Rethinking the Bystander Effect in Violence Reduction Training Programs

Introduction

Violence reduction in general, and interpersonal violence reduction in particular, has been a key aim of global public health policies for the last three decades (Krug, Dahlberg, Mercy, Zwi, & Lozano, 2002). Psychological theories have played a key role in attempts to ensure a robust evidence base for the development of violence reduction interventions (Hollin, 2016; WHO, 2009). Early work drew on ideas about the psychology of perpetrators (Holtzworth-Munroe & Stuart, 1994) or the psychology of victims (Seligman, 1972). More recently, however, there has been a recognition of the key role that can be played by bystanders and third parties in violence prevention (Banyard, 2011; Coker et al., 2011). These programs draw, in part, from classic work on the social psychology of bystander behavior (Latané & Darley, 1970). The inclusion of the bystander in violence prevention has led to a change in focus in intervention initiatives. Attention is focused as much on the wider community as it is on challenging perpetrator behavior and supporting victims. In this paper, we will argue that, despite the clear advances that the bystander turn has brought to the development of violence reduction initiatives, these programs are in danger of being undermined by a misunderstanding of research in the bystander tradition—particularly as it applies to aggression and violence. To the best of our knowledge, all violence reduction initiatives that include a bystander component draw uncritically on the idea of the “bystander effect” (Darley & Latané, 1968). This is the notion that the more people who are present during an emergency incident, the less likely any single bystander is to intervene. Bystander intervention programs often have a component designed to counteract this effect. Based on the theory which underpins the classic bystander effect, these programs seek to inoculate the individual from the supposed negative effect of the presence of others and seek to empower the individual to act in spite of the group.

Recent theoretical and empirical work on bystander behavior has shown that the classic view of the bystander effect is not supported by the evidence—particularly in the context of aggression and violence. Meta-analyses (Fischer et al., 2011, including a reanalysis of the original meta-analysis in this field; Stalder, 2008) show that the bystander effect does not hold in violent or dangerous emergencies and that people are more likely to be helped when more bystanders are present. In fact, Fischer et al. (2011) propose a “reverse bystander effect” (the greater the number of bystanders, the greater the likelihood of intervention) when emergencies are less ambiguous and it is clear what bystanders should do. Moreover, recent research which uses CCTV footage to study real-life bystander behavior in violent or dangerous emergencies in public spaces (as opposed to work using laboratory or self-report measures) shows that bystander intervention is actually the norm in these kinds of incidents (Philpot, Liebst, Levine, Bernasco, & Lindegaard, 2019). Of course, not everybody intervenes, but intervention levels are very high, and the more bystanders are present, the more likely it is that the victims receive help. At the same time, there has been significant development in theoretical tools for understanding the behavior of bystanders. Work drawing on the social identity approach (SIA: Haslam, 2001; Reicher, Spears, & Haslam, 2010) by Levine and colleagues (Levine, Prosser, Evans, & Reicher, 2005; Levine & Manning, 2013) has demonstrated that it is the social identity relationships between bystander, victim, and perpetrator (rather than the simple presence or absence of bystanders) that is key to understanding bystander behavior.

The aim of this paper therefore is to challenge the taken-for-granted inclusion of the bystander effect in violence reduction programs, to review the evidence base for the role of bystanders in violent and dangerous emergencies, and to make recommendations for how this new body of evidence should inform violence reduction initiatives in the future.

Bystander Informed Violence Reduction Programs

Over the last three decades, there has been a significant rise in the number of violence reduction programs that have been developed and deployed. Much of the impetus for developing these initiatives came in part from changes in the law in the United States. For example, the Violence Against Women Act (VAWA) was signed into federal law in 1994. The VAWA was the culmination of decades of activism around the prevalence of domestic violence and sexual assault. The legislation changed the landscape in a way that made it incumbent on states and institutions to adopt violence reduction initiatives. At first, these focused mainly on trying to identify and change the characteristics of perpetrators and victims of gender-based violence. However, the relative ineffectiveness of this kind of approach led to the development of programs designed to try and encourage and empower third parties/bystanders to intervene. This shift in emphasis coincided with increasing concern about sexual assault and violence in American colleges and university campuses. For example, Koss and colleagues (Koss, Gidycz, & Wisniewski, 1987; Koss, Dinero, Seibel, & Cox, 1988) began publishing surveys of what they described as “stranger and acquaintance rape” based on samples from college students. This new focus on campus-based sexual violence leveraged existing education legislation (including Title IX of the Education Amendment Act prohibiting discrimination or social exclusion based on sex) and developed bespoke legislation (including the Cleary Act (1990) and the more recent Campus SaVE (Campus Sexual Violence Elimination) Act (2013)). This body of legislation has meant that colleges and universities in America have increasingly been required to report on incidents of violence and to deliver violence reduction training programs.

It is against this background that the form and content of violence reduction programs began to emerge. For example, the programs were designed and implemented without time for pilot work or systematic program evaluation prior to deployment (Coker et al., 2011; Jessup-Anger, Lopez, & Koss, 2018). They were constructed out of existing theoretical approaches that were assumed a priori to be relevant. While a number of different programs were developed in different places, they all shared a similar hybrid approach and tended to draw on similar theoretical tools (Jouriles, Krauss, Vu, Banyard, & McDonald, 2018). These included work packages which challenged norms and values around violent or sexual assault, work packages designed to improve intervention skills and feelings of personal and collective efficacy, and work packages tackling the bystander effect.

Although there are a large number of these intervention programs in existence, for the purposes of this paper, we will focus on just three. These will include the best-known campus and community-based interventions from the United States and the United Kingdom. These are the Green Dot program (Green Dot, 2010), Bringing in the Bystander (Banyard, Moynihan, & Plante, 2007) and The Intervention Initiative (Fenton & Mott, 2018; Fenton, Mott, McCartan, & Rumney, 2014). We choose these as they are programs that have been subjected to systematic review, and thus we can explore evidence for their efficacy.

Green Dot (2010) is a sexual violence prevention program where prevention efforts are focused on changing participants’ violence acceptance and engagement as potential bystanders in their community. The program builds awareness of sexual assault and bystander intervention as a manageable activity. Participants learn about the categories of influence based on the ecological model—personal, relationship, and general issues (audience inhibition, diffusion of responsibility, etc.) Participants are then trained to notice the situation and identify it as a problem (“red dot”), to identify bystander behaviors they can do individually and collectively (“green dots”), and to generate the confidence to perform them. The three “D's” Direct (approaching the situation yourself)—Delegate (asking others to help)—Distract (diverting people's attention) outline the options of direct and indirect intervention in situations of high risk.

Bringing in the Bystander (Banyard et al., 2007; Banyard, Moynihan, & Crossman, 2009; Inman, Chaudoir, Galvinhill, & Sheehy, 2018; Moynihan & Banyard, 2008; Moynihan, Banyard, Arnold, Eckstein, & Stapleton, 2010) is a multicomponent violence reduction curriculum that combines two psychological models: the transtheoretical model of change (Prochaska, DiClemente, & Norcross, 1992) and the community readiness model (Edwards, Jumper-Thurman, Plested, Oettin, & Swanso, 2000). The program utilizes the idea of achieving greater helping behavior through a greater sense of community (Banyard, 2008). The curriculum includes empathy building, changing beliefs about rape myths, increasing sense of responsibility, and finding solutions to intervene through empathy-building exercises and brainstorming. Participants then sign pledges to be active bystanders.

The Intervention Initiative (Fenton & Mott, 2018; Fenton et al., 2014) is an intervention program designed for the United Kingdom context, which draws on existing evidence from programs developed in the US. The eight-session toolkit is underpinned by theories of classic bystander intervention, theory of change, and social norms. The program equips participants with knowledge about bystander intervention and the negative effect of predictors, sexual and domestic violence, law, and sexism and is aimed at increasing responsibility, empathy, and knowledge about bystander intervention strategies. Participants then rehearse behavior and attitudes through role-play and individual tasks.

Effectiveness of Existing Violence Reduction Programs

There have been a number of reviews of the effectiveness of these kinds of bystander intervention programs (e.g., DeGue et al., 2014; Jouriles et al., 2018; Katz & Moore, 2013; Labhardt, Holdsworth, Brown, & Howat, 2017; Mujal, Taylor, Fry, Gochez-Kerr, & Weaver, 2019; Storer, Casey, & Herrenkohl, 2016). The reviews recognize that there are considerable challenges in being able to determine the effectiveness of such programs. There are often practical, ethical, and logistical hurdles which make conclusive evaluations problematic. For example, the gap between the delivery of intervention initiatives and the recording of incidents of sexual assaults allows for correlational rather than causal conclusions (Jouriles et al., 2018). At the same time, conditions of anonymity and confidentiality during data collection mean that it is often not possible to follow the same individuals through repeated data collection points (e.g., Coker, Bush, Brancato, Clear, & Recktenwald, 2019). This means that it is often not possible to see changes in individuals exposed to treatments, and only cohort effects are available for analysis (Coker et al., 2011, 2015). On a structural level, funding for long-term evaluations is difficult to secure and so pre- and postintervention measures tend to be close together (Banyard et al., 2007; Powell, 2011, 2014). This makes it very difficult to tell if any of these programs have long-term impacts on rates of sexual violence and assault. Given that the intervention programs are all multifaceted, there is also the difficulty of disambiguating which elements of the “treatments” might be effective, and which might not. Even if a program might be deemed to be effective, it is important to know whether some or all of the elements are effective, and whether any might be counterproductive. Finally, the programs tend to collect self-report rather than behavioral data, so even where there might be evidence of changes in attitude or expressed intention to act, it remains uncertain that these intentions will be manifested in actual behavior (e.g., Cissner, 2009; Coker et al., 2019; Fouber, Langhinrichsen-Rohling, Brasfield, & Hill, 2010; Gidycz, Orchowski, & Berkowitz, 2011; Moynihan et al., 2011).

Having acknowledged these caveats, we can now review the evidence for the effectiveness of existing programs. There are clear positive effects on some attitude dimensions, and on some self-reported behaviors. For example, Coker and colleagues (2011) carried out a large-scale evaluation of the Green Dot program and found that training had a positive effect on rape myth reduction. They also found that the training had a positive effect on self-reported observations of, and participation in, active bystander behaviors. A second study (Coker et al., 2015) found evidence that violence victimization and violence perpetration rates were significantly lower in a college that had delivered the Green Dot program than controls who had not. More recently, Coker and colleagues (2019) evaluated dating and sexual violence acceptance among high school students in a 5-year randomized control trial and found significant reductions in dating and sexual violence myth acceptance in years 3 and 4 for both genders on the school level. Interestingly, this result included those who did not directly receive the training. The researchers suggest that the effect of the program diffused to those who did not receive the training via those who received training. This suggests that Green Dot–trained participants have a positive impact on others/bystanders and can promote prosocial behavior in their groups. What remains unclear is the relative contribution of different aspects of the program to the overall treatment effect, the mechanisms by which the program might be working, and the causal link from the program to independently recorded behavior rather than self-reported behavior.

The Bringing in the Bystander program has also been subject to several evaluations (Banyard et al., 2007; Inman et al., 2018; Moynihan et al., 2010, 2015; Senn & Forrest, 2016). These evaluations tend to find decreases in rape myth acceptance and increases in self-reported feelings of efficacy, willingness to intervene, and acts of intervention. There is also some evidence of the perseverance of these effects at 4 month and 12 month follow-ups. What remains unresolved is the relationship to actual rather than self-reported behavior, whether the intervention works beyond the confines of a single university (but see Inman et al., 2018), and what the mechanisms by which the intervention might work actually are.

The Intervention Initiative is a more recent program for a UK context and has had less evaluative attention. However, Fenton and Mott (2018) show in a sample of university students that exposure to The Intervention Initiative led to a reduction in rape myth acceptance, domestic abuse myth acceptance, and a significant decrease in expression of denial. At the same time, self-reported efficacy, willingness to help, and feelings of responsibility increased. However, self-reported prosocial bystander behavior did not increase significantly from pretest to posttest immediately after taking part in the program. In common with the limitations of the US programs, the relationship to actual behavior rather than self-report behavior, the mechanisms by which the intervention might actually work, and the longer-term impact of the intervention are still without an evidence base.

It is clear from this brief review of reviews that the evaluation of existing bystander programs reveals a mixed picture in respect of the evidence for practical effectiveness. It seems programs can change norms and attitudes (around “rape myths” and the acceptability of perpetrator behavior), and self-reported feelings of efficacy, empowerment, willingness to intervene, and intentions to intervene. However, the evidence that this actually impacts on real-life intervention; or is responsible for a reduction in sexual assaults and violence, is much less strong. Similarly, there is a failure to explore how the elements of these programs might work (singularly or in combination) and what the mechanisms for effectiveness might be. The fact that none of the intervention programs have been concerned with the relative effectiveness of their constituent parts leads to some existential difficulties for all these violence reduction programs. This is because, if some elements of the program turn out to be built on weaker theoretical foundations, that may in turn undermine the effectiveness of the programs as a whole.

The Bystander Effect in Violence Reduction Programs

One element that we will argue is a shared weakness of all these violence reduction programs is their uncritical adoption from social psychology of research on the “bystander effect.” The bystander effect (Darley & Latané, 1968; Latané & Darley, 1970) argues that the presence of others in public space leads to the activation of concepts like “diffusion of responsibility,” “pluralistic ignorance,” or “audience inhibition,” which in turn inhibit the likelihood of intervention in emergency situations. This has meant that policy makers and intervention designers have focused on the question of how to overcome the forces which seem to prevent bystanders from helping others.

Darley and Latané’s work (Darley & Latané, 1968; Latané & Darley, 1970) on the bystander effect amplifies the idea (developed in tandem with other iconic American social psychologists like Milgram, 1970, and Zimbardo,1969a,1969b) that threats to public safety can be located in the presence or absence of others. Taking inspiration from a report in the New York Times in March 1964, which claimed that 38 witnesses had watched a man stalk and kill Kitty Genovese without intervening, they set out to study the case empirically. Their experimental work focused on the presence or absence of others as the key variable in explaining bystander decision making. The laboratory analogues regularly manipulated the number of bystanders and the ambiguity of whether action was required in a variety of helping situations (although rarely violent or dangerous emergencies—see Cherry, 1995) and began to show a consistent effect of the inhibition of helping in the presence of others. This effect was believed to be robust in a 1980s meta-analysis of the bystander studies (Latané & Nida, 1981). The findings seemed to confirm that the bystander effect was a clear impediment to public safety in cases of public aggression and violence.

However, a reanalysis of Latané and Nida's meta-analysis (1981), correcting for a mathematical error, found that a victim was in fact more likely to be helped with more bystanders present—at least under conditions in which the bystanders could not all see one another (i.e., the Kitty Genovese incident; Stalder, 2008). As Fischer et al.’s (2011) later meta-analysis would go on to show, while there is empirical support for the idea of a bystander effect in general, there is no evidence that it applies in the kinds of aggressive or violent emergencies that prompted research into the phenomenon in the first place. In fact, Fischer et al describe a “reverse bystander effect,” where the presence of bystanders can increase the likelihood of intervention. Recent reexaminations of the Kitty Genovese murder (see Lurigio, 2015; Manning, Levine, & Collins, 2007) have conclusively demonstrated that the notion of the 38 inactive bystanders in this tragic case was a media story without substance. At most, a handful of people saw Winston Mosely and Kitty Genovese together. Moreover, the idea that there was no intervention was also incorrect—although the attempts at intervention were ultimately unsuccessful. It turns out that, in the iconic Genovese case, the bystander effect did not hold.

Social Identity and the Bystander Effect

The social identity tradition has also developed a literature on the nature of the bystander effect. For example, Levine and colleagues, through a combination of laboratory experiments, virtual reality experiments, field experiments, interviews, focus groups, and the examination of historical and legal documents, have explored the role of social identity in bystander behavior (for review, see Levine & Manning, 2013). This includes the study of social identity relationships between bystanders and other bystanders, bystanders and victims, and bystanders and perpetrators. Studies show that the presence of others influences behavior to the extent that we feel shared identity with other bystanders. This shared identity, and the norms and values associated with the identity, can sometimes promote and sometimes inhibit intervention (Levine & Crowther, 2008). For example, male participants in a study who watched CCTV footage of a domestic violence attack by a man towards a woman, were more likely to say they would intervene in the presence of other women than in the presence of other men. However, women who watched the same video were less likely to intervene in the presence of other men than in the presence of women. In other words, the influence of the presence of others is not uniform but interacts with the social identity relationships among the bystanders. In some circumstances (and combination of identity relationships), the presence of others can facilitate bystander intervention, in others it can inhibit it.

Social identity research also demonstrates that we are more likely to intervene when we see signs of common group identity with a stranger in distress. In a field experiment using a confederate who wore different football shirts (Manchester United, Liverpool FC, or an unbranded red shirt), Manchester United fans were more likely to help the stranger who had had an accident when he wore the Manchester United shirt than when he wore the out-group shirt or the unbranded shirt (Levine et al., 2005). Signs of shared identity with a stranger increased the likelihood of help being offered. More importantly, in a second study using the same design, but this time making a more inclusive “football fan identity” salient, Manchester United fans were as likely to help the victim when he wore the Liverpool FC shirt as the Manchester United shirt—but helping levels remained low if the victim showed no signs of being a football fan. In this way, the study demonstrates that being an in-group member is important for being helped in public places—but that the boundaries of who is seen as in-group are not fixed. The more open and inclusive the boundaries of the shared identity, the greater the range of people who are likely to be afforded care and protection.

Shared categorization also affects the willingness of bystanders to challenge perpetrators of violence in public places. For example, Lowe, Levine, Best, and Heim (2012) show that women are more likely than men to intervene to stop female–female fights—particularly when they fight in front of bystanders that include men. Female bystanders talk about the feelings of shame and embarrassment that they experience when watching women fight. They feel that male observers will see this kind of behavior as an opportunity to sexualize or trivialize women in general—and that this threat to identity needs to be challenged. In similar fashion, Stott, Hutchison, and Drury (2001) describe an intervention by Scottish football fans against a fellow Scot who behaved aggressively towards a Tunisian fan at a camp site during the World Cup in France in 1998. Here again, intervention against aggressive in-group members was prompted by the danger to the reputation of the group. Stott et al. argue that Scottish fans feared that the actions of the in-group member would make people think that Scottish fans are no better than English football fans—who already have a reputation for racist and violent behavior. There is nothing more aversive and identity threatening for a Scot than to be mistaken for somebody who is English. This willingness of in-group bystanders to challenge in-group antinormative behavior which threatens the positive identity of the wider group has been referred to as “the black sheep effect” (Marques, Abrams, Páez, & Hogg, 2001) and explains situations where group members are prepared to be even more punitive towards a poorly performing in-group member than they are to an out-group member. Conversely, shared group membership with a perpetrator may also result in a positive in-group bias, in which the adverse behavior of an in-group member may be more readily justified or permitted than similar behavior by a non-group member (Bennett & Banyard, 2016; Otten, 2009).

Taken together, there is clear evidence that, when it comes to understanding the behavior of bystanders to aggressive or emergency events, the rather simple story, from classical bystander theory about the negative impact of the presence of others, lacks explanatory power. Bystander behavior is shaped by the social identity relationships to other bystanders, as well as the relationships to the victim and the perpetrator.

What Do We Know About the Actual Behavior of Bystanders in the Context of Aggression and Violence?

In the light of this critique of the traditional bystander effect story, it seems we might need to rethink the role of the bystander effect in the development of violence reduction initiatives. However, before we can do so, it is important to establish an evidence base of what we actually know about the behavior of bystanders in aggressive and violent emergencies. This is not as straightforward a task as it may appear. Despite the extensive body of literature that has been produced on the subject, there is very little work that has actually studied the behavior of bystanders to aggression and violence directly. Most of the existing research has tended to employ self-report methods (surveys, questionnaires, interviews, focus groups—Hart & Miethe, 2008; Heinskou & Liebst, 2017; Levine, Lowe, Best, & Heim, 2012) or other proxy measures (Anderson & Bushman, 2002; Finkel, 2014)—rather than the direct study of aggression and violence. On the occasions when aggression and violence has been studied directly (e.g., during bar fights—Parks, Osgood, Felson, Wells, & Graham, 2013; Wells & Graham, 1999) the data recording tends to be in the form of ethnographic field notes or in situ observations—which limits the behavioral richness and reliability of the data (see Philpot, Liebst, Møller, Lindegaard, & Levine, 2019). However, an emerging body of work examining CCTV footage of real-life public altercations has begun to transform what we know about actual bystander behavior during these kinds of events.

Rethinking the Bystander Effect in Violence Reduction Programs

The first element of a redesigned “bystander effect” component of violence reduction programs would be to challenge the commonly held notion that the presence of others makes people less safe and reduces the likelihood of receiving help. The program would begin by pointing out that in violent or dangerous emergencies the presence of multiple bystanders is at worst neutral, and at best a positive influence on the likelihood of intervention (Fischer et al., 2011; Levine et al., 2011). Furthermore, from the victim's perspective, the more people that are present, the better. This is because greater numbers mean that the overall likelihood of somebody intervening is higher (Philpot et al., 2019; Stalder, 2008). And the evidence from the analysis of CCTV footage of actual events shows this to be the case.

The third element of this reconfiguring of bystander effect training comes in the way the presence of others is related to the appraisal of risk. At present, the classic bystander effect gives the impression that victims are more likely to be helped when there are fewer bystanders. In other words, that the risk for victims is higher in situations where there are many bystanders rather than when bystanders are alone. In fact, the evidence shows that the opposite is the case. From the perspective of the victim, fewer bystanders increase the risk that no intervention will be attempted. Moreover, it is now clear that the presence of other bystanders is not the most important factor in the likelihood of any single bystander being prepared to intervene. It is important therefore that violence reduction program training reflects this in their education and messaging. Overall, the presence of multiple bystanders reduces the risk for victims.

The fourth element of a redesigned bystander effect component of violence reduction programs centers on the importance of the impact of social identities on bystander behavior. In arguing for the importance of social identity, we recognize that existing bystander programs do already orientate to the importance of bystanders at the group level. For example, Green Dot talks about “community mobilization.” Bringing in the Bystander promises a “community responsibility approach,” and The Intervention Initiative encourages “active bystander communities.” However, these approaches tend to treat the concept of community as both homogeneous and stable. In other words, when they invoke the idea of the importance of the community, they seem to be imagining a single entity with a stable set of values. The appeal is to a single community, and the aim is often to try and change community values such that people are more likely to intervene.

Given the evidence that the bystander effect is contingent upon how identities play out in the immediate context of the emergency, the fifth recommendation centers on the importance of taking social context seriously when considering how to leverage the power of bystanders to act. The power of the social context could be harnessed for more than just changing the social identity relations. It could also be used to maximize the impact of the training itself. For example, existing intervention training tends to be done in “classroom” environments in colleges, universities or workplaces (e.g., Banyard et al., 2007; Fouber, Godin, & Tatum, 2010; Gidycz et al., 2011). These places afford particular types of identities and social relationships, which might not overlap with the kinds of social identities which are salient during the emergency event. The reason why the mode of delivery for existing bystander programs happened in the way that it did can be traced back to the history of the emergence of these programs. The programs were developed in response to the “discovery” of sexual assault and violence in colleges and universities, and the programs were designed to fit into existing curriculum delivery. The very nature of a college or university as a place of learning—and with curriculum delivery mechanisms already in place—made it logical to deliver classroom-based interventions. However, the kinds of events that were being challenged were not (for the most part) taking place in the classroom. They were at the parties, in the bars, and the clubs in the leisure zones of the towns and cities. If intervention programs are to take the importance of social context seriously, then it is worth considering whether there are forms of intervention that can be delivered in the places where the events happen—rather than in the decontextualized and sanitized space of a classroom.

There is precedent for this kind of approach in public health campaigns in sexual health and disease prevention. For example, there is a history of delivering interventions around the prevention of STDs/HIV/AIDS transmission in pubs and clubs, rather than in more general information campaigns (Kalichman, 2010; Pitpitan & Kalichman, 2016). These programs combine social and structural interventions which are delivered in situ. For example, the programs engage people in bars and at other venues in an attempt to influence norms and values—as well as dispensing condoms and offering opportunities for different kinds of medical testing. One method for delivering interventions in this way is the popular opinion leader (POL) model (Kelly et al., 1997). The POL model identifies key opinion leaders who emerge from the community and then offers them training to act as agents of change. The POL intervention develops a small cadre of people who are already regulars in the nighttime economy scene, who are trained to initiate conversations and endorse safer sex practices. The active ingredient in the intervention is therefore informal casual conversations with peers, rather than formal education.

This public health intervention model has also been applied to violence reduction directly. The cure violence model (Slutkin, Ransford, Decker, & Volker, 2015) also delivers in situ interventions based on an epidemiological “contagion” model of the spread of violence. Trained violence interrupters, recruited from the community and who have the requisite skills and credibility, deliver interventions on the streets, in community spaces and in hospitals in the aftermath of shootings. This public health model of violence reduction has had significant impact (compared to controls) in several American cities and in other places around the world. While the theoretical utility of the “contagion” idea at the heart of this model has come in for sustained criticism (Drury, 2018), the principle of delivering interventions through credible group members, and doing so in the places where the emergencies are taking place, remains a powerful model.

There is also a small literature on delivering interventions to tackle sexual assault and other forms of violence in bars and clubs. In some instances, trained bystanders have formal leadership and responsibilities to intervene as part of their job. One such program is Safe Bars (Graham et al., 2004), which applies a situational crime prevention model to barroom scenarios in Toronto, Canada. Managers and staff are trained to recognize, respond to, and manage alcohol-related problem behavior. Independent observers are trained separately to evaluate the program without any involvement. Despite low rates of aggression in bars overall, the program proved to be somewhat effective in reducing incidents of severe and moderate aggression (Graham et al., 2004). In similar fashion, Powers and Lelli (2018) describe the evaluation of a community-based bystander intervention program for bar staff delivered in the workplace. Initial evaluation suggested that the intervention program led to a decrease in rape myths, a reduction in barriers to intervention, and an increase in bartenders’ willingness to intervene.

Conclusion

These recommendations are underpinned by a robust evidence base of actual bystander behavior in real-life emergencies. They are supported by a significant body of theoretical and empirical work in the social identity tradition. Taken together they offer the opportunity for existing violence reduction programs (that include a bystander component) to reconfigure their approach to the presence of bystanders in emergencies. In doing so, they will increase the overall efficacy of these violence reduction programs in delivering a safer society.