Empirical Evaluation of Hazard Anticipation ... - SAGE Journals

8 downloads 0 Views 898KB Size Report
already changing the face of driver education. Empirical Evaluation of Hazard Anticipation. Behaviors in the Field and on. Driving Simulator Using Eye Tracker.
Empirical Evaluation of Hazard Anticipation Behaviors in the Field and on Driving Simulator Using Eye Tracker Donald L. Fisher, Anuj K. Pradhan, Alexander Pollatsek, and Michael A. Knodler, Jr. (e) efficacy of novice (6, 7) and older driver training programs. Driving simulators are used most often when considerations of the safety of participants is paramount. Contrary to studies in the field, on a simulator participants can be put in scenarios that do not compromise their safety or the safety of other drivers. The focus in this paper will be on the generalizability to the field of simulator studies in which the driver is exposed to potential hazards and eye movements are used as an index of drivers’ hazard anticipation and perception skills. Three examples of simulator studies in which eye movements are used as a window into hazard anticipation are discussed below. First, eye movements were measured in a recent study of the effect of additional cognitive load (such as cell phone conversations) on drivers’ scanning patterns (8, 9). Drivers were asked to navigate a virtual world and in the experimental conditions to simultaneously execute tasks that increased either their verbal or spatial load. Results indicated that the mean saccade distance is reduced by about 23% for the verbal task and 33% for the spatial task when compared with the control task (no additional load). Such simulator studies may soon affect legislatures that must decide whether to ban cell phones outright or to restrict their use in certain highly dangerous situations, for example, work zones. Second, eye movements have been used to determine whether older drivers look appropriately for oncoming vehicles in scenarios in which such vehicles may be hidden until the last several seconds before the driver makes a turn (10). For example, in one scenario presented to drivers in Romoser et al. (10), the participant driver approached a T-intersection and was instructed to turn to the right at the intersection. Just after the participant driver initiated the turn, a car rounded a corner or came up over a hill from the left, potentially colliding with the participant driver if he or she did not speed up appropriately. Consistent with crash statistics, it was found that the older drivers made the appropriate eye movements less often than younger drivers. Such studies could eventually be central to the determination of whether older drives are fit to drive. Third, eye movements have been used to evaluate whether novice drivers trained with a PC-based hazard anticipation program would look for information in areas of the roadway that should reduce the likelihood of a crash (11). For example, in one scenario presented to drivers in Pollatsek et al. (11), a truck is parked in front of a crosswalk (the truck crosswalk scenario, Figure 1). A pedestrian could potentially emerge from behind the truck. The participant driver should look to the right as he or she passes the truck. Indeed, the trained drivers were more likely to do so. Studies such as these are already changing the face of driver education.

Eye behaviors have been used with driving simulators to evaluate the effectiveness of novice and older driver training programs. Driving simulators are often favored when drivers must be placed in risky situations. Because there was no study of whether eye behaviors observed on a driving simulator in risky scenarios were also observed in the field, the authors had both trained and untrained novice drivers maneuver a controlled set of 10 scenarios on a driving simulator. The scenarios were similar to a set of scenarios that a different, matched set of trained and untrained drivers had navigated in the field. Drivers in this simulator study were trained with the same PC program used by drivers in the field study. Five of the scenarios that the trained drivers saw on the simulator and in the field were similar to those seen in training on a PC (near transfer); the other five were similar in concept to those in training but different in surface features (far transfer). A fixation on the region of a scenario that had information relevant to identifying a risk was scored as recognizing the risk. On the simulator, trained drivers recognized the risk 41.7% more often than untrained drivers in the near-transfer scenarios and 32.6% more often in the far-transfer scenarios. In the field, trained drivers recognized the risk 38.8% more often in the near-transfer and 20.1% more often in the far-transfer scenarios. Both effects were highly significant, and the difference between them was not close to significant. Thus results from tests on a simulator have a close correspondence with those in the field.

Driving simulator scenarios are becoming key elements in state and federal efforts to probe (a) how using electronic devices in the car affects drivers’ situation awareness, especially as that affects their safety (1); (b) whether patients with performance decline (e.g., aging, mild cognitive impairment, traumatic brain injury, neurodegenerative disorders such as Alzheimer’s and Parkinson’s diseases, or sleep disturbances) are fit to drive (2, 3); (c) acute and chronic effects of many medications such as analgesics, antidepressants, psychostimulants, antidepressants, and cancer chemotherapy agents (4); (d ) effectiveness of alternative signs, signals, and pavement markings (5); and D. L. Fisher and A. K. Pradhan, Department of Mechanical and Industrial Engineering, 220 Engineering Laboratory, 160 Governor’s Drive; A. Pollatsek, Department of Psychology, Tobin Hall, 135 Hicks Way; and M. A. Knodler, Jr., Department of Civil and Environmental Engineering, 216 Marston Hall, 130 Natural Resources Road, University of Massachusetts, Amherst, MA 01003. Corresponding author: A. K. Pradhan, [email protected]. Transportation Research Record: Journal of the Transportation Research Board, No. 2018, Transportation Research Board of the National Academies, Washington, D.C., 2007, pp. 80–86. DOI: 10.3141/2018-11

80

Fisher, Pradhan, Pollatsek, and Knodler

81

FIELD STUDY: METHOD The field study is described in depth by Pradhan et al. (19); briefer discussions are provided by Fisher et al. (17) and Pollatsek et al. (18). The most important details are discussed below.

Participants

Participant Driver

FIGURE 1 Plan view of truck parked in front of crosswalk scenario. (Pedestrian may emerge from behind front of truck parked on right shoulder. Participant drivers should look to right as they pass truck.)

In summary, results from studies such as those above are finding their way into state (12) and national policies (13), federal guidelines (14), and business (15) and foundation (16) efforts. Yet in none of these studies on the driving simulator in which the scenarios were potentially hazardous has it been confirmed that participants would behave similarly on the open road. Perhaps participants are simply less vigilant in the driving simulator because hazards cannot harm them there. Thus, for example, in the truck crosswalk scenario (Figure 1) the participant in a simulator experiment may not be as vigilant as he or she might be on the open road because there would be no real-world consequences if a simulated pedestrian emerged from behind the truck. At this point it is critical to know whether the eye behaviors of drivers in the simulator in a potentially hazardous scenario closely resemble the behaviors that one would observe on the open road in a similarly hazardous scenario. Obviously it is difficult to collect information on all hazardous scenarios in both environments. Still, there are scenarios that allow for the collection of relevant data. Given such data, there are then two ways to approach the issue of generalization. First, one can ask whether the relevant dependent variables on the simulator and in the field are absolutely identical. In this case, the values of the relevant dependent variable will be eye behaviors. Second, when evaluating the efficacy of training programs (or devices designed to make driving safer), one can ask whether the difference between the simulator results and the field results is similar. One can ask both questions overall (average over all the different scenarios) as well as on a scenario-by-scenario basis. A study on the driving simulator is now being undertaken that replicates, in part, a study that was previously undertaken by the authors in the field. Specifically, 10 of the scenarios that were built for the driving simulator for the study and reported below are similar to (but not identical to) the situations to which drivers were exposed in the field (17–19). In both experiments trained and untrained drivers between the ages of 18 and 21 were evaluated primarily by measuring the percentage of scenarios in which each driver looked at a critical area: an area that had information relevant to reducing the likelihood of a crash. In each study, the trained and untrained drivers were compared on these measures. In the following there will be (a) a brief description of the experimental design of the field study, (b) a fuller description of the design of the simulator study, and then (c) a comparison of the findings of the two studies.

The 24 participants were all recruited from the student body of the University of Massachusetts–Amherst campus. They were between 18 and 21 years old and all had held a valid U.S. driver’s license for at least 1 year. The 12 male and 12 female participants were separately, randomly assigned to the trained group or the untrained group, so that there were six male and six female participants in each group. Because of difficulties with the eye-tracker calibration, it was not possible to recruit participants wearing eyeglasses; thus all participants either had normal vision or vision corrected to normal with contact lenses. The mean ages of the control group and the experimental group were 19.27 and 19.87, respectively, with standard deviations of 0.64 and 0.75.

Design

Training Program Participants in the experimental group were trained on PCs using Version 3 of the Risk Awareness and Perception Training (RAPT-3) program developed at the University of Massachusetts Amherst. RAPT-3 was designed to illustrate different categories of not-so-obviously hazardous scenarios and to train drivers to focus their attention on critical regions that, if scanned, reduced the likelihood of a crash. The training program contained nine driving scenarios, in each of which there was an inherent risk of a collision with another vehicle or pedestrian. RAPT-3 is described in more detail in the discussion of the method used for the simulator study.

Field Driving Route The route driven by participants was a 16-mi course plotted in and around Amherst, Massachusetts, that included major arterials; a variety of intersections; and covered rural, residential, city, and highway driving situations. It was designed to include 10 situations of interest (scenarios) that would be analyzed. They were all embedded naturally in the driving course so that the participant had no indication that these were the primary areas of interest to the researchers. Five of the open road scenarios, the near-transfer scenarios, were similar in concept to five of the scenarios that were in the RAPT-3 training program; the remaining five open road scenarios, the far-transfer scenarios, were different from the scenarios that were seen during training. Twelve measures were extracted from 10 scenarios, five from the near-transfer and seven from the far-transfer scenarios.

Apparatus A portable lightweight eye tracker (Mobile Eye developed by Applied Science Laboratories) was used to collect the eye-movement data for each driver during the on-road drives. It has a lightweight optical system consisting of an eye camera and a color scene camera mounted on a pair of safety goggles (see Figure 2). The images from these two

82

Transportation Research Record 2018

Participants Optics

Scene Camera

Safety Glasses

A total of 12 participants, six men and six women, were all recruited from the student body of the University of Massachusetts–Amherst campus. They were between 18 and 21 years of age and all had held a valid U.S. driver’s license for at least 1 year. Equal numbers of men and women were assigned to the trained and untrained groups. Again, because of difficulties with the eye-tracker calibration, no participants were recruited who wore eyeglasses. The mean ages of the control and the experimental group were 20.28 and 19.71, respectively, with standard deviations of 0.57 and 0.85.

Reflecting Mirror

Equipment: Driving Simulator and Eye Tracker FIGURE 2

ASL mobile eye tracker.

cameras are interleaved and recorded on a remote recording system, thus ensuring no loss of resolution. The interleaved video can then be transferred to a PC on which the images are separated and processed. The eye movement data are converted to a crosshair, representing the driver’s point of gaze, which is superimposed on the scene video recorded during the drive. This provides a record of the driver’s point of gaze on the driving scene while on the on-road driving course. The remote recording system is battery-powered and capable of recording up to 90 min of eye and scene information at 60 Hz in a single trial. Each participant drove the field test with a four-door sedan with automatic transmission (a 2002 Chevy Prizm or a 2000 Chevrolet Cavalier). The vehicles were rented from a local area driving school and had a secondary braking system that could be operated by a certified driving instructor (who was sitting in the front passenger seat solely for safety reasons).

Procedure The RAPT-3 training program took about 30 to 45 min to complete. Participants in the control group did not take part in the training program. The trained and the control drivers were fitted with the eye tracker and the necessary calibration process was carried out, which took about 5 min. The participant then drove through the course with the driving instructor in the front passenger seat and a researcher in the backseat. The researcher provided the participant with directional information at appropriate points in the course. The drive through the entire course took about 45 to 55 min to complete. To control for timeof-day effects and traffic conditions, the drives were all at 9 a.m. or 10 a.m. on weekdays. The eye-tracking system recorded the point of gaze data, which served as the primary dependent measure, along with a video record of the driver’s view of the roadway during the entire drive.

An advanced fixed-base driving simulator was used for the driver evaluation. The simulator has a fully equipped 1995 Saturn sedan placed in front of three screens on which the virtual environment is projected. The screens subtend 135 degrees horizontally, and the virtual world is displayed on each screen at a resolution of 1,024 × 768 pixels at a frequency of 60 Hz (see Figure 3). Participants sit in the car and operate the controls, moving through the virtual world according to their inputs to the car. The sound is controlled by another computer, the Acoustetron, which consists of two mid- to high-frequency speakers located on the left and right sides of the car and two subwoofers located under the hood of the car. The system provides realistic road, wind, and other vehicle noises with appropriate direction, intensity, and Doppler shift. The same eye tracker was used in this study as was used in the field study.

Experimental Design and Procedure

Training Program As in the field study, participants in the simulator study were trained with RAPT-3. Here details of the training program are discussed in greater depth. Recall that RAPT-3 contained nine scenarios, each of which included a potential risk. In the first set of risks, the risks were due to vehicles or pedestrians being hidden from view until the last moment, either because of the geometry of the roadway or the presence of an obscuring vehicle or object. In the second set, risks were due to visible elements, either cars that plausibly might change lanes

SIMULATOR EXPERIMENT: METHOD The simulator experiment used the same PC training program as the field experiment (RAPT-3). Furthermore, the simulator test (which had 18 scenarios) contained 10 scenarios that were similar to the 10 field scenarios, five of which were classified as neartransfer simulator scenarios and five of which were classified as far-transfer simulator scenarios. Details follow.

FIGURE 3

University of Massachusetts–Amherst driving simulator.

Fisher, Pradhan, Pollatsek, and Knodler

FIGURE 4

83

RAPT-3 training program for hidden sidewalk scenario.

abruptly or pedestrians that might enter a crosswalk suddenly and cause a lead car to brake suddenly. The scenarios were selected from a set used in previous studies (11, 17 ), but because perspective views had to be photographed, safety issues made it necessary to select only those that did not directly involve any moving vehicle as the inherent risk in a scenario. In addition, to portray several of the scenarios accurately, some staging with other vehicles was necessary so that all elements in the scenario would appear in the snapshots. The hidden sidewalk scenario described below should illustrate the general idea of the training. In this scenario, the driver is approaching an intersection with a stop sign. There is a pedestrian crosswalk at the intersection located after the stop line. The stop line and crosswalk are themselves relatively distant from the intersection with the road on which cross traffic travels. On the right just beyond the stop line there is a high hedge that hides a sidewalk that emerges onto the crosswalk (see Figure 4). The risk is that a bicyclist or a pedestrian, hidden behind the hedge, could potentially emerge onto the crosswalk. The scenario is one that is difficult for drivers to predict as hazardous. When the test course

was set up, this particular test intersection in downtown Amherst was studied. Of the 20 drivers observed, all failed to stop at the stop line and look to the right as they passed by the bushes, instead proceeding directly up to the boundary with the crossroad. The training program (RAPT-3) started with instructions and an initial practice section to familiarize participants with the displays and the tasks they were to perform. That was followed by the three main sections of the training: pretest, training, and posttest. In the pretest each scenario was presented as a sequence of snapshots displaying the driver’s view from a vehicle traversing through a particular driving situation. A scenario contained five to 12 snapshots depending on the length and complexity of the situation (a snapshot used to train the hidden sidewalk scenario is presented in Figure 5). Each snapshot was displayed for 3 s. Participants used the mouse to click on areas of each snapshot in which they would have to pay particular attention if they were actually driving through the scenario (see red circles in Figure 5). The coordinates of the click and response time for it were internally recorded by the program. In this section, participants received no feedback on their performance. Generally, the snapshots were straight ahead from the car, but in situations in which it was necessary for a driver to look to the left or the right (e.g., at an intersection), the participant could click on buttons provided on the left or right margins of the snapshot, which would show the corresponding left or right views. The side views materialized only for situations in which the driver would need to have a view of the left or right; in other cases, the side view buttons were still displayed but clicking on them did not change the view. The training came next. The user was first shown a top-down schematic view of a scenario accompanied by explanations about the risky aspects of the particular scenario. For example, the explanation that accompanied Figure 4 appears below: This is an example of a situation in which a potential risk is obscured by bushes. In this scenario, there is a crosswalk as indicated by the pavement striping. On the left, you can easily see approaching pedestrians or bicyclists. On the right, however, any approaching pedestrian or bicyclist is hidden by the bushes. 1. It is clear from the diagram that when you are at Position 1 you cannot see to the right. As you approach the stop sign, you, the driver,

1

2

FIGURE 5

Hidden sidewalk snapshot.

84

should glance frequently at the area where the crosswalk/sidewalk disappears behind the bushes so that you can slow immediately if someone were to appear in the crosswalk or sidewalk. 2. When you are passing the actual crosswalk, you should look far to the right, turning your head if necessary. A bicyclist riding fast might not see you and could be at risk if you were to accelerate through the crosswalk without looking to the right.

After these explanations, the user was presented again with the sequence of perspective view snapshots for that scenario and was given up to four opportunities to correctly identify the areas of risk on the sequence of snapshots using the mouse. If the user could successfully identify the areas, then the program moved on to the next scenario. If not, the user was taken back to the training part of the scenario with the schematic view and corresponding explanations (until the fourth time, after which the next scenario was automatically presented). Finally, in the posttest section, the user was once again presented with the nine sequences of photographs and asked to use mouse clicks to identify areas of potential risk. As in the pretest, the click coordinates and response times were recorded for this section and no feedback was provided to the user. (The response times were not used in the scoring, however.) The training program was presented on a laptop computer running Microsoft Windows XP, and a mouse was used as the pointing device. The program was developed using Macromedia Director and was designed to operate on any Microsoft Windows operating PC. Although the program was a single executable file and can be deployed on CD-ROMs or over the Internet, it was administered on the same computer in the driving laboratory to all trained participants.

Driving Simulation Participants in both groups were evaluated on the simulator. They were given written instructions and verbal instructions with respect to driving in the simulator at the beginning of the session. They were then fitted with the eye-tracking system, and the calibration process was completed. Once the eye tracker had been calibrated, drivers were asked to drive a practice scenario. This scenario was designed to contain all elements of the virtual environment that a driver would experience during the actual experimental scenarios, including intersections, traffic situations, and other elements. This practice scenario also served to accustom the participant to the specific handling characteristics of the vehicle used in the driving simulator. Participants were encouraged to drive the practice scenario as many times as necessary until they were comfortable with the handling of the vehicle, especially left and right turns and braking situations. There were 18 experimental scenarios in the driving simulation. These scenarios were laid out in three blocks of six scenarios each. The blocks were counterbalanced evenly within the groups. The 18 scenarios were composed of nine of the scenarios that were used for the PC training, that is, the near-transfer scenarios, and nine other scenarios that were different from those used in the training, the far-transfer scenarios. The far-transfer scenarios were used to test for possible generalization of the PC training. The databases were designed such that each block contained three near-transfer and three far-transfer scenarios. Participants drove through the three blocks with a rest between blocks. During the entire drive, eye movements of participants and various vehicle parameters were constantly being recorded. Of the 18 scenarios that were presented to participants in the driving simulator, 10 of the scenarios were similar to the scenarios in the field study, five near-transfer and five far-transfer. The scenar-

Transportation Research Record 2018

ios were all embedded in the three blocks with stretches of regular driving between scenarios. The experimental scenarios were all located and designed in such a manner that it would not be obvious to participants that they were traversing an area of interest to the researcher.

RESULTS: SIMULATOR AND FIELD The key data in both studies were whether a participant fixated on the appropriate region during a scenario within an appropriate time window (e.g., whether a participant fixated on the region of the circle on the right in the scenario shown in Figure 5 in a time window such that there would be adequate time to react to the presence of a hazard if it appeared). The data (i.e., the record of the crosshair indicating the fixation point against the view of the scene) were scored by several independent raters who were “blind” to which group the participant was in. There were few “judgment calls” because the difference between fixating appropriately and continuing to fixate straight ahead was usually close to 10 degrees of visual angle. Averaged over all 18 scenarios in the new simulator study, there was a 37.4% overall effect of the RAPT-3 training, t(22) = 4.12, p < .001 (see Table 1). The training effect was 42.2% for near transfer and 32.6% for far transfer, t(22) = 4.88, p < .001, and t(22) = 2.88, p < .01, respectively. Although the training effect appeared to be somewhat larger for the near-transfer scenarios, the 9.6% difference between the two training effects was far from significant, t(22) = 1.06, p > .10. These training effects were somewhat larger than those observed in the field study; the overall training effect there was 27.1%, and the near- and far-transfer effects were 38.8% and 20.1%, respectively. All three training effects in the field study were also significant, and again, although the near-transfer effect was larger than the far-transfer effect, the difference was only marginally significant (p < .10). To summarize the above, results of the two studies appear to be quite similar. However, the training effect appears to be somewhat larger in the simulator data for the far-transfer tests. The comparison above, however, compares all 18 scenarios in the simulator study with the 10 scenarios in the field study. Thus, a better comparison would be of the data from the scenarios that the two studies shared and for which the very similar scoring criteria were used. Although 10 of the scenarios in the simulator study had been devised to be similar to the scenarios in the field study, when the data were being scored, it was realized that two of the far-transfer scenarios were really not comparable between the studies—the

TABLE 1 Percentage of Time Critical Region Was Scanned in Simulator and Field Studies as Function of Training Group

Study Simulator study Field study

Test Condition Near transfer Far transfer Average Near transfer Far transfer Average

Trained (%)

Untrained (%)

Difference (effect of training) (%)

77.9 76.8 77.4 79.2 58.3 64.4

35.7 44.2 40.0 40.4 38.2 37.3

42.2 32.6 37.4 38.8 20.1 27.1

Fisher, Pradhan, Pollatsek, and Knodler

85

reason being that the field geometries of these two particular scenarios differed just enough from their counterparts in the simulator to make it difficult to compare them using the same objective scoring rules. However, the pattern of data from this “purified” set of eight scenarios is quite similar to that of the overall data above because the near-transfer effect was about the same in the field and simulator studies and the far-transfer advantage was somewhat bigger in the simulator study. The near-transfer effects averaged over the five scenarios were 37.4% (71.0% versus 33.6%) for the field study and 41.3% (70.6% versus 29.3%) for the simulator study; the far-transfer effects averaged over the three remaining scenarios were 26.5% (51.7% versus 25.2%) for the field study and 41.7% (77.8% versus 36.1%) for the simulator study. To assess whether any of these differences between the field study and simulator study were reliable, another analysis was done using the means for individual participants on one of the following: (a) all eight scenarios, (b) the five near-transfer scenarios, or (c) the fartransfer scenarios. In these “purified” analyses, the 95% confidence intervals for the differences in the training effect (simulator study minus field study) were 12.4% ± 25.0% over all eight scenarios, 3.9% ± 26.4% for the near-transfer scenarios, and 28.9% ± 36.6% for the far-transfer scenarios. Finally, a comparison was made of the training effects in the field and on the simulator, scenario by scenario, for the neartransfer scenarios. (The focus was on just the near-transfer scenarios because there were only three comparable far-transfer scenarios.) The training effects corresponded quite well in each scenario (see Figure 6). The standard error of the difference between each pair of data points in Figure 6 is about 16%. The averages of the individual scenario data are somewhat different from the average near-transfer data reported above because the latter were first averaged over participants for each scenario and there are some missing data.

DISCUSSION OF RESULTS As the cost of driving simulators has come down, their use has increased dramatically. They can provide information on drivers’ anticipation of hazards that is not as easy to gather in the field. Results of these simulator studies are even now informing national and state policies. Yet, there has been no study to date that makes it possible to determine whether the hazard anticipation behaviors observed on a driving simulator are also observed in the field. The study reported above indicates that there is a close correspondence between the effect of training on hazard anticipation behaviors of drivers in the field and on a simulator and a reasonably close correspondence in their overall level of hazard anticipation. The correspondence is rather remarkable given the large number of differences between the environment in the laboratory (the driving simulator study) and the environment on the road (the field study), not the least of which is the real danger present when a driver is actually maneuvering a vehicle on the open road. Not only is there a close similarity overall between the hazard anticipation behavior of trained and untrained novice drivers in the field and on the simulator, but this correspondence extends to actual scenarios. The scenarios in the simulator and the field were, for the most part, not chosen so that they were identical to one another down to the smallest detail. Again, the correspondence is remarkable. Several caveats are, of course, in order. First, the hazard anticipation behavior of only two groups of novice drivers was examined, those who received and those who did not receive hazard anticipation training. Other groups of drivers may perform differently in the field and on the driving simulator. Second, there are many other eye behaviors that one could have measured besides hazard anticipation behavior. Again, these behaviors may differ in the field and on the driving simulator. And third, only drivers who did not wear glasses took part. We recommend that the study be repeated with drivers who do wear glasses (the eye tracker can now accommodate such a population).

70

Percent Scanning Critical Region

60

Simulator Field

50 40 30 20 10 0 Incoming Right Turn Vehicle from with Reveal Left Fork Point

Left Turn with Reveal Point Scenario

Hidden Sidewalk

Abrupt Lane Change

FIGURE 6 Percentage scanning critical area in simulator and field studies as function of scenario.

86

Having said this, these results are still encouraging in general and especially for this group of drivers. Specifically, newly licensed drivers are more likely to be involved in a crash than drivers with more driving experience. This cohort suffers 9.3 fatal crashes per 100 million vehicle miles as compared with 1.4 for drivers between 45 and 54 years of age (20). Recent analyses of police accident reports suggest that it is poor hazard anticipation skills that are a major cause of the inflated crash rates of the newly licensed drivers (21). It has been confirmed on a driving simulator that newly licensed drivers’ hazard anticipation skills are indeed compromised (22). Further studies indicate that training on the PC-based program described above (RAPT) can dramatically improve novice drivers’ hazard anticipation skills on the simulator (11) and in the field (17), and both immediately after training and 3 to 5 days after training (23). Finding a very close correspondence on the simulator and in the field between the hazard anticipation skills of trained and untrained novice drivers suggests that one can continue to gather critical information in risky scenarios on a driving simulator about the causes of crashes and the procedures needed to remediate those crashes without putting the novice driver at risk.

ACKNOWLEDGMENTS Portions of this research were funded by grants from the Link Foundation for Simulation and Training, the National Highway Traffic Safety Administration, and the National Science Foundation. The authors thank the Pioneer Valley Driving School for its help with the research and especially Tom Marlow for his help in keeping the driving simulator always operating at its fullest potential.

REFERENCES 1. Lee, J. D., D. V. McGehee, T. L. Brown, and M. L. Reyes. Collision Warning Timing, Driver Distraction and Driver Response to Imminent Rear-End Collisions in a High-Fidelity Driving Simulator. Human Factors, Vol. 44, 2002, pp. 314–334. 2. Staplin, L. Simulator and Field Measures of Driver Age Differences in Left-Turn Gap Judgments. In Transportation Research Record 1485, TRB, National Research Council, Washington, D.C., 1995, pp. 49–55. 3. Skaar, N., M. Rizzo, and L. Stierman. Traffic Entry Judgments by Aging Drivers. Proc., 2nd International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design, Park City, Utah, 2003, pp. 92–97. 4. Friedel B., S. Joo, K. Reker, W. Kading, P. Klostermann, K. S. Saternus, and V. Schenider. Test Drivers in the Daimler-Benz Driving Simulator with Drivers Under Diphenhydramine. US-DOT-HS-807-668. NTIS No. PB91-230896. National Highway Traffic Safety Administration, Washington, D.C., 1991. 5. Knodler, M. A., and D. A. Noyce. Tracking Driver Eye Movements at Permissive Left-Turns. Proc., 3rd International Driving Symposium on Human Factors in Driver Assessment, Training and Vehicle Design, Rockport, Maine, 2005, pp. 134–142. 6. Regan, M. A., T. J. Triggs, and S. T. Godley. Simulator-Based Evaluation of the Drivesmart Novice Driver CD-ROM Training Product. Proc., Road Safety Research, Policing and Education Conference. 2000. www.rsconference.com/pdf/RS000060.pdf. Accessed April 19, 2006.

Transportation Research Record 2018

7. Allen, R. W., T. J. Rosenthal, G. Park, M. Cook, D. Fiorentino, and E. Viirre. Experience with a Low Cost, PC Based System for Young Driver Training. Paper No. 614. Systems Technology, Hawthorne, Calif., 2003. 8. Hammel, K., D. L. Fisher, and A. K. Pradhan. Verbal and Spatial Loading Effects on Eye Movements in Driving Simulators: A Comparison to Real World Driving. Proc., Human Factors and Ergonomics Society’s 46th Annual Meeting, Baltimore, Md., 2002, pp. 2174–2178. 9. Victor T. W., J. L. Harbluk, and J. A. Engström. Sensitivity of EyeMovement Measures to In-Vehicle Task Difficulty. Transportation Research Part F, Traffic Psychology and Behavior, Vol. 8, No. 2, 2005, pp. 167–190. 10. Romoser, M., D. L. Fisher, R. Mourant, J. Wachtel, and K. Sizov. The Use of a Driving Simulator to Assess Senior Driver Performance: Increasing Situational Awareness Through Post-Drive One-on-One Advisement. Proc., 3rd International Driving Symposium on Human Factors in Driver Assessment, Training, and Vehicle Design (CD-ROM), 2005, Rockport, Maine. 11. Pollatsek, A., V. Narayanaan, A. Pradhan, and D. L. Fisher. The Use of Eye Movements to Evaluate the Effect of PC-Based Risk Awareness Training on an Advanced Driving Simulator. Human Factors, Vol. 48, 2006, pp. 447–464. 12. Joshua Brown Foundation. Novice Driving Workshop. A. L. Burruss Institute, KSU Center, Kennesaw, Ga., June 28, 2006. 13. National Research Council and Institute of Medicine. Workshop on the Contributions from the Behavioral and Social Sciences in Reducing and Preventing Teen Motor Crashes. National Academy of Sciences, Washington, D.C., May 15–16, 2006. 14. Knodler, M. A., Jr., D. A. Noyce, K. C. Kacir, and C. L. Brehmer. Evaluation of Flashing Yellow Arrow in Traffic Signal Displays with Simultaneous Permissive Indications. In Transportation Research Record: Journal of the Transportation Research Board, No. 1918, Transportation Research Board of the National Academies, Washington, D.C., 2005, pp. 46–55. 15. Winston, F. K., and T. M. Senserrick. Competent Independent Driving as an Archetypal Task of Adolescence. Injury Prevention, Vol. 12, Supplement I, 2006, pp. i1–i3. 16. Northport Associates. Guidelines for Evaluating Driver Education Programs. AAA Foundation for Traffic Safety and BMW, Washington, D.C., 2004. 17. Fisher, D. L., A. Pollatsek, and A. Pradhan. Can Novice Drivers Be Trained to Scan for Information That Will Reduce Their Likelihood of a Crash? Injury Prevention, Vol. 12, Suppl. 1, 2006, pp. i25–i29. 18. Pollatsek, A., D. L. Fisher, and A. K. Pradhan. Identifying and Remediating Failures of Selective Attention in Younger Drivers. Current Directions in Psychological Science, Vol. 15, No. 5, 2006, pp. 255–259. 19. Pradhan, A. K., D. L. Fisher, A. Pollatsek, M. Knodler, and M. Langone. Field Evaluation of a Risk Awareness and Perception Training Program for Younger Drivers. Proc., Human Factors and Ergonomics Society 50th Annual Meeting, San Francisco, Calif., Oct. 2006, pp. 2388–2391. 20. Insurance Institute for Highway Safety. Fatality Facts, Older People. 2004. www.iihs.org/research/fatality_facts/olderpeople.html. Accessed March 20, 2006. 21. McKnight, J. A., and S. A. McKnight. Young Novice Drivers: Careless or Clueless. Accident Analysis and Prevention, Vol. 35, 2003, pp. 921–925. 22. Pradhan, A. K., K. R. Hammel, R. DeRamus, A. Pollatsek, D. A. Noyce, and D. L. Fisher. Using Eye Movements to Evaluate Effects of Driver Age on Risk Perception in a Driving Simulator. Human Factors, Vol. 47, 2005, pp. 840–852. 23. Pradhan, A. K., D. L. Fisher, and A. Pollatsek. Risk Perception Training for Novice Drivers: Evaluating Duration of Effects on a Driving Simulator. Presented at 85th Annual Meeting of the Transportation Research Board, Washington, D.C., 2006. The Simulation and Measurement of Vehicle and Operator Performance Committee sponsored publication of this paper.