1, Oxford Street False Alarm, 24th November 2017 Collective false alarm incidents, where crowds of people flee from what they think is a shooter or similar hostile threat, are politically, socially, and psychologically impactful. These incidents have been neglected by researchers, despite an upsurge in them in recent years. False alarms seem anomalous, as the usual public response to warnings of an emergency (such as a fire alarm) is to discount them. A first question is: under what conditions does a crowd perceive an ambiguous signal as a hostile threat, and what are the social processes involved? False alarm flight incidents have been used rhetorically to suggest that crowds are prone to panic, even though the concept has been largely discredited across the social sciences as an explanation for collective behaviour in real emergency events. Yet there is a lack of research evidence into how people actually behave in false alarms. Therefore a second question is what do people actually do and why? To examine the usefulness of a number of social-cognitive models (signal detection, social appraisal, social identity) in answering these questions, we interviewed 39 people who were involved in the false alarm incident in Oxford Street, London, on Black Friday 2017. In this event, thousands of people over an area of over one square kilometre ran and many hid from what they thought was a terrorist attack. Thematic analysis of the transcripts suggested the following pattern of behaviour and experience. First, there had been a large number of genuine terrorist attacks in London and other areas of the UK and Europe that same year, and for many people this fact operated as a framing through which ambiguous signals (including alarms and banging noises) were interpreted. People were also aware that parts of London were vulnerable to such attacks. In short, using the immediate historical context as a relevant reference point meant that ‘we’ (i.e., people on Oxford Street) were a possible target for a further terrorist attack. Second, there appeared to be a social appraisal process whereby the sight of others’ emotions and repeated (flight) behaviour added to the accumulating evidence of a hostile threat. Within this, the appearance of armed police offered only partial reassurance, as it added to the impression that an attack was in progress. In many cases, people’s interpretation was not immediate but (similar to the usual pattern of discounting emergency signals) developed gradually and only after repeated and separate pieces of ‘evidence’. The messages some received (by phone, social media, or shouts in the street), stating that there was a terrorist attack, contributed further to the evidence base. Third, in terms of behaviour, while half of the participants reported running, others walked, or delayed their self-evacuation. A large number hid. While there were a number of instances of supportive behaviour, particularly in pockets (e.g. those sheltering together), this wasn’t always the case, and there were also instances of pushing and trampling. Rather than the common fate that arises in those emergencies when there is a clear and well-defined threat, in this case there was a fragmented experience across the crowd as whole, and so a lack of shared social identity. Our analysis suggests a new way of thinking about collective responses to perceived hostile threats, and has implications for how the authorities communicate with and prepare the public in case of such emergencies. 2, Virtual Reality experiments: Social influence in collective responses to perceived threats Collective flight incidents, often referred to as ‘stampedes’, have been and still are mistakenly associated with self-interest and panic. However, a growing body of research using post-incident analysis of crowd flights strongly suggest that flight phenomena are underlaid by complex psychological processes involving social identity, shared fate, and appraisal. A series of 5 experiments were undertaken to further our understanding of the mechanisms underlying collective responses to perceived threats. We used virtual reality technology to create realistic scenarios of an urban environment where a potential threat is triggered and the crowd reacts to it. In the first experimental series (2 studies, N ~ 800), we used animation as a vignette and measured participants’ reported perception of threat, perception of other crowd members, and intentions to run in response to a potential threatening noise, and a crowd in the animation that either ran or ignored that noise. Consistent with the social identity approach, the self-reported data indicate a key role of shared social identity amongst crowd members in explaining collective behaviours. The shared experience of facing and responding to a potential threat brought people together and increased shared social identification with other crowd members, especially when their response was consistent with the situation. Moreover, and challenging a ‘contagion’ model of collective behaviour in evacuations, the behaviour to run was mediated by the perceived threat, shared fate and shared social identity with other crowd members. In the second experimental series (3 studies, N ~ 1200) participants were immersed in an online desktop version of the virtual reality scenarios, being able to decide how and where to move, at what speed, etc. Across experiments, we manipulated the presence or absence of the noise triggering the crowd behaviour, the terror threat level, and the crowd behaviour. The behavioural data showed social influence processes, whereby participants were more likely to run away from a potential threat when other crowd members fled. Importantly, they were not equally influenced by all crowd members. Instead, they relied more on those with whom they identified (i.e, ingroup members) for guidance, both when appraising the situation and deciding how to act. These results provide empirical evidence against a ‘contagion’ model of collective response to threat, by showing that several factors interact with each other to impact participants’ running behaviour, and the extent to which they are likely to follow other crowd members. Moreover, they highlight the importance of social-psychological variables, and in particular of the social identity approach, to improve models of collective behaviours and inform the intervention of emergency services.This research addresses the question of how people respond collectively to acute states of perceived emergency. Both in the UK and globally, the occurrence of terrorist attacks has been associated with an increase in collective flight incidents - often called 'stampedes'. Most have been false alarms. In the case of the incidents in Oxford Street, in 2017, for example, hundreds ran from what they thought was gunshots, and many more joined in when they saw people running, in a cascade of secondary effects. These events are often distressing for those involved. Some of them result in injury and even deaths. 'Stampedes' in response to perceived hostile threats raise the following questions: When and how is a signal perceived as threatening? When and how do people flee? When do they follow (or ignore) others? What is the role of other groups (authorities, emergency services) in communicating threat? When do these incidents become disorderly? Yet despite their social and psychological significance, these incidents are poorly understood, and an adequate theoretical understanding is currently lacking. Explanations in terms of hypervigilance are necessary but insufficient. They don't explain either how threat perceptions are shared or the collective nature of 'stampedes'. Explanations in terms of 'crowd panic' are also inadequate, despite dominating public discourse. They cannot account for the evidence of discrimination and restraint in these evacuation events. This project provides the first systematic evidence on the nature and dynamics of human 'stampedes' in response to perceived hostile threats. We developed a new analysis, based on the social identity approach (a theory of our psychological group memberships) and social appraisal theory (which focuses on the social and emotional information we infer from others' responses). We argue that 'who we are' - which can vary across contexts - shapes perceptions of threat, emotional appraisal, and social influence. A distinctive claim is that crowd events such as these are intergroup relationships. Therefore, as well as examining crowd members' perceptions and behaviour, we examined the relationship between responders' communicative acts and public behaviours. Previous limitations in understanding 'stampedes' in response to hostile threats are due to a lack of appropriate methodology as much as the paucity of theory. We addressed these limitations through a programme of research comprising three strands. A first strand examines historical cases of false alarms. This provides a systematic review of false alarm incidents going back to 2010, as well as a detailed analysis, using interview and archive data, of a contemporary and a historical incident. Second, experiments using Virtual Reality technology allow us to manipulate and control variables, such as group relationships and norms, and to measure direction and speed of flight response, in order to test systematically our hypotheses about the roles of identity and appraisal. Third, we worked with Transport for London on CCTV footage of an actual bladed instrument attack to analyses the patterns of public responses, including interventions, in detail. Together these studies enable us to determine when and how a signal is perceived as threatening, whether and how people flee, when people follow (or ignore) others, and the role of the emergency services in this. Understandings of public responses to perceived emergencies have implications for emergency policy and practice and for public debate. Therefore, as well as a new theoretical model, outcomes from this project developed new emergency guidance on this topic.

1, Oxford Street False Alarm, 24th November 2017: The criteria for recruitment were individuals who witnessed the Oxford Street false alarm, 24th November 2017, using: passive recruitment via advertisements on researchers’ pages on social media sites Twitter, Facebook, LinkedIn and WhatsApp; active recruitment where researchers directly contacted people on Twitter who identified themselves as being at the incident; a call out on BBC Radio 4’s ‘All In The Mind’ programme; in person recruitment where a researcher went to Oxford Street and recruited public and shop workers; and snowballing from these initial contacts. 39 semi structured interviews were conducted via Zoom. Thematic analysis (Braun & Clarke, 2006) was used to analyse the data. 2, Virtual Reality experiments: Social influence in collective responses to perceived threats Virtual reality technology is increasingly being used to study public behaviour in emergency evacuations and similar incidents, as a way of combining experimental control with psychological immersion. Working with Make Real, a company specializing in immersive technologies, we constructed a virtual reality animation of a street scene (based on Oxford Street). At one point in the animation, members of the public flee apparently to get away from something. The animation allowed for the manipulation of a number of relevant features, including the source of the flight (e.g., an ambiguous or unambiguous noise), the visible appearance of characters in the animation, and the speed and coordination of the crowd flight. This scenario was used in two series of experiments. In the first series of experiments, we first presented participants with video recordings of the virtual scenario in which we manipulated several factors, and next assessed their behavioural intentions, perception of the scenario (e.g., perceived threat) and perception of the other crowd members (e.g, shared social identity with the crowd) in a survey. Specifically, this series comprises 2 studies: the ambiguity study (N = 634) and the ‘shared social identity’ study (N = 198). In the second experimental series, we used a desktop online playable version of the virtual reality scenario. Participants were, therefore, immersed in a VR-based scenario and able to decide where to move to, at what speed, etc. Again, we manipulated several factors to create different scenarios and assess participants’ spontaneous behaviours in these scenarios. This experimental series comprised three studies: the crowd behaviour study (N = 248), the noise study (N = 520), and the threat study (N = 464). In all three studies, participants were led to perceive half of the crowd as ingroup members and the other half as non-ingroup members.