Factors That Affect Drivers’ Perception of Closing and an Immediate Hazard

Abstract

Objective

To measure the looming threshold for when drivers perceive closing and an immediate hazard and determine what factors affect these thresholds.

Background

Rear-end collisions are a common type of crash. One key issue is determining when drivers first perceive they need to react. The looming threshold for closing and an immediate hazard are critical perceptual thresholds that reflect when drivers perceive they need to react.

Method

Two driving simulator experiments examined whether engaging in a cell phone conversation and whether the complexity of the roadway environment affect these thresholds for the perception of closing and immediate hazard. Half of the participants engaged in a cognitive task, the last letter task, to emulate a cell phone conversation, and all participants experienced both simple and complex roadway environments.

Results

Drivers perceived an immediate hazard later when engaged in a cell phone conversation than when not engaged in a conversation but only when the driving task was relatively less demanding (e.g., simple roadway, slow closing velocity). Compared to simple scenes, drivers perceived closing and an immediate hazard later for complex scenes but only when closing velocity was 30 mph (48.28 km/h) or greater.

Conclusion

Cell phone conversation can affect when drivers perceive an immediate hazard when the roadway is less demanding. Roadway complexity can affect when drivers perceive closing and an immediate hazard when closing velocity is high.

Application

Results can aid accident analysis cases and the design of driving automation systems by suggesting when a typical driver would respond.

Keywords

driver behavior perception-action driver distraction accident analysis threshold

Introduction

In 2018, there were approximately 2,175,000 rear-end collisions, accounting for 32% of all crashes in the United States (National Highway Traffic Safety Administration [NHTSA], 2020). Of these crashes, 594,000 resulted in injury, and 2439 were fatal (NHTSA, 2020). Among these fatal crashes, 457 involved a distracted driver (NHTSA, 2020). Prior research found more than 70% of rear-end collisions involved a stopped or slow-moving lead vehicle (Knipling et al., 1993), and when closing velocity was greater than 15 mph (24.14 km/h), the danger of the crash in terms of loss of life or property increased (Young et al., 1995). There is an unequivocal need to examine the driver’s response to a lead vehicle and what factors affect that response, especially for cases in which the lead vehicle is slow-moving or stopped.

Muttart et al. (2005) identified five phases that occur during a driver’s response to a lead vehicle: (1) detection of the presence of the lead vehicle, (2) detection of closing between the driver’s vehicle and the lead vehicle, (3) detection of an immediate hazard, (4) cognitive response, and (5) motor response. Determining when the driver first perceives they need to react is a crucial aspect of the driver’s response to a lead vehicle. Hoffmann and Mortimer (1996) suggested that drivers may react to a lead vehicle once they perceive they are closing on it, which occurs when the lead vehicle’s optical expansion rate reaches .003 radians per second (rad/s). This optical expansion rate is called the looming threshold for closing. However, Muttart et al. (2005) suggested that most drivers will not initiate a response to a lead vehicle before they perceive an immediate hazard, which occurs when the lead vehicle’s optical expansion rate reaches .006 rad/s. This optical expansion rate is called the looming threshold for an immediate hazard. However, our prior research (Weaver et al., 2019 b) indicated that the time of the driver’s first response occurred when the optical expansion rate was .00035 rad/s, an optical expansion rate about 10 times smaller than either of the two previous thresholds.

One reason for the discrepancy observed in optical expansion rates may be attributable to a difference in stimuli: Muttart’s et al. (2005) roadway scenes were more complex than our simpler scenes (Weaver et al., 2019b). The complexity of the roadway environment may affect the looming threshold for closing and an immediate hazard because drivers glance away from the forward roadway as the driver approaches a hazard in a complex environment but not in a simple environment, resulting in smaller optical expansion rates for simpler driving environments (J. W. Muttart, personal communication, August 10, 2018). Another reason for the discrepancy in observed optical expansion rates may be due to differences in methods. For instance, Hoffmann and Mortimer (1996) instructed participants to determine which of two film clips showed a higher closing velocity to a lead vehicle, and Muttart et al. (2005) instructed participants to press a button when they first perceived an immediate hazard. However, we provided no such instructions and instead recorded when drivers first initiated a driving response (Weaver et al., 2019b). Given Hoffmann and Mortimer’s assumption that drivers react to a lead vehicle once they perceive they are closing, our method of recording when drivers first initiate a driving response should theoretically produce the same looming threshold for closing after perception-response time is taken into consideration. Similarly, given Muttart et al.’s (2005) assumption that most drivers will not initiate a driving response until they perceive an immediate hazard, our method should produce the same looming threshold for an immediate hazard after perception-response time is taken into consideration. However, given that we did not obtain either of these looming thresholds, we suggested that a driver’s response to a lead vehicle goes through phases as they assess the situation on the roadway (Weaver et al., 2019b). Thus, our criterion for first driving response—when drivers began to release the accelerator—may have reflected detection of the presence of the lead vehicle, and stronger driving responses, such as releasing the accelerator or braking, may have been associated with the looming threshold for closing and an immediate hazard. Hoffmann and Mortimer (1996) suggested that drivers need to quickly respond once they perceive closing, and Muttart et al. (2005) suggested that drivers need to initiate an emergency avoidance response once they perceive an immediate hazard. This means that both the looming threshold for closing and an immediate hazard are important to the driver’s response to a lead vehicle because they reflect when drivers perceive stronger driving responses are required, and therefore, factors that affect these thresholds are of interest to understand a driver’s response to a lead vehicle in various situations.

One factor that may affect these thresholds is engagement in a hands-free cell phone conversation. A recent meta-analysis by Caird et al. (2018) analyzed 93 studies on this issue and found that engaging in a hands-free cell phone conversation while driving led to moderate performance decrements on brake response time to emergency roadway events, accuracy and response time for detecting targets or secondary probes (e.g., sign, light emitting diode), and rate of collisions. These performance decrements putatively occur because the driver’s attention is diverted away from the driving task toward the cell phone conversation (Strayer & Johnston, 2001; Strayer et al., 2003). Interestingly, Caird et al. (2018) found engaging in a cell phone conversation did not affect lateral position, though this finding can be accounted for by hierarchical control theory (Cooper et al., 2013). Conversely, naturalistic driving research showed no significant adverse effects of cell phone conversation on driving performance, and this was attributed to the fact that drivers did not have to remove their eyes from the forward roadway (Olson et al., 2009). Prior research has not addressed whether engaging in a cell phone conversation affects the looming threshold of closing and an immediate hazard. This is important for understanding whether such a distracting task affects these perceptual thresholds. Our prior research (Weaver et al., 2019b) tried to examine how a cell phone conversation affects the looming threshold for closing and an immediate hazard, but as discussed previously, it unsatisfactorily addressed how a cell phone conversation affects these thresholds because the expansion rates observed likely reflect detection of the presence of the lead vehicle.

In the current work, our goal was to reconcile the discrepant findings between our prior work (Weaver et al., 2019b) and prior work on drivers’ perception of closing (Hoffmann & Mortimer, 1996) and an immediate hazard (Muttart et al., 2005). Additionally, we aimed to build upon the prior research by examining the effect of engaging in a cell phone conversation on when drivers perceive closing and an immediate hazard. To assess the effect of hands-free cell phone conversations on the looming threshold for closing and an immediate hazard, we conducted a pair of experiments with instructions that more closely resemble Muttart et al.’s (2005) instructions. Experiment 1 focuses on the looming threshold of closing, and Experiment 2 focuses on the looming threshold for an immediate hazard. Furthermore, we manipulated scene complexity in these experiments to examine the effect this factor has on the thresholds. We hypothesized that drivers engaged in a cell phone conversation would perceive closing and an immediate hazard later than drivers not engaged in a cell phone conversation. In the current studies, we tested our hypotheses using optical expansion rate because drivers putatively base their perceptual judgments of closing and an immediate hazard upon optical expansion rate (Hoffmann & Mortimer, 1996; Muttart et al., 2005). Additionally, optical expansion rate of an approaching object increases monotonically with time provided closing velocity remains constant. This means that a larger optical expansion rate reflects a later response (when the driver is closer to the hazard). Accordingly, drivers engaged in a cell phone conversation should exhibit a larger mean optical expansion rate. For roadway complexity, we hypothesized that drivers would perceive closing and an immediate hazard later in a complex roadway environment compared to a simple roadway environment because drivers glance away from the forward roadway as the driver approaches a hazard in a complex environment but not in a simple environment. Thus, complex scenes should result in a larger mean optical expansion rate.

Experiment 1

Method

Participants

Forty undergraduate students at Rice University (25 female) participated for partial course credit. Age ranged from 18 to 24 years (M = 19.34, SD = 1.37). All participants held a driver’s license for between 1 and 6 years (M = 3.00, SD = 1.32). All participants reported normal or corrected visual acuity. This research complied with the American Psychological Association Code of Ethics and was approved by the Institutional Review Board at Rice University. Informed consent was obtained from each participant.

Apparatus and displays

The experiment was conducted in a fixed-base STISIM Drive simulator equipped with three 20.75-in. (52.71 cm) screens (measured horizontally) that provided a 135° field-of-view. To achieve a 135° field-of-view, the outer screens were angled 45° toward the participant, partially surrounding the participant. The participants were positioned so that they had a viewing distance of approximately 25 in. to each screen. Each screen had a 1920 × 1080 pixel resolution and 60 Hz refresh rate. STISIM Drive has been described as a medium-fidelity driving simulator that provides a realistic roadway environment (Braly et al., 2018; Freund et al., 2005; Levulis et al., 2015).

By crossing the two within-subjects factors (two scene complexities and four closing velocities) and having three replicates for each unique scene, we created and randomly ordered 24 traffic scenes that consisted of a flat, one-way road that was 23 ft (7.01 m) wide, consistent with scenes used by Muttart et al. (2005). The participant’s vehicle was completely automated and moved at a constant velocity of 60 mph (96.56 km/h). In all scenes, a slow-moving or stopped lead vehicle first came into sight 2500 ft (762 m) ahead of the participant. To prevent participants from timing their responses on the basis of the lead vehicle’s motion, the lead vehicle traveled at four speeds in separate trials: 0, 15, 30, and 45 mph (0, 24.14, 48.28, and 72.42 km/h). This resulted in four closing velocities of 60, 45, 30, and 15 mph (96.56, 72.42, 48.28, and 24.14 km/h) respectively, that are similar to prior research (Hoffmann & Mortimer, 1996; Muttart et al., 2005). The lead vehicle appeared at a randomly selected time between 10 s and 35 s after the scene started. Each scene ended when the participant’s vehicle reached the lead vehicle.

Cell phone conversation task

We used the last letter task to emulate a hands-free cell phone conversation, consistent with prior research (Gugerty et al., 2004; Laberge, 2003; Strayer & Johnston, 2001). A target word was read to the participant who then generated a word that started with the last letter of the target word. For example, a correct response to the word train is a word that starts with the letter n such as net. To ensure consistent timing, target words were prerecorded and presented at a rate of one word every 3 s, a rate consistent with prior research (Gugerty et al., 2004). Participants were permitted to repeat words because pilot testing showed that accuracy degraded significantly when no repetition was allowed.

Although the last letter task lacks some of the ecological validity of a naturalistic conversation, it shares several key characteristics with an actual conversation. In particular, the last letter task requires participants to listen to speech and produce an appropriate verbal response (Gugerty et al., 2004). We chose to use a cognitive task to emulate a cell phone conversation instead of a naturalistic conversation because the words in the cognitive task can be consistently repeated across participants and scored for performance, whereas naturalistic conversations have more variability and are harder to evaluate (Laberge, 2003). Employing such a cognitive task produces a conservative estimate for the effects of cell phone conversations on driving performance compared to naturalistic conversations (Caird et al., 2018).

Scene complexity

To reconcile the discrepant findings between Muttart et al. (2005) and our prior work (Weaver et al., 2019b), pedestrians were used to manipulate scene complexity in a manner consistent with Muttart et al. (2005). In simple scenes, there were no pedestrians (Figure 1). In complex scenes, there were rows of pedestrians (i.e., lines of pedestrians perpendicular to the roadway) similar to the scenes used by Muttart et al. (2005). As shown in Figure 2, there were six pedestrians in each row (three pedestrians on each side of the road). In the current study, the participant’s vehicle traveled at constant velocity of 60 mph (96.56 km/h) and passed a row of pedestrians every 3 s. Pedestrians started to move only when they were in the next row the participant’s vehicle would pass. When pedestrians started to move, some randomly selected pedestrians walked into the roadway while other pedestrians walked alongside or away from the roadway; for those that did walk into the roadway, some did not safely reach the other side (i.e., they presented a collision hazard). This random selection resulted in a pedestrian walking into the road on a collision course with the participant’s vehicle every 18 s, on average. When the three pedestrians on each side of the road started to move, they were 2.5, 5.0, and 7.5 degrees of visual angle to the left and right of the driver’s line of sight, respectively, assuming a forward gaze towards the focus of expansion. Our scenes differed from Muttart et al.’s in that we did not cap the maximum number of rows of pedestrians shown at one time to four, as Muttart et al. did, because we wanted the rows of pedestrians to extend to the horizon. Similarly, pedestrians moved at variable speeds in our study (randomly selected speed from 3 to 12 ft/s [0.91 to 3.66 m/s]), whereas they moved at a constant speed in Muttart et al.’s study. We chose variable speeds so that the pedestrians in each position could present a collision hazard.

Figure 1

Snapshot of a simple driving scene with a lead vehicle 1000 ft ahead.

Figure 2

Snapshot of a complex driving scene with a lead vehicle 1000 ft ahead.

Procedure and design

Participants were randomly assigned to either engage in a cell phone conversation or not. All participants experienced both simple and complex scenes. To disentangle the perception of the hazard from the decision of what driving action to take, participants were instructed to press a button as soon as they perceived they were getting closer to either the lead vehicle or a pedestrian. They did not take any action to avoid the hazard (e.g., brake or steer). In other words, they were instructed to press a button when they first perceived closing; we did not instruct participants to identify the actual closing speed. A different button was used to report closing for a pedestrian versus a vehicle; button assignment was counterbalanced. Participants completed a short practice drive, completed the 24 driving scenes, and then completed a post-study questionnaire. The time at which participants pressed the button was recorded and used to derive the lead vehicle’s optical expansion rate, which is the dependent variable.

Results

Button press task

There is a delay between the moment drivers perceive an event (e.g., closing, immediate hazard) and the moment the driver presses the button; this is known as the perception-response time (PRT; see Muttart et al., 2005). Muttart et al. (2005) suggested a PRT of 1.6 s based on the average response time from their experiment. Muttart et al.’s estimate is most suitable for the current study because of the similarities between studies. Both studies instruct participants to press a button when they perceive a particular perceptual event in a driving scenario with a lead vehicle. Additionally, our aim was to reconcile discrepant findings between our prior work (Weaver et al., 2019b) and Muttart et al.’s (2005), and we designed our complex scenes with the aim of reproducing the cognitive demands imposed by the presence of pedestrians entering the roadway in Muttart et al.’s (2005) study. For these reasons, we adopted a 1.6 s PRT. We worked backward from the time of the button press by subtracting the PRT from the time of the button press before calculating the optical expansion rate (Muttart et al., 2005, Equation 1).

Results were analyzed with a mixed analysis of variance (ANOVA). For all analyses in this paper, the significance level was set to .05, and when Mauchly’s test was significant, indicating a violation of the sphericity assumption, a Greenhouse–Geisser correction was applied to the degrees of freedom. Effect size estimates are given using generalized eta squared (η_G^2; Bakeman, 2005; Olejnik & Algina, 2003) for all analyses, for which .010 is a small effect size, .059 is a medium effect size, and .138 is a large effect size (Cohen, 1988). Initial analyses showed that the button assignment was not significant; thus, results were averaged across this factor.

A 2 (cell phone conversation: not engaged, engaged) × 2 (scene complexity: simple, complex) × 4 (closing velocity: 15, 30, 45, 60 mph [24.14, 48.28, 72.42, and 96.56 km/h]) mixed ANOVA showed that no effect involving cell phone conversation was significant: cell phone conversation, F(1, 38) = 1.72, p = .197, η_G² = .036; cell phone conversation × scene complexity, F(1, 38) = .54, p = .468, η_G² = .001; cell phone conversation × closing velocity, F(2.47, 93.98) =.68, p = .534, η_G² = .001; cell phone conversation × scene complexity × closing velocity, F(2.26, 86.06) = 1.83, p = .145, η_G² = .003. However, the main effect of scene complexity was significant, F(1, 38) = 14.83, p < .001, η_G² = .021, such that the mean optical expansion rate was larger when scenes were complex (M = .0027 rad/s, SD = .0030) compared to when they were simple (M = .0017 rad/s, SD = .0034). The main effect of closing velocity was also significant, F(2.47, 93.98) = 4.90, p = .006, η_G² = .008. Tukey’s honestly significant difference (HSD) tests indicated that the mean optical expansion rate when the closing velocity was 45 mph (72.42 km/h; M = .0026 rad/s, SD = .0034) was significantly larger than when the closing velocity was 15 mph (24.14 km/h; M = .0020 rad/s, SD = .0033) or 30 mph (48.28 km/h; M = .0019 rad/s, SD = .0033); no other pairwise comparisons were significant (60 mph [96.56 km/h]: M = .0024 rad/s, SD = .0028).

As shown in Figure 3, the interaction between scene complexity and closing velocity was significant, F(2.26, 86.06) = 8.15, p < .001, η_G² = .014. When this interaction was broken down by scene complexity, closing velocity was not significant for simple scenes, F(1.29, 50.27) = 1.83, p = .181, η_G² = .003, but was significant for complex scenes, F(2.86, 111.69) = 8.76, p < .001, η_G² = .040. Breaking down closing velocity for complex scenes further, Tukey’s HSD tests indicated that the mean optical expansion rate when the closing velocity was 45 mph (72.42 km/h) was significantly larger than when the closing velocity was 15 or 30 mph (24.14 or 48.28 km/h). Additionally, Tukey’s HSD indicated that the mean optical expansion rate when the closing velocity was 60 mph (96.56 km/h) was significantly larger than when the closing velocity was 15 mph (24.14 km/h). No other pairwise comparisons were significant. When the interaction between scene complexity and closing velocity was broken down by closing velocity, the mean optical expansion rate for complex scenes were significantly greater than simple scenes only when closing velocity was 30 mph (48.28 km/h) or greater: 15 mph (24.14 km/h), F(1, 39) = .11, p = .745, η_G² = .001; 30 mph (48.28 km/h), F(1, 39) = 13.36, p = .001, η_G² = .011; 45 mph (72.42 km/h), F(1, 39) = 31.48, p < .001, η_G² = .072; 60 mph (96.56 km/h), F(1, 39) = 12.80, p = .001, η_G² = .052. In short, faster closing velocities had higher mean optical expansion rates compared with slower closing velocities but only when scenes were complex.

Figure 3

Optical expansion rate as a function of closing velocity for each scene complexity. Note. Error bars represent ± standard error of the mean. Table of means and standard errors is available online in the Supplemental Materials.

Cell phone conversation task

The primary analysis on the button press showed engaging in a cell phone conversation did not affect the looming threshold for closing. This could mean engaging in a cell phone conversation did not interfere with when drivers perceived closing or that drivers were sacrificing performance on the cell phone conversation task to maintain performance on the button press task. This was evaluated in the current analysis.

The Shapiro–Wilk test for mean response accuracy was significant, W = .77, p < .001, indicating that mean response accuracy was not normally distributed and violated the assumption of normality of an ANOVA. However, “much of the research on the normality assumption has been consistent in noting the relative insensitivity of the F test to departures from normality” (Lix et al., 1996, p. 582). In other words, prior research has indicated departures from normality have very little effect on the level of significance or power of an ANOVA (Cochran, 1947; Glass et al., 1972). Thus, results were analyzed with a repeated measures ANOVA.

A 2 (scene complexity: simple, complex) × 4 (closing velocity: 15, 30, 45, 60 mph [24.14, 48.28, 72.42, and 96.56 km/h]) repeated measures ANOVA showed the main effect of complexity, F(1, 19) = 2.29, p = .146, η_G² = .017, the main effect of closing velocity, F(2.18, 41.39) = 1.13, p = .337, η_G² = .014, and the interaction between scene complexity and closing velocity, F(2.07, 39.28) = 1.45, p = .247, η_G² = .010, were all not significant. Overall, the mean response accuracy was 97.3% (SD = 3.9). Thus, cell phone conversation did not interfere with when drivers perceived closing.

Discussion

Our hypothesis that drivers who are engaged in a cell phone conversation would perceive closing later than those not engaged in a cell phone conversation was not supported because all of the effects involving cell phone were not significant. The cell phone conversation task analysis indicated this was not because drivers were sacrificing performance on the cell phone conversation to maintain driving performance. Our hypothesis that drivers would perceive closing later in a complex roadway environment compared to a simple roadway environment was partially supported. A significant interaction between scene complexity and closing velocity indicated that the mean optical expansion rate was significantly larger for complex scenes compared to simple scenes but only when the closing velocity was 30 mph (48.28 km/h) or greater, as shown in Figure 3. Drivers may have perceived closing later when scenes were complex and closing velocity was 30 mph (48.28 km/h) or greater because the drivers’ glances away from the forward roadway during complex scenes were for the same duration for all closing velocities, and therefore the glances away from the forward roadway were more detrimental when closing velocity was higher because the lead vehicle loomed more. Another factor that may have contributed to this effect is how far away the lead vehicle’s projected image on the retina is from the fovea, which prior research found affected brake reaction time (Summala et al., 1998).

Experiment 2

Method

Everything was identical to Experiment 1 except for the instructions. Participants were instructed to press a button as soon as they perceived that an emergency avoidance response was required to avoid a collision with either the lead vehicle or a pedestrian (i.e., when drivers first perceived an immediate hazard).

Participants

Forty undergraduate students (28 female) at Rice University participated for partial course credit. Age ranged from 18 to 36 years (M = 19.23, SD = 2.86). All participants held a driver’s license for between less than 1 year and 19 years (M = 2.40, SD = 2.94). All participants reported normal or corrected visual acuity. This research complied with the American Psychological Association Code of Ethics and was approved by the Institutional Review Board at Rice University. Informed consent was obtained from each participant.

Results

Button press task

Consistent with Experiment 1, a PRT of 1.6 s was subtracted from the time of the button press before calculating the optical expansion rate. Results were analyzed with a mixed ANOVA. Initial analyses showed that button assignment was not significant; results were averaged across this factor.

A 2 (cell phone conversation: not engaged, engaged) × 2 (scene complexity: simple, complex) × 4 (closing velocity: 15, 30, 45, 60 mph [24.14, 48.28, 72.42, and 96.56 km/h]) mixed ANOVA showed that the main effect of cell phone conversation was not significant, F(1, 38) = 2.96, p = .094, η_G² = .058. The main effect of scene complexity also was not significant, F(1, 38) = .44, p = .512, η_G² = .001. The main effect of closing velocity was significant, F(2.43, 92.32) = 3.00, p = .045, η_G² = .005. Tukey’s HSD tests indicated that the mean optical expansion rate when the closing velocity was 45 mph (72.42 km/h; M = .0052 rad/s, SD = .0038) was significantly larger than when the closing velocity was 60 mph (96.56 km/h; M = .0045 rad/s, SD = .0033); no other pairwise comparisons were significant (15 mph [24.14 km/h]: M = .0052 rad/s, SD = .0045; 30 mph [48.28 km/h]: M = .0050 rad/s, SD = .0044).

As shown in Figure 4, there was a significant interaction between cell phone conversation and scene complexity, F(1, 38) = 6.87, p = .013, η_G² = .009. For drivers not engaged in a cell phone conversation, mean optical expansion rate was significantly larger for complex scenes compared to simple scenes, F(1, 19) = 8.51, p = .009, η_G² = .028; however, there was no significant difference between simple and complex scenes for drivers engaged in a cell phone conversation, F(1, 19) = 1.33, p = .264, η_G² = .004. When the interaction was broken down by complexity, results showed that for simple scenes, mean optical expansion rate was significantly larger for drivers engaged in a cell phone conversation compared to those who were not, F(1, 38) = 4.42, p = .042, η_G² = .104; however this was not true for complex scenes, F(1, 38) = 1.37, p = .249, η_G² = .034. It is possible that drivers prioritized the driving task over the cell phone conversation to maintain their driving performance when the scenes were complex (Figure 4 for what appears to be a ceiling effect for drivers engaged in a cell phone conversation).

Figure 4

Optical expansion rate as a function of scene complexity for each cell phone group. Note. Error bars represent ± 1 standard error of the mean. Table of means and standard errors is available online in the Supplemental Materials

As shown in Figure 5, there was a significant interaction between cell phone conversation and closing velocity, F(2.43, 92.32) = 4.60, p = .008, η_G² = .008. For drivers not engaged in a cell phone conversation, mean optical expansion rate was not significantly different among closing velocities, F(2.14, 40.71) = 2.61, p = .082, η_G² = .013; however, for drivers engaged in a cell phone conversation, mean optical expansion rate was significantly different among closing velocities, F(1.63, 30.99) = 4.55, p = .025, η_G² = .016. Tukey’s HSD tests indicated the mean optical expansion rate for 15 mph (24.14 km/h) was significantly larger than 60 mph (96.56 km/h) for drivers engaged in a cell phone conversation, but no other pairwise comparisons of closing velocity were significant. When the interaction was broken down by closing velocity, results showed that drivers engaged in a cell phone conversation had a significantly larger mean optical expansion rate than drivers not engaged in a cell phone conversation when closing velocity was 15 mph (24.14 km/h), F(1, 38) = 5.29, p = .027, η_G² = .121; however, this was not true for any other closing velocity (30 mph [48.28 km/h]: F[1, 38] = 3.29, p = .078, η_G² = .080; 45 mph [72.42 km/h]: F[1, 38] = 1.22, p = .277, η_G² = .033; 60 mph [96.56 km/h]: F[1, 38] = 1.69, p = .202, η_G² = .041). It is possible that drivers prioritized the driving task over the cell phone conversation when closing velocity exceeded 15 mph [24.14 km/h].

Figure 5

Optical expansion rate as a function of closing velocity for each cell phone group. Note. Error bars represent ± standard error of the mean. Table of means and standard errors is available online in the Supplemental Materials.

As shown in Figure 6, there was a significant interaction between scene complexity and closing velocity, F(2.13, 80.86) = 19.21, p < .001, η_G² = .048. Mean optical expansion rate was significantly different among several of the closing velocities for both simple scenes, F(1.69, 65.94) = 10.04, p < .001, η_G² = .044, and for complex scenes, F(2.41, 94.09) = 16.31, p < .001, η_G² = .056. Tukey’s HSD tests indicated that when scenes were simple, the mean optical expansion rate was significantly larger for the closing velocity 15 mph (24.14 km/h) compared to all the other closing velocities, but when scenes were complex, the mean optical expansion rate was significantly smaller for the closing velocity 15 mph (24.14 km/h) compared to all the other closing velocities. Additionally, the mean optical expansion rate when closing velocity was 45 mph (72.42 km/h) was significantly larger than when the closing velocity was 60 mph (96.56 km/h) for complex scenes; no other pairwise comparisons were significantly different. When the interaction was broken down by closing velocity, results showed that the mean optical expansion rate was significantly different between simple and complex scenes for all closing velocities: 15 mph ([24.14 km/h), F(1, 39) = 11.72, p = .002, η_G² = .076; 30 mph (48.28 km/h), F(1, 39) = 8.58, p = .006, η_G² = .018; 45 mph (72.42 km/h), F(1, 39) = 22.68, p < .001, η_G² = .049; and 60 mph (96.56 km/h), F(1, 39) = 5.18, p = .028, η_G² = .012. In particular, the mean optical expansion rate was significantly larger for complex scenes compared to simple scenes for all closing velocities except for 15 mph (24.14 km/h), which was significantly smaller for complex scenes compared to simple scenes. In brief, when closing velocity was 15 mph (24.14 km/h), the effect differed compared to the higher closing velocities such that complex scenes did not have a larger mean optical expansion rate than simple scenes, and instead complex scenes had a smaller mean optical expansion rate than simple scenes when closing velocity was 15 mph (24.14 km/h).

Figure 6

Optical expansion rate as a function of closing velocity for each scene complexity. Note. Error bars represent ± 1 standard error of the mean. Table of means and standard errors is available online in the Supplemental Materials.

The interaction among cell phone conversation, scene complexity, and closing velocity was not significant, F(2.13, 80.86) = 2.92, p = .057, η_G² = .008.

Cell phone conversation task

The primary analysis on the button press task suggested that drivers may be sacrificing performance on the cell phone conversation to maintain driving performance when scenes were complex or closing velocity exceeded 15 mph (24.14 km/h). We evaluated this explanation in the current analysis. Mean response accuracy was not normally distributed as indicated by the Shapiro–Wilk test , W = .83, p = .003. Consistent with Experiment 1, results were analyzed with a repeated measures ANOVA.

A 2 (scene complexity: simple, complex) × 4 (closing velocity: 15, 30, 45, 60 mph [24.14, 48.28, 72.42, and 96.56 km/h]) repeated measures ANOVA showed that the main effect of complexity was not significant, F(1, 19) = 3.36, p = .082, η_G² = .013. The main effect of closing velocity was significant, F(3, 57) = 6.66, p = .001, η_G² = .041. However, Tukey’s HSD indicated that none of the pairwise comparisons for closing velocity were significant. As shown in Figure 7, the interaction between scene complexity and closing velocity was significant, F(3, 57) = 4.58, p = .006, η_G² = .034. When this interaction was broken down by scene complexity, mean response accuracy was significantly different between closing velocities for both simple scenes, F(2.09, 39.7) = 7.54, p = .001, η_G² = .151, and complex scenes, F(3, 57) = 2.85, p = .045, η_G² = .026. Tukey’s HSD showed that none of the pairwise comparisons were significant for complex scenes, but for simple scenes, mean response accuracy was significantly higher when the closing velocity was 15 or 30 mph (24.14 or 48.28 km/h) than when closing velocity was 45 mph (72.42 km/h); no other pairwise comparisons were significant. When the interaction between scene complexity and closing velocity is broken down by closing velocity, mean response accuracy was significantly higher for simple scenes compared to complex scenes when the closing velocity was 15 mph (24.14 km/h), F(1, 19) = 10.60, p = .004, η_G² = .123, but not for any other closing velocity (30 mph [48.28 km/h]: F[1, 19] = 2.99, p = .100, η_G² = .034; 45 mph [72.42 km/h]: F[1, 19] = .77, p = .392, η_G² = .007; 60 mph [96.56 km/h]: F[1, 19] < .01, p = .985, η_G² < .001). Therefore, drivers performed best on the cell phone conversation task when there was a slow closing velocity and a simple driving environment.

Figure 7

Response accuracy as a function of closing velocity for each scene complexity. Note. Error bars represent ± 1 standard error of the mean. Table of means and standard errors is available online in the Supplemental Materials.

Discussion

Our hypothesis that drivers who are engaged in a cell phone conversation would perceive an immediate hazard later than those not engaged in a cell phone conversation was partially supported. A significant interaction between cell phone conversation and scene complexity showed that the mean optical expansion rate was significantly larger for drivers engaged in a cell phone conversation compared to those not engaged but only for simple scenes. Additionally, a significant interaction between cell phone conversation and closing velocity showed that the mean optical expansion rate was significantly larger for drivers engaged in a cell phone conversation compared to those not engaged but only for the closing velocity 15 mph (24.14 km/h). These results suggest that drivers perceived an immediate hazard later when engaged in a cell phone conversation than when not engaged but only when the driving task was relatively less demanding (e.g., simple roadway environment or low closing velocities, which putatively impose lower cognitive demands). As the driving task became more demanding, drivers may have prioritized the driving task at the expense of the cell phone conversation. This explanation was supported by the analysis of the mean response accuracy of the cell phone conversation task, which showed that drivers performed significantly worse on the cell phone conversation when the roadway was complex or closing velocity was high. However, for this explanation to fully account for the results, we would have expected drivers to perceive an immediate hazard later when the roadway is simple and closing velocity is low, instead of just one or the other. This effect actually emerges in the next analysis that compares the experiments. Altogether, these results are consistent with Fuller’s (2005) model of task difficulty regulation for driving in which the driver strives to stay within a preferred range of task difficulty. According to Fuller (2005), drivers will sacrifice low priority tasks, such as a cell phone conversation, when task demands begin to exceed capability to achieve that goal.

Our hypothesis that drivers would perceive an immediate hazard later in a complex roadway environment compared to a simple roadway environment was partially supported as well. A significant interaction between cell phone conversation and scene complexity indicated that the mean optical expansion rate was significantly larger for complex scenes compared to simple scenes but only for drivers not engaged in a cell phone conversation. Additionally, a significant interaction between scene complexity and closing velocity indicated that the mean optical expansion rate was significantly larger for complex scenes compared to simple scenes for all closing velocities except 15 mph (24.14 km/h), which resulted in the opposite pattern. In summary, our results suggest that drivers perceived an immediate hazard later when driving in a complex driving environment compared to a simple driving environment but only when closing velocity was high or drivers were not engaged in a cell phone conversation. The subsequent analysis that compares experiments helps to explain these results because it reveals that drivers engaged in a cell phone conversation while closing on a lead vehicle at 15 mph (24.14 km/h) perceived an immediate hazard later for simple scenes compared to complex scenes.

Comparing Experiment 1 and Experiment 2

To compare Experiment 1 and Experiment 2, we conducted a 2 (cell phone conversation: not engaged, engaged) × 2 (experiment: 1, 2) × 2 (scene complexity: simple, complex) × 4 (closing velocity: 15, 30, 45, 60 mph [24.14, 48.28, 72.42, and 96.56 km/h]) mixed ANOVA. Given we found effects involving cell phone conversation were significant in Experiment 2 but not in Experiment 1, we expected cell phone conversation to have a differential effect on the looming threshold for closing and an immediate hazard. Additionally, we expected experiment to be an important factor because the experiments’ instructions reflect the distinction between the looming threshold for closing and an immediate hazard.

The results of the four-way mixed ANOVA are shown in Table 1. Focusing on the highest level interaction, the four-way interaction between cell phone conversation, experiment, scene complexity, and closing velocity was significant as shown in Figure 8. When this four-way interaction is broken apart by cell phone conversation, the three-way interaction between experiment, complexity, and closing velocity was not significant for drivers not engaged in a cell phone conversation, F(2.32, 88.10) = 1.11, p = .339, η_G² = .002, but was significant for drivers engaged in a cell phone conversation, F(2.06, 78.30) = 7.14, p = .001, η_G² = .018. Breaking apart this significant three-way interaction for drivers engaged in a cell phone conversation by experiment, the two-way interaction between complexity and closing velocity was not significant for Experiment 1, F(2.15, 40.90) = 2.69, p = .076, η_G² = .011, but was significant for Experiment 2, F(1.93, 36.70) = 12.50, p < .001, η_G² = .064. Breaking apart this two-way interaction for Experiment 2 by closing velocity, mean optical expansion rate was significantly larger for simple scenes compared to complex scenes when the closing velocity was 15 mph (24.14 km/h), F(1, 19) = 12.10, p = .002, η_G² = .154, and significantly smaller for simple scenes compared to complex scenes when the closing velocity was 45 mph (72.42 km/h), F(1, 19) = 9.10, p = .007, η_G² = .021. However, the effect of complexity was not significant for the other closing velocities: 30 mph (48.28 km/h), F(1, 19) = 1.78, p = .198, η_G² = .009; 60 mph (96.56 km/h), F(1, 19) = .43, p = .522, η_G² = .002. Breaking apart this two-way interaction between complexity and closing velocity for Experiment 2 by complexity, closing velocity was significant for both simple scenes, F(1.56, 29.6) = 9.88, p = .001, η_G² = .087, and complex scenes, F(1.60, 30.5) = 9.57, p = .001, η_G² = .056. However, Tukey’s HSD indicated none of the pairwise comparisons between closing velocities was significant for simple or complex scenes.

Table 1

Four-Way Analysis of Variance to Compare Experiments

Effect	F Ratio	Df	p	η_G²
CPC	0.27	1, 76	.606	.003
Exp	12.85	1, 76	.001	.119
SC	8.81	1, 76	.004	.006
CV	3.52	2.51, 190.63	.022	.003
CPC × Exp	4.68	1, 76	.034	.047
CPC × SC	6.17	1, 76	.015	.004
Exp × SC	3.80	1, 76	.055	.003
CPC × CV	3.74	2.51, 190.63	.017	.003
Exp × CV	3.87	2.51, 190.63	.015	.003
SC × CV	25.75	2.20, 167.09	<.001	.029
CPC × Exp × SC	2.40	1, 76	.126	.002
CPC × Exp × CV	2.62	2.51, 190.63	.063	.002
CPC × SC × CV	1.04	2.20, 167.09	.362	.001
Exp × SC × CV	6.13	2.20, 167.09	.002	.007
CPC × Exp × SC × CV	4.16	2.20, 167.09	.015	.005

Note. CPC = cell phone conversation; CV= closing velocity; Exp = experiment; SC = scene complexity

Figure 8

Four-way interaction between cell phone, experiment, scene complexity, and closing velocity. Note. Error bars represent ± 1 standard error of the mean. Table of means and standard errors is available online in the Supplemental Materials.

This analysis showed that drivers engaged in a cell phone conversation perceived an immediate hazard later when the driving environment was simple and they were closing on the lead vehicle at 15 mph (24.14 km/h; the far left filled circle along the solid line in the bottom half of Figure 8) compared with when the driving environment was complex and closing at 15 mph (24.14 km/h; the far left filled triangle along the unfilled line in the bottom half of Figure 8), but this was not the case for higher closing velocities. This finding is important because it is the effect that was not found in Experiment 2. Now that it has emerged, the explanation that drivers perceived an immediate hazard later when engaged in a cell phone conversation than when not engaged but only when the driving task was relatively less demanding fully accounts for the results of Experiment 2. The four-way interaction also revealed that this effect found in Experiment 2 was not found in Experiment 1, supporting the conclusion drawn in Experiment 1 that cell phone conversation does not affect the perception of closing (compared to perception of an immediate hazard in Experiment 2). Finally, this three-way interaction between experiment, complexity, and closing velocity was not present when drivers were not engaged in a cell phone conversation, underscoring the finding that drivers regulated performance on the cell phone conversation task to maintain driving performance. Similarly, Gugerty et al. (2003) found that drivers regulated their performance on a cell phone conversation task by slowing their responses to maintain driving performance that was comparable to when drivers conversed with an in-vehicle passenger.

General Discussion

The objectives of the current studies were to examine the effects of a hands-free cell phone conversation (operationalized by the last letter task) and scene complexity (operationalized as the presence or absence of pedestrians entering the roadway) on the looming threshold for closing (Experiment 1) and an immediate hazard (Experiment 2). In Experiment 1, we found engaging in a cell phone conversation had no significant effect on the looming threshold for closing and thus no support for the hypothesis on cell phone conversation. Drivers’ performance on the cell phone conversation task did not differ across conditions, and the overall mean response accuracy was comparable with the condition in Experiment 2 with the best mean response accuracy. This means that drivers did not sacrifice performance on the cell phone conversation task when they were instructed to indicate when they first perceived closing.

In Experiment 2, we found medium effects such that drivers perceived an immediate hazard later when engaged in a cell phone conversation than when not engaged but only when the driving task was relatively less demanding such as when the roadway environment was simple (η_G² = .104) or closing velocity was relatively low (η_G² = .121), partially supporting the hypothesis on cell phone conversation. We found that drivers performed worse on the cell phone conversation task when the roadway was complex or closing velocity was relatively high. Hence, drivers sacrificed performance on the cell phone conversation task to maintain driving performance when the driving task became more demanding. This is consistent with Fuller’s (2005) model of task difficulty regulation for driving in which the driver strives to stay within a preferred range of task difficulty and is willing to sacrifice lower priority tasks to meet that goal. This contrasts with the findings by Caird et al. (2018) that drivers do not exhibit compensatory behavior when engaging in a cell phone conversation, though Caird et al. did not specifically examine drivers’ perceptual judgments. However, if drivers always sacrificed performance on the cell phone conversation task when task demands exceeded capability, we would not expect drivers to perceive an immediate hazard later in any condition, but we observed medium effect sizes when the roadway was simple and closing velocity was 15 mph (24.14 km/h). From the perspective of Fuller’s (2005) model, one possible explanation for this is that task difficulty exceeded the driver’s capability, yet the drivers were not aware of this because the degree to which task difficulty exceeded their capability was insufficient for them to recognize it. This explanation assumes that drivers are approximately aware of task difficulty and their capability but cannot assess them perfectly.

The fact that a cell phone conversation affected only the looming threshold for an immediate hazard and not the looming threshold for closing suggests that looming is used differently in these two judgments. Cell phone conversations may not interfere with the looming threshold for closing because it requires so little cognitive resources that it can be easily time-shared with a cell phone conversation. This may be because drivers directly perceive closing from optic flow (Gibson, 1979). However, cell phone conversations may affect an immediate hazard judgment because it also involves cognitive processes similar to how some have characterized time-to-collision judgments (Kiefer et al., 2006). Furthermore, the effect of cell phone conversation on the looming threshold for an immediate hazard but not on the looming threshold for closing suggests that the two judgments rely on different underlying mechanisms. The underlying mechanism for the looming threshold for an immediate hazard may rely on an evidence accumulation process (Markkula et al., 2016, 2020), whereas the looming threshold for closing may rely on a discrete looming threshold. Regardless, it appears that the degradation in driving performance that occurs when drivers engage in a cell phone conversation is due to an effect on the looming threshold for an immediate hazard but not due to effects on the looming threshold for closing.

For both experiments, we found partial support for the hypothesis regarding scene complexity because drivers responded later for complex scenes compared to simple scenes but only when closing velocity was 30 mph (48.28 km/h) or greater. However, most of these effect sizes were small, meaning that this effect may be of less consequence relative to the effect of cell phone conversation which had larger effect sizes. One possible explanation for this is that the duration of drivers’ glances away from the forward roadway remained constant regardless of closing velocity, delaying when the driver perceived closing more for faster closing velocities. Future research should use an eye tracker to further investigate this explanation.

When we compared our results with prior research, we found that drivers not engaged in a cell phone conversation while driving in a simple environment with a closing velocity of 15 mph (24.14 km/h) first perceived closing when the mean optical expansion rate was .0028 rad/s, which is reasonably consistent with comparable conditions from prior research indicating the mean optical expansion rate was .003 rad/s (Hoffmann & Mortimer, 1996). Our results were also consistent with those by Dinakar et al. (2018). For the looming threshold for an immediate hazard, we found that drivers not engaged in a cell phone conversation while driving in a complex environment with a closing velocity of 45 mph (72.42 km/h) first perceived an immediate hazard when the mean optical expansion rate was .0057 rad/s, which is reasonably consistent with comparable conditions from prior research indicating the mean optical expansion rate was .006 rad/s (Muttart et al., 2005). Hence, our results provide corroborating evidence for these looming thresholds and support Muttart et al.’s (2005) framework for a driver’s response to a lead vehicle. Our results do not corroborate looming thresholds from analyses of crash and near-crash data (e.g., Maddox & Kiefer, 2012; Markkula et al., 2016), which found thresholds near .02 rad/s. Such analyses have been previously criticized (Mortimer et al., 2014) because they are missing all of the cases in which drivers acted early enough to avoid a crash or near crash. Our results also highlighted the importance of considering the circumstances the driver is in, including roadway complexity and closing velocity, when applying the looming thresholds. For instance, drivers who were engaged in a cell phone conversation perceived an immediate hazard much later (η_G² = .154) in a simple driving environment compared to a complex driving environment when closing velocity was 15 mph (24.14 km/h).

Limitations

The results of the current study should be interpreted in light of its limitations. First, we used a cognitive task to emulate a cell phone conversation, which is not the same as an actual conversation and thus is limited in external validity. Even though we carefully chose the cognitive task to mirror the demands of a cell phone conversation, the cognitive task still lacks some components found in some cell phone conversations. Second, we manipulated roadway complexity by the presence or absence of pedestrians, but in the real world roadway complexity is made up of more than this and includes, for instance, traffic density, intersection density, and scenery (Lee et al., 2001). It is possible other forms of scene complexity, such as traffic signs (Caird et al., 2018), are not as important to the driver, meaning that the driver may not regulate their cell phone conversation as much. Further research should consider a broader definition of roadway complexity and examine how different categories of roadway hazards affect the looming threshold for an immediate hazard. Third, our method of working backward from a button press using a perception-response time to infer when perception took place requires more research. For instance, our method constrains the task by instructing participants to take a specific action (press a button) when they perceive a particular perceptual event (e.g., closing on a lead vehicle; immediate hazard). On an actual roadway, drivers would not have such instructions and may have a higher workload from basic vehicle handling tasks such as lane keeping and speed control (but see Levulis, 2018; Zhang et al., 2014). Future research should endeavor to measure when drivers naturally respond without being given instructions and while controlling the vehicle.

Practical Implications

The current research has important practical implications for accident analysis. For example, in forensic cases, an important question to consider when analyzing a collision is the time at which a reasonable driver should react to a hazard (Green, 2018; Krauss, 2015). For cases involving a slow-moving or stopped lead vehicle, the results of the current study suggest that scene complexity and the use of a cell phone can affect this response to a hazard. Future research should examine what particular driving action is associated with the looming threshold for closing and an immediate hazard (e.g., releasing the accelerator, moderate braking, hard braking). The current research also could be applied to the driving behavior of driving automation systems. For example, a driving automation system should not initiate a driving response to a roadway hazard later than a typical human driver. Otherwise, the occupants in the vehicle may begin to distrust the driving automation system, and a late response by the automated driving system may evoke the human driver to take over vehicle control unnecessarily.

Key Points

Drivers perceived an immediate hazard later only when the demands of the driving task were relatively low, such as when the roadway was simple (operationalized by the absence of pedestrians entering the roadway) or closing velocity was low, because drivers prioritize performance on the cell phone conversation task (operationalized as the last letter task) when the driving task was less demanding. Conversely, drivers sacrificed performance on the cell phone conversation task when the driving task was more demanding to maintain driving performance.

There was no indication that cell phone conversations affect the looming threshold for closing.

Drivers perceived both closing and an immediate hazard later during complex scenes compared to simple scenes but only when closing velocity was 30 mph (48.28 km/h) or greater.

Our results corroborate the previously published button press looming thresholds of .003 rad/s (closing) and .006 rad/s (immediate hazard).

Supplemental Material

Online supplementary file 1 - Supplemental material for Factors That Affect Drivers’ Perception of Closing and an Immediate Hazard

Supplemental material, Online supplementary file 1, for Factors That Affect Drivers’ Perception of Closing and an Immediate Hazard by Bradley W. Weaver, Patricia R. DeLucia and Jason Jupe in Human Factors: The Journal of Human Factors and Ergonomics Society

Footnotes

Acknowledgments

This work built upon work summarized in the proceedings (Weaver et al., 2019a). We thank Viola Yu for her assistance with building the scenes and collecting the data. We also thank Adam Braly for guidance on SAS programming and the undergraduate research assistants in our laboratory who served as pilot subjects.

Supplemental Material

The online supplemental material is available with the manuscript on the HF website.

ORCID iDs

Bradley W. Weaver

Patricia R. DeLucia

Author Biographies

Bradley W. Weaver is a PhD student in the Human Factors and Human-Computer Interaction Program at Rice University. He obtained his MA in psychological sciences at Rice University in 2020.

Patricia R. DeLucia is a professor in the Department of Psychological Sciences at Rice University. She obtained her PhD in experimental psychology at Columbia University in 1989.

Jason Jupe, P.E. is a forensic engineer at Rimkus Consulting Group, Inc. He obtained his MS in industrial engineering from Texas Tech University.

References

Bakeman

. (2005). Recommended effect size statistics for repeated measures designs. Behavior Research Methods, 37, 379–384.doi:10.3758/BF03192707

16405133

Braly

A. M.

Parent

M. C.

DeLucia

P. R

. (2018). Do threats to masculinity result in more aggressive driving behavior? Psychology of Men & Masculinity, 19, 540–546.doi:10.1037/men0000135

Caird

J. K.

Simmons

S. M.

Wiley

Johnston

K. A.

Horrey

W. J

. (2018). Does talking on a cell phone, with a passenger, or dialing affect driving performance? An updated systematic review and meta-analysis of experimental studies. Human Factors: The Journal of the Human Factors and Ergonomics Society, 60, 101–133.doi:10.1177/0018720817748145

29351023

Cochran

W. G

. (1947). Some consequences when the assumptions for the analysis of variance are not satisfied. Biometrics, 3, 22–38.doi:10.2307/3001535

20240415

Cohen

. (1988). Statistical power analysis for the behavioral sciences. Lawrence Erlbaum Associates.

Cooper

J. M.

Medeiros-Ward

Strayer

D. L

. (2013). The impact of eye movements and cognitive workload on lateral position variability in driving. Human Factors: The Journal of the Human Factors and Ergonomics Society, 55, 1001–1014.doi:10.1177/0018720813480177

24218908

Dinakar

Muttart

J. W.

Garrison

Gernhard

Marr

. (2018). Influence of taillight brightness on the ability to recognize closing speed, closing distance, and closing vs. separating. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 62, 1868–1872.doi:10.1177/1541931218621425

Freund

Gravenstein

Ferris

Burke

B. L.

Shaheen

. (2005). Drawing clocks and driving cars. Journal of General Internal Medicine, 20, 240–244.doi:10.1111/j.1525-1497.2005.40069.x

15836527

Fuller

. (2005). Towards a general theory of driver behaviour. Accident Analysis & Prevention, 37, 461–472.doi:10.1016/j.aap.2004.11.003

15784200

10.

Gibson

J. J

. (1979). The ecological approach to visual perception. Houghton Mifflin.

11.

Glass

Peckham

P. D.

Sanders

J. R

. (1972). Consequences of failure to meet assumptions underlying the fixed effects analyses of variance and covariance. Review of Educational Research, 42, 237–288.doi:10.3102/00346543042003237

12.

Green

. (2018). Roadway human factors: From science to application. Lawyers and Judges Publishing.

13.

Gugerty

Rakauskas

Brooks

. (2004). Effects of remote and in-person verbal interactions on verbalization rates and attention to dynamic spatial scenes. Accident Analysis & Prevention, 36, 1029–1043.doi:10.1016/j.aap.2003.12.002

15350880

14.

Gugerty

Rando

Rakauskas

Brooks

Olson

. (2003). Differences in remote versus in-person communications while performing a driving task. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 47, 1855–1859.doi:10.1177/154193120304701605

15.

Hoffmann

E. R.

Mortimer

R. G

. (1996). Scaling of relative velocity between vehicles. Accident Analysis & Prevention, 28, 415–421.doi:10.1016/0001-4575(96)00005-X

8870768

16.

Kiefer

R. J.

Flannagan

C. A.

Jerome

C. J

. (2006). Time-to-collision judgments under realistic driving conditions. Human Factors: The Journal of the Human Factors and Ergonomics Society, 48, 334–345.doi:10.1518/001872006777724499

16884053

17.

Knipling

R. R.

Mironer

Hendricks

D. L.

Tijeripa

Everson

Allen

J. C.

Wilson

. (1993). Assessment of IVHS countermeasures for collision avoidance: Rear-end crashes (No. DOT-HS-807-995). National Highway Traffic Safety Administration. https://rosap.ntl.bts.gov/view/dot/4276

18.

Krauss

. (2015). Forensic aspects of driver perception and response (4th ed.). Lawyers and Judges Publishing.

19.

Laberge

. (2003). The effect of passenger and cellular phone conversations on driver distraction (Unpublished master’s thesis). University of Calgary.

20.

Lee

J. D.

Caven

Haake

Brown

T. L

. (2001). Speech-based interaction with in-vehicle computers: The effect of speech-based e-mail on drivers’ attention to the roadway. Human Factors: The Journal of the Human Factors and Ergonomics Society, 43, 631–640.doi:10.1518/001872001775870340

12002011

21.

Levulis

S. J

. (2018). Effects of cell phone use on left turn across path judgments: Identifying critical informational sources and attentional resources necessary for safe driving [Doctoral dissertation, Texas Tech University]. Texas Digital Library. https://ttu-ir.tdl.org/bitstream/handle/2346/73834/LEVULIS-DISSERTATION-2018.pdf?sequence=1

22.

Levulis

S. J.

DeLucia

P. R.

Jupe

. (2015). Effects of oncoming vehicle size on overtaking judgments. Accident Analysis & Prevention, 82, 163–170.doi:10.1016/j.aap.2015.05.024

26080078

23.

Lix

L. M.

Keselman

J. C.

Keselman

H. J

. (1996). Consequences of assumption violations revisited: A quantitative review of alternatives to the one-way analysis of variance “F” test. Review of Educational Research, 66, 579–619.doi:10.2307/1170654

24.

Maddox

M. E.

Kiefer

. (2012). Looming threshold limits and their use in forensic practice. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 56, 700–704.doi:10.1177/1071181312561146

25.

Markkula

Engström

Lodin

Bärgman

Victor

. (2016). A farewell to brake reaction times? Kinematics-dependent brake response in naturalistic rear-end emergencies. Accident Analysis & Prevention, 95, 209–226.doi:10.1016/j.aap.2016.07.007

27450793

26.

Markkula

Uludağ

Wilkie

Billington

. (2020). Evidence accumulation, not a perceptual threshold, constrains neural and behavioural responses in human collision threat detection. PsyArXiv.

27.

Mortimer

Hoffmann

Kiefer

. (2014). The psychological and accident reconstruction “thresholds” of drivers’ detection of relative velocity (No. 2014-01-0437). SAE Technical Paper.

28.

Muttart

J. W.

Messerschmidt

W. F.

Gillen

L. G

. (2005). Relationship between relative velocity detection and driver response times in vehicle following situations (No. 2005-01-0427). SAE International. https://doi.org/10.4271/2005-01-0427

29.

National Highway Traffic Safety Administration . (2020, September 1). Fatality and Injury Reporting System Tool (Version 3.2) [Database Query Tool]. Department of Transportation. https://cdan.dot.gov/query

30.

Olejnik

Algina

. (2003). Generalized ETA and omega squared statistics: Measures of effect size for some common research designs. Psychological Methods, 8, 434–447.doi:10.1037/1082-989X.8.4.434

14664681

31.

Olson

R. L.

Hanowski

R. J.

Hickman

J. S.

Bocanegra

. (2009). Driver distraction in commercial vehicle operations (No. FMCSA-RRR-09-042). Federal Motor Carrier Safety Administration. https://rosap.ntl.bts.gov/view/dot/17715

32.

Strayer

D. L.

Drews

F. A.

Johnston

W. A

. (2003). Cell phone-induced failures of visual attention during simulated driving. Journal of Experimental Psychology: Applied, 9, 23–32.doi:10.1037/1076-898X.9.1.23

12710835

33.

Strayer

D. L.

Johnston

W. A

. (2001). Driven to distraction: Dual-task studies of simulated driving and conversing on a cellular telephone. Psychological Science, 12, 462–466.doi:10.1111/1467-9280.00386

11760132

34.

Summala

Lamble

Laakso

. (1998). Driving experience and perception of the lead CAR’s braking when looking at in-car targets. Accident Analysis & Prevention, 30, 401–407.doi:10.1016/S0001-4575(98)00005-0

9666236

35.

Weaver

B. W.

DeLucia

P. R.

Jupe

. (2019a). Effects of cell phone conversation and roadway complexity on the perception of an immediate hazard while driving. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 63, 2098.doi:10.1177/1071181319631376

36.

Weaver

B. W.

DeLucia

P. R.

Jupe

. (2019b). Effects of cell phone conversation on the first driver response to a stopped lead vehicle. Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 63, 2031–2035.doi:10.1177/1071181319631360

37.

Young

S. K.

Eberhard

C. A.

Moffa

P. J

. (1995). Development of performance specifications for collision avoidance systems for lane change, merging, and backing. Task 2 interim report: Functional goals establishment. US Department of Transportation, National Highway Traffic Safety Administration.

38.

Zhang

Kaber

D. B.

Rogers

Liang

Gangakhedkar

. (2014). The effects of visual and cognitive distractions on operational and tactical driving behaviors. Human Factors: The Journal of the Human Factors and Ergonomics Society, 56, 592–604.doi:10.1177/0018720813495279

24930178

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.09 MB