Running head: POLICING, COGNITION AND THREAT
Enhanced Threat Detection in Experienced Riot Police Officers: Cognitive Evidence from the Face in the Crowd Effect
1Department of Psychology, University of Chester, Chester, United Kingdom
Amy E. Pinkham2
2Department of Psychology, Southern Methodist University, Dallas, Texas, United States of America
1Department of Psychology, University of Chester, Chester, United Kingdom
1Department of Psychology, University of Chester, Chester, United Kingdom
*Correspondence should be addressed to: Ljubica Damjanovic, Department of Psychology, University of Chester, CH1 4BJ, Chester, United Kingdom
Acknowledgments: Funded by the Small Research Grants Scheme (SRG/SP-S) awarded to the first and last authors from the Research and Knowledge Transfer Office at the University of Chester http://www.chester.ac.uk/business-support-services/rkt-office The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Declaration of conflicting interests: None. We thank the participants for their participation in the research, as well as police personnel at Cheshire Constabulary for facilitating this research. The authors wish to thank Panos Athanasopoulos for helpful discussions on earlier versions of the manuscript.
We explored how varying levels of professional expertise in hostile crowd management could enhance threat detection capabilities as assessed by the face in the crowd paradigm. Trainee police officers and more experienced police officers specialized in, and having extensive experience with, riot control, were compared with participants with no experience in hostile crowd management on their search times and accuracy levels in detecting angry and happy face targets against a display of emotional and neutral distracter faces. The experienced officers relative to their trainee counterparts and non-police controls showed enhanced detection for threatening faces in both types of display along with a greater degree of inhibitory control over angry face distracters. These findings help to reinforce the ecological validity of the face in the crowd paradigm and provide a new theoretical link for the role of individual differences on the attentional processing of socially relevant stimuli.
Keywords: visual search, facial emotion, attention, threat detection, policing
August 2011 marked a distressingly familiar return to city wide rioting across the United Kingdom, prompting many senior police officers to declare that “the game has changed” for crowd control policing (BBC News, 2010; Her Majesty’s Inspectorate of Constabulary [HMIC], 2011). As media coverage continues to show, such spontaneous acts of public disorder are frequently experienced across the globe, highlighting an international need for an effective ‘game plan’ in the policing of hostile crowd behaviour (e.g., Beene, 2006; Jobard, 2009; Lewis, 2007; Poynting, 2006; Stol & Bervoets, 2002; Waddington & King, 2005). Thus, an understanding of the processes involved in the early detection and response to individuals within a crowd who may have violent intent can have far-reaching practical implications to help safeguard a peaceful protest.
In crowd control policing work, snap decisions need to be made, often under complex, dynamic and limited viewing conditions, to rapidly determine not only the threat level of a single individual within the crowd, but also the behaviour of the crowd itself for further signs of violent discord (Waddington, 2007). Police training manuals advise front line policing staff on the complexities of crowd behaviour, such that “members of a crowd do not necessarily get carried away by ‘crowd hysteria’” (Manual of Guidance on Keeping the Peace [ACPO], 2001, p.26). Police officers are required to develop a crowd engagement plan to incorporate a number of considerations of the crowd’s behaviour, including how: “Individual crowd participants have a different perception of events, which may affect the behaviour of the individual, or influence the group. When crowd members become highly emotional or aggressive, their visible behaviour can spread and become regarded as acceptable by those around them. It can be an opportunity for a minority to influence and mobilise others around them” (ACPO, 2010, p. 88). Further unpredictable elements of a crowd’s behaviour relate to the increasing trend for celebrations to turn into violent riots, a phenomenon referred to in the social psychology literature as ‘celebratory’ riots, and often documented in highly charged sporting events (Lewis, 2007). Therefore, police officers are trained to engage in effective monitoring of both the individual and the crowd itself, for highly charged emotional cues, that can change in an instant from signs of jubilations to cues encouraging violence and public disorder.
At a biological level at least, personnel working in highly charged and hostile environments demonstrate a greater neural tuning in response to viewing facial expressions of threat during an expression matching task, in a region associated with enhanced performance under high-stress conditions, the right insular cortex, relative to controls (Paulus et al., 2010). This particular neural pattern coupled with longer decision times in response to viewing happy and fearful face targets when trying to match them to another face that shows the same expression, suggests that effective visual scanning for threat specific cues in one’s environment involves both heightened neural tuning towards threat along with conservation of processing resources for non-threatening facial cues (Paulus et al., 2010). However, as noted by Paulus and colleagues, it remains to be established whether the effects of expertise in working in such environments extend to cognitive processes when more performance-based tasks are used. In the present study, we sought to investigate this specific issue by applying a robustly developed paradigm from cognitive psychology, the face in the crowd effect (FICE), to determine cognitive characteristics associated with effective and efficient assessments of crowds within a context relevant to policing and to determine the extent to which experience in professional policing can influence performance on threat perception tasks.
The FICE provides an index of the attentional demands (Posner & Petersen, 1990; Treisman & Gelade, 1980) involved in searching our way through crowds of faces (e.g., Fox & Damjanovic, 2006; Hansen & Hansen, 1988; Öhman, Juth & Lundqvist, 2010; Pinkham, Griffin, Baron, Sasson & Gur, 2010; Schmidt-Daffy, 2011). The task is presented as a visual search paradigm in which individuals are shown a group of facial stimuli and are asked to determine if all faces are displaying the same expression or if one face is different. The task has been refined over the years since its initial inception by Hansen and Hansen (1988), with the latest variant of the task using search displays consisting of a newly validated set of photographic stimuli of faces that closely resemble real ‘crowds’ (Pinkham et al., 2010). Previous efforts to generate search ‘crowd’ environments in the laboratory have kept identity constant whilst manipulating emotional valence between the target and distractor faces, thus creating unrealistic ‘clones’ for participants to search through for the entire experiment (e.g., Fox et al., 2000; Fox & Damjanovic, 2006; Lipp, Price & Tellegen, 2009).
In order to replicate the kinds of crowds we experience more realistically in our social encounters with others, Pinkham et al., (2010) maximized the heterogeneity of the search display, and increased the demands placed on attentional resources, by including faces of nine different identities. Whilst this is certainly a step in the right direction in the development of ecologically valid visual search tasks, and helps to address the explicit aspects of each visual search trial, that is, distracter homogeneity and target-distracter discriminability (e.g., Duncan & Humphreys, 1989), other studies have focused their ecological efforts on implicit aspects of overall visual search performance. This has recently been achieved by Öhman et al., (2010) by manipulating the sum total of all stimulus individuals a participant is exposed to during the task, otherwise known as the task’s implicit set size. Small implicit set sizes involve exposing participants to only six different identities, whereas with large implicit set sizes a participant is exposed to the identities of 60 different individuals over a given testing session. Whilst using a large implicit set size has its obvious advantages in capturing the richness in crowd diversity, and the constraints this can exert on attentional performance, the current study is concerned more with establishing how explicit aspects of search performance, namely the detection of angry and happy face targets against competing distracter faces, may vary as a function of real-world experience in handling hostile crowds.
The primary finding from the FICE as utilized in Pinkham et al’s study is that participants are faster and more accurate in detecting, or engaging their attentional resources towards, a discrepant angry face in a crowd of distracter faces than a discrepant happy face in a crowd of distracters. This difference in response time and accuracy is known as the threat superiority effect (TSE: Fox, 2008; Fox & Damjanovic, 2006) and is thought to be due to an evolutionary advantage for quickly detecting environmental threat (Öhman, Lundqvist, & Esteves, 2001), which is supported by a neural ‘alarm’ network consisting of the amygdala at its hub (Liddell et al., 2005; Öhman & Mineka, 2001).
However, what poses a theoretical hurdle for an automatic, evolutionary-based account of threat performance is the finding that the detection of angry faces is often affected by the emotionality of the surrounding distracter faces (Frischen, Eastwood & Smilek, 2008). For example, some findings indicate that a bias for processing threat is more pronounced when emotional distracter faces, as opposed to neutral distracters, are used (e.g., experiment 5 in Juth, Lundqvist, Karlsson & Öhman, 2005; Pinkham et al., 2010), whereas other FICE paradigms have shown a stronger threat bias with neutral rather than emotional distracters (e.g., Öhman et al., 2001). As detailed below, we investigate here whether the influence of the presence or absence of emotionality in distracter faces may differ across individuals based on varying levels of experience with hostile crowd management.
Also of interest from the FICE but somewhat distinct from the TSE, when the discrepant item in the search display is a happy face, surrounding angry faces can serve as particularly potent distracters, slowing down the detection of a happy face in angry distracters relative to neutral distracter trials (Byrne & Eysenck, 1995; Derakshan & Eysenck, 2009; Fox et al., 2000; Juth et al., 2005). Thus, on trials in which a happy face is embedded among angry faces, individuals must inhibit threat-related attentional capture in order to perform efficiently.
To date, the theoretical distinction made between threat detection (i.e., TSE) and threat inhibition performance from the FICE task has mainly been modeled on performance patterns observed in anxious samples. For example, individuals with generalized social phobia show facilitatory engagement of attentional resources towards angry face targets (Gilboa-Schechtman, Foa, & Amir, 1999) as do individuals with high levels of self-reported trait anxiety (Byrne & Eysenck, 1995). High levels of self-reported trait anxiety also affect the disengagement of attentional resources to a greater extent than low levels of self-reported trait anxiety, impeding the detection of happy face targets when the surrounding distracters consist of angry faces (Byrne & Eysenck, 1995). The nature of these anxiety-related patterns on threat processing have recently been accommodated in Attentional Control Theory (ACT; Derakshan & Eysenck, 2009; Eysenck, Derakshan, Santos & Calvo, 2007),which proposes that disproportional involvement of a stimulus-driven (e.g., bottom-up) attentional control system relative to that of a goal-directed (e.g., top-down) attentional system (e.g., Corbetta & Schulman, 2002) may result in this pattern of performance. The increased involvement of the stimulus-driven control system means that attentional control is compromised as its resources are drawn to task-irrelevant information (Derakshan & Eysenck, 2009; Eysenck et al., 2007).
From this view, the FICE provides a unique methodological resource to inform both engagement and disengagement of threat processing within the same paradigm (Derakshan & Koster, 2010; Fox et al., 2000; Pinkham et al., 2010). Despite this utility, its application to real-life contexts has yet to be empirically determined (Frischen, et al., 2008; Weierich, Treat & Hollingworth, 2008). Indeed, obtaining measures of cognitive performance from individuals who are regularly exposed to working in dangerous and life-threatening environments provides an essential element to the validity of key models of human cognition and performance, such as ACT (see Baddeley, 1997). Whilst such developments in the past have been limited by practical and ethical constraints in replicating highly charged emotional events under experimental conditions, recent work has made a substantial first step in the validity of ACT, by considering the consequences of this extensive exposure and training on an individual’s motor co-ordination, such as shooting accuracy and measures of self-reported anxiety (Nieuwenhuys & Oudejans, 2010; Nieuwenhuys, Savelsbergh & Oudejans, 2011). We extend this important work to the conceptual validation of the FICE task, by considering the cognitive aspects that are shaped by experience and training in hostile crowd management.
Public policing offers an ideal opportunity for this application as it not only involves making correct decisions, but also making these decisions in a time effective manner under varying levels of task complexity (e.g., Bechtoldt, Rohrmann, De Pater & Beersma, 2011; Mann, Vrij & Bull, 2004; Waddington, 2007), and the FICE offers both indices of interest to determine efficient searching of crowd behaviour. Testing the FICE in experienced police officers also allows for examination of whether real-world knowledge as developed from expertise and training in responding to hostile situations can influence the allocation of attention to threat-specific information. To date, attempts to simulate the effects of responding to emotionally-charged environmental cues have involved laboratory studies with student populations, who are instructed to engage in complex simulations, such as imagining that they were being ambushed while in a foreign city (Becker et al., 2011). In this rare opportunity to focus on the real-world value of FICE, the current study also speaks directly to the concerns raised about developing theories based on data produced almost exclusively by Undergraduate psychology participants (Henrich, Heine & Norenzayan, 2010). Thus, the aims of the study are to establish the ecological validity of FICE by applying it to a policing context and by doing so establish how expertise-based factors can be incorporated alongside an anxiety-based framework of attentional cognitive control (Byrne & Eysenck, 1995; Derakshan & Eysenck, 2009; Eysenck, Derakshan, Santos & Calvo, 2007; Nieuwenhuys, Savelsbergh & Oudejans, 2011).
The effects of expertise in responding to threat in real-life can be operationalized by the FICE in two important ways by focusing on the responses made from the discrepant display trials (Fox et al., 2000). First, extensive experience in hostile crowd control may equip experts with a perceptually robust template of their object of interest, that is, impending signals of aggression as conveyed by an angry face (Brosch, Sander, Pourtois & Scherer, 2008). Such a template would allow experts to search efficiently (Jackson & Raymond, 2008; Paulus et al; 2010; Tong & Nakayama, 1999) for the angry face target in a range of different contexts, thus making them less susceptible to the effects of the surrounding distracter faces that have been previously reported with control samples (e.g., Juth et al., 2005; Öhman et al., 2001; Pinkham et al., 2010).
Second, experienced officers may also possess enhanced inhibitory control over task-irrelevant material as a direct consequence of their training and expertise in routinely evaluating and updating a range of behavioral responses during their threat appraisal of the crowd’s behaviour (ACPO, 2001, 2010; Anshel, 2000; Beilock & Carr, 2004; Paulus et al., 2010; Waddington & King, 2005). As such, experienced officers’ response times for happy target faces should not be significantly impeded by the presence of non-target threatening distracter faces. Support for these hypotheses will be tested by comparing threat detection performance between three groups: participants with no experience in hostile crowd management (controls), police officers in training for this type of task (trainees), and police officers specialized in, and having extensive experience with, riot control (experts). The controls may lack this fine grained perceptual template due to low levels of exposure to threat-specific facial cues in their social environment (Whalen, 1998), thus displaying a threat superiority effect that is stronger in one type of crowd context than another, in addition to showing interference from angry face distracters on happy face target trials.
Given that visual search performance has been shown to vary with individual levels of intelligence (Nakagawa, 1996), and that the modulating capacities of ACT are affected by trait-based anxiety differences (Byrne & Eysenck, 1995; Derakshan & Eysenck, 2009; Eysenck et al., 2007), we aimed to equate our two police groups with our non-police controls on these factors by administering the Raven’s Standard Progressive Matrices (RSPM; Raven, 1958) as an index of cognitive ability and the trait component scale of the State-Trait Anxiety Inventory (STAI; Spielberger, Gorsuch, Lushene, Vagg & Jacobs, 1983). Additionally, detecting and responding to highly charged, threat-specific situations can trigger a range of adverse emotional responses, including anticipatory stress at the start of duty (Anderson, Litzenberger & Plecas, 2002; Anshel, 2000; Asmundson & Stapleton, 2008). These changes in emotional responsiveness reported in police officers have been shown to affect performance on memory (Yuille, Davies, Gibling, Marxsen & Porter, 1994) and motor co-ordination tasks (Nieuwenhuys & Oudejans, 2010), and may also yield strong implications for the efficient functioning of inhibitory attentional control processes (Byrne & Eysenck, 1995; Derakshan & Eysenck, 2009; Derakshan, Smyth & Eysenck, 2009; Juth et al., 2005; Nieuwenhuys & Oudejans, 2010; Nieuwenhuys et al., 2011 ). As such, we compared changes in the state anxiety component of the STAI in order to examine how the three groups responded to searching for threat in an experimental context, in addition to correlating changes in state anxiety for the entire sample with visual search performance to determine its facilitatory and inhibitory contributions on threat processing.
This study was approved by the Department of Psychology Ethics Committee at the University of Chester, United Kingdom. Consenting participants gave written informed consent.
A total of 78 participants were recruited for the study, from which 76 contributed data to the FICE analysis (see Results section for exclusion details). The control group (female = 6, male = 18) consisted of non-academic staff recruited from the University of Chester. The police trainee group (female = 7, male = 17) were recruited from a Foundation Degree Programme in Community Policing at the University of Chester and had received both cognitive and affective-experiential training (e.g., baton and hand-cuff arrests in real crime intervention settings) at the time of testing (Zacker & Bard, 1973). The experts consisted of experienced police officers (female = 5, male = 19), known as the VECTORS within the Cheshire Constabulary, who were experienced fire arms officers deployed for crowd control, high-risk combat and public order policing. Their ongoing training regime consisted of intensive weekly combat exercises.
Permission to recruit the VECTORS was granted by the Chief Inspector in the first instance, whereas permission to recruit the trainee sample was obtained from the Foundation Degree’s Programme Leader. All participants were recruited through the internal e-mail of their organization and were assigned a randomized code for their participation and a password to ensure confidentiality. Participants were compensated £20 each for their time. Table 1 shows the full details for those participants who contributed data to the FICE task.
Insert Table 1 about here
Groups differed significantly in age H(2) = 19.78, p < .001 and in length of service in their respective professions, H(2) = 46.62, p < .001. The VECTORS were significantly (Bonferroni adjustment, p < .02) older than the trainees (U = 53.50, r = -.70), but no group differences were found between controls and trainees (U = 229.50, r = -.17), or between controls and VECTORS (U = 178.00, r = -.33). Trainees served the shortest time from both controls (U = 119.50, r = -.52) and VECTORS (U = 3.50, r = -.88), who also had a longer length of service than the controls, (U = 53.00, r = -.70).
Stimuli and Apparatus
Visual search task presentation and data collection were conducted with an Intel Core PC desktop computer with a 2.93GHz processor and 18-inch monitor. A refresh rate of 60 Hz and a resolution of 1024 x 768 were used. Presentation version 12.1 software (http://www.neurobs.com) delivered stimuli and recorded responses and reaction times (RT). Manual responses to the visual search task were collected from designated response keys on the computer’s keyboard. The face stimuli and the visual search task were those used and developed by Pinkham et al., (2010), and consisted of 9 individuals posing angry, happy and neutral facial expressions. Each photo measured 5.3cm (width) x 5.3 cm (height). The expressions made by each poser were validated using the facial action coding system (Ekman, Freisen & Hager, 2002), and equated for mean luminance and contrast across faces and expressions (see Pinkham et al., 2010, supplemental materials).
Threat detection performance was measured using reaction time (RTs) recorded from the onset of each matrix to participant response and accuracy on discrepant trials (Fox et al., 2000). These dependent variables were analyzed using two 3 (group: control, trainees vs VECTORS) x 2 (target: angry and happy) x 2 (distracter: neutral and emotional) ANOVAs with repeated measures on the last two factors. The FICE incorporated varied mapping to ensure that all six combinations of distracters and targets were utilized, thus promoting a controlled rather than automatic search strategy (Schneider & Shiffrin, 1977). Even though neutral faces appeared as targets, they did not form an explicit part of our hypotheses, and as such were not included in the analysis (see also, Juth et al., 2005; Öhman et al., 2001). Emotional reaction to completing the visual search task was examined by a one-way independent groups ANOVA on participants’ pre and post test change in state anxiety (STAI-S) scores from phase 2 (described below) of the study. The STAI is rated on a 4-point Likert scale ranging from 1 (never) to 4 (almost always), with higher scores indicating a greater tendency to experience anxiety. Furthermore, correlational analyses on the entire sample’s state anxiety change scores were carried out to assess their relationship to performance on threat detection (i.e., angry target-happy distracters, angry target-neutral distracters) and inhibitory control over threat distracters on happy target discrepant trials.