, the number of persons visible in each clip—ex- cept for the antagonists and recorders—to estimate the potential pool of intervening third parties. Third, whether physical violence was of low intensity (only one push, shove, punch or slap) or was absent altogether. Fourth, how many antagonists were involved in the incident, their gender and whether they displayed or used weapons, including knives, machetes, sticks, clubs or improvised weapons such as bottles (our data do not include incidents involving firearms). Fifth, we distin- guished open spaces (streets, squares, parks etc.) from confined spaces such as pubs, restaurants and shops as well as buses, trains and train platforms. We suspected the lack of escape routes and limited manoeuvring room in confined spaces to deter bystanders from intervening (Dhattiwala 2016). Sixth, as prior research pointed to the importance of group identification between victims and bystanders for the latter to intervene (Levine and Manning 2013; Slater et al. 2013; Liebst et al. 2019), we observed: (1) whether a third party and an antagonist remained close (at touching distance) to each other as they moved through space; (2) whether they called each other by their names; (3) whether they touched each other, other than to de-escalate the conflict (for instance, by holding hands) and (4) whether they shared a vehicle or other object (for instance, a bag or skateboard). We discussed each case in which we observed one or more of these behaviours to determine whether third parties had pre-existing ties with one of the antagonists. Finally, we noted whether the event took place during day- or nighttime, whether the clip was recorded in English- language countries or elsewhere and the duration of the clip.

2 We attempted to code bystanders’ mutual monitoring, but it turned out we could hardly verify to whom bystanders were looking exactly and by whom they were looked at, especially when larger numbers of bystanders were present.

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

Circles of Peace • 23

The three authors first coded the above indicators separately. We then discussed the ratings we did not agree on. To estimate the external reliability of our coding, assistants coded the in- dicators in 59 (four assistants) and 38 clips (three assistants). Krippendorff ’s alpha scores were calculated with the ReCal OIR webservice (Freelon 2013). They were 0.66 for whether by- standers and antagonists had prior ties; 0.74 for whether the violence was moderate or not; 0.77 for the appearance of a circular formation; 0.79 for the appearance of a line formation; 0.86 for whether a weapon was displayed and again 0.86 for whether a space was confined or open; 0.90 for whether the incident also involved female antagonists; and 1.0 for whether the event took place during day- or nighttime, or whether this could not be identified, for instance when the incident took place in an underground carriage. To assess the inter-rater reliability of our obser- vations of bystanders’ vocalizations and pointing gestures, the authors and an assistant coded 25 clips, resulting in intra class correlations coefficients of, respectively, 0.91 and 0.92. The number of antagonists, intervening third parties and people present were counted by the authors separ- ately and then compared until agreement was reached.

To assess the relationship between situational group formation and collective third-party intervention, we used a Poisson regression model (as the outcome variable is a count of the number of individuals who took de-escalatory action). Because our data are drawn from an unknown population with an unknown distribution, we refrain from inferential interpretation of the statistical results. Therefore, we ignore and do not report the p-values, standard errors and confidence intervals of our results, and instead focus on the effects of the variables that are found for the data that we have.

As the highest variance inflation factor (VIF) value was 2.46 and the mean VIF was 1.58, issues of multicollinearity are negligible. We also checked for observations with large differ- ences between predicted and observed values. Re-estimated models, without ‘offending’ obser- vations, did not produce substantially different results than shown in Table 2.

After conducting the statistical analyses, we created storylines of three contrasting cases that reveal commonalities observed in our data. The storylines are transcriptions of the action sequences of bystanders in relation to the unfolding incident, allowing us to qualitatively under- stand processes of bodily alignment and bodily de-escalatory action. In the first case, situational group formation and bodily de-escalatory action did not occur; in the second, situational group formation occurred and bystanders engaged in simultaneous collective bodily de-escalatory ac- tion; in the third, situational group formation occurred but it took a relatively long time before collective third-party intervention took place.

Ethical considerations The use of videos for research purposes poses distinct ethical challenges, largely due to the non-anonymous content of the videos. For repurposed videos, ethical considerations become even less clear. Generally speaking, ethical guidelines for digital spaces tend to be less restrict- ive (Corneil 2012), with the consent of participants being less stringent for data acquired from the public domain including the internet. Even so, we ensured the anonymization of people’s names, when known, and blurred the faces in the stills (Figures 1–6). While our video corpus and code log are open to use and inspect by other researchers upon request, we require that they take the same measures to ensure the anonymity of the persons portrayed as we did here. To give an impression of the video material without revealing the identity of the persons captured on the recordings, we ‘cartoonized’ the clips we transcribed into storylines (see below). They can be viewed at https://www.violent-interactions.org/research-material/video-data/phone- recorded-footage-of-antagonistic-encounters-in-urban-public-spaces.

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

24 • The British Journal of Criminology, 2022, Vol. 62, No. 1

S I T UAT I O N A L G RO U P F O R M AT I O N A N D CO L L ECT I V E T H I R D - PA RT Y I N T E RV E N T I O N

Before we discuss the relationship between situational group formation and collective third- party intervention, we consider descriptive statistics of our data first (see Table 1).

Third parties engaged in bodily de-escalatory action in nearly three quarter of the situations in our data.3 On average, two bystanders engaged in de-escalatory action, which is over 10 per cent of all people present in each case (except for the antagonists and the recorder). While we cannot know to what extent these numbers are representative for all forms of street violence, they do provide an optimistic view of the behaviour of third parties—all the more so given that in over half of the cases, they were most likely strangers to the antagonists. However, we can compare these results with prior research into third-party intervention based on CCTV footage. First, Liebst et al. (2018: 37; 2019: 601) noted averages of 15.8 and 18.3 bystanders present at violent incidents, close to our count of 18.9. Second, Philpot et al. (2019: 71) found that in 90 per cent of situations, at least one bystander physically intervened and that the average number of third parties intervening was 3.76.

We observed the appearance of situational groups (circular formations) in over 27 per cent of the cases; they comprised 12 bystanders on average. In over half of our clips, we saw people stand- ing in a line watching the incident. As we will describe below, situational groups could emerge from such line formations. Bystanders abundantly vocalized their attentiveness to the incident; we heard over 20 of such utterances on average. They were, however, much less likely to use their arms to gesture that something noteworthy occurred; we observed only 1.1 of such gestures on average. As for the severity of violence, weapons were used or displayed in over a quarter of the

3 Third-party escalation occurred in 29 per cent of the 131 cases.

Table 1 Descriptive statistics, N = 131 video clips

Binary indicators Frequency Percentage

Bodily de-escalatory action 98 74.8 Situational group formation 36 27.5 Line formation 69 52.7 Moderately or non-violent incidents 26 19.8 Weapons 35 26.7 Confined space 60 45.8 Prior relationship between bystanders and antagonists 62 47.3 Incident involved females 9 6.9 Incident occurred at nighttime 29 22.1 Incident occurred at daytime 69 52.7 Time of incident unclear 33 25.2 English speaking country 110 84

Count indicators Mean SD Min. Max.

# third parties who took de-escalatory action 2.0 1.9 0 10 # third parties in situational group 12.2 8.5 3 39 # third-party vocalizations 20.6 21.6 0 105 # third-party pointing gestures 1.1 2.2 0 15 # people present 18.9 16.7 2 118 # antagonists involved in incident 2.6 1.3 2 8

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

Circles of Peace • 25

cases. On the other hand, violence was limited to just one punch, slap or push (6 cases), or absent altogether (20 cases) in about one-fifth of cases. On average 2.6 persons were part of the conflict. Nearly 46 per cent of the incidents occurred in confined spaces, such as pubs, restaurants and shops (10 cases, 8 per cent) and buses, trains and train platforms (50 cases, 38 per cent), whereas the remaining 54 per cent took place in streets, squares, parks and beaches.

In about a quarter of the cases, we could not determine whether it was night- or daytime, as these incidents took place in underground carriages or platforms. Of the remaining cases, 22 per cent occurred during the night and 53 per cent at daytime. More specifically, only 12 out of our 131 cases could be related to the nighttime economy, as they occurred in or near pubs, bars and cafés at night. As a result of our data selection decisions, 84 per cent of our clips was recorded in an English- speaking country and only 6.9 per cent of the incidents involved both females and males. Finally, the average length of our video footage was one minute and 57 seconds (not shown in Table 1).

We can now consider the relationship between situational group formation and the collect- ivization of bodily de-escalatory action. To start with, the average number of intervening third parties was 3.58 in situational groups, while it was 1.40 in all other situations. Table 2 displays the results of a multivariable Poisson regression analysis of the number of third parties who took de-escalatory action.4

4 Instead of the formation of situational groups, the number of participants in such groups might be more influential on the collectivization of de-escalatory action. A model in which we had replaced the former variable by our count of the number of participants in situational groups produced a Bayesian information criteria (BIC) value that was 3.6 higher as compared to the model displayed in Table 2. It was 19.7 higher in a model that only included the count of the number of participants versus one that only included the situational group variable. As lower BIC values indicate more explained variation (Kass and Raftery 1995: 777), this test offers support to the situational group variable.

Table 2 Poisson regression analysis of the number of third parties who took bodily de-escalatory action

IRR

Situational group 2.15 Line formation 1.17 # third-party vocalizations 1.00 # third-party pointing gestures −1.04a

# people present 1.01 Moderately or non-violent incidents −1.57 Weapons 1.07 Prior relationship between bystanders and antagonists 1.33 Incident involved females 1.22 # antagonists involved in the incident 1.02 Confined space 1.11 Time of incident (ref.: daytime) Nighttime 1.24 Unclear 1.03 English-speaking country −1.11 Clip duration 1.00 Constant 1.07

Observations 123b

Pseudo-R2 0.15

aWe transformed IRRs lower than 1 into negative values for ease of interpretation. bEight cases had to be dropped from the dataset because we could not determine the number of vocalizations in them.

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

26 • The British Journal of Criminology, 2022, Vol. 62, No. 1

Table 2 shows that the incidence rate ratio (IRR) of the number of intervening bystanders was over two times larger in situations where a situational group had formed as compared to that of situations without a situational group, controlling for the other variables in the model. Situations in which we observed line formations had an IRR of intervening bystanders of only 1.17 as compared to situations without such formation. Bystanders’ vocalizations virtually did not affect the IRR of people who took de-escalatory action, and their gesturing towards the incident did so only marginally. We conclude that collective intervention is much more facili- tated through circular spatial-bodily formations that allow bystanders to notice each other’s monitoring than through the extent to which they express their attentiveness to the incident as such.

The total number of people present hardly affected the IRR of intervening bystanders. In line with prior work emphasizing ‘emergency awareness’ as a prerequisite for intervention (Fischer et al. 2006: 276; 2011: 533), we found that incidents involving just one push, slap or punch or altercations, in which physical violence was absent altogether, reduced the IRR of intervening third parties with nearly 60 per cent as compared to cases in which more severe forms of vio- lence were committed.5 Strikingly, the presence of weapons hardly reduced the IRR at which bystanders intervened. We observed third parties using buckets, sticks and chairs to get between antagonists who wielded knives or machetes.

In situations where we noted the existence of prior social ties between bystanders and antag- onists, the IRR was nearly one-third larger. This result is in accordance with earlier work which found that group identification and prior ties between victims and bystanders increased the likelihood of intervention (Levine and Manning 2013; Slater et al. 2013; Liebst et al. 2019). The results also show (slightly) higher rates of intervening bystanders when female antagonists were involved, when they occurred at night,6 when the incident took place in a confined space and when the incident occurred in a non-English-speaking country. Lastly, the number of antagon- ists involved in the incident, whether the time of the day was unclear and the duration of the clip hardly affected the IRR of intervening bystanders.

One remaining issue is what makes situational group formation conducive to collective inter- vention. We suggested two processes: feelings of group membership and individual calculus. If the former is at work, we expect that third parties act as one, with immediate matching of de-escalatory action. If collective intervention works through individual calculus, we expect lower levels of synchrony as bystanders consider and await the actions of others present. In the 28 cases in which we observed situational group formation and collective intervention, the average time gaps between the first, second and third intervening individuals were four to five seconds (standard deviations four seconds).7 In five cases, we did not observe time lags of over one second between the first three intervening third parties. Here, bystanders had developed a strong focus towards the incident, their vocal responses were equivocal and signaled urgency (especially in knife fights). While these results should be taken as tentative given the low num- ber of cases, they seem to suggest that situational group formation mostly works to reduce the perceived costs and risks of de-escalatory action.

5 Estimating them separately, our indicators of non-violent versus moderately violent situations turned out to yield virtually similar effect sizes. 6 This finding can be related to the higher de-escalation rates noted by the aforementioned studies based on CCTV footage: given the prevalence of surveillance cameras in inner city leisure areas, these studies are more likely to include incidents related to urban nightlife; groups of friends going out at night might be more inclined to intervene. 7 In 6 out of all 36 cases in which we observed situational group formation, there was only one intervening bystander, and in 2 cases intervention did not occur at all. We only included the time lags between the first three intervening individuals, given the low number of cases in which over three individuals took de-escalatory action.

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

Circles of Peace • 27

B O D I LY A L I G N M E N T A N D D E -E S C A L ATO RY   A CT I O N Our qualitative analysis further explores the relationship between bodily alignment and de- escalatory action. We will consider contrasting types of situations: situations in which bodily alignment did not occur and situations in which circular formations appeared. In the latter cases, third parties not only signalled their awareness of both the incident and one another, but also demarcated the site of the incident from the surrounding space.

Disalignment In 57 clips, we did not observe bodily alignment. To understand how third parties typically react to the unfolding conflict in these situations, we provide the storyline of such a case below. It contains common features of disalignment situations that help us understand the conditions under which bodily alignment develops or not. The incident took place in a subway and was witnessed by at least eight third parties, the recorder excluded. The numbers in square brackets indicate the time elapsed since the start of the clip.

Storyline of clip 20 [0:00] Two men are exchanging provocations in a subway carriage. Antagonist A is standing in front of B who is sitting on a bench. B says: ‘Touch me’. A male bystander (1) sitting next to B swiftly turns his head to look at the incident. Another male bystander (2) sitting on the same bench stares ahead. A warns B: ‘[…unclear] I will beat you up good’. B responds: ‘Touch me’. [0:03] They repeat this sequence two times. Meanwhile, third parties are silently watching the scene or are looking away from the incident (Figure 1). [0:05] A moves closer to B and states: ‘Call me a nigger again’, to which B replies: ‘I won’t fucking touch you’. [A is Afro-American and B probably Pacific Islander]. A reacts: ‘Call me a nigger’. B replies: ‘Touch me’. [0:10] Bystander 1, sitting close to B, looks away, surveying the train. Bystander 2 seated next to him maintains the same staring posture, away from the incident. A continues: ‘Give me the opportunity… tough guy’. [0:12] A  steps back. B loudly says: ‘I will waste you, fucking bitch’. Bystander 2 continues to look away from the antagonists, two bystanders seated further away watch the incident. During the cycle of provocations, all third parties remain quiet. [0:13] A reacts to B’s insult by nodding dismissively. A male bystander (3) walks away, positioning himself further

Fig. 1 Still taken from clip 20 [0:03]. As one of the antagonists (standing, striped vest, dark cap) threatens his opponent (on bench with coat, hand between legs), two bystanders seated close by silently stare ahead.

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

28 • The British Journal of Criminology, 2022, Vol. 62, No. 1

back in the train. [0:14] B loudly says: ‘Touch me!’. A reacts by heaving up his arms. [0:16] As A lowers his arms again, B says: ‘Nigger’. [0:18] A swiftly slaps B in the face and continues to throw a series of hard punches at him. Now all third parties look at the fight but remain silent. [0:24]. A exclaims ‘What?! What?!’, energetically moving away and towards B. At this point a male bystander stands up from his seat and is recording the incident on his phone. Other third parties do not react. Bystander 1 looks at A, holding his phone to his chest. [0:27] A says: ‘I told you’ and nods his head. Bystander 1 pulls up his shoulders and quickly turns his head and trunk away as if to protect himself, right before [0:28] A launches a hard punch and a loud slap at B’s face, saying ‘I told you’. [0:30] Now third parties start to make sounds: a male, seated some steps away, yells ‘You got him…’ [unclear] at A and a female (not visible) adds: ‘OK, all right, all right, all right, all right’. [0:31] A steps away from B, saying ‘I told you’ while pointing his finger at B.

This case highlights three features common to disalignment situations. The first is the infre- quency of vocal responses. While the storyline indicates that third parties vocalized their con- cern after the second round of violence, they remained silent during the tense half minute that preceded it, from the altercation until the end of the second series of punching and slapping. Thirty seconds of silence is a long period of time, seen from the perspective of our clips.

The second common feature is that third parties prevent others from grasping their experi- ence, that they indeed notice the altercation. When bystander 1 was scanning the environment [0:10], he did not make eye contact with other third parties in the subway carriage. During the entire clip, he did not even face bystander 2, who was sitting close to him, as the latter main- tained his fixed gaze away from the incident until the violence broke out. The behaviour of by- stander 3 [0:13] is telling as well: by moving to the back of the carriage, he signalled that he did not want to be part of the situation, thus undermining potential attempts by other third parties, such as those of bystander 1, to build up a shared experience. Similarly, when a man got up from his seat to record the incident [0:24], others did not acknowledge this. More generally, disaligned third parties pass by or evade the incident, often glancing at it quickly before looking away. Bystanders in buses and trains keep their eyes on their phones, newspapers or books, gaze outside or pretend to do so. Some get up from their seats and walk away while the train or bus is still moving. In all of these instances, bystanders do not signal to the others present that they notice what is happening, or signal that they do not want to be part of the situation.

The third common feature has to do with the form of intervention. As third parties do not develop a shared sense of being in the same situation, bodily de-escalatory action either does not come off or is taken by single individuals mostly rather than as a collective endeavour. Thus, the average number of de-escalating third parties in situations of disalignment was 1.19, while it was 2.62 in the remaining situations.

Bodily alignment: circular formations We observed circular formations in 36 clips. To understand bodily alignment patterns and bod- ily de-escalatory action in these cases, we first present the storylines of clips 18 and 27—con- trasting cases that reveal commonalities observed in our data. We counted 29 bystanders in clip 18 and 51 in clip 27.

Storyline clip 18 [0:00] Antagonists A and B are arguing on a subway platform. Some bystanders look at the antagonists; others keep talking or look away. The bystanders are either alone or form lines of two or three standing next to each other. [0:02] A male bystander (1) extends his arm to the antagonists and yells: ‘Do it! Do it!’. Another male bystander shouts: ‘Pussy! Pussy!’ and a third male bystander loudly says: ‘Don’t do it. Don’t do it’. A female bystander adds: ‘No. No. No’. [0:04] The antagonists engage in aggressive posturing with their faces very close to

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

Circles of Peace • 29

each other. Most bystanders are now looking at the altercation. They form a ring around the antagonists with their bodies facing the incident. Some bystanders again shout ‘Don’t do it! Don’t do it!’ and ‘No, No’. Others yell: ‘Do it! Do it!’. Two male bystanders are recording the incident. [0:10] B gives A two hard blows to the face. A male bystander comments: ‘World Star!’ Another male bystander says: ‘Yeaaah!’ [0:14] Bystanders move closer to the fight. They make a lot of noise, ranging from ‘Beat his ass!’ to ‘Stop!’ and ‘Ooooohw’. [0:20] A  and B push and shove each other. Bystanders again move closer to the antagonists, encircling them. [0:21] Although most bystanders are looking at the incident, they assume postures that signal non-involvement: arms crossed over the chest, hands in pockets and hands folded over the belly (Figure 2). [0:24] A is on his knees. B gives him a series of hard punches to the back of his head. [0:26] A female bystander (2) standing behind the turn tiles shouts: ‘Stop! Let it go!’. [0:28] Antagonist B pulls A to the ground and they start to wrestle. [0:39] A bystander sits down near the antagonists, smiles to his companion’s camera and raises his fingers in a V-shape. [0:43] A now sits on top of B and tries to choke him. Two bystanders stand only a meter away, one filming the incident. The other bystanders remain a little further away, maintaining their non-involved posturing. The wrestling continues. [0:47] Bystanders encourage the antagon- ists: ‘Do everything you can!’ [0:50] A bystander suggests: ‘Head butt! Head butt!’. We hear bystanders sniggering. A female bystander shouts: ‘Break it up! Break it up!’ [0:55] Bystander 2 moves towards the antagonists. She holds a phone in her hands. [0:58] Standing near the wrestling antagonists, bystander 2 swings her arm in a wide movement, nearly touching the face of A. [0:59] She then makes a swift, swiping gesture with her lower arm, pointing at B, but she does not touch the antagonists. She shouts: ‘Stop!’ (Figure 3). [1:01] A punches B in the face. Bystander 2 says ‘No!’ again. She puts her arms between the antagonists shortly and then pulls them away. [1:03] A and B continue to wrestle on the ground. [1:04] Bystander 2 steps back. She interrupts her intervention temporarily, looks around and yells: ‘Get off him’. Then she touches her hair. Male bystanders begin to move around indecisively, sidestepping. [1:05] Antagonist A frees himself from B’s grip and stands up. As A is about to kick B who is lying on the ground, bystander 2 extends her arms forward and steps in between them, shouting ‘Stop!’ This does not prevent the antagonists from kicking each other. [1:10] A and B are on their feet, with their backs turned at bystander 2. Bystander 2 extends her underarms, hand palms

Fig. 2 Still taken from clip 18 [0:21]. Bystanders in a circular formation around the antagonists (not visible, behind the bystander in white shirt and the bystander in dark shirt with cap right to the pillar in the middle). Their posturing signals non-involvement. The bystander in the left-hand corner has his arms crossed over his chest; the bystander left to the pillar in the middle near the turnstile has his hands in his pockets; the two bystanders in the right-hand corner have their hands folded over their bellies.

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

30 • The British Journal of Criminology, 2022, Vol. 62, No. 1

downward. She shouts: ‘Stop it!’. Her hands go up and down, emphasizing her words. The antagonists do not react. [1:12] A male bystander (3) steps forward to intervene, immediately followed by another male bystander (4). Bystander 3 touches the back of antagonist A. He pushes him aside and then steps in between the antagonists, separating them by extending his arms, hand palms open. [1:14] Bystander 4 joins the intervention. He pats A on the back and then touches his arm. Bystander 1, who has been standing close to the fight all this time, pulls bystander’s 3 arm down. [1:15] Bystander 2 yells ‘Go home!’ to B. [1:16] Bystander 3 gestures to B and bystander 1, then turns his back at them, separating them from A. Bystander 4 extends his underarm, as to prevent A from moving to B. [1:18] A leaves the platform.

Storyline clip 27 [0:00] Antagonist A is phone recording B at an airport check-in area. B walks aggressively to A. Bystanders are standing in queues, waiting for their turns; only a few are attentive to the an- tagonism. [0:07] B tries to grab A’s phone while making sexual gestures close to A. Bystanders look at the altercation but remain in their queues and keep quiet. Airport officials are busy assisting customers at their desks. [0:10] B tries to grab A’s phone twice. [0.13] A remains at his position near a desk. B steps away from A and then approaches him again, bending his head close to B’s face. [0:18]. A male bystander (1) walks towards the antagonists and stops about three meters away. He says something to B, who turns around and gives him the finger. [0:20] As the camera moves, two other bystanders (2, 3) appear to approach the antagonists. They form a line, standing next to each other with their bodies facing the antagonists, about four to five meters away. [0:21] B turns his back to the bystanders and continues to argue with A. [0:23] B again turns around, saying something to bystander 1 who is standing closest to him. [0:24] Right after that, bystanders 1–3 simultaneously move two to four steps in the direction of B (Figure 4). [0:27] B moves away from A, getting closer to the three bystand- ers. [0:29] B drops his coat [signalling his willingness to fight]. [0.31] Bystander 3 tries to calm down A. He extends his under arms, palms open and facing upwards. [0:35] B walks towards bystander 3. Bystander 3 says: ‘You don’t want to go to jail’ [0:36] Other bystanders move closer to B. A  is again recording B, from only a meter away. [0:38] Bystander 3 asks again: ‘Do you want to go to jail?’. B shrugs his shoulders. He says: ‘I don’t care’ [? unclear] [0:42] B points at A. He says: ‘I don’t give a damn’. Four bystanders form a circular formation about two

Fig. 3 Still taken from clip 18 [0:59]. Bystanders who had encircled the wrestling antagonists watch a female bystander intervene. After her attempt, the bystander in the dark shirt with the white star in the left-hand corner, the bystander with the bottle in his right hand in the left-hand corner and the bystander in the white shirt behind the intervening female will make indecisive movements.

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

Circles of Peace • 31

meters from the antagonists (Figure 5). [0:43] B tries to grab A’s phone. One bystander says: ‘Don’t do it’. [0:44] B points at A. He says: ‘I don’t give shit’. [0:55] Bystander 3 asks: ‘What are you upset about?’ [1:03] B again points aggressively at A, now extending two arms. B says: ‘What I am upset about is this faggot right here’. [1:05] B acts as if he holds a camera in his hands. He says: ‘This photographer is taking pictures of me’. [1:09] B suddenly turns to A and says: ‘Shove it up your ass man’ and gives A the finger. [1:10] A is smiling at B. He stands with his arms crossed over his chest. A female bystander joins the circular formation. Other by- standers watch while they remain in their queues. [1:12] B continues to A: ‘See how that feels’. [1:13] B abruptly pushes his hips upward in a sexual gesture. A maintains the same posture, still smiling. [1:14] A female bystander says: ‘No! No!’ [1:16] B says ‘Yeah’ and suddenly kicks A in the legs. A bends downwards and B smacks him hard on the back of his head. Bystanders

Fig. 4 Still taken from clip 27 [0:24]. Three bystanders (from right to left: male, checkered shirt, dark pants, far right-hand corner; male standing next to pillar with backpack; male with dark vest holding bag and paper in hand) moved simultaneously towards an aggressive antagonist (standing behind male with checkered shirt and backpack near the white line). Two of them (male, checkered shirt, dark pants, far right-hand corner; male standing next to pillar with backpack) had formed a line seconds before.

Fig. 5 Still taken from clip 27 [0:42]. An antagonist (seen from the back, light shirt with dark lines) is pointing at his opponent (dark vest, dark pants, hands next to his hip). Four third parties encircle them, facing the incident. One of third parties (male, checkered shirt, dark pants) stands close to the pointing antagonist and tries to calm him down.

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

32 • The British Journal of Criminology, 2022, Vol. 62, No. 1

yell: ‘Oh no!’, ‘Stop it!’, ‘Oh my God’ and ‘Don’t do it!’ [1:18] Five bystanders rush at B and grab him (Figure 6). [1:22] B is pushed to the ground and the five bystanders pile upon him.

Clips 18 and 27 reveal two common features that require detailed discussion: (1) the presence of vocal responses and (2) the positioning of third parties into spatial-bodily formations.

First, vocal responses can contribute to the development of a shared definition of the situ- ation. In many clips, we heard responses similar to those in the storyline of clip 27. They took the form of simple, repetitive startle responses which signalled an urge to stop the fight: ‘Hey, hey, hey’, ‘Stop it, stop it, stop it’ and ‘No, no, no’. Third parties also appealed to the antagonists to walk away from conflict, telling them to ‘Relax’, ‘Chill out, chill out’, ‘Let it go’, ‘Come on’, ‘Walk away, man’ and ‘Ignore it, ignore it’. Other third parties advised weighting the costs and benefits of pursuing a violent line of action: ‘It’s not worth it’, ‘Not worth it guys’, ‘It’s just rub- bish’, ‘Do you want to go to jail?’ and ‘Go to your family!’.

In other cases, vocal responses were more equivocal. The storyline of clip 18 shows how third parties’ vocal responses proposed contrasting definitions of the situation. On the one hand, the giggling, sniggering and commentary underscoring violent actions clearly suggested that bystanders were enjoying the spectacle, with slaps, punches to the face and moves that felled an opponent eliciting ‘Oohw!’, ‘Aahw!’ and ‘Yooohw!’. Such exclamations of excitement also appeared in verbal form when bystanders yelled the name of the website where clips of vio- lence can be uploaded: ‘World Star shit happening right here!’ or ‘YouTube!’ Bystanders also encouraged the fighters, saying things like ‘Come on man’ and ‘Do it, do it’. Clearly, the giggling, excited outcries and explicit encouragements to fight presented opposing definitions of the situ- ation and proposed contrasting lines of action to the startle responses and verbal de-escalatory interventions of other third parties. However, even in situations of equivocality vocal responses signal that the incident has caught the attention of third parties, thus opening up their experi- ence of the situation to others.

However, without the alignment of bodies into a distinct spatial form, the mutual recognition of sharing a focus of attention is less likely to develop as it is harder for third parties to see who is being attentive to the incident and who is not; as Table 2 shows, our coding of vocalizations and pointing gestures hardly increased the IRR of the number of intervening bystanders, while our

Fig. 6 Still taken from clip 27 [1:18]. Immediately after the aggressive antagonist (seen from the back, white shirt, dark lines) kicks and smacks his opponent, three third parties rush forward to grab him. Two other third parties come running from further away (male, short-sleeved shirt, dark pants, right hand corner; male, dark short-sleeved shirt, light pants, left-hand corner.)

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

Circles of Peace • 33

observation of situational group formation did. The storyline of clip 18 shows how third parties created a circular formation as they turned their bodies to watch the incident on the platform. They remained in this formation and moved closer to the antagonists as the incident unfolded. In other cases, we observed such formations developing out of prior line formations. Once third parties halt or change their prior positions to watch the incident, they tend to position them- selves next to each other, as if on an imaginary line. This happened in clip 27 [0:20]. Rather than occupying dispersed positions in space to view the incident, bystanders often gravitate towards other bystanders, adjoining their bodies and standing with their faces oriented towards the in- cident.

When line formations remained stable during the incident and third parties moved closer to the antagonists, they tended to curve around them, creating crescent shapes or more completely rounded circular formations (see Figures 3 and 5). Compared to line formations, this encircling provides optimal opportunities for each third party to notice their monitoring of one another. We thus only consider circular formations as full-fledged situational groups. Another import- ant difference between line and circular formations is that the latter more clearly separate third parties who are attentive to the incident from others who are just passing by. As they face the unfolding incident, third parties in circular formations literally turn their backs to others who are not part of the formation and to whatever else may then be happening in the public space.

CO N CLU S I O N A N D D I S C U S S I O N Our work advances prior research by revealing the situational process that is conducive to the collectivization of de-escalatory behaviour: bodily aligned bystanders generate situational groups which increase the number of intervening third parties. Given the time lags between their interventions, we suggest that situational group formation reduces individual bystand- ers’ risk perceptions—although this latter finding is based on a small number of cases only. Individual calculus is very likely to come into play where the environment is unfamiliar, where the bystander is wary of retreat if caught in the fight (Darley and Latane 1970; Dhattiwala 2016).

Our qualitative observations emphasize the relationship between bodily alignment and by- standers’ opening up their experience of the situation to others. Circular formations not only provide more opportunities for each third party to notice their monitoring of one another, but they also more clearly separate third parties who are attentive to the incident from others who are not. Both features facilitate the development of a shared sense of being together in the same situation.

It could be argued that the formation of situational groups is itself already a form of interven- tion, and that there may be a selection process in which third parties who may de-escalate form circular arrangements. But if the individual propensity to intervene were to be the actual causal force, we would expect that the number of people taking de-escalatory action would increase proportionally when more people are present in the situation, instead of being related to situ- ational group formation. However, our results show that the number of intervening bystanders is unrelated to how many people are present.

To our knowledge, these findings have not been described earlier. We were able to gener- ate them after a careful inductive analysis of naturally occurring incidents captured on video material. Theoretically, our study contributes to a criminological understanding of assemblies as being capable of generating purposive, self-regulatory collective action that enhances public safety. Physical co-presence—incidental as it may be—can produce a collective self-awareness that can circumvent the uncertainty of a risky environment. Such conceptual move is needed, because outdated ideas about the dangerous or stultifying influences of gatherings on the be- haviour of individuals still inform public debate and policy making (Stott and Drury 2017).

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

34 • The British Journal of Criminology, 2022, Vol. 62, No. 1

One remaining question is why the formation of situational groups results in conciliatory rather than escalatory behaviour. Experimental studies show that strong identification with an in-group leads to more in-group support of attackers’ behaviour, while weaker identi- fication leads to less support, greater feelings of anger towards perpetrating group members (Mummendey and Otten 1993; Gordijn et al. 2006) and a greater tendency to defuse the situ- ation (Slater et  al. 2013: 10). These studies suggest that third-party behaviour is more likely to be de-escalatory when they are part of groups sharing weaker, momentarily identities, for instance when they are among strangers in public urban spaces.

The broader implications of our findings become clear once we acknowledge that situational groups cannot exist in a social vacuum. Recent events of police brutality in the United States urge us to scrutinize the relationship between third-party situational group formation and the structural asymmetry between antagonists. Would situational groups generate the same ten- dency to intervene where power relations between antagonists are skewed? In the recent police killing of George Floyd in Minneapolis, the hesitation of bystanders to ‘call the police on the police’ would be understandable because the police possess a mandate that legitimates and ob- ligates violence (Manning 1980). Intervening a police officer conducting obligatory duty can lead to an arrest or even potential death.

The police–civilian encounter is a relation of extreme asymmetry. But asymmetry underlines situations of any violent conflict even when it is relatively more tractable. Gaining tractability, in other words moving towards de-escalation, implies a reduction in asymmetry between not only the antagonists but also between third parties and the antagonists.

Two contemporary developments likely contribute to a declining tolerance of explicit power and status asymmetries between people in urban public spaces, providing third parties with greater leeway to intervene. First, Wouters (2007: 217–9) argues that the moral regulation of demeanour in urban public spaces in the western world has become more informal and less hier- archical, with overt expressions of superiority and inferiority seen as offensive and shameful. Indeed, research shows that most complaints about the behaviour of strangers in urban public spaces involve the denial of mutuality and equality (Horgan 2019). People complain about those who claim too much space (e.g. bumping into others), too much attention (e.g. prolonged staring), or demean others based on status differences, thus making hierarchy explicit (e.g. cat- calling and racial slurs). Second, while the recording and mass dissemination of video footage of (street) violence is undoubtedly related to voyeurism, recording and watching these videos can also spark engagement (Tester 2001: 133, quoted in Carrabine 2012: 467). More specifically, psychological research into ‘morbid curiosity’ shows that people watch such footage to prepare themselves to being confronted with similar situations, prompting the question ‘what would I do’ (Niehoff and Oosterwijk 2020). Moreover, the awareness of visibility itself probably con- tributes to a further levelling of relations in urban public space (including those between civil- ians and police officers, see Goldsmith 2010).

Despite the pervasive presence of the visual in society, its analysis only takes up a marginal slice of criminological inquiry (but see Carrabine 2012). We hope that our paper sparks crim- inological interest in video material, not just as data, but also how its recording and dissemin- ation change the relationships between civilians and between civilians and the state in matters of crime, social control and legitimacy.

F U N D I N G This work was supported by the European Research Council under the European Union’s Horizon 2020 research and innovation programme, Consolidator Grant no. 683133 awarded to Don Weenink. This article also benefitted from a fellowship at the Paris Institute for Advanced

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

Circles of Peace • 35

Study (France) granted to Don Weenink, with the financial support of the French State, pro- gramme ‘Investissements d’avenir’ managed by the Agence Nationale de la Recherche (ANR- 11-LABX-0027-01 Labex RFIEA+).

A CK N O W L E D G E M E N TS We would like to thank the following (PhD) students who assisted us in this research: Sander Brinkhuijsen, Laura Keesman, Marlou Koch, Rozalie Lekkerkerk, Max Noback, Amber MacDonald, Boje Moers, Floris Mosselman, Sarina Nooitgedacht, Phie van Rompu, Rein de Sauvage Nolting, Kariem el Sharkawy and Melissa Sexton. Also we thank Michael Biggs and our colleagues at the Department of Sociology at the University of Amsterdam and the two anonymous reviewers for providing us with valuable comments that have improved the paper.

R E F E R E N CE S Baumann, Z. (1993), Postmodern Ethics. Blackwell. Bloch, C., Liebst, L. S., Poder, P., Christiansen, J. M. and Heinskou, M. B. (2018), ‘Caring Collectives and

Other Forms of Bystander Helping Behavior in Violent Situations’, Current Sociology, 66: 1049–69. Briggs, P., Churchill, E., Levine, M., Nicholson, J., Pritchard, G. and Olivier, P. (2016), ‘Everyday Surveillance

Conference Paper’, Workshop Summary CHI’16 Extended Abstracts, 7–12 May, San Jose, CA. Carrabine,  E. (2012), ‘Just Images: Aesthetics, Ethics and Visual Criminology’, The British Journal of

Criminology, 52: 463–89. Collins, R. (2004), Interaction Ritual Chains. Princeton University Press. Corneil, M. K. (2012), Handbook of Participatory Video. AltaMira Press. Darley, J. M. and Latane, B. (1970), The Unresponsive Bystander: Why Doesn’t He Help? Appleton Century

Crofts. Darley,  J.  M., Teger,  A.  I. and Lewis,  L.  D. (1973), ‘Do Groups Always Inhibit Individuals’ Responses to

Potential Emergencies?’, Journal of Personality and Social Psychology, 26: 395–9. Dhattiwala,  R. (2016), ‘The Ecology of Ethnic Violence: Attacks on Muslims of Ahmedabad in 2002’,

Qualitative Sociology, 39: 71–95. Durkheim, E. (1995/1912), The Elementary Forms of Religious Life. The Free Press. Felson, R. B. and Tedeschi, J. T. (1993), Aggression and Violence: Social Interactionist Perspectives. American

Psychological Association. Fischer, P., Greitemeyer, T., Pollozek, F. and Frey, D. (2006), ‘The Unresponsive Bystander: Are Bystanders

More Responsive in Dangerous Emergencies?’, European Journal of Social Psychology, 36: 267–78. Fischer, P., Krueger, J. I., Greitemeyer, T., Vogrincic, C., Kastenmüller, A., Frey, D., Heene, M., Wicher, M.

and Kainbacher, M. (2011), ‘The Bystander-Effect: A Meta-analytic Review on Bystander Intervention in Dangerous and Non-dangerous Emergencies’, Psychological Bulletin, 137: 517–37.

Freelon,  D. (2013), ‘ReCal OIR: Ordinal, Interval, and Ratio Intercoder Reliability as a Web Service’, International Journal of Internet Science, 8: 10–6.

Goffman, E. (1961), Encounters: Two Studies in the Sociology of Interaction. Bobbs-Merrill Company. Goldsmith, A. J. (2010), ‘Policing’s New Visibility’, The British Journal of Criminology, 50: 914–34. Goodwin, C. (1994), ‘Professional Vision’, American Anthropologist, 96: 606–33. Gordijn,  E.  H., Yzerbyt,  V., Wigboldus,  D. and Dumont,  M. (2006), ‘Emotional Reactions to Harmful

Intergroup Behavior’, European Journal of Social Psychology, 36: 15–30. Greitemeyer, T. and Mügge, D. O. (2015), ‘When Bystanders Increase Rather Than Decrease Intentions to

Help’, Social Psychology, 46: 116–9. Harari, H., Harari, O. and White, R. V. (1985), ‘The Reaction to Rape by American Male Bystanders’, The

Journal of Social Psychology, 125: 653–8. Horgan, M. (2019), ‘Everyday Incivility and the Urban Interaction Order: Theorizing Moral Affordances in

Ritualized Interaction’, Journal of Language Aggression and Conflict, 7: 32–55. Jones, N. and Raymond, G. (2012), ‘“The Camera Rolls” Using Third-Party Video in Field Research’, The

Annals of the American Academy of Political and Social Science, 642: 109–23. Kass, R. E. and Raftery, A. E. (1995), ‘Bayes Factors’, Journal of the American Statistical Association, 90: 773–95.

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

36 • The British Journal of Criminology, 2022, Vol. 62, No. 1

Latane, B. and Darley, J. M. (1968), ‘Group Inhibition of Bystander Intervention in Emergencies’, Journal of Personality and Social Psychology, 10: 215–21.

Levine,  M. and Manning,  R. (2013), ‘Social Identity, Group Processes, and Helping in Emergencies’, European Review of Social Psychology, 24: 225–51.

Levine, M., Taylor, P. J. and Best, R. (2011), ‘Third Parties, Violence, and Conflict Resolution: The Role of Group Size and Collective Action in the Microregulation of Violence’, Psychological Science, 22: 406–12.

Liebst, L. S., Heinskou, M. B. and Ejbye-Ernst, P. (2018), ‘On the Actual Risk of Bystander Intervention: A Statistical Study Based on Naturally Occurring Violent Emergencies’, Journal of Research in Crime and Delinquency, 55: 27–50.

Liebst, L. S., Philpot, R., Bernasco, W., Dausel, K. L., Ejbye‐Ernst, P., Nicolaisen, M. H. and Lindegaard, M. R. (2019), ‘Social Relations and Presence of Others Predict Bystander Intervention: Evidence from Violent Incidents Captured on CCTV’, Aggressive Behavior, 45: 598–609.

Lindegaard, M. R. and Bernasco, W. (2018), ‘Lessons Learned from Crime Caught on Camera’, Journal of Research in Crime and Delinquency, 55: 155–86.

Manning, P. K. (1980), ‘Violence and the Police Role’, The Annals of the American Academy of Political and Social Science, 452: 135–44.

Manning, R., Levine, M. and Collins, A. (2007), ‘The Kitty Genovese Murder and the Social Psychology of Helping: The Parable of the 38 Witnesses’, American Psychologist, 62: 555–62.

McPhail,  C. (1994), ‘From Clusters to Arcs and Rings: Elementary Forms of Sociation in Temporary Gatherings’, Research in Community Sociology, 4(suppl. 1): 35–57.

Milgram, S. and Toch, H. (1969), ‘Collective Behavior: Crowds and Social Movements’, in G. Lindzey and E. Aronson, eds., Handbook of Social Psychology, Vol. IV, 2nd ed, 507–610. Addison-Wesley.

Mogan, R., Fischer, R. and Bulbulia, J. A. (2017), ‘To Be in Synchrony or Not? A Meta-analysis of Synchrony’s Effects on Behavior, Perception, Cognition and Affect’, Journal of Experimental Social Psychology, 72: 13–20.

Mummendey, A. and Otten, S. (1993), ‘Aggression: Interaction between Individuals and Social Groups’, in R.  B.  Felson and J.  T.  Tedeschi, eds., Aggression and Violence: Social Interactionist Perspectives, 145–67. American Psychological Association.

Niehoff,  E. and Oosterwijk,  S. (2020), ‘To Know, to Feel, to Share? Exploring the Motives That Drive Curiosity for Negative Content’, Current Opinion in Behavioral Sciences, 35: 1–6.

Philpot,  R., Liebst,  L.  S., Levine,  M., Bernasco,  W. and Lindegaard,  M.  R. (2020), ‘Would I  Be Helped? Cross-National CCTV Footage Shows That Intervention Is the Norm in Public Conflicts’, American Psychologist, 75: 66–75.

Slater, M., Rovira, A., Southern, R., Swapp, D., Zhang, J. J., Campbell, C. and Levine, M. (2013), ‘Bystander Responses to a Violent Incident in an Immersive Virtual Environment’, PLoS ONE, 8: 1–13.

Stott, C. and Drury, J. (2017), ‘Contemporary Understanding of Riots: Classical Crowd Psychology, Ideology and the Social Identity Approach’, Public Understanding of Science, 26: 2–14.

Tester, K. (2001), Compassion, Morality and the Media. Open University Press. Thompson, John B. (2005), ‘The New Visibility’, Theory, Culture & Society, 22: 31–51. van Bommel, M., van Prooijen, J., Elffers, H. and van Lange, Paul A. M. (2014), ‘Intervene to be Seen: The

Power of a Camera in Attenuating the Bystander Effect’, Social Psychological and Personality Science, 5: 459–66.

Wouters, Cas. (2007), Informalization: Manners and Emotions Since 1890. Sage.

D ow

nloaded from https://academ

ic.oup.com /bjc/article/62/1/18/6299950 by Jnls C

ust Serv on 03 February 2022

© 2022 Centre for Crime and Justice Studies (formerly ISTD). Copyright of British Journal of Criminology is the property of Oxford University Press / USA and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.