Applied Ergonomics 48 (2015) 263e272

Contents lists available at ScienceDirect

Applied Ergonomics journal homepage: www.elsevier.com/locate/apergo

Persistence of threat-induced errors in police officers' shooting decisions ^ ul R.D. Oudejans b Arne Nieuwenhuys a, b, *, Geert J.P. Savelsbergh b, c, Rao a

Behavioural Science Institute, Radboud University Nijmegen, The Netherlands MOVE Research Institute Amsterdam, Faculty of Human Movement Sciences, VU University Amsterdam, The Netherlands C Academy of Physical Education, Amsterdam University of Applied Sciences, The Netherlands b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 22 May 2014 Accepted 15 December 2014 Available online

This study tested whether threat-induced errors in police officers' shooting decisions may be prevented through practice. Using a video-based test, 57 Police officers executed shooting responses against a suspect who rapidly appeared with (shoot) or without (don't shoot) a firearm. Threat was manipulated by switching on (high-threat) or switching off (low-threat) a “shootback canon” that could fire small plastic bullets at the officers. After an initial pretest, officers were divided over four different practice groups and practiced their shooting decisions for three consecutive weeks. Effects of practice were evaluated on a posttest. On the pretest, all groups experienced more anxiety and executed more falsepositive responses under high-threat. Despite practice, these effects persisted on the posttest and remained equally strong for all practice groups. It is concluded that the impact of threat on police officers' shooting decisions is robust and may be hard to prevent within the limits of available practice. © 2014 Elsevier Ltd and The Ergonomics Society. All rights reserved.

Keywords: Anxiety Decision making Representative practice

1. Introduction Gaining control over automated, fear-related, responses is not easy (e.g., Bargh, 1999). Yet, in several professions, being able to manage your anxieties is essential for performance. For instance, police officers are often confronted with the aggressive behavior of civilians. In such situations, it is important that officers make the right decisions and do not let feelings of fear and anxiety influence their operational performance (e.g., Anderson et al., 2002). Under anxiety, people generally tend to show increased attention for threat (Easterbrook, 1959; Eysenck et al., 2007) and are more likely to interpret situations in a threat-related manner (Bishop, 2007). In addition, anxiety is believed to facilitate behavioral responses to threat (e.g., avoidance), which make it harder to efficiently execute goal-directed action (Frijda, 1988; Zajonc, 1980). Although police officers are required to perform well under stressful circumstances, several studies have shown that anxiety does not leave them e or their performances e unaffected (e.g.,

* Corresponding author. Behavioural Science Institute, Radboud University Nijmegen, Montessorilaan 3, 6525 HR Nijmegen, The Netherlands. Tel.: þ31 24 3612094. E-mail addresses: [email protected] (A. Nieuwenhuys), g.j.p. [email protected] (G.J.P. Savelsbergh), [email protected] (R.R.D. Oudejans). http://dx.doi.org/10.1016/j.apergo.2014.12.006 0003-6870/© 2014 Elsevier Ltd and The Ergonomics Society. All rights reserved.

Hulse and Memon, 2006; Nieuwenhuys et al., 2009; Nieuwenhuys and Oudejans, 2010; Nieuwenhuys et al., 2012b; Shipley and Baranski, 2002; Vickers and Lewinski, 2012). In the current study we follow-up on a previous experiment in which we showed how threat-induced increases in anxiety negatively influence police officers' shooting decisions (Nieuwenhuys et al., 2012b). Based on this finding, the aim of the current study was to explore the extent to which the making of such errors may be prevented through practice. In our previous experiment (Nieuwenhuys et al., 2012b), we asked police officers to take shooting decisions (i.e., shoot or don't shoot) in relation to video-images of a suspect that rapidly appeared with or without a firearm. If the suspect appeared with a firearm, officers were supposed to shoot at the suspect. If the suspect appeared without a firearm, officers were supposed not to shoot at the suspect. Threat was manipulated by switching on (high-threat) or switching off (low-threat) a so-called ‘shootbackcanon’ that could fire small plastic bullets at the officers' legs. It appeared that officers were more anxious and showed a larger bias towards shooting in the high-threat compared to the low-threat condition. That is, shooting responses became faster and the percentage of unarmed suspects that was accidentally shot at, almost doubled (i.e., from 11.8% to 18.3%). While the officers' gaze patterns (as measured with a mobile eye-tracker) did not differentiate

264

A. Nieuwenhuys et al. / Applied Ergonomics 48 (2015) 263e272

between correct and incorrect shooting responses (i.e., officers fixated the same locations and detected the suspect equally fast), incorrect shooting responses (at unarmed suspects) were made almost 20% (88 ms) faster than correct shooting responses (at armed suspects). These results indicated that when the officers made a shooting error, they tended not to wait for visual information about the suspect's firearm but immediately responded to the suspect's appearance, which e on average e appeared 100 ms earlier in the video recordings. Because shooting errors occurred almost twice as often in the high-threat than in the low-threat condition, it was concluded that threat-induced increases in anxiety may have biased officers towards responding on the basis of threat-related inferences and expectations (i.e., expecting that the suspect would appear with a firearm) rather than actual visual information about the presence of a firearm (see Correll et al., 2011; Fleming et al., 2010 for similar findings; see Nieuwenhuys and Oudejans, 2012; Payne, 2006; for a more theoretical discussion on this type of effects).1 Because the ability to make good decisions under stressful circumstances is critical for police officers, the current study aimed to test the extent to which threat-induced errors in shooting decisions may be prevented through practice. Although erroneous shooting decisions by police officers have great societal impact, to our knowledge there are no studies directly addressing this topic. However, related work on unintended stereotyping and weapon identification (Correll et al., 2007; Plant and Peruche, 2005; Plant et al., 2005), as well as some of our own experiments on the effects of anxiety on police officers' shot accuracy (Oudejans, 2008; Nieuwenhuys and Oudejans, 2011), indicate that this might be possible. In addition, recent work by Vickers and Lewinski (2012) showed that experienced police officers outperformed lessexperienced police officers on a high-threat shooting decision task. Although these authors did not control for the anxiety that was experienced by participants, this finding indirectly suggests that high-threat shooting decisions are sensitive to practice.

1.1. Weapon identification, unintended stereotyping and representative design When people sit in front of a computer screen and are asked to rapidly judge images of a person holding a gun or another (nonthreatening) object, they are more likely to falsely report guns in relation to Black rather than White individuals (e.g., Correll et al., 2002; Payne, 2001). According to Payne (2006) the key mechanism here is that intuitively, Black individuals are more strongly associated with violence. Seeing a Black individual automatically triggers this stereotype and causes people to more often respond on the basis of threat-related associations, rather than the actual sight of a gun (see also Correll et al., 2011). Effects of unintended stereotyping on weapon identification are widespread and are very hard to willfully suppress (e.g., Payne et al., 2002). However, there are indications that with practice, the observed bias can be eliminated within a relatively short period of time (i.e., by performing an additional 80 trials on the same task; e.g., Correll et al., 2007; Plant and Peruche, 2005; Plant et al., 2005; but also see Luini and Marucci, 2013).

1 It is of interest to note that in a related study by Mitchell and Flin (2007) the priming of officers' threat-related expectations (by means of neutral vs. high-threat briefing information) did not significantly affect shooting decisions and accompanying response times in a subsequent shoot vs. don't-shoot scenario. Although this study has some clear experimental weaknesses (e.g., limited control over scenarios, participants executing only one single trial) it does indicate that increased threatrelated expectations may not always bias officers towards shooting.

Although the effects of unintended stereotyping have clear parallels with the threat-induced shooting bias shown by Nieuwenhuys et al. (2012b; i.e., the bias to respond on the basis of threat-related inferences and associations rather than visual information about the presence of a gun), it is important to note that pressing a button on a computer keyboard is not the same as actually shooting at another person. Similarly, implicit threatrelated associations that are connected to the ethnicity of a suspect are different from the actual possibility of getting hit, which remains uninvestigated in the weapon identification literature (Nieuwenhuys et al., 2012b). In this respect, several studies have shown that such differences in the reality of a task and the specificity and nature of responses can have large consequences for the detection of visual information, decision making, and the eventual action that is undertaken by participants e with experienced participants showing greater sensitivity to threat (Nieuwenhuys et al., 2012a), more accurate perception and better decision making under more representative circumstances (e.g., Dicks et al., 2010; Mann et al., 2010; see Pinder et al., 2011; for a more theoretical discussion of this topic). As a consequence, it remains to be seen whether the positive effects of practice, as reported by Correll et al. (2007) and Plant and Peruche (2005), would hold under more representative circumstances, where police officers' can actually get hit and indicate their shooting decisions based on actual shooting responses. 1.2. Training with anxiety: shot accuracy vs. shooting decisions Under high threat, police officers are more inclined to shoot (Nieuwenhuys et al., 2012b) and tend to shoot with lower accuracy (Nieuwenhuys and Oudejans, 2010). Recently, however, we showed that by training with anxiety, police officers can improve their shot accuracy under stressful circumstances (Nieuwenhuys and Oudejans, 2011; see also Oudejans, 2008). Nieuwenhuys and Oudejans (2011) had two groups of police officers practice their shot accuracy e either against opponents that carried a handgun loaded with colored-soap cartridges (high-threat; experimental group) or against opponents that carried an imitation handgun (low-threat; control group). Importantly, this study did not involve decision making (i.e., making the decision to shoot or not shoot): participants always shot at the suspect and only practiced their shot accuracy (i.e., hit the suspect). Before training, high threat caused a decrease in shot accuracy for both groups. After training, the experimental group was able to maintain accuracy under high threat. Shot accuracy of the control group, on the other hand, was still negatively affected by threat (see Oudejans, 2008; Oudejans and Pijpers, 2009, 2010, for similar results). Analyses of gaze behavior indicated that improved shot accuracy was likely related to improved goal-directed attention (i.e., maintaining relatively long fixations on the target while reducing distraction from other sources of information). These results are comparable with effects of visual attention training (e.g., Harle and Vickers, 2001; Vine and Wilson, 2011; Wood and Wilson, 2011) and indicate that by training with anxiety participants can learn to maintain visual attentional control, also under high threat circumstances (Oudejans and Nieuwenhuys, 2009; see also Eysenck et al., 2007). Importantly, studies on training with anxiety, as well as studies on visual attention training, have always focused on some form of shot accuracy (e.g., darts throwing, basketball free throw shooting, soccer penalty kicks, and police officers' handgun shooting). In the current study, however, we focused on decision making. Clearly, being able to accurately shoot at a target is not the same as deciding to shoot or not shoot at another person. That is, shooting clearly is a motor task, while taking the decision to shoot is arguably more cognitive (Nieuwenhuys and Oudejans, 2012). Because different

A. Nieuwenhuys et al. / Applied Ergonomics 48 (2015) 263e272

types of tasks involve different control mechanisms, it remains to be investigated whether training with anxiety is equally effective for shooting decisions as it is for shot accuracy. More specifically, it may be relatively easy to learn how to keep one's eyes on a target under anxiety, while it is more difficult to objectify one's interpretation of a stressful situation such that it allows unbiased selection of shooting vs. non-shooting responses (see also Payne et al., 2002). 1.3. Current study Our experiments on training with anxiety, as well as the broader literature on representative practice, suggest that practice is most effective when the circumstances under which individuals practice closely match the circumstances under which they are tested (see Oudejans and Nieuwenhuys, 2009; Pinder et al., 2011). That is, if people need to perform well under high-threat circumstances, then practice should also occur under high-threat circumstances (Oudejans and Nieuwenhuys, 2009). Similarly, if people need to perform well under real-life circumstances, then practice should also occur under real-life circumstances (e.g., as opposed to videosimulation; Pinder et al., 2011). In the current study we tested this assumption, as we explored to what extent threat-induced errors in police officers' shooting decisions can be prevented through ‘representative’ practice. To this end, we adopted our previous shoot-don't shoot paradigm (Nieuwenhuys et al., 2012b) and again asked police officers to take shooting decisions under low- and high-threat conditions. After a pretest, in which we expected to replicate our previous findings (i.e., more shooting errors under high-threat, potentially because of an increased tendency to respond on the basis of threat-related inferences and expectations), the officers were divided into four different practice groups, which differed with respect to the affective (high-threat vs. low-threat) and visual (video-based vs. real-life) circumstances under which practice took place. Following previous experiments on training with anxiety (Nieuwenhuys and Oudejans, 2011, 2008; Oudejans and Pijpers, 2009, 2010), practice consisted of three 1-h sessions that were executed over a period of three weeks (i.e., one session per week). At the end of the practice period, effects of practice were evaluated on a posttest. In line with our representative practice assumption (Oudejans and Nieuwenhuys, 2009; Pinder et al., 2011), and provided that threat-induced errors in shooting decisions can be prevented through practice, we predicted that positive effects of practice (i.e., less threat-induced shooting errors) would depend on the degree to which practice circumstances would be representative of (i.e., match with) the circumstances under which participants performed the tests. As such, we predicted positive effects for the group that practiced under circumstances that were completely (i.e., both affectively and visually) representative of the test and smaller or no effects for the practice groups that practiced under circumstances that were partially representative (i.e., affective or visual) or not representative (no practice control) of the test. Findings were expected to bear implications for the training of police officers, which (a) is limited in terms of the time that is available for practice, and (b) often occurs under non-representative circumstances (e.g., on a shooting lane). Also, depending on the extent to which threatinduced errors in shooting decisions are sensitive to practice, findings were expected to bear implication for the legal investigation of officer-involved shootings. 2. Methods The experiment was formally approved by the Ethics Committee of the research institute. Given the involvement of firearms and a

265

‘shootback canon’ (see Section 2.2), the experiment took place at the premises of the local police academy and was executed under the responsibility of certified police firearms instructors, thereby taking into account all of their normal safety procedures. Before the experiment started, all participants provided written informed consent. 2.1. Participants 57 police officers (52 men, 5 women) volunteered to participate in the experiment. All participants had a full license to carry a firearm on duty. After performing a pretest (see ‘Experimental Setup, Task, and Conditions’) participants were divided over four practice groups: a video-based practice under high-threat (VBP-HT) group, a video-based practice under low-threat (VBP-LT) group, a real-life practice under high-threat (RLP-HT) group, and a nopractice control group (see ‘Experimental Setup, Task and Conditions’ for a detailed explanation of the different practice circumstances for each group). The VBP-HT group consisted of 16 participants (14 men, 2 women), with a mean age of 38.07 years (SD ¼ 8.66) and a mean working experience of 15.73 years (SD ¼ 9.76). The VBP-LT group consisted of 18 participants (17 men, 1 woman), with a mean age of 37.41 years (SD ¼ 10.72) and a mean working experience of 14.29 years (SD ¼ 9.33). The RLP-HT group consisted of 13 participants (12 men, 1 woman), with a mean age of 36.54 years (SD ¼ 11.38) and a mean working experience of 13.19 years (SD ¼ 12.24). The control group consisted of 10 participants (9 men, 1 woman), with a mean age of 37.20 years (SD ¼ 8.90) and a mean working experience of 12.90 years (SD ¼ 9.09). Among groups, there were no differences regarding the age and working experience of participants, F(3,46) ¼ 1.30, p ¼ .285. Overall, participants' trait anxiety scores were significantly lower than the norm (i.e., 36.1; with M ¼ 30.86, SD ¼ 6.51, t(50) ¼ 5.70, p < .001; STAI A-Trait Scale; Van der Ploeg et al., 1980), indicating that they had no extraordinary tendency to respond to specific situations with large elevations in state anxiety. Again, no significant differences were observed among groups (F < 1, p > .05). 2.2. Experimental setup, task and conditions The experiment consisted of a pretest, three practice sessions, and a posttest (see Fig. 1).2 2.2.1. Test sessions During the pretest and posttest, participants performed a test exercise that was identical to that of Nieuwenhuys et al. (2012b). Participants took position in an Applied Interactive Systems (‘AIS’) PRISim® shooting simulator (AIS-solutions Ltd., Wrecclesham, United Kingdom) and responded to life-size video images of a dangerous suspect. In the current setup, video images were projected on a 6 m  2.5 m projection screen. Participants stood at a fixed position 5 m away from the screen (see Fig. 2) and could shoot at the suspect with a blank firing Walther P5 handgun (similar to their duty weapon) that was fitted with a trigger sensor and a laser diode that emitted a single laser pulse (5 ms) upon firing. In case of a shot, the laser pulse was detected by an infrared sensor (60 Hz

2 For the VBP-HT and VBP-LT groups, pretest results (pooled over both groups) have been published previously (Nieuwenhuys et al., 2012b). In that publication immediate effects of anxiety on police officers' shooting decisions were investigated, with a specific focus on the importance of visual information (assessed by measuring participants' gaze patterns). Due to technical problems, gaze data was not available at the posttest and for the RLP-HT and control groups.

266

A. Nieuwenhuys et al. / Applied Ergonomics 48 (2015) 263e272

Fig. 1. Congruency of test and practice circumstances (visual and affective) for each practice group (Left Panel) and overview of the experimental design (Right Panel). VBPHT ¼ Video-Based Practice under High-Threat; VBP-LT ¼ Video-Based Practice under Low-Threat; RLP-HT ¼ Real-Life Practice under High-Threat; Control ¼ No Practice.

Fig. 2. Overview of the Experimental Setup (Left Side) and Video Image of the Suspect on GUN trials (Top Right) and NO-GUN trials (bottom right). On Each Trial (GUN and NO-GUN) the Suspect Could Appear in the Left as well as in the Right Window.

sampling rate) that was used to record the timing (±12 ms) and accuracy (±2 cm) with which the gun was fired. As in Nieuwenhuys et al. (2012b), the pretest and posttest consisted of a low-threat (LT) and a high-threat (HT) condition, which were counterbalanced among participants. In both conditions, participants performed 4  12 trials (i.e., 48 in total) of the same experimental task. Each trial started with showing an image of two empty windows (see Fig. 2). Then, after a randomized period of 1.5, 2, or 2.5 s, a suspect would rapidly appear in one of the windows. If the suspect appeared with a firearm (i.e., GUN trial, Fig. 2, top right) participants were supposed to shoot him as fast and accurately as possible. When participants responded too late (i.e., used more than ~500 ms for their response) or responded inaccurately (i.e., failed to hit the suspect), the suspect would shoot back (see below). If the suspect appeared without a firearm (i.e., NO-GUN trial, Fig. 2, bottom right) participants were supposed not to shoot at the suspect. In this case, they placed a shot in a black taped square (40 cm  40 cm) that was positioned well below the two windows in the middle of the screen (see Fig. 2). Participants were allowed only one shot per trial. Between trials, the screen went black for 3 s. In the LT condition, participants only saw and heard the opponent's shots, thereby creating a relatively harmless experience. In the HT condition, however, a so-called “shootback-canon” (AISsolutions Ltd., Wrecclesham, United Kingdom) was activated. The shootback-canon is a manually aimed air-pressure system that was positioned at the bottom-right corner of the projection screen (see Fig. 2), and which could be used to shoot small plastic bullets (15 mm diameter) at the participants' legs, at the exact moment of the suspect's shots. Being hit with such a bullet caused a sensation of pain, the threat of which has been found to cause an increase in participants' state anxiety (see also Nieuwenhuys et al., 2012b).

Given the speed of the scenarios, participants inevitably failed to be faster than the suspect on ~50% of the GUN trials, thereby providing enough opportunity to shoot back with the shootback canon. To maintain threat and minimize physical inconvenience, the shootback canon was used independent of participants' performance on a fixed number of HT GUN trials (i.e., 5e7 times in total, randomly divided over trials; Nieuwenhuys et al., 2012b). On other trials the shootback-canon was aimed slightly off target, thereby preventing the participants from being hit.3 2.2.2. Practice sessions In between the pretest and posttest, all groups (with exception of the control group) performed three 1-h practice sessions. During each practice session, participants individually performed seven extended scenarios in which they practiced their shooting decisions. As in previous experiments (Nieuwenhuys and Oudejans, 2011, 2008; Oudejans and Pijpers, 2009), practice scenarios were , a house or situated in a wide range of different settings (e.g., a cafe a car park) and e due to an extended period of verbal interaction

3 It is important to note that the applied manipulation of threat (i.e., the threat of being hit) was specifically selected on the basis of its high ecological validity for the current target group (i.e., police officers). In its current form, the manipulation was an existing functionality of the Applied Interactive Systems shooting simulator (AIS-solutions Ltd., Wrecclesham, United Kingdom), which e at the police academy in question e was being used in regular police officers' firearms practice. In our study, participants were informed about the manipulation before agreeing to participate. Operation of the system always occurred under supervision of certified police firearms instructors. In comparable forms, the applied manipulation has been used successfully in several other studies that investigated the impact of threat on police officers' shooting behavior (e.g., Nieuwenhuys and Oudejans, 2010, 2011; Nieuwenhuys et al., 2012b; Shipley and Baranski, 2002; Vickers and Lewinski, 2012).

A. Nieuwenhuys et al. / Applied Ergonomics 48 (2015) 263e272

267

Fig. 3. Example of the Structure of a Practice Trial. Participants Enter a Bar Where the Suspect Sits at a Table (Left Picture). Initially the Gun is Not Visible to Participants (On the Table, Next to the Suspect). After an Initial Conversation with the Participant, the Suspect Suddenly Grabs the Gun and Shoots (A) or Raises his Hands in Order to Surrender (B).

with the suspect e generally lasted longer than the test trials (i.e., ~60 s vs. ~5 s). This was done to increase ecological validity and to make sure that positive effects of practice would not be restricted to (repeated experience with) the test exercise. To maintain tasksimilarity, however, shooting decisions (i.e., shoot or do not shoot) were always taken within less than a second, as at the end of each scenario the suspect would suddenly reach for his gun (shoot) or raise his hands (not shoot; see Fig. 3 for an example). In general, practice scenarios were similar for all groups. However, following our representative practice hypothesis (Oudejans and Nieuwenhuys, 2009; Pinder et al., 2011), practice circumstances (affective and visual) were different for each group and showed full (VBP-HT), partial (VBP-LT and RLP-HT) or no (control) overlap with test circumstances (see Fig. 1). The VBP-HT and VBT-LT groups practiced in the AIS video-simulator with the shootback canon switched on or off, respectively. The RLP-HT group practiced against real-life opponents that shot back with colored-soap cartridges. 2.3. Dependent variables 2.3.1. Manipulation check To analyze the effect of our threat manipulation during the pretest and posttest, as well as during the respective practice sessions, we assessed participants' subjective ratings of anxiety and mental effort (in each test condition and practice session) by using two distinctive visual-analog scales (i.e., the ‘anxiety thermometer’, Houtman and Bakker, 1989, and the 'Rating Scale for Mental Effort' [RSME], Zijlstra, 1993). Furthermore, we continuously assessed participants' heart rate by using a Polar wristwatch and chest band. 2.3.2. Shooting decisions Shooting decisions (pretest and posttest only) were measured using the AIS and e following Nieuwenhuys et al. (2012b) e concentrated on participants' shooting responses only. As such, we calculated the percentage of correct responses on GUN trials (i.e., shooting the suspect when he appeared with a gun) and the percentage of incorrect responses on NO-GUN trials (i.e., accidentally shooting an unarmed suspect) for each condition. 2.3.3. Response times Response times (pretest and posttest only) were measured using the AIS and e in line with the analysis of shooting decisions e also

concentrated on participants' shooting responses (i.e., correct responses on GUN trials and incorrect responses on NO-GUN trials). Response times (in milliseconds) were measured from the first moment that the hands of the suspect became visible in the video scenarios (i.e., GUN/NO-GUN stimulus onset) until the firing of the gun by the participant. 2.3.4. Shot accuracy Shot accuracy (pretest and posttest only) was analyzed on the basis of the AIS data and operationalized as the percentage of target (suspect) hits on GUN trials. On NO-GUN trials, shot accuracy was not assessed. 2.4. Statistical analysis Anxiety scores, mental effort scores, and heart rate values on the pretest and posttest, were collectively analyzed by using a 4  2  2 (practice group  test  condition) mixed-design MANOVA, with repeated measures on test (pretest and posttest) and condition (LT and HT) and with practice group (VBP-HT, VBP-LT, RLP-HT, control) as a between-subject factor. Shooting decisions and associated response times were analyzed using univariate 4  2  2  2 (practice group  test  condition  decision quality) mixed design ANOVAs, with repeated measures on test (pretest and posttest), condition (LT and HT) and decision quality (correct and incorrect shooting responses), and with practice group (VBP-HT, VBP-LT, RLPHT, control) as a between-subject factor. Shot accuracy was analyzed by using a univariate 4  2  2 (practice group  test  condition) mixed-design ANOVA, with repeated measures on test (pretest and posttest) and condition (LT and HT) and with practice group (VBP-HT, VBP-LT, RLP-HT, control) as a between-subject factor. To verify our manipulation of threat during the practice sessions, anxiety scores, mental effort scores and heart rate values (averaged over the different practice sessions) were collectively analyzed by using a MANOVA, with practice group (VBP-HT, VBP-LT, RLP-HT) as a between-subject factor. For the MANOVAs, significant multivariate effects were followed-up by conducting univariate ANOVAs on each of the respective variables (anxiety, mental effort and heart rate). For the ANOVAs, significant main effects or interactions were followed-up using post-hoc pairwise comparisons with Bonferroni correction.

268

A. Nieuwenhuys et al. / Applied Ergonomics 48 (2015) 263e272

Table 1 Overview of the Results for our Manipulation Check at the Test Sessions (Low-Threat [LT] and High-Threat [HT] Condition; Pretest and Posttest) and at the Practice Sessions, for Each Practice Group. Variable

Pretest

Practice

LT

HT

M (SD)

M (SD)

Anxiety score (0e10) VBP-HT 4.1 (2.5) 6.1 VBP-LT 3.7 (2.0) 4.3 RLP-HT 4.4 (2.2) 6.5 Control 3.3 (1.9) 5.1 Mental effort score (0e150) VBP-HT 54.0 (27.0) 67.4 VBP-LT 57.2 (24.3) 61.4 RLP-HT 75.5 (20.8) 82.4 Control 49.9 (23.5) 58.1 Mean heart rate value (b/min) VBP-HT 92.5 (17.2) 94.8 VBP-LT 95.9 (11.3) 98.3 RLP-HT 102.9 (13.4) 105.8 Control 97.8 (16.3) 103.9

M (SD)

Posttest LT

HT

M (SD)

M (SD)

(2.3) (2.3) (2.3) (2.1)

5.7 (1.6) 5.3 (1.8) 6.1 (1.1)

4.2 4.2 5.0 3.6

(1.9) (1.7) (2.3) (1.6)

5.7 5.6 6.5 5.2

(1.7) (1.9) (1.4) (1.8)

(27.9) (22.2) (19.5) (24.9)

62.3 (18.8) 58.4 (16.9) 74.9 (12.9)

59.1 58.4 74.9 62.9

(21.5) (16.7) (22.2) (21.1)

68.3 69.4 85.2 69.7

(20.2) (16.6) (19.3) (25.3)

(16.9) (14.3) (12.9) (18.3)

88.6 (12.0) 89.7 (13.8) 118.8 (19.4)

89.4 90.2 93.4 95.7

(11.0) (10.9) (13.1) (11.0)

92.3 91.5 96.9 99.3

(12.7) (12.5) (11.4) (11.9)

For each analysis, effect sizes (h2) were calculated. The alpha level for significance was set at p < .05. 3. Results 3.1. Manipulation checks Table 1 shows an overview of the anxiety scores, mental effort scores, and mean heart rate values at the pretest, practice sessions, and posttest. 3.1.1. Test sessions The MANOVA that was executed for the anxiety scores, mental effort scores and mean heart rate values at the pretest and posttest showed significant multivariate effects of test (pretest vs. posttest), l ¼ .755, F(3,50) ¼ 5.43, p ¼ .003, h2 ¼ .245, and condition (LT vs. HT), l ¼ .280, F(3,50) ¼ 42.97, p < .001, h2 ¼ .720, and no significant multivariate effect of practice group or any of the interactions (all ps > .11). For the effect of test (pretest vs. posttest), follow-up univariate ANOVAs showed that while mental effort scores were a little higher on the posttest than on the pretest, F(1, 52) ¼ 5.25, p ¼ .023, h2 ¼ .096, mean heart rate values were a little lower, F(1, 52) ¼ 5.25, p ¼ .004, h2 ¼ .145 (see Table 1). Anxiety scores did not change significantly from pretest to posttest, F(1, 52) ¼ 2.11, p ¼ .15, h2 ¼ .039. Importantly, for the effect of condition (LT vs. HT), follow-up univariate ANOVAs showed that anxiety scores, mental effort scores, and mean heart rate values were all significantly higher in the HT than in the LT condition, with F(1,52) ¼ 114.67, p < .001, h2 ¼ .688, F(1,52) ¼ 41.48, p < .001, h2 ¼ .444 and F(1,52) ¼ 17.96, p < .001, h2 ¼ .257, respectively (see Table 1). Taken together, these findings indicate that our manipulation of threat was equally successful for all practice groups and on both tests. 3.1.2. Practice sessions The MANOVA that was executed for the anxiety scores, mental effort scores and mean heart rate values during the practice sessions showed a strong multivariate effect of practice group (VBPHT, VBP-LT, RLP-HT), l ¼ .477, F(6,84) ¼ 6.27, p < .001, h2 ¼ .309.

Follow-up, univariate ANOVAs showed a non-significant effect for anxiety score, F(2,44) ¼ 1.97, p ¼ .15, h2 ¼ .082, and significant effects for mental effort score and mean heart rate value, with F(2,44) ¼ 3.91, p ¼ .027, h2 ¼ .151 and F(2,44) ¼ 18.32, p < .001, h2 ¼ .454, respectively. Post-hoc analyses showed that mental effort scores and mean heart rate values of the RLP-HT group were significantly higher than those of the VBP-HT and VBP-LT groups, with p ¼ .046 and p ¼ .009 for mental effort and ps < .001 for mean heart rate (see Table 1). There were no significant differences between the VBP-HT and VBP-LT group (all ps > .50). Overall, these effects indicate that during practice the RLP-HT experienced slightly higher levels of threat than the other groups, which may be entirely subscribed to the fact that in this group, participants were shooting at live opponents rather than video simulated assailants. However, it should be noted that for all practice groups, anxiety scores, mental effort scores and mean heart rate values during practice, were at or above what was measured during the HT test sessions (see Table 1). 3.2. Shooting decisions Because participants always responded correctly on GUN trials (see Table 2) the analysis of shooting decisions was confined to the percentage of incorrect responses on NO-GUN trials (i.e., accidentally shooting at unarmed suspects). The ANOVA that was executed showed a significant main effect of test (pretest vs. posttest), F(1,53) ¼ 4.24, p ¼ .044, h2 ¼ .074, and condition (LT vs. HT), F(1,53) ¼ 53.93, p < .001, h2 ¼ .504, but no significant effect of practice group or any of the interactions (all ps > .31; see Fig. 4). The main effect of test showed that for all practice groups (including the control group), and regardless of condition, the percentage of incorrect responses in NO-GUN scenarios was slightly, but significantly lower on the posttest than on the pretest (i.e., M ¼ 12.20%, SD ¼ 1.26 vs. M ¼ 14.07%, SD ¼ 1.52%; see Table 2). The main effect of condition showed that for all practice groups (and regardless of test) the percentage of incorrect responses on NO-GUN trials was significantly higher in the HT than in the LT condition (i.e., M ¼ 16.51%, SD ¼ 1.68 vs. M ¼ 9.74%, SD ¼ 1.04; see Table 2 and Fig. 4). 3.3. Response times The ANOVA that was executed for the response times showed significant effects of decision quality (correct vs. incorrect shooting responses), F(1,47) ¼ 290.44, p < .001, h2 ¼ .861, test (pretest vs. posttest), F(1,47) ¼ 22.21, p < .001, h2 ¼ .299, and condition (LT vs. HT), F(1,47) ¼ 35.89, p < .001, h2 ¼ .408, which were overruled by a significant interaction between decision quality and test, F(1,47) ¼ 9.38, p ¼ .004, h2 ¼ .166, and a marginally significant interaction between decision quality and condition, F(1,47) ¼ 3.87, p ¼ .055, h2 ¼ .076. The main effect of practice group and all other interactions were non-significant (ps > .27). Post-hoc pairwise comparisons showed that incorrect shooting responses (i.e., accidentally shooting an unarmed suspect) were made much faster (i.e., 80 ms, 17%) than correct shooting responses (ps < .001, Fs > 96.70; M ¼ 381 ms, SD ¼ 7 vs. M ¼ 461 ms, SD ¼ 6; see Table 2). In addition, while the speed with which incorrect shooting responses were made remained equally fast from pretest to posttest, F(1,47) ¼ .17, p ¼ .68, h2 ¼ .004 (M ¼ 383 ms, SD ¼ 9 vs. M ¼ 380 ms, SD ¼ 7) and did not differ significantly between LT and HT conditions, F(1,47) ¼ 3.15, p ¼ .083; M ¼ 375 ms (SD ¼ 8 vs. M ¼ 388 ms, SD ¼ 7), correct shooting responses were made significantly faster on the posttest than on the pretest, F(1,47) ¼ 23.95, p < .001, h2 ¼ .338 (M ¼ 450 ms, SD ¼ 7 vs. M ¼ 473 ms, SD ¼ 6) and were significantly faster in the HT than in

A. Nieuwenhuys et al. / Applied Ergonomics 48 (2015) 263e272

269

Table 2 Overview of the results for shooting decisions, response times, and shot Accuracy at the test sessions (low-threat [LT] and high-threat [HT] condition; pretest and posttest), for each practice group. Variable

Shooting decisions GUN correct (%)

NO-GUN incorrect (%)

Response times GUN correct (ms)

NO-GUN incorrect (ms)

Shot accuracy Suspect hits (%)

Pretest

Posttest

LT

HT

LT

HT

M (SD)

M (SD)

M (SD)

M (SD)

VBP-HT VBP-LT RLP-HT control

100 100 100 100

100 100 100 100

100 100 100 100

100 100 100 100

VBP-HT VBP-LT RLP-HT control

11.72 (10.67) 12.27 (12.08) 10.58 (8.09) 8.33 (5.56)

17.19 19.91 19.05 13.33

7.81 (7.43) 10.02 (7.01) 9.66 (8.75) 7.52 (5.82)

17.49 17.13 15.49 12.50

VBP-HT VBP-LT RLP-HT control

499 490 480 491

(58) (44) (46) (39)

468 464 456 468

(58) (45) (43) (45)

465 479 461 470

(42) (56) (54) (47)

440 455 435 451

(44) (49) (52) (66)

VBP-HT VBP-LT RLP-HT control

413 389 384 365

(59) (52) (62) (42)

381 381 354 371

(59) (60) (85) (20)

387 386 374 402

(72) (57) (56) (66)

377 383 382 372

(57) (70) (75) (61)

VBP-HT VBP-LT RLP-HT control

82.50 84.32 90.48 88.75

(16.02) (14.83) (5.60) (7.31)

the LT conditions F(1,47) ¼ 36.78, p < .001, h2 ¼ .439 (M ¼ 448 ms, SD ¼ 7 vs. M ¼ 475 ms, SD ¼ 7; see Table 2). 3.4. Shot accuracy The ANOVA that was executed for shot accuracy (correct responses on GUN trials only) showed a significant main effect of condition, F(1,48) ¼ 9.49, p ¼ .003, h2 ¼ .165, and no other main effects or interactions (all ps > .09). In general, shot accuracy was lower in the HT than in the LT condition (M ¼ 80.10%, SD ¼ 2.00 vs. M ¼ 85.16%, SD ¼ 1.66; see Table 2). 4. Discussion When police officers are anxious to get hit (i.e., under highthreat circumstances) they are more inclined to shoot and make more incorrect shooting decisions than when they are not anxious

Fig. 4. Incorrect Shooting Decisions (NO-GUN incorrect; % of trials) in the Low-Threat (LT) and High-Threat (HT) Condition on the Pretest (Before Practice) and Posttest (After Practice), for Each Practice Group. *p < .05; ***p < .001.

73.43 81.90 83.36 81.61

(11.87) (16.20) (13.74) (9.38)

(19.98) (14.78) (15.01) (10.98)

84.84 83.26 81.07 86.06

(12.75) (14.53) (11.22) (16.27)

82.21 83.46 73.36 81.44

(14.69) (12.45) (12.83) (7.35)

(13.08) (13.61) (19.39) (23.71)

to get hit (Nieuwenhuys et al., 2012b; see also Payne, 2001; Correll et al., 2002; Fleming et al., 2010). Using a pretesteinterventioneposttest design, the current study aimed to explore the extent to which the occurrence of these threat-induced errors in police officers' shooting decisions (i.e., accidentally shooting at unarmed suspects) may be prevented through practice. Based on previous work, which suggests that practice is more effective when test circumstances are adequately represented (see Pinder et al., 2011; Oudejans and Nieuwenhuys, 2009; for reviews), officers were divided over four different practice groups, which differed with respect to the overlap between test and practice circumstances (i.e., complete, partial, or not at all). We predicted that e if threat-induced increases in shooting errors can be prevented through practice e positive effects of practice (i.e., a reduction of threat-induced shooting errors) should be most pronounced in the group that practiced under the most representative circumstances. Results showed that at the pretest, all groups experienced more anxiety and made more shooting errors under high-threat than under low-threat circumstances, thereby confirming the findings of Nieuwenhuys et al. (2012b; see also Payne, 2001; Correll et al., 2002; Fleming et al., 2010). Overall, the percentage of incorrect shooting decisions was around 10% in the low-threat condition and well over 16% in the high-threat condition (see Fig. 4), indicating that in this case 1 out of 7 unarmed suspects was accidentally shot at. In addition to this effect, officers tended to shoot faster under high threat, which, even when the selected response (shoot) was correct, significantly affected their shot accuracy (i.e., which decreased from 85% in the low-threat to 80% in the high-threat condition; see also Nieuwenhuys et al., 2012b; Causer et al., 2011; Nieuwenhuys and Oudejans, 2010, 2011; Vickers and Williams, 2007). After practice (i.e., on the posttest), all practice groups appeared to make slightly fewer shooting errors than on the pretest, as was evidenced by a significant main effect of test (see Fig. 4). However, this improvement also applied to the control group, which did not practice, suggesting that the observed decrease in shooting errors

270

A. Nieuwenhuys et al. / Applied Ergonomics 48 (2015) 263e272

did not emerge as a result of practice but was likely due to situation-specific experience obtained during the tests. Although the overall percentage of shooting errors thus seemed to decrease slightly from pretest to posttest, the impact of threat (low-threat vs. high-threat condition) on police officers' shooting decisions remained equally strong for all practice groups (Fig. 4). Additional analyses comparing each group's difference scores between both tests (i.e., decision errors made under high-threat minus decision errors made under low-threat) confirmed this observation, showing that there was no significant main effect of test, p ¼ .85, h2 < .01, and no significant test  group interaction of p ¼ .31, h2 ¼ .07 (mixed-design ANOVA). Based on these findings we conclude that our practice intervention did not help to significantly reduce threat-induced errors in police officers' shooting decisions. Of course it is possible that we would have found positive effects of practice if participants had practiced more often, or if practice sessions had involved more repetitions (e.g., Correll et al., 2007). However, we selected the number of practice sessions based on the time that police officers currently have available for practice (Oudejans, 2008), thereby providing a functional account of the possibility to improve police officers' shooting decisions under high-threat circumstances. As such, whether additional practice indeed leads to more positive effects remains a topic for future studies. Furthermore, with an average group size of N ¼ 14.5 (see Methods), another issue that might be raised is that our experimental power may have been too low to observe positive effects of practice. Nevertheless, our group sizes were larger than those obtained in other studies which tested similar types of practice using comparable designs and which showed positive effects of practice for police officers' shot accuracy (Oudejans, 2008; Nieuwenhuys and Oudejans, 2011; see also Harle and Vickers, 2001; Oudejans and Pijpers, 2009, 2010; Vine and Wilson, 2011; Wood and Wilson, 2011). As such, we do not believe that the absence of a positive effect of practice is the simple result of a lack of power. Instead, we believe that our findings indicate that effects of threat on police officers' shooting decisions are relatively robust and may be hard to prevent within the limits of available practice. In addition, they also highlight the importance of distinguishing between different skill types (e.g., decision making vs. shot accuracy) when investigating the impact of threat and anxiety on performance (Nieuwenhuys and Oudejans, 2012; see also DeCaro et al., 2011). 4.1. Decision making vs. shot accuracy Being able to shoot accurately at a target is not the same as taking the decision to shoot (or not shoot) at another person. That is, while shooting itself is clearly a motor task, taking the decision to shoot is typically more cognitive. While shot accuracy has been shown to rely heavily on visual attentional control (e.g., to align movements with the location of the target) and appears to be sensitive to threat-induced visual distraction (e.g., Causer et al., 2011; Nieuwenhuys and Oudejans, 2010; Vickers and Williams, 2007; Wilson et al., 2009), taking the decision to shoot relies on adequate judgment and interpretation and is more sensitive to increased threat-related expectations (e.g., expecting the suspect to appear with a firearm; Correll et al., 2002; Nieuwenhuys et al., 2012b; Payne, 2001). The idea that threat-induced errors in shooting decisions are indeed the result of increased threat-related expectations is indirectly confirmed by the current findings. In line with Nieuwenhuys et al. (2012b), incorrect shooting responses (accidentally shooting at unarmed suspects) were consistently executed much faster (i.e., ~80 ms or 17%) than correct shooting responses (correctly shooting at an armed suspect; see Table 2). Although a difference of 80 ms may appear trivial at first sight, in light of the current experimental task it

reveals that when the officers' made an incorrect response they did not to wait for visual information about the gun but, instead, immediately reacted as soon as they saw the suspect appear from behind one of the two windows e that is, they fired because they expected the suspect to appear with a gun (see also Nieuwenhuys et al., 2012b). Because shooting errors occurred almost twice as often in the high-threat than in the low-threat condition (see Fig. 4) we conclude that high-threat may have biased officers towards responding on the basis of threat-related inferences and expectations (i.e., expecting the suspect to appear with a gun) rather than actual visual information about the presence of a firearm (see Correll et al., 2011; Fleming et al., 2010; Payne, 2001; Nieuwenhuys et al., 2012b). Although speculative, this effect may be explained by a threat-induced increase in anxiety, which strengthens the output of emotional centers in the brain (e.g., the amygdala) while inhibiting pre-frontal control, thereby causing individuals to have more attention for threat (Eysenck et al., 2007), be more likely to interpret specific situations as threatening (Bishop, 2007) and be more likely to respond in a threat-related manner (e.g., Frijda, 1988; see also Nieuwenhuys and Oudejans, 2012). Also, based on our finding that the impact of threat on police officers' shooting decisions remained unchanged from pretest to posttest we conclude that, apparently, reducing the impact of threat on shooting decisions (i.e., shoot or don't shoot) is much harder than reducing the impact of threat on shot accuracy (i.e., being able to hit a target). That is, practice did not lead to better decision making in the current study, while in previous work, similar amounts and similar types of practice did lead to more accurate shooting (e.g., Nieuwenhuys and Oudejans, 2011, Oudejans, 2008). In light of the discussion about different control mechanisms, this means that while it may be relatively easy to learn how to keep one's eyes on a target under anxiety (which is important for shot accuracy), it is very hard to objectify one's interpretation of a stressful situation or to gain control over automated fear-related responses (which is important for judgment and decision-making). 4.2. Representative design Although we were unable to show positive effects of practice in the current study, previous work on weapon identification and unintended stereotyping indicated that in other contexts it is possible to improve judgment and decision making under challenging circumstances (e.g., Correll et al., 2007; Plant and Peruche, 2005; Plant et al., 2005). In these studies, however, practice consisted of repeated execution of the test exercise, which arguably limits transfer across different situations. Another distinction between this work and the current study concerns differences in task design. That is, in the current study, police officers could actually get hit by the suspect and indicated their shooting decisions on the basis of actual shooting responses, whereas in the typical weaponidentification experiment officers cannot get hit and indicate their ‘shooting’ decisions by pressing a button on a keyboard (Correll et al., 2007; Plant and Peruche, 2005; Plant et al., 2005). Such differences in the reality of a task and the specificity and nature of responses can have large consequences for the detection of visual information, decision making, and the eventual action that is undertaken by participants (e.g., Dicks et al., 2010; Mann et al., 2010; see also Pinder et al., 2011). In addition, there is also evidence to suggest that the more representative nature of our shooting task (albeit video-based) may have induced higher levels of anxiety and, hence, more persistent changes in shooting decisions as a result of threat (see Nieuwenhuys et al., 2012a, for an example of this effect). In light of these arguments, it remains to be seen whether the positive effects of practice observed by Correll et al. (2007) and

A. Nieuwenhuys et al. / Applied Ergonomics 48 (2015) 263e272

Plant and colleagues (Plant and Peruche, 2005; Plant et al., 2005) would hold under more representative circumstances. Importantly, we do not suggest that it is impossible to improve police officers' shooting decisions under high threat. After all, we did not explore whether increases in practice volume (e.g., more practice sessions and/or more repetitions) would lead to more positive effects of practice. In addition, while the impact of threat remained unaltered, officers did make slightly fewer decision errors on the posttest than on the pretest (see Fig. 4). Although this was a general effect which was also shown by the control group (i.e., no practice), this indicates that shooting decisions are sensitive to improvement. What we do suggest, however, is that e in our experiment e practice did not reduce the impact of threat on police officers' shooting decisions. Apparently, the effect of threat on shooting decisions is relatively robust and, hence, may be harder to prevent than the effect of threat on shot accuracy, which e as previous studies have indicated e does improve as a result of similar amounts and similar types of practice (Nieuwenhuys and Oudejans, 2011, 2008; see also Harle and Vickers, 2001; Oudejans and Pijpers, 2009, 2010; Vine and Wilson, 2011; Wood and Wilson, 2011). 4.3. Conclusion If the impact of threat on police officers' shooting decisions is indeed robust, then this has important consequences. That is, threat and anxiety are an inherent part of police work (Anderson et al., 2002). As such, police officers should be made aware that under specific circumstances which elicit these experiences, their performance may be affected. In this light, it seems reasonable to also look at individual differences in the ability of officers to maintain performance under high threat (e.g., Jostmann and Koole, 2007), as this might help to prevent accidents and improve police effectiveness. Also, with respect to the legal investigation of officer-involved shootings, the impact of threat and anxiety should be considered. That is, if the impact of threat on police officers' shooting decisions is persistent, in specific situations, the extent to which officers can personally be held responsible for their actions might be limited (see also Payne, 2006). All in all, the current study confirmed that high threat causes officers to experience more anxiety and make more shooting errors. Notwithstanding our representative practice intervention e which involved several reality-based practice sessions e this effect appeared to be persistent and remained equally strong from pretest to posttest. Compared with previous work that showed positive effects of similar forms of practice for police officers' shot accuracy (Nieuwenhuys and Oudejans, 2011, 2008; see also Harle and Vickers, 2001; Oudejans & Pijpers, 2009, 2010; Vine and Wilson, 2011; Wood and Wilson, 2011), these results highlight the importance of distinguishing between different skill types (e.g., aiming vs. decision making) in predicting the effects of threat and anxiety on performance (Nieuwenhuys and Oudejans, 2012; see also DeCaro et al., 2011). Given the importance of successful police performance, more research is needed to substantiate these findings and to develop more effective methods for practice. Author note This study was funded by the Police Research Program of the Netherlands (www.politieenwetenschap.nl). We thank Quirine Goedhart, Maarten Melief, Gerard Willemsen, Hans Pieren, and the Amsterdam-Amstelland Police Department for their help in conducting the experiment. With respect to two of the four different practice groups involved in the current study, pretest data (pooled over both

271

groups) have been published elsewhere (Nieuwenhuys, A., Savelsbergh, G.J.P., & Oudejans, R.R.D. [2012]. Shoot or don't shoot? Why police officers are more inclined to shoot when they are anxious. Emotion, 12, 827e833. http://dx.doi.org/10.1037/a0025699). In that publication, the psychological mechanisms underlying immediate effects of anxiety on police officers' shooting decisions were investigated. The current study involved additional participants and, using a pretesteinterventioneposttest design, investigated the trainability and persistence of these effects over time. References Anderson, S., Litzenberger, R., Plecas, D.B., 2002. Physical evidence of police officer stress. Polic. Int. J. Police Strateg. Manag. 25, 399e420. http://dx.doi.org/ 10.1108/13639510210429437. Bargh, J.A., 1999. The cognitive monster: the case against the controllability of automatic stereotype effects. In: Chaiken, S., Trope, Y. (Eds.), Dual-process Theories in Social Psychology. Guilford, New York, pp. 361e382. Bishop, S.J., 2007. Neurocognitive mechanisms of anxiety: an integrative account. Trends Cogn. Sci. 7, 307e316. http://dx.doi.org/10.1016/j.tics.2007.05.008. Causer, J., Holmes, P.S., Smith, N.C., Williams, A.M., 2011. Anxiety, movement kinematics, and visual attention in elite-level performers. Emotion 11, 595e602. http://dx.doi.org/10.1037/a0023225. Correll, J., Park, B., Judd, C.M., Wittenbrink, B., 2002. The police officer's dilemma: using race to disambiguate potentially threatening individuals. J. Personal. Soc. Psychol. 83, 1314e1329. http://dx.doi.org/10.1037/0022-3514.83.6.1314. Correll, J., Park, B., Judd, C.M., Wittenbrink, B., Sadler, M.S., Keesee, T., 2007. Across the thin blue line: police officers and racial bias in the decision to shoot. J. Personal. Soc. Psychol. 92, 1006e1023. http://dx.doi.org/10.1037/00223514.92.6.1006. Correll, J., Wittenbrink, B., Park, B., Judd, C.M., Goyle, A., 2011. Dangerous enough: moderating racial bias with contextual cues. J. Exp. Soc. Psychol. 47, 184e189. http://dx.doi.org/10.1016/j.jesp.2010.08.017. DeCaro, M.S., Thomas, R.D., Albert, N.B., Beilock, S.L., 2011. Choking under pressure: multiple routes to skill failure. J. Exp. Psychol. General 140, 390e406. http:// dx.doi.org/10.1037/a0023466. Dicks, M., Button, C., Davids, K., 2010. Examination of gaze behaviors under in situ and video simulation task constraints reveals difference in information pickup for perception and action. Atten. Percept. Psychophys. 72, 706e720. http:// dx.doi.org/10.3758/APP.72.3.706. Easterbrook, J.A., 1959. The effect of emotion on the utilization and organization of behaviour. Psychol. Rev. 66, 183e201. http://dx.doi.org/10.1037/h0047707. Eysenck, M.W., Derakshan, N., Santos, R., Calvo, M.G., 2007. Anxiety and cognitive performance: attentional control theory. Emotion 7, 336e353. http://dx.doi.org/ 10.1037/1528-3542.7.2.336. Fleming, K.K., Bandy, C.L., Kimble, M.O., 2010. Decisions to shoot in a weapon identification task: the influence of cultural stereotypes and perceived threat on false positive errors. Soc. Neurosci. 5, 201e220. http://dx.doi.org/10.1080/ 17470910903268931. Frijda, N.H., 1988. The laws of emotions. Am. Psychol. 43, 349e358. http:// dx.doi.org/10.1037/0003-066X.43.5.349. Harle, S.K., Vickers, J.N., 2001. Training quiet eye improves accuracy in the basketball free throw. Sport Psychol. 15, 289e305. Houtman, I.L.D., Bakker, F.C., 1989. The anxiety thermometer: a validation study. J. Personal. Assess. 53, 575e582. http://dx.doi.org/10.1207/s15327752jpa5303_ 14. Hulse, L.M., Memon, A., 2006. Fatal impact? The effects of emotional arousal and weapon presence on police officers' memories for a simulated crime. Leg. Crim. Psychol. 11, 313e325. http://dx.doi.org/10.1348/135532505X58062. Jostmann, N.B., Koole, S.L., 2007. On the regulation of cognitive control: action orientation moderates the impact of high demands in stroop interference tasks. J. Exp. Psychol. General 136, 593e609. http://dx.doi.org/10.1037/00963445.136.4.593. Luini, L.P., Marucci, F.S., 2013. Effect of emotions on cognitive processes: are proficient decision-makers inoculated against the influence of affective states? In: Chaudet, H., Pellegrin, L., Bonnarel, N. (Eds.), Proceedings of the 11th International Conference on Naturalistic Decision Making (NDM 2013), Marseille, France, 21e24 May 2013. France: Arpege Science Publishing, Paris, pp. 87e94. Mann, D.L., Abernethy, B., Farrow, D., 2010. Action specificity increases anticipatory performance and the expert advantage in natural interceptive tasks. Acta Psychol. 135, 17e23. http://dx.doi.org/10.1016/j.actpsy.2010.04.006. Mitchell, L., Flin, R., 2007. Shooting decisions by police firearms officers. J. Cogn. Eng. Decis. Mak. 1, 375e390. Nieuwenhuys, A., Caljouw, S.R., Leijsen, M.R., Schmeits, B.A.J., Oudejans, R.R.D., 2009. Quantifying police officers' arrest and self-defence skills: does performance decrease under pressure? Ergonomics 52, 1460e1468. http://dx.doi.org/ 10.1080/00140130903287981. ~ al-Bruland, R., Oudejans, R.R.D., 2012a. Effects of threat on Nieuwenhuys, A., Can police officers' shooting behavior: anxiety, action specificity, and affective influences on perception. Appl. Cogn. Psychol. 26, 608e615. http://dx.doi.org/ 10.1002/acp.2838.

272

A. Nieuwenhuys et al. / Applied Ergonomics 48 (2015) 263e272

Nieuwenhuys, A., Oudejans, R.R.D., 2010. Effects of anxiety on handgun shooting behavior of police officers: a pilot study. Anxiety Stress Coping 23, 225e233. http://dx.doi.org/10.1080/10615800902977494. Nieuwenhuys, A., Oudejans, R.R.D., 2011. Training with anxiety: short- and longterm effects on police officers' shooting behavior under pressure. Cogn. Process. 12, 277e288. http://dx.doi.org/10.1007/s10339-011-0396-x. Nieuwenhuys, A., Oudejans, R.R.D., 2012. Anxiety and perceptual-motor performance: toward an integrated model concepts, mechanisms, and processes. Psychol. Res. 76, 747e759. http://dx.doi.org/10.1007/s00426-011-0384-x. Nieuwenhuys, A., Savelsbergh, J.P., Oudejans, R.R.D., 2012b. Shoot or don't shoot: why police officers are more inclined to shoot when they are anxious. Emotion 12, 827e833. http://dx.doi.org/10.1037/a0025699. Oudejans, R.R.D., 2008. Reality based practice under pressure improves handgun shooting performance of police officers. Ergonomics 51, 261e273. http:// dx.doi.org/10.1080/00140130701577435. Oudejans, R.R.D., Nieuwenhuys, A., 2009. Perceiving and moving in sports and other high-pressure contexts. Prog. Brain Res. 174, 35e48. http://dx.doi.org/10.1016/ S0079-6123(09)01304-1. Oudejans, R.R.D., Pijpers, J.R., 2009. Training with anxiety has a positive effect on expert perceptual-motor performance under pressure. Q. J. Exp. Psychol. 62, 1631e1647. http://dx.doi.org/10.1080/17470210802557702. Oudejans, R.R.D., Pijpers, J.R., 2010. Training with mild anxiety may prevent choking under higher levels of anxiety. Psychol. Sport Exerc. 11, 44e50. http:// dx.doi.org/10.1016/j.psychsport.2009.05.002. Payne, B.K., 2001. Prejudice and perception: the role of automatic and controlled processes in misperceiving a weapon. J. Personal. Soc. Psychol. 81, 181e192. http://dx.doi.org/10.1037//0022-3514.81.2.181. Payne, B.K., Lambert, A.J., Jacoby, L.L., 2002. Best laid plans: effects of goals on accessibility bias and cognitive control in race-based misperceptions of weapons. J. Exp. Soc. Psychol. 38, 384e396. http://dx.doi.org/10.1016/S00221031(02)00006-9. Payne, B.K., 2006. Weapon bias: split-second decisions and unintended stereotyping. Curr. Dir. Psychol. Sci. 15, 287e291. http://dx.doi.org/10.1111/j.14678721.2006.00454.x.

Pinder, R.A., Davids, K., Renshaw, I., Araújo, D., 2011. Representative learning design and functionality of research and practice in sport. J. Sport Exerc. Psychol. 33, 146e155. Plant, E.A., Peruche, B.M., 2005. The consequences of race for police officers' responses to criminal suspects. Psychol. Sci. 16, 180e183. http://dx.doi.org/ 10.1111/j.0956-7976.2005.00800.x. Plant, E.A., Peruche, B.M., Butz, D.A., 2005. Eliminating automatic racial bias: making race non-diagnostic for responses to criminal suspects. J. Exp. Soc. Psychol. 41, 141e156. http://dx.doi.org/10.1016/j.jesp.2004.07.004. Shipley, P., Baranski, J.V., 2002. Police officer performance under stress: a pilot study on the effects of visuo-motor behavior rehearsal. Int. J. Stress Manag. 9, 71e80. http://dx.doi.org/10.1023/A:1014950602826. Van der Ploeg, H.M., Defares, P.B., Spielberger, C.D., 1980. Handleiding bij de zelfbeoordelings vragenlijst ZBV (Manual for the Dutch version of the State-Trait Anxiety Inventory). Swets & Zeitlinger, Lisse, the Netherlands. Vickers, J.N., Lewinski, W., 2012. Performing under pressure: gaze control, decision making and shooting performance of elite and rookie police officers. Hum. Mov. Sci. 31, 101e117. Vickers, J.N., Williams, A.M., 2007. Performing under pressure: the effects of physiological arousal, cognitive anxiety, and gaze control in biathlon. J. Motor Behav. 39, 381e394. Vine, S.J., Wilson, M.R., 2011. The influence of quiet eye training and pressure on attention and visuo-motor control. Acta Psychol. 136, 340e346. http:// dx.doi.org/10.1016/j.actpsy.2010.12.008. Wilson, M.R., Vine, S.J., Wood, G., 2009. The influence of anxiety on visual attentional control in basketball free throw shooting. J. Sport Exerc. Psychol. 31, 152e168. Wood, G., Wilson, M.R., 2011. Quiet-eye training for soccer penalty kicks. Cogn. Process. 12, 257e266. http://dx.doi.org/10.1007/s10339-011-0393-0. Zajonc, R.B., 1980. Feeling and thinking: preferences need no inferences. Am. Psychol. 35, 151e175. http://dx.doi.org/10.1037/0003-066X.35.2.151. Zijlstra, F.R.H., 1993. Efficiency in Work Behaviour: a Design Approach for Modern Tools. Delft University Press, Delft.