Effect of Secondary Tasks on Police Officer Cognitive Workload and Performance Under Normal and Pursuit Driving Situations

Problem Statement

Motor vehicle crashes are a leading cause of officers’ line of duty injuries and deaths (NLEOMF, 2018). These crashes have been mainly attributed to high-demand driving situations and use of in-vehicle technologies while driving (Yager et al., 2015). Previous driving simulation studies in this domain have been limited to a single secondary task and were conducted under normal driving conditions. However, observations of police daily activities revealed that officers were involved in multi-tasking situations while driving in high-demand emergency and pursuit conditions. Therefore, the objective of this study was to evaluate the effect of single and multiple secondary tasks using the MCT and radio on officers’ cognitive load and driving performance under normal and pursuit driving conditions to provide a more comprehensive representation of police daily operations and the impact on officer safety as compared to prior studies. The findings of this study might be useful for police agencies and manufacturers to modify their policies regarding the use of in-vehicle technologies while driving and improve future technologies.

Participants

Eighteen police officers (age: M = 36.29 yrs., SD = 6.52 yrs.; experience as a primary patrol officer: M = 7.92 yrs., SD = 5.67 yrs.) from different departments in Texas participated in the experiment. All participants were driving police vehicles on a regular basis (average of 14 h per shift), had high experience using in-vehicle technologies such as MCT while driving (technology experience: M = 82.34%, SD = 16.01%), and had 20/20 vision or corrected vision. Participants were provided with a unidimensional visual analog rating scale to identify their level of experience with the MCT and other police in-vehicle technologies. They were asked to give a subjective rating by marking a point on a continuous (100 mm) scale with anchors of “no experience” and “high experience.” The distance from the left anchor to the marking was measured (with millimeter accuracy) and this distance was transformed to a percentage. Participants were either a traffic or highway patrol officer. All officers except one had prior experience driving in pursuit situations and following a perpetrator vehicle. In addition, 7 out of 18 participants mentioned that they completed additional trainings beyond the police academy and regular biannual trainings specifically for pursuit driving conditions. For the officer that did not have prior experience in pursuit driving, we closely monitored his driving performance in data post-processing and removed four data points as outlier data based on Cook’s D criteria due to this reason. All other data related to this participant were within the range and were not identified as outliers. Participants were compensated $60 for their time.

Apparatus

The experiment was conducted using the STISIM fixed-based driving simulator setup (System Technology, Inc., Hawthorne, CA) located at Texas A&M University (Figure 1). A set of full-size driving controls, including accelerator, brake pedal, and steering wheel provided real-time feedback for speed control and lane maintenance. The simulator recorded driver performance data (i.e., vehicle speed and lane position), with a sampling frequency of 30 Hz. A 15-inch laptop was located to the right side of the driver’s forward-view to simulate the MCT (Figure 1). The location of the laptop and its angle was adjusted based on officers’ preferences and the recommendation from McKinnon et al. (2011) to minimize any potential physical discomfort. Pupil Labs eye-tracking glasses were used to measure participants’ cognitive workload while driving. The eye-tracking glasses included a world camera with 100 degrees field of view (FOV) and two eye cameras (focused on officers’ eyes to capture pupillometry data). The frequency of data collection was 60 Hz. Markers were placed on eight locations around the driving simulator screen and four locations on the MCT to facilitate data analysis from different areas of interests (AOIs).

OPEN IN VIEWER

Figure 1

Driving simulator setup.

Independent Variables and Experiment Design

The independent variables manipulated in this study included: (1) secondary task (no secondary task, single secondary task: plate number check task with the MCT or radio communication task, and multiple secondary tasks: MCT and radio communication task simultaneously); and (2) driving condition (normal vs. pursuit driving). The MCT and radio communication tasks were selected based on the findings of Zahabi and Kaber (2018b) as they are the most important and frequently used in-vehicle technologies for the officers. The study followed a within-subjects design. All participants experienced eight driving trials in a random order. Each trial included two data blocks; and therefore, the total number of observations were 288 (i.e., 4 × 2 × 2 × 18 = 288 ). In scenarios without any secondary task, the officers were only focused on the primary task of driving. In trials where the MCT was on, the participants were required to perform a plate number check task using the MCT. In trials where the radio was on, a real recorded police radio communication was played and the officers were asked to listen to the radio and remember important pieces of information (e.g., description of the driver or vehicle). In trials with both MCT and radio commutation tasks, the participants were asked to listen to the radio communication and use the MCT to perform the plate number check task simultaneously. The driving condition included two levels of normal and pursuit driving. The normal driving condition represented a standard patrol situation for the officers in which all the roadway regulations had to be followed (e.g., traffic lights, stop signs) and drivers were asked to follow the speed limit, which was 40 mph (17.88 m/s). In the pursuit driving condition, participants were asked to follow a lead vehicle (similar to real police pursuit situations) with a maximum speed of 65 mph (29.06 m/s). The lead vehicle was initially driving at the speed limit (40 mph). Once the siren went off, the lead vehicle increased the speed to reach 65 mph and was driving at a constant speed afterwards.

Dependent Variables

The dependent variables included driving performance (speed and lane deviation), eye-tracking and subjective measures of cognitive workload, secondary task reaction time (RT), and accuracy. Speed deviation in normal driving scenarios was defined as the absolute deviation in driver speed from the posted limit (i.e., 40 mph). In pursuit situations, speed deviation was calculated as the absolute deviation in driver speed from 65 mph. Lane deviation was defined as the absolute lateral vehicle deviation from the center of a lane. The eye-tracking measure of cognitive workload was the average percentage change in pupil size (PCPS) across the two eyes. Prior studies have found that increase in pupil size is associated with higher cognitive load (Iqbal et al., 2004). The lighting condition of the laboratory was kept constant for all participants to avoid any potential effect of light on pupil size. Baseline pupil size was captured for 2 min. before and after the completion of the experiment by asking the participant to relax while seated in the driving simulator cab. The average of these two baselines was used as the baseline pupil size for each participant. This procedure was used in prior studies using the PCPS measure (White et al., 2017). The eye-tracking measure was combined with the subjective workload measure using the driving activity load index (DALI) questionnaire (Pauzié, 2008) to provide a more comprehensive assessment of officers’ cognitive workload. The DALI questionnaire is based on a six-point Likert scale (0–5), in which zero indicates a “low” workload and five indicates a “high” level of workload. DALI is a method to measure subjective workload, specifically in a driving context. The basic concept of DALI is the same as the NASA-Task Load Index (NASA-TLX), however, some demand components (e.g., “physical demand”) have been changed to promote applicability to the driving domain. DALI has six workload dimensions including effort of attention, visual demand, auditory demand, temporal demand, interference, and situational stress. DALI was found as a reliable and valid measure of driver mental work (Zakerian et al., 2018) and has been used in prior driving simulation studies (Shupsky et al., 2020; Tretten et al., 2009). The composite workload score (or overall DALI score) was calculated based on a weighted average of ratings on six subscales (i.e., ∑ W i C i ∑ W i , where w i is the weight of each workload component captured from the pairwise comparisons, and C i is the subscale rating provided by the participant after each trial).

In trials where there was a secondary task (using MCT, or radio, or both), the participants were asked to listen to the radio communication and related questions and to verbally indicate the answer. The experimenter recorded the answers as being “correct” or “incorrect.” Secondary task RT was determined as the time from when a question was presented (event 1) until the participant provided an answer (event 2) and was measured by marking the events using keystrokes during the experiment and by reviewing the video recordings of the experiment sessions afterwards.

Driving Scenarios

The study consisted of eight simulated driving scenarios. Each scenario was approximately 3 min. and simulated an urban environment with moderate traffic density (on both sides of the roadway). The scenarios presented a six-lane roadway condition with three lanes on either side of a double-yellow line with opposite directions of travel. Each scenario included four intersections. All roadways were designed as straight roads without any horizontal or vertical curvature. The traffic lights at the intersections were all set to be green when the police vehicle was approaching them. There was no cross traffic at the intersections. To increase the realism of the scenarios, several parked vehicles were located on either side of the road. The speed limit was 40 mph on all roadways. The driving simulation environment and features (e.g., traffic density, number of parked vehicles, number of intersections, and scenario duration) was similar among both normal and pursuit driving conditions to ensure similar difficulty level. To ensure that the average luminance across driving trials were approximately equal, the simulation environment and road conditions were consistent across all driving trials. All driving simulations presented clear day conditions with no adverse weather. Furthermore, the lighting condition of the scenarios and the roadway, pavement markings, on-road signage, and sidewalk colors were consistent across all driving trials.

In the normal driving condition, the officers were asked to follow the speed limit and all roadway regulations. In addition, they were asked to stay in the middle lane at all times. In the pursuit condition, a vehicle appeared at a fixed distance in front of the driver, and the participant was asked to follow the lead vehicle at a maximum speed of 65 mph. Roadway regulations could be overridden in the pursuit driving condition similar to actual police operations. The officers were not asked to keep any specific distance from the lead vehicle and they could choose that themselves. To increase the realism of the pursuit driving condition, the lead vehicle changed its lane several times and the officers had to follow the path and make sudden lane changes. Furthermore, having a moderate traffic density level in the streets made the process of car following more challenging for the officer to avoid crashes, which is similar to actual police pursuit situations. In addition, the route included several intersections that required the officers to perform sudden road maneuvers and to check their surroundings to avoid crashes with other vehicles.

Secondary Tasks

In scenarios including the secondary tasks, at two pre-determined points (also called data blocks) along the drive, the officers were asked to perform the plate number check task and/or listen to a radio communication. In the single secondary task condition, there was only one radio or MCT task in each data block. In the multiple secondary tasks condition, MCT and radio communication tasks were presented simultaneously in each data block. In scenarios without any secondary task, the officers were only performing the primary task (i.e., driving) in both data blocks. For the plate number check task, an automated voice from the simulator provided a question regarding a vehicle (e.g., “what is the plate status?”). The questions were designed based on our prior studies and interviews with police officers (Shupsky et al., 2020; Zahabi & Kaber, 2018b). Once the participant heard the question, they searched for the information on the MCT (by pressing the arrow keys to go to different information pages and reading the information on each page). The task completed once the officer verbally provided the answer to the experimenter. The MCT interface prototype was designed from the actual MCT interface used by Texas police departments to ensure all officers were familiar with the layout (Figure 2). For the radio task, a recorded police radio communication was played at two points in each scenario (each radio communication was less than 60 sec.). Subsequently, an automated voice from the simulator provided a question regarding that radio communication (e.g., “what was the description of the suspect?”). The task completed once the officer verbally provided the answer to the experimenter.

OPEN IN VIEWER

Figure 2

The MCT screen with a sample plate number check task.

Procedure

Upon arrival at the lab, participants were asked to complete the informed consent form and the demographic questionnaire. Upon completion of the forms, participants were provided with three training scenarios to get familiar with the driving simulator (without using the MCT or radio). Each training scenario took approximately 3 min. to complete and presented simulation of an urban driving environment similar to the actual experiment trials. The traffic density level, road geometry, and scenario duration were similar to the actual scenarios. However, to avoid learning effects from the training to experimental scenarios, participants were asked to follow a different path in the training scenarios. At the end of the three training scenarios, to ensure sufficient training with the simulator, participants’ average speed and lane deviations were calculated and compared with established performance criteria (i.e., absolute lane deviation ≤1.37 ft (.42 m), and absolute speed deviation ≤1 mph (.45 m/s)). These criteria have been used and validated in our prior driving simulation studies with civilian drivers and police officers using similar simulator software and setup (Zahabi & Kaber, 2018a; Zahabi et al., 2017). If the participants did not pass the criteria in three training scenarios, they were provided with additional training scenarios until the average performance in the last three trials were within the established criteria. Only two participants required an additional training scenario beyond the initial three training scenarios. Upon successful completion of the driving simulator training, the participants were provided with instructions regarding the MCT interface and plate number check task (without driving). Subsequently, they were presented with a sample radio communication and an example question to get familiar with the secondary tasks. Once the participants were comfortable with the secondary tasks, they were provided with an electronic version of the simulator sickness questionnaire (SSQ; Kennedy et al., 1993) and were asked to identify if they had developed any symptoms of simulator sickness. Next, participants provided relative ratings of different workload contributors using the DALI. Participants were provided with a 2 min break between trials. SSQ and DALI questionnaires were administered after each trial. After the completion of the eight experimental trials, participants were debriefed and compensated for their time. The experiment took approximately 2 h to complete. The Texas A&M institutional review board reviewed and accepted the study procedure.

Hypotheses

A set of research hypotheses were formulated based on the literature review as shown in Table 1. The objectives of hypotheses 1–3 were to assess the effect of the type and number of secondary tasks on officers’ driving performance, workload, and secondary task performance. These hypotheses were formulated on the findings of Lansdown et al. (2004),Williams et al. (2013), Zahabi and Kaber (2018a), and multiple resource theory (Wickens, 2002). Although studies with civilian drivers have found negative effects of secondary tasks on driver performance (Kaber et al., 2012; Salvucci et al., 2007), findings of Zahabi and Kaber (2018a) and Williams et al. (2013) revealed that police officers were protective of their primary task performance even in situations where they were involved in secondary tasks. A possible explanation for these findings might be that officers in those studies were expert police vehicle drivers who completed professional driver training. Furthermore, officers in Zahabi and Kaber (2018a) were asked to follow a consistent lane position and maintain their speed at the posted speed limit (similar to the instruction given in this study). The findings did not indicate any significant effect of secondary tasks on driving performance, which confirms officer compliance with the driving instructions. The objectives of hypotheses 4–6 were to assess the impact of different police driving situations on officer driving performance, workload, and task performance. These hypotheses were formulated based on the results of Shupsky et al. (2020).

OPEN IN VIEWER

Table 1

Experiment Hypotheses (Hypotheses Number in Parentheses)

	Driving performance	Workload	Secondary task performance
Secondary task(s)	None=Radio=MCT=Both (H1)	Both >MCT > Radio>None (H2)	Radio >MCT > Both (H3)
Driving condition	N > P (H4)	P > N (H5)	N > P (H6)

Note. N = normal driving condition; P = pursuit driving condition.

Data Analysis Approach

Initially, a data screening process was conducted to identify any outliers due to participants not following the instructions (e.g., not driving in the middle lane, speeding) or eye-tracking calibration issues. The number of outliers were different depending on the dependent variable.

No outliers were identified for the DALI, speed deviation, and secondary task accuracy responses. Regarding the PCPS data, for the right eye, 40 data points were identified as outliers due to eye-tracking calibration issues or quality of data (the data confidence level was lower than the 90% threshold). For the left eye PCPS data, 56 data points were identified as outliers due to similar eye-tracking calibration or data quality issues. For the secondary task reaction time response, 27 data points were identified as outliers. These outliers were due to some participants providing the answers after the scenario had been completed or video recording issues. For the lane deviation response, 23 outliers were identified due to participants changing lane despite the provided instructions to stay in the middle lane at all times. All the identified outliers were removed from the dataset.

Subsequently, diagnostics were conducted on all dependent measures to ensure parametric assumptions of variance homoscedasticity and residual normality are met using Bartlett’s tests and inspection of normal probability plots and the Shapiro-Wilk normality test respectively. Parametric assumption were violated for speed deviation, lane deviation, PCPS, and secondary task RT responses and therefore Box-Cox transformation were applied to the responses. For lane deviation and secondary task RT, the Box-Cox transformation could not resolve the assumption violations and therefore ranked observations were submitted to parametric tests. In those cases, since the nonparametric results were similar to ANOVA results on untransformed measures, analyses on the untransformed responses were considered valid and reported instead (Montgomery, 2017). Logistic regression was used for secondary task accuracy due to the binary response. The Geisser-Greenhouse epsilon correction was applied to correct violations of the sphericity assumption for variables with more than three levels and the corrected degrees of freedom were reported instead. Tukey’s Honest Significant Difference (HSD) post-hoc procedure was used to identify differences among levels of any significant effects. The significance groupings are shown by letters A, B, and C on the bottom of bar chart. A significance level of p ≤ .05 was set as a criterion for the study. All error bars in the figures represent standard errors.

Driving Performance

An ANOVA on log transformed speed deviation response indicated a significant effect of secondary task (F(2.70,45.88) = 6.87, p < .001, η p 2 = .40 ), driving condition (F(1,17)=13.27, p = .002, η p 2 = .44 ), and a significant two-way interaction between the secondary task and driving condition (F(2.56,43.45) = 3.41, p = .03, η p 2 = .06 ). Post-hoc analysis showed that performing any secondary task while driving (MCT, radio, or both) significantly increased officers’ speed deviation as compared to driving without any secondary task (Figure 3). In addition, on average, officers had higher speed deviation in the pursuit condition (M = 1.55 mph [.69 m/s], SD = .96 mph [.43 m/s]) as compared to the normal driving condition (M = 1.12 mph [.50 m/s], SD = .75 mph [.33 m/s]).The interaction effect revealed that although the officers had higher speed deviation in the pursuit condition as compared to the normal driving condition, the difference was higher when they were engaged in secondary tasks.

OPEN IN VIEWER

Figure 3

Effect of secondary task on speed deviation (letters A and B indicate Tukey’s HSD significance groupings).

An ANOVA on lane deviation data revealed a significant effect of driving condition (F(1,16.12) = 92.29, p < .0001, η p 2 = . 70 ) and secondary task (F(1.18,17.69) = 72.94, p < .0001, η p 2 = . 14 ). There was no significant interaction between driving condition and secondary task (F(1.77,26.48) = 2.59, p = .10, η p 2 = . 12 ). Post-hoc analysis showed that performing a single secondary task (either MCT or radio) while driving significantly increased officers’ lane deviations as compared to driving without any secondary task or with multiple secondary tasks (Figure 4). In addition, officers had significantly higher lane deviations in the pursuit driving condition (M = 2.52 ft. [.77 m], SD = 1.46 ft. [.44 m]) as compared to the normal condition (M = .63 ft. [.19 m], SD = .36 ft. [.11 m]).

OPEN IN VIEWER

Figure 4

Effect of secondary task on lane deviation (letters A and B indicate Tukey’s HSD significance groupings).

Cognitive Workload

Driver activity load index (DALI)

An ANOVA on overall DALI score revealed a significant effect of secondary task (F(1.98,33.71) = 30.86, p < .0001, η p 2 = . 44 ) and driving condition (F(1,17)=45.06, p < .0001, η p 2 = . 45 ). There was no significant two-way interaction between the secondary task and driving condition (F(2.74,46.63) = .69, p = .57, η p 2 = . 01 ). Post-hoc analysis showed that interacting with both MCT and radio while driving significantly increased cognitive workload as compared to using the MCT alone or without any secondary task (Figure 5). In addition, officers perceived higher cognitive load when they were driving in the pursuit situation (M = 2.98, SD = 1.10) as compared to the normal driving condition (M = 2.03, SD = 1.09).

OPEN IN VIEWER

Figure 5

Effect of secondary task on overall DALI score (Letters A, B, and C indicate Tukey’s HSD significance groupings).

Percentage change in pupil size (PCPS)

An ANOVA on log transformed PCPS data revealed a significant effect of driving condition (F(1,11.58) = 5.80, p = .03, η p 2 = . 35 ) and secondary task (F(1.58,14.20) = 5.16, p = .03, η p 2 = . 21 ). There was no interaction between the secondary task and driving condition (F(1.64,14.76) = 2.32, p = .14, η p 2 = . 03 ). The average PCPS increased (i.e., higher cognitive load) in pursuit driving condition (M = 10.99, SD = 6.99) as compared to the normal driving condition (M = 9.22, SD = 6.27). In addition, officers had higher cognitive load (i.e., higher average PCPS values) when they were performing multiple secondary tasks as compared to driving without any secondary tasks (Figure 6).

OPEN IN VIEWER

Figure 6

Effect of secondary task on average PCPS (letters A and B indicate Tukey’s HSD significance groupings).

Secondary Task Performance

An ANOVA on secondary task RT indicated a significant effect of secondary task (F(1.51,21.09) = 8.99, p < .001, η p 2 = . 32 ). However, there was no effect of driving condition (F(1,19.09) = .65, p = .43, η p 2 = .00 ) or two-way interaction between the secondary task and driving condition (F(1.59,22.32) = .05, p = .95, η p 2 = . 00 ) on the RT. Post-hoc analysis revealed that performing multiple secondary tasks or the MCT task alone significantly increased officers’ RT as compared to the radio communication task (Figure 7).

OPEN IN VIEWER

Figure 7

Effect of secondary task on reaction time (letters A and B indicate Tukey’s HSD significance groupings).

The logistic regression model for secondary task accuracy revealed a significant effect of secondary task ( χ ( 2 ) 2 = 25.45 , p < . 0001 ) and a significant two-way interaction between the driving condition and secondary task ( χ ( 2 ) 2 = 8.66 , p = . 013 ). However, there was no significant effect of driving condition on the response ( χ ( 1 ) 2 = . 43 , p = . 51 ). Pairwise Chi-Squared tests revealed that accuracy significantly increased when the officers were only interacting with the MCT as compared to the other conditions. In addition, the interaction effect revealed that the radio task accuracy was significantly reduced when the officers were driving in the pursuit situation as compared to the normal driving condition (Figure 8).

OPEN IN VIEWER

Figure 8

Interaction effect of secondary task and driving condition on accuracy (letters A and B indicate Tukey’s HSD significance groupings).

Effect of Secondary Tasks

Hypothesis 1 (H1) posited that there would be no difference in officers’ driving performance when they interact with the radio and/or MCT as compared to the baseline condition (i.e., driving without any secondary task). This hypothesis was not supported by the data. The officers had higher speed deviation when they were performing single or multiple secondary tasks while driving as compared to the baseline condition. In addition, the lane deviation significantly increased when the officers were performing the secondary task with the radio or MCT. The results are not aligned with the findings of Zahabi and Kaber (2018a) and Williams et al. (2013), who did not find any degradation in officers’ driving performance when they were interacting with the MCT while driving. However, it is important to note that these studies were conducted in normal driving situations and the MCT was the only secondary task for police officers. Our study simulated a more demanding situation by representing the pursuit driving condition and use of radio in addition to the MCT. Our findings are coincident with Louw et al. (2013), who found driving performance decreased with concurrent secondary tasks irrespective of the modality as compared to the baseline, and with Lansdown et al. (2004), who found a significant reduction in civilian drivers’ driving performance in single and multiple secondary task conditions. It is also important to point out that although the differences in speed and lane deviation among different secondary task conditions were statistically significant, they might not be practically significant (i.e., mean difference in speed deviation between the conditions < 1 mph and mean difference in lane deviation between the conditions < 1 ft.) per criteria identified by Green et al. (2003) (mean speed deviation: 2.43 mph, mean lane deviation: 1.02 ft. for driving and performing secondary tasks). Therefore, from a practical implications perspective, the findings of this study broadly agree with previous investigations that revealed that officers are protective of their driving performance while they are engaged in single or multiple secondary activities. However, it is important to note that in multiple secondary task condition (i.e., both radio and MCT), officers had statistically better lane deviation as compared to the single secondary task conditions. This might have been due to officers, prioritizing the primary task of driving over the secondary tasks, which is supported by the findings of secondary task accuracy and cognitive workload discussed below. This observation is similar to the findings of De Waard (1996) who compared different measures of driver’s mental workload. Based on De Waard (1996), lane deviation was the most sensitive measure of driver performance as a result of additional task load. De Waard (1996) found that driver information overload improved lane-keeping performance due to drivers’ increased effort on the primary task during high workload situations.

Hypothesis 2 (H2) posited that cognitive workload would increase as the demands of the secondary tasks increases (i.e., Both >MCT > Radio>None). This hypothesis was partially supported. The findings of both DALI questionnaire and PCPS indicated that workload increased as officers were performing secondary tasks as compared to the baseline condition. The findings are similar to Lansdown et al. (2004) who found drivers perceived higher workload when they were performing multiple or single secondary tasks as compared to the baseline. Our findings also further extend the findings of prior studies in the police domain (Zahabi & Kaber, 2018a) in that although performing either single or multiple secondary tasks using radio and MCT significantly increased workload as compared to the baseline (i.e., no secondary task), there was no significant difference between secondary task types/modalities.

Hypothesis 3 (H3) posited that officers should have better secondary task performance (i.e., faster RT and higher accuracy) with the radio as compared to the MCT and the multiple in-vehicle task condition. This hypothesis was not supported by the data. Although officers were faster in responding to the radio, they were less accurate in their responses as compared to the MCT task. The finding might have been due to the assumption of higher urgency for the radio task. During the experiment, it was observed that a majority of participants gave higher priority to the radio task as compared to the MCT (i.e., they first answered the radio question and then provided the answer for the MCT task). However, they were not provided with any instruction regarding the order or priority of secondary tasks and were asked to perform these tasks as they would normally do in their vehicles. This observation was also aligned with Lansdown et al. (2004) who found that drivers gave higher priority (faster responses) to urgent tasks, but were less accurate in performing them. Another reason might be due to the difference in “knowledge in-the-world” and “knowledge in-the-head” and limitations of working memory (Norman, 1989). The radio task required the officers to remember several pieces of information at the same time, but in the MCT task, the information was presented in consistent locations on the screen and the officers had to review the information and provide their answer. Therefore, they were more accurate in performing the MCT task as compared to the radio task.

Effect of Police Driving Situations

We expected the pursuit driving condition to degrade officers’ driving performance (H4), increase cognitive workload (H5), and decrease secondary task performance (H6) as compared to the normal driving condition. While H4 and H5 were supported by the data, H6 was not. Regarding driver performance, the trends of speed deviation and lane deviation in the pursuit condition were higher as compared to the normal driving condition. The findings are in line with Shupsky et al. (2020) who found officers had worse driving performance in tactical driving conditions as compared to operational driving conditions. Pursuit driving involves both operational (e.g., lane keeping, following the lead car) and tactical driving (e.g., passing maneuvers or overtaking) which is more demanding than the normal condition that only requires operational driving behavior. In addition, the pursuit driving condition required officers to drive at high speed (i.e., 65 mph), change lanes, and perform sudden maneuvers as compared to the normal condition. The results call for more training of police officers in handling pursuit situations. The findings also suggested that officers experienced significantly higher cognitive load under the pursuit driving situation as compared to the normal condition. This is in line with the results of Kaber et al. (2012) that workload increases under tactical driving as compared to the operational driving level. In police pursuits, officers are engaged in a prolonged hazard (e.g., following the victim’s vehicle) which requires driving in high speed, close following behavior, sudden road maneuvers, and complex decision making situations, which all can increase driver workload (Crundall et al., 2003).

There was no significant effect of driving condition on secondary task RT or accuracy except when the officers were performing the radio task. Officers were more accurate in their responses to the radio task under normal driving condition as compared to the pursuit situation. Lower accuracy in the radio task under pursuit driving might have been due to the sounds of the siren and higher noise level overall, although the officers were allowed to adjust the radio sound level based on their preferences. Therefore, the findings of this study suggest dispatchers not rely only on the radio for communicating critical information to the officers in pursuit driving situations. The other explanation might be that officers are under higher workload in pursuit driving and therefore have less residual attentional resources to perform the radio task, which can lead to reduced accuracy.

Limitations and Future Work

One of the limitations of this study was the use of a fixed-based, single-screen driving simulator. Although the simulator provided real-time feedback on speed and lane control, it did not convey motion cues to the officers during the simulation. In addition, the use of a single-screen simulator might limit the generalizability of study findings due to field of view limitations and fidelity. Future studies should validate the findings of this study using high-fidelity simulators and a wider field of view. Second, due to the number of study manipulations and to reduce the total duration of the experiment and avoid potential simulator sickness issues, the duration of each scenario was limited to 3 min. However, police officers spend most of their shift in the vehicle, experience high-demand situations, and might be engaged in secondary tasks for longer durations. Finally, although the radio recordings were selected from real police broadcasts available online from different states (except Texas to ensure all participants had never heard that broadcast previously), some of the participants mentioned that the voice/accent of some of the dispatchers in the audio clips were not clear to them. This issue might have negatively affected their comprehension of the message. Future studies should investigate the impact of in-vehicle technology interactions on officers’ safety using naturalistic studies of longer durations.