PROJECT BACKGROUND, OBJECTIVES, AND SCOPE
Table of Content


The single most pronounced area of difficulty for older drivers, as documented extensively in recent crash analyses, is the approach to and negotiation of intersections; concurrent work has linked such difficulties to a number of diminished functional capabilities known to decline with normative aging. This body of work is exhaustively reviewed in Volume II of this report, and an overview is presented several pages below. Based on our review, gaps in knowledge were identified, and priorities for continuing research in this area were defined. Accordingly, the overall goal of the present project was, first, to objectively derive a comprehensive classification of the types of specific driving errors evidenced at intersections by elderly motorists suffering functional decline; and, second, to advance understanding of the relationships between alternative measures of functional capability and an accepted index of crash risk.


Historically, attempts to directly relate measures of functional capability to motor vehicle crashes, as modeled in Figure 1, have shown associations that are modest at best, typically accounting for well under ten percent of the variability in the criterion measure (crash rates). This includes both sensory (visual) performance measures such as acuity and contrast sensitivity, plus measures of perceptual and cognitive skills including immediate memory span, complex reaction time, discrimination of embedded figures, and an array of additional functional capabilities, using various testing techniques.


The reasons for this failure are many, as reported by Peck (1993) and others. Most importantly: since crashes are rare, most drivers remain crash-free for many years, thus restricting the range for this variable in any analysis; and, crashes are not a direct and inevitable result of unsafe driving behaviors, but are the consequences of interactions between a driver's behavior, situational factors, and the actions of other motorists.


The most successful of the efforts simplistically modeled in Figure 1 has examined the relationship between involvement in selected intersection crash types and measures of attentional and pre-attentional behavior, most notably research addressing the functional or "useful" field of view (UFOV). This body of work has predominantly considered crashes retrospectively, however, and with samples who have been selected specifically on the basis of prior crash involvement. Under these methodological constraints, the crash variance accounted for has been reported to exceed 25 percent (cf. Ball, Owsley, Sloane, Roenker, and Bruni, 1993). In contrast, another related study by the California Department of Motor Vehicles (CA DMV) using 3,669 randomly-selected license renewal applicants showed correlations between UFOV measures and crashes for drivers age 70 and older that were statistically significant, after adjusting for gender, age, and driving exposure, but the percentage of crash variance accounted for fell to just over 4 percent (Hennessy, 1995).

The importance of this work, regardless of specific outcomes, is that few now accept sensory (visual) ability alone as necessary and sufficient for safe driving. Instead, a broader focus incorporating attentional stages of information processing has gained acceptance among researchers and practitioners alike, and appears to hold promise for both screening and diagnostic tests to identify high-risk drivers.


A further evolution of thinking in this area of research has been to broaden criterion measures to focus upon driving competency, apart from the outright occurrence of a crash. This construct-valid approach offers several distinct advantages. First, measures of competency may be developed which are directly observable. Second, the instances of incompetency, manifested as driving errors in a particular performance context with describable physical attributes, level of task demand, degree of familiarity/expectancy for the vehicle operator, etc., occur with a much higher frequency than crashes do. Gebers (1990), in applying a theoretical (Newbold-Cobb) model to 3-year crash rates for the California driving population, calculated that the maximum correlation that could be obtained between an infallible test battery or predictor variable and crash rates was 0.33; this reflects the restriction of range and variability in crash occurrence that were noted above. Using directly observable measures of driver performance deemed to be acceptable surrogates for crash risk (i.e., significantly correlated with crashes), this limiting factor in testing hypothesized consequences on safety of drivers' diminished functional capability is removed.


The next logical step is to determine how (age-related) diminished functional capabilities may predict driving errors, particularly critical errors that a strong consensus among traffic safety experts would characterize as direct antecedents of crashes. A recent study which has followed this paradigm has been reported by Janke and Hersch (1997). As modeled in Figure 2, clear associations between one or more measures of functional ability and driving competency could provide the strongest argument to date that this approach to prediction of crash risk will ultimately be fruitful. At the same time, the identified functional measures would assume priority as candidates for subsequent research studies and pilot programs by licensing agencies.

What remains to be accounted for in the simple models diagrammed thus far is the contribution of "situational factors," as underscored earlier. These factors control when a specific driving behavior--even including some blatantly extreme examples such as crossing the highway centerline--results in a crash instead of a near-miss, or a non-event entirely (e.g., when driving on a road completely free of other traffic, or raised objects). Situational factors thereby mediate the relationship between functional status and the occurrence and criticality of driving errors. Key among these, based upon reviews conducted by members of this research team (Staplin, Lococo, and Sim, 1990, 1993; Staplin, Harkey, Lococo, and Tarawneh, 1997; Staplin, Ball, Park, Decina, Lococo, Gish, and Kotwal, 1997; Staplin, Lococo, McKnight, McKnight, and Odenheimer, in press), are assumed to be (1) the driving task demands, particularly those requiring "effortful" (serial) information processing, and (2) the driver's expectancies--reflecting familiarity or prior knowledge accessible in memory--relating to potential hazard sources encountered and vehicle control actions required along the to-be-traveled route. The relationships hypothesized to be mediated by these variables are modeled in Figure 3.

Together, the variables identified in Figure 3 provided the theoretical framework for this study, and led to a definition of its scope. While the focus on intersection negotiation dictated a field study, the primary objective to develop a classification of driving difficulties suggested the need for an objective means of marking the occurrence of driving errors to complement the traditional approach of examiners' scoring of performance during an on-road test drive. A differentiation of task demands and route familiarity, at least on an ordinal basis, was also desired to explore the influence of these hypothesized mediating factors. In addition, a sample selection strategy which guaranteed that varying aspects of functional decline, including significant degrees of impairment, was crucial to this study. Finally, in this research it was essential to thoroughly document the functional status of all study participants, drawing upon as many reliable alternative measures and assessment techniques as could practically be implemented.


To address these requirements, a study was conducted in cooperation with the CA DMV, using a sample of drivers over the age of 60 who had been referred to the Department for reexamination. A within-subjects research design was applied, calling for two test drives by each subject: one drive on a standard route presumed to be of relatively lower familiarity, common to all study participants; and a second drive over a route of relatively higher familiarity that was unique to each individual, in the immediate area of the person's residence. Field measures of driving competency were obtained, using a modified driver performance evaluation protocol with demonstrated interrater reliability, scored by examiners who were specially-trained in its use and in the testing of older, frail individuals. Complementary, observational data were obtained during test drives through miniature on-board videocameras recording the view of traffic through the windshield, the rear window view, and a view of the driver's head and eyes. These were reduced to yield descriptive logs of driving errors during later analyses. Before completing the first test drive, each subject completed a battery of vision measures and other tests of attentional and perceptual skills developed for this research project.


In the following sections of this document, a brief overview of older driver functional deficits and problems in the use of intersections is presented, the study methodology is described, and data analysis procedures and results are reported, closing with a discussion of the study's findings.



OVERVIEW OF AGE DIFFERENCES AFFECTING SAFE DRIVING AT INTERSECTIONS
Table of Content


One of the principal concerns surrounding older road users is the ability of these persons to safely maneuver through intersections. Hauer (1988) reported that 37 percent of the fatalities and 60 percent of the injuries experienced by older drivers, occur at intersections. For drivers age 80 and older, more than half of fatal crash involvements occur at intersections, compared to 25 percent or less for drivers up to age 45 (Insurance Institute for Highway Safety, 1988). These data reinforce a long-standing recognition that driving situations involving complex speed-distance judgments under time constraints--the typical scenario for intersection operations--are more problematic for older drivers than for their younger counterparts (Waller, House, and Stewart, 1977).


AGE-RELATED FUNCTIONAL DEFICITS
Table of Content


An examination of the characteristics of older road users that affect intersection use reveals that this population differs from their younger counterparts in a number of important ways. This group may experience greater difficulties at intersections as the result of diminished capabilities, which limit both response initiation and movement execution.


The safety and mobility of older road users at intersections are overwhelmingly vision-dependent. Static, geometric features and traffic control devices (TCDs), plus a wide array of dynamic targets, are relevant to drivers and pedestrians at intersections; these must be detected and recognized in a timely fashion to allow for the subsequent cognitive processing preceding response selection and action. Deficits in vision and vision-dependent processes that probably have the greatest impact on older road users at intersections include diminished capabilities in spatial vision, the functional or "useful" field of view (UFOV), and depth and motion perception.


Spatial visual functions, including acuity and contrast sensitivity, are probably the most important functions for detection/recognition of downstream features at intersections. Tests of visual acuity--measuring response to high spatial frequency stimuli at contrast levels far above threshold--show a slow decline, beginning during the forties, which accelerates markedly during the sixties (Richards, 1972). Shinar and Schieber (1991) have argued that dynamic visual acuity--the ability to resolve targets by a moving driver, or moving targets by a standing pedestrian--should correlate more strongly with crash involvement, especially among older individuals. Though the loss of sensory response is greatest for high-frequency (more than 24 cycles/deg) information, older road users' sensitivity to visual contrast at lower and middle-range spatial frequencies (i.e., for 6-, 12-, and 18-cycle/deg targets) also declines steadily with increasing age over 40 (Owsley, Sekuler, and Siemsen, 1983).


Next, the UFOV measure addresses the detection, localization, and identification of targets against complex visual backgrounds, i.e., the earliest stage of visual attention used to quickly capture and direct attention to the most salient events in a driving scene. Most importantly, tests assessing the useful field of view appear to be better predictors of problems in driving than are standard visual field tests. In one study, drivers with restrictions in UFOV had 15 times more intersection crashes than those with normal visual attention (Owsley, Ball, Sloane, Roenker, and Bruni, 1991).


Finally, age differences in the use of visual cues for depth and motion perception deserve emphasis. Researchers have found that the angle of stereopsis (seconds of arc) required for a group of older drivers age 75+ to discriminate depth using a commercial vision tester was roughly twice as large as that needed for an 18- to 55-year-old group to achieve the same level of performance (Staplin, Lococo, and Sim, 1993). Also, it has been shown that older persons require up to twice the rate of movement to perceive that an object's motion-in-depth is approaching, and require significantly longer to perceive that a vehicle is moving closer at a constant speed (Hills, 1975). The Staplin et al. (1993) study investigating causes of older driver over-involvement in turning crashes at intersections, building on the previously reported decline for detection of angular expansion cues, did not find evidence of overestimation of time-to-collision. At the same time, a relative insensitivity to the speed of an approaching vehicle was shown for older versus younger drivers; this result supports the notion that older drivers rely primarily or exclusively on perceived distance to perform gap-acceptance judgments, reflecting a reduced ability to integrate time and distance information with increasing age. Thus, a principal source of risk at intersections is the error of an older, turning driver in judging gaps in front of fast vehicles.


Compounding the varied age-related deficits in visual performance, an overall slowing of mental processes has been postulated as individuals continue to age into their seventies and beyond (Cerella, 1985), and a decline has been demonstrated in a number of specific cognitive activities with high-construct validity in the prediction of driver and pedestrian safety (Stelmach and Nahom, 1992). The cognitive functions included in this processing stage perform attentional, decisional, and response-selection functions crucial to maintaining mobility. Complementary functions essential to the safe and effective use of intersections are selective attention, attention switching, and divided attention, which together comprise the core of what is often termed "situational awareness." Older drivers appear to benefit disproportionately from interventions that compensate for divided attentional deficits during a high-workload task such as negotiating an intersection, for example, cuing drivers with advanced notice of protected versus permissive movement regulations through a redundant upstream posting of advisory signs (Staplin and Fisk, 1991). Related studies suggest that if older drivers must increase their attention to inconspicuous or confusing features to make appropriate maneuver decisions during an intersection approach, a deficit in the discrimination of peripheral targets (e.g., other vehicles or pedestrians) is likely (Brouwer, Ickenroth, Ponds, and Van Wolffelaar, 1990).


Finally, the execution of vehicle control movements by an older driver, or walking movements by an older pedestrian, is likely to be slowed due to a number of factors. A study by Goggin, Stelmach, and Amrhein (1989) linked response slowing by older individuals to abbreviated stimulus exposure times and interstimulus intervals. Also, these researchers have shown that older persons will have greater difficulty in situations where planned actions must be rapidly altered, and corrections during movement execution are slower and much less efficient. The spacing of vehicle control movements required of drivers to negotiate intersections, therefore, may be expected to strongly influence the ability of older individuals to respond in a safe and timely manner; thus, the potential for older driver difficulties at sites which require weaving or successive lane changes within a restricted timeframe increases substantially. In Simon and Pouraghabagher's (1978) study, older adults demonstrated slower reaction times than younger adults when faced with response uncertainty, indicating greater risk when older road users are faced with two or more choices of action. This exacerbates intersection negotiation problems in any situation where older road users are called upon to execute multiple responses in quick succession.


Perhaps most common is the age-related decline in head and neck mobility. Joint flexibility has been estimated to decline by approximately 25 percent in older adults, due to arthritis, calcification of cartilage, and joint deterioration. This restricted range of motion reduces an older driver's ability to effectively scan to the rear and sides of his/her vehicle to observe blind spots, and can also hinder the timely recognition of conflicts during turning and merging maneuvers at intersections (Ostrow, Shaffron, and McPherson, 1992). Reduced neck flexibility also penalizes older pedestrians who must detect potential conflicts without unreasonable delay to accomplish intersection crossings within a protected signal phase.


IDENTIFIED PROBLEMS WITH INTERSECTION USE
Table of Content


Other studies within the large body of evidence showing dramatic increases in intersection crash involvements as driver age increases have revealed detailed patterns of data associating specific crash types and vehicle movements with particular age groups, and in some cases have linked such patterns to the driving task demands in a given maneuver situation (see Campbell, 1993; Council and Zegeer, 1992; Staplin and Lyles, 1991).


Another approach to characterizing older driver problems at intersections was employed by Brainin (1980), who used in-car observations of driving behavior with 17 drivers ages 25 to 44, 81 drivers ages 60 to 69, and 18 drivers age 70 and older, on a standardized test route. The two older age groups showed more difficulty making right and left turns at intersections and negotiating traffic signals. The left-turn problems resulted from a lack of sufficient caution and poor positioning on the road during the turn. Right-turn difficulties were primarily a result of failing to signal. Errors demonstrated at STOP signs included failing to make complete stops, poor vehicle positioning at STOP signs, and jerky and abrupt stops. Errors demonstrated at traffic signals included stops that were either jerky and abrupt, failure to stop when required, and failure to show sufficient caution during the intersection approach.


Complementing crash analyses and observational studies with subjective reports of intersection driving difficulties, a statewide survey of 664 senior drivers by Benekohal, Resende, Shim, Michaels, and Weeks (1992) found that the following activities become more difficult for drivers as they grow older (with proportion of drivers responding in parentheses):


• Reading street signs in town (27 percent).

• Driving across an intersection (21 percent).

• Finding the beginning of a left-turn lane at an intersection (20 percent).

• Making a left turn at an intersection (19 percent).

• Following pavement markings (17 percent).

• Responding to traffic signals (12 percent).

Benekohal et al. (1992) also found that the following highway features become more important to drivers as they age (with proportion of drivers responding in parentheses):


• Lighting at intersections (62 percent).

• Pavement markings at intersections (57 percent).

• Number of left-turn lanes at an intersection (55 percent).

• Width of travel lanes (51 percent).

• Concrete lane guides (raised channelization) for turns at intersections (47 percent).

• Size of traffic signals at intersections (42 percent).


Comparisons of responses from drivers ages 66 to 68 versus those age 77 and older showed that the older group had more difficulty following pavement markings, finding the beginning of the left-turn lane, and driving across intersections. Similarly, the level of difficulty for reading street signs and making left turns at intersections increased with increasing senior driver age. Turning left at intersections was perceived as a complex driving task. This was made more difficult when raised channelization providing visual cues was absent, and only pavement markings designated which were through versus turning lanes ahead. For the oldest age group, pavement markings at intersections were the most important item, followed by the number of left-turn lanes, concrete guides, and intersection lighting. A study of older road users completed in 1996 provides evidence that the single most challenging aspect of intersection negotiation for this group is performing left turns during the permitted (green ball) signal phase (Staplin, Harkey, Lococo, and Tarawneh, 1997).


During focus group discussions conducted by Benekohal et al. (1992), older drivers reported that intersections with too many islands are confusing, that raised curbs that are unpainted are difficult to see, and that textured pavements (rumble strips) are of value as a warning of upcoming raised medians, approaches to (hidden or flashing red) signals, and the roadway edge/shoulder lane boundary. Regarding traffic signals, study subjects indicated a clear preference to turn left on a protected arrow phase, rather than making "permitted phase" turns. When turning during a permitted phase (green ball) signal operation, they reported waiting for a large gap before making a turn, which frustrates following drivers and causes the following drivers to go around them or blow their horns at the older drivers. A general finding here was the need for more time to react.


Additional insight into the problems older drivers experience at intersections was provided by focus group responses from 81 older drivers in the Staplin et al. study (1997). The most commonly reported problems are listed below:


• Difficulty in turning head at skewed (non-90) angles to view intersecting traffic.

• Difficulty in smoothly performing turning movements at tight corners.

• Hitting raised concrete barriers such as channelizing islands in the rain and at night due to poor visibility.

• Finding oneself positioned in the wrong lane--especially a "turn only" lane--during an intersection approach, due to inadequate design or poor visibility (maintenance) of pavement markings or the obstruction of roadside signs informing drivers of intersection traffic patterns.

• Difficulty at the end of an auxiliary (right) turn lane, at channelized intersections, in seeing potential conflicts well and quickly enough to smoothly merge with adjacent-lane traffic.

• Merging with adjacent-lane traffic when a lane drop occurs near the intersection (e.g., when two lanes merge into one lane).


Although these problems are by no means unique to older drivers, the various functional deficits associated with aging appear to exacerbate difficulties for this user group.




RESEARCH METHODOLOGY
Table of Content


This section of the report first presents an overview of the test situation in which data collection was performed, while explaining the project's relationship to a concurrent CA DMV study utilizing a significant number of shared resources, and without which this research would not have been feasible. The subject selection procedures and sample characteristics are described next. The section closes with a comprehensive description of the study variables, performance measures, instrumentation and apparatus, and procedures for data collection and scoring.


OVERVIEW OF TEST SITUATION
Table of Content


This study included a combination of laboratory (office) measures of vision and perceptual skills developed by the Contractor, labeled MultiCAD (Multiple Competency Assessment for Driving), plus on-road drive tests. The MultiCAD test battery was administered by a graduate research assistant working temporarily for CA DMV, and the drive tests were administered by specially-trained examiners employed by CA DMV; in each case, these persons were dedicated to a concurrent project performed under a cooperative agreement with NHTSA (DTNH22-93-Y-05330), directed by Dr. Mary Janke. The test site for both research efforts was the Santa Teresa, CA, DMV facility.


The subjects in this study were all referred to the DMV for reexamination by physicians, family, the law enforcement or judicial system, or other DMV staff. All were compelled to complete the functional test battery and two, on-road tests of driving skill: (1) a standard test route, common to all subjects, that began and ended at the Santa Teresa DMV facility; and (2) a route developed in the area near each subject's home, including travel to frequently visited destinations. The two routes, which represent a contrast in the level of familiarity and expectancy for vehicle control requirements during the test drives, both required roughly equal travel times, but did not necessarily include intersections with closely-matched characteristics or traffic conditions.


All subjects first completed the MultiCAD test battery, then the standard drive test, then, on a following day, the home area drive test. The examiners conducting the drive tests could discontinue and conclude testing at any time, during either on-road test, due to safety concerns.


STUDY VARIABLES AND PROCEDURES
Table of Content


The independent (predictor) variables examined in this study included a set of functional status indicators describing age-related changes in vision and perceptual skills, plus contrasting (relatively lower versus higher) levels of familiarity of subjects with the test routes on which their driving competency was evaluated. The dependent (criterion) variables included a weighted error score derived from the structured observations of trained DMV examiners during the test drives, plus probabilities of occurrence of specific behavioral errors derived from videotapes recorded in the subjects' own vehicles as they drove over each test route.


The description of the MultiCAD battery of functional tests is presented first below. For convenience, this discussion integrates the test contents and test scoring, data collection apparatus, and protocol for administration of the test battery. Similarly, the following discussion of the drive test procedures includes a description of test route development, test protocol, and scoring by the DMV examiners. Finally, the equipment and procedures used to obtain videotaped observations of driver behaviors and views of the road and traffic conditions on each test route are described.


Measuring the Functional Status of the Study Sample


A PC-based tabletop testing system was delivered to the CA DMV by the Contractor and was used to conduct limited functional assessments of all 82 subjects. The test battery used a combination of video clips of driving scenes and computer-generated images to maintain a high level of face validity for everyday driving situations. It was constructed to assess a restricted set of candidate "minimum qualifications thresholds," rather than to obtain a precise psychophysical measurement of capability level. The battery of nondriving tests in this research was designed to measure a variety of competencies deemed critical to safe vehicle operation, based on the review of previous research findings conducted earlier in the project.


The MultiCAD protocol displays dynamic, suburban arterial driving scenes on a 27-inch screen capable of accepting both video (NTSC TV standard) and computer graphics (SVGA) inputs. As described below, these measures required subjects to respond to the test scenes using a hand-held, 3-button response pad in a 3-alternative, forced-choice paradigm, to identify stimulus features as requested in a series of audio/visual instructions. A brake and accelerator pedal assembly was used for stop-and-go decisions, and brake reaction measures.


Static Acuity. This test used MultiCAD to measure drivers' ability to resolve fine detail on a stationary target under high contrast conditions. The subject was shown a driver's eye view of travel along a suburban arterial, approaching and then stopping at an intersection with a traffic signal in plain view. The image centered and then zoomed on the signal until it filled the screen, while the subject was instructed to use the 3-button response pad to identify which face on the (conventional, 3-face) signal looked different than the other two. Instead of solid red, yellow, and green circles, however, the signal faces contained acuity test stimuli. Square wave gratings with vertical bars were used, such that one signal face

contained a high contrast test stimulus (90 percent contrast) and the other two faces showed a uniform luminance (without bars). The ability to discriminate which two signal faces were "blank" versus which one contained the vertical bars defined the subject's static acuity level. An example acuity stimulus target is shown in Figure 4. Three levels of testing were conducted--20/40 (15 cycles per degree), 20/80 (7.5 cycles per degree), and 20/200 (3 cycles per degree)--with a pass/fail score assigned at each level. A passing score was defined as at least two correct responses out of the three presentations for each level tested. Mean response time was also calculated for correct responses at each level. Three replications of each measurement were performed.


Dynamic Acuity. This test used MultiCAD to measure drivers' visual acuity, for a target that was moving relative to the observer, under high contrast conditions. The same type of stimuli as described for MultiCAD static acuity were shown again while moving at a predetermined rate from one side of the screen to the other. The same "which signal is different?" discrimination was required of the subject, using the 3-button response pad. The rate of movement across the screen (12 degrees per second) corresponded to a driver trying to read a street sign posted at roadside while passing by at a moderate (40 to 64 km/h [25 to 40 mi/h]) rate of speed. Three replications of each measurement were performed--20/40, 20/80, and 20/200--with a pass/fail score assigned at each level. A passing score was defined as at least two correct responses out of the three presentations for each level tested. Mean response time was also calculated for correct responses at each level.


Static Contrast Sensitivity. This test used MultiCAD to measure drivers' sensitivity to differences in brightness, as required to detect edges between adjacent lighter and darker areas in the roadway environment. The subject was asked to use the 3-button response pad to indicate which of three signal faces contained a test pattern. The traffic signal image remained stationary during this test. The test patterns were the same as used for the static acuity test for 20/40 (15 cycles per degree) and 20/80 (7.5 cycles per degree), and were presented at 2 contrast levels (high contrast=20.6 percent; low contrast = 4.9 percent). Three replications of each measurement were performed with a pass/fail score assigned at each level. A passing score was defined as at least two correct responses out of the three presentations for each level tested. Mean response time was also calculated for correct responses at each level.


Dynamic Contrast Sensitivity. This test used MultiCAD to measure drivers' contrast sensitivity for a target that is moving relative to the observer. Immediately following the MultiCAD static contrast sensitivity test, exactly the same type of stimuli were shown while moving at a predetermined rate (12 degrees per second) from one side of the screen to the other. The same "which signal face is different?" discrimination was required of the subject, using the 3-button response pad. Three replications of each measurement were performed with a pass/fail score assigned at each level. A passing score was defined as at least two correct responses out of the three presentations for each level tested. Mean response time was also calculated for correct responses at each level.


Angular Motion Sensitivity. This test used MultiCAD to measure drivers' ability to rapidly detect changes in the relative motion of their own versus other vehicles. A video of suburban driving scenes was used which presented a driver's eye view of travel along an arterial route with light traffic, following a lead vehicle (that the subject was told to pay attention to) at varying distances. Subjects were required to depress the brake in the MultiCAD assembly whenever the vehicle directly ahead in the same lane applied its brakes or at any other time it would be advisable to stop or slow down under actual driving conditions (e.g., an adjacent-lane driver encroaches into the lane of travel). The lead vehicle brake lights were illuminated when it slowed for 12 of the angular motion sensitivity trials. For three other angular motion sensitivity trials, the lead vehicle's brake lights were disabled during filming of the video, so that the subject was required to detect the change in headway without the additional brake light cue. These three trials were intermixed with the trials in which the brake lights were illuminated.


Measures of effectiveness were: (1) mean brake reaction time across 12 trials, to slowing/stopping lead vehicle with brake light activation, for correct responses; (2) percent error for these trials (e.g., percent of the trials where the vehicle ahead slowed and the brake lights were clearly visible, but the subject did not press the brake pedal); (3) mean brake reaction time across three trials, to slowing/stopping lead vehicle with no brake light activation, for correct responses; and (4) percent error for these three trials.


Useful (Functional) Field of View. This divided attention test used MultiCAD to measure drivers' ability to remain vigilant and respond in a timely and appropriate manner to events that occurred directly ahead, in the travel path, while also detecting unexpected events of a safety-critical nature that occur in the areas of peripheral vision. After angular motion sensitivity data were obtained, the same driving video continued to use the lead vehicle target as a "foveal task" (i.e., located centrally along the driver's line of sight). At predetermined intervals in relation to a (lead vehicle) brake light stimulus, vehicles and pedestrians were introduced unexpectedly in the periphery of the driver's forward vision, offset at angles of approximately 15 degrees and 30 degrees to the left and right sides. The motion of these peripheral targets brought them into potential conflict with the driver within several seconds' travel time.


For threats intersecting from the periphery at approximately a 15-degree angle of eccentricity (2 trials), the measures of effectiveness were: (1) mean reaction time for correct response to (a) a vehicle pulling out from behind a building on the right side of the scene and (b) a vehicle backing out of a parking space from behind a (blocking) U-Haul van on the left side of the scene; and (2) percent error for these two trials.


For threats intersecting from the periphery at approximately a 30-degree angle of eccentricity (1 trial), the measures of effectiveness were: (1) mean reaction time for correct response to a pedestrian stepping off the curb and entering the driver's path; and (2) percent error.


In addition to the tests described above, a manual measure of neck flexibility was obtained using a goniometer to measure degrees of head/neck rotation to the left and to the right.


Data collection using the MultiCAD functional test driving simulator proceeded in the following manner. After an initial greeting, the participant was escorted to the simulator and asked to sit in the simulator chair ("driver's seat"). The test administrator obtained demographic information, and then measured the subject's neck flexibility. A goniometer attached to the back of the driver's seat was lowered to a position slightly above the participant's head. The participant was told to look straight ahead and the goniometer was set at 0 degrees. The participant was then asked to turn his/her head left to the maximum point where there was no discomfort and then to the right with the same instructions. The maximum head turning angles were recorded. The goniometer was then taken out of the back of the driver's seat.


The driver's seat was adjusted horizontally so that the individual could easily reach the accelerator and brake pedals. The seat was adjusted vertically so that the individual's eye height was at the mid-point of the monitor. A 30 inch eye-to-screen distance was set by moving the monitor assembly along a track.


The MultiCAD battery contains multimedia (audio and visual) instructions, presented on-screen through pre-recorded video of a "talking head" (a staff member employed by the Contractor). This allowed for identical delivery of instructions to each participant. The test administrator informed the participant to follow along with the instructions presented to him/her on the monitor by the moderator from the video program. The test administrator then initialized the video program. Before the visual and perceptual tests were presented, the "talking head" requested the participant to press each of the response switches (3-button response pad, accelerator pedal, brake pedal) to ensure that responses were being recorded by the data collection computer. The correct operation of these response devices was confirmed by the test system before the automated data collection protocol was allowed to continue.


Measuring Driving Competency on Less Familiar and More Familiar Test Routes


Standard Route Drive Test. As noted earlier, the CA DMV Field Operations Division office in the Santa Teresa area of southern San Jose was used as the test site. The driving exam for the standard (low familiarity) route began and ended at this DMV office.


A reconnaissance of the surrounding area within a 0.5-hour radius from the DMV was performed to identify potential intersection locations suitable for inclusion in the standard route driving exam. The intersection types sought were those providing examples of high demand intersection geometric and operational situations identified in an intersection negotiation task analysis performed earlier in the project (see Volume II of this report).


A preliminary test route was then developed that incorporated as many of the high-criticality intersection maneuver/geometry/operation types as possible. A second traversal was conducted to refine the test route to accommodate the requirements of DMV driver exam testing (i.e., route must be no more than 0.5 hours from start to finish, and must include all DMV driver exam scoring situations). In addition to the maneuvers performed in the parking lot, the DMV requires that the driver performance examination include:


• Four left turns.

• Four right turns.

• Eight non-turn (through) intersections (stop-controlled, signal-controlled, and uncontrolled).

• Left lane change.

• Right lane change.


These stipulations limited the ability for the study to include all of the intersection maneuver/operational situations originally planned. As the route was finalized, site characteristics essential for later data reduction and analysis purposes were noted.


Table 1 identifies high criticality intersection types, selected on the basis of the maneuver requirements (e.g., right turn, left turn, straight through), geometry and operational factors (e.g., presence of left-turn bays and auxiliary right-turn lanes, number of through lanes), and traffic control elements (e.g., traffic signal, stop sign, pavement markings). It should be noted that this list is not exhaustive; other intersection types (such as T-intersections) and maneuvers (lane changes, parking, backing) occurred along the standard exam route, which are not listed here.



Table 1. Characteristics of high-criticality intersections included on the standard test route.
Table of Content

Maneuver Requirement Intersection Type* Traffic Control
Right Turn

(on green or red after stop)

4-lane by 4-lane with opposing dual left-turn lanes (Type 13)


•2 left-turn only lanes

•1 through lane

•1 right-turn only lane


•pedestrian refuge island between right-turn only lane and through lane.

•3 vertical 3-lens signals with solid red, solid yellow, solid green balls

•2 vertical 3-lens signals, containing solid red ball, solid yellow ball, and left-turn green arrows

•2 pedestrian signals with "hand" for Don't Walk and "walking person" for Walk.

•Turn-only lanes (2 left and 1 right) marked with "ONLY" pavement markings

Through

(on steady green ball)

Divided 4-lane by 4-lane with opposing left-turn lanes (Type 14)


•1 channelized left-turn bay

•1 right-turn bay

•2 through lanes


•pedestrian refuge island between right-turn only lane and through lane


•3 vertical 3-lens signals with solid red, solid yellow, solid green balls

•2 vertical 3-lens signals, containing solid red ball, solid yellow ball, and left-turn green arrows

•2 pedestrian signals with "hand" for Don't Walk and "walking person" for Walk.

•Turn-only lanes (1 left and 1 right) marked with "ONLY" pavement markings

* See Volume II for definition and schematic drawing of each intersection Type.
Left Turn

(on steady green ball)

2-lane by 2-lane with opposing dual left-turn lanes (Type 7)


•1 left-turn lane

•1 combined through and right-turn lane

•4 vertical 3-lens signals containing solid red, solid yellow, and solid green balls

•2 pedestrian signals with "hand" for Don't Walk and "walking person" for Walk.

•Left-turn only arrow pavement marking in left-turn lane; no markings in right/through lane

Right Turn

(on steady green ball or red ball after stop)

Divided 4-lane by 4-lane with opposing left-turn lanes (Type 14)


•1 left-turn lane

•1 right-turn lane

•2 through lanes


•There is a bike lane and the right-turn lane is divided by an island

•4 vertical 3-lens signals containing solid red, solid yellow, and solid green balls

•2 vertical 3-lens signals containing red arrow, solid green ball, solid yellow ball

•Left-turn only arrow pavement marking in left-turn lane

Through 2-lane by 2-lane with no auxiliary lanes (Type 4) •4-way stop signs
Left Turn 2-lane by 2-lane with opposing left-turn lanes (Type 7) •4-way stop signs

•Left-turn only pavement marking

Left Turn

(on steady green ball)

2-lane by 2-lane with right turn lane (Type 5)


•1 shared left/through lane (unmarked)


•1 right-turn only lane (marked)

•3 vertical 3-lens signals containing solid red, solid yellow, and solid green balls

•2 pedestrian signals with "hand" for Don't Walk and "walking person" for Walk.

• Right turn only pavement marking

Through

(on steady green ball)

4-lane by 4-lane with opposing left-turn lanes (Type 14)


•1 left-turn bay

•3 through lanes

•3 vertical 3-lens signals containing solid red, solid yellow, and solid green balls

•1 vertical 3-lens signal with left-turn green arrow, left-turn yellow arrow, solid red ball

* See Volume II for definition and schematic drawing of each intersection Type.
Right Turn

(on steady green ball or red ball after stop)

2-lane by 2-lane on driver's side (Type 4)

2-lane by 2-lane with opposing left-turn lane on opposite side (Type 7)


•no lane lines were marked on pavement

•4 vertical 3-lens signals containing solid red, solid yellow, and solid green balls

•2 pedestrian signals with "hand" for Don't Walk and "walking person" for Walk.

•Sight distance to left is blind; applicant must creep forward if light is red, to see to the left.

Left Turn

(green arrow)

4-lane by 4-lane with opposing left- turn lanes (Type 14)


•1 left-turn bay

•1 channelized right-turn lane

•3 through lanes

•4 vertical 3-lens signals containing solid red, solid yellow, and solid green balls

•2 vertical 3-lens signals, with leading left-turn green arrow, solid yellow ball, solid red ball


Left-turn only pavement marking

* See Volume II for definition and schematic drawing of each intersection Type.


The examination protocol used in this study was, of necessity, the same as for the concurrent project (DTNH22-93-Y-5330). This protocol was a modification of the Driver Performance Examination (DPE), which California plans to adopt as its standard road testing instrument for both novice drivers and some experienced, but functionally impaired drivers. It employs a fixed number of maneuvers that are scored at specific locations ("structured maneuvers"), resulting in a fixed number of possible errors and objective scoring criteria. The road test was conducted by two DMV examiners who have extensive training in the effects of functional impairments on driving, and much experience in administering special drive tests in California.


During the on-road exam, a test examiner sat beside the subject driver and used a score sheet to indicate whether the structured maneuvers at predesignated points on the route were performed unsatisfactorily (e.g., inadequate traffic check, poor lane position). If so, a driving error was recorded. In addition to the simple occurrence of a driving error, two additional categories of errors could also be denoted: critical errors and hazardous errors. These are defined as follows:


• Critical Errors -- Errors which would in normal circumstances cause test termination. These included: driver strikes object; drives up/over curb/sidewalk; drives in oncoming traffic lane; disobeys sign/signal; inappropriate reaction to school bus; inappropriate reaction to emergency vehicle; inappropriate speed; inappropriate auxiliary equipment use; and turn from improper lane.


• Hazardous Errors -- A subset of critical errors, such that any critical error which is judged to involve a dangerous maneuver or that required examiner intervention, is also termed hazardous.


As noted above, the number of possible errors was fixed, with the exception of critical driving errors, which were recorded as they occurred. Hazardous errors are a subset of critical errors, and critical errors a subset of total errors.


In the analyses performed by Janke and Hersch (1997), a weighted error score was calculated as the total number of errors committed during a subject's drive test, plus twice the sum of critical and hazardous errors. This weighting scheme is discussed further in the next section of the report.


Home Area Drive Test. The second on-road driving performance examination was conducted in a subject's home area, usually the day following the standard route drive test. Unlike the standard exam, the home area test route was not pre-planned; structured maneuvers could not be assigned to specific points on the route. However, other than arranging to meet at the subject's home or at some other location convenient to subjects, protocols and procedures for the second drive test paralled those implemented during the first drive test. A count of total errors, critical errors, and hazardous errors was obtained from examiner scoresheets for the home area drive test in a manner identical to the standard route exam, and the home area test was conducted by the same DMV examiners.


Obtaining Videotaped Observations of Driver Behavior and Situational Factors


Drivers' performance on both on-road exams was videotaped to obtain more precise information about the subjects' errors in two broad categories: maneuver errors and errors of observation. A video recording that documents where a subject looked, and the roadway view in front of and behind the subject's vehicle, was selected to meet this objective data requirement.


The videotape data collection system was comprised of the following components: 3 mini-cameras; a portable mini-monitor; 3 12-volt, AC compatible video cassette recorders (VCRs) attached together with a bracket; power supply (car battery); photographic flash attachment; and an accelerometer with display unit. The system was designed to be portable and able to be installed in any vehicle within 10 minutes.


Subjects performed the on-road exams after completion of the MultiCAD battery. Each participant was asked to park his/her vehicle near the back doorway of the DMV building, to permit installation of the video data collection system. This was performed by the same research assistant who had administered the functional test battery.


The mini-cameras covered three fields of view: (1) the forward roadway view; (2) a view of the driver's face; and (3) a view out of the rear window of following traffic and of an accelerometer placed on the rear dashboard. The forward roadway view mini-camera was clamped onto the passenger-side visor and pointed out toward the center of the road. Camera wires were concealed above the visor and were run across the top of the visor to the floor of the passenger side and then to the rear floor of the vehicle, where they were connected to a VCR unit and the power supply. The driver head/eye view mini-camera was wedged into the top of the dashboard at the bottom of the windshield toward the left side of the steering wheel. The mini-camera pointed toward the subject's face. Camera wires were wedged along the edge of the bottom of the windshield across the front of the vehicle to the passenger side and then to the rear floor where they were connected to another VCR unit and the power supply. The third mini-camera was set up in the rear center of the subject's vehicle. This camera was clamped to a metal rod connected to the accelerometer, which was positioned along the back compartment of the rear seats. The camera pointed out the back window, viewing the accelerometer display and the road behind the subject's vehicle. Camera wires were connected to the VCR unit and the power supply. Accessories and additional clamps were available to accommodate affixing the mini-cameras to a variety of dashboards, visors, and compartments behind the rear seats. Prior to each drive test, proper aiming of the cameras was confirmed using a mini-monitor that was easily connected and disconnected to each camera.


The VCRs were positioned in the rear floor of the participant's vehicle. The VCRs were powered by a 12-volt auto battery, which was recharged each night after use. VHS 60-min videotapes were used with each VCR. These tapes were loaded into the VCR units before each run. Each participant required three tapes to record performance on the standard exam and three tapes to record performance on the home area exam on the following day, for a total of six videotapes per subject.


Once the mini-cameras and VCRs were installed, powered on, and had started recording, all three cameras were aimed at a photographic flash attachment and the flash attachment was triggered. The flash provided a discreet moment visible to all cameras that could be used to synchronize data from all tapes. This procedure was necessary to enable data reduction and coding of events. Once this procedure was completed, all cameras were aimed using the mini-monitor, all wires were concealed, and the VCRs and power supply were covered with a black dropcloth.


At that point, the study participant and the CA DMV driver examiner were requested to enter the participant's vehicle to begin the standard on-road exam. The data collector informed the participant that the equipment set up in the vehicle was used to monitor the view of the road. Then, the driver examiner explained where he wanted the participant to drive. Once the road exam was completed the videotape data collection system was disassembled and removed from the vehicle. The home area driving performance examination was then scheduled, usually for the following day. The same procedure was followed for setting up the videotape equipment in the subject's vehicle for the home area drive test.


After each on-road exam was completed by each subject, the tapes were rewound, then were labeled (by view) and mailed to the project staff who would subsequently reduce and code the observational data for planned analyses in this study.


CHARACTERISTICS AND FUNCTIONAL STATUS OF TEST SAMPLE
Table of Content


Test Sample Characteristics


Participants were selected from individuals who were referred to the CA DMV Driver Safety Office for reexamination because of a medical condition, a series of license failures, a flagrant driving error, or some other indicator of driving-related problems. CA DMV staff further screened individuals to include only those who were over the age of 60, with English literacy. These individuals were then scheduled for their driving examinations.


The test sample in this study included 82 individuals (54 males and 28 females) over the age of 60. The age range was 61 to 92; the mean age was 77 and the median was 78. Ninety-six percent of the participants were 65 years of age or older. Sixty-five percent of the group was 75 years of age and older. The study sample distribution by gender, and by five-year age groups (61-65, 66-70, 71-75, 76-80, 81-85, and 86+) is shown in Figure 5.



The average number of years of driving reported by the group was approximately 56 years and the average number of miles driven annually was 6,150. The participants reported that they drove most of the time (89 percent) in daylight hours; and that most of their driving (76 percent) was on local (non-freeway) roads.


None of the participants, when asked, mentioned consuming any alcoholic beverages within the 4 hours preceding their examinations. However, 56 percent of the participants had taken prescription medication within the past 8 hours.

Participants were referred to the CA DMV Driver Safety Office for several reasons. Police referral was the most common reason (24 percent of the group). The police had either investigated an crash or stopped the driver for a violation or some other erratic maneuver and decided (based on observation or noticing a physical impairment) to send a referral. The next most common reasons for referral were related to the person recently having a stroke (20 percent of the group); the person showing some form of dementia (20 percent of the group); or the person having a vision problem (18 percent of the group). In these cases, physicians, family members, or driver examiner referrals were made. Other conditions of the participants also provided reasons for referral, including neurological disorders, musculoskeletal conditions, and endocrine-related disorders (e.g., diabetes mellitus, hypoglycemia, hyperthyroidism).


Functional Status of Test Sample


In this section, the performance capabilities of two older driver populations are summarized. Results are presented for the group of 29 subjects ages 61 to 74 (termed "young-old") and for the group of 53 subjects age 75 and older (termed "old-old"). This division was chosen in light of the fact that crash-involvement and fatality rates increase sharply per mile driven, as drivers reach age 75. Supporting graphics are presented in Appendix A, in Figures A-1 through A-11.


Static Acuity. The sample's static acuity capabilities are presented in Figures A-1 and A-2 for each of the three levels of test stimuli that were presented--20/40, 20/80, and 20/200. Figure A-1 presents the percent of the sample which meets the indicated performance level for young-old vs. old-old drivers, while Figure A-2 presents the mean response times for correct responses for these age groups. These data indicate, surprisingly, that fewer subjects correctly discriminated the 20/200 targets than the 20/40 or 20/80 targets. However, this is interpreted as a practice effect: static acuity measurement was the initial test procedure, and as in conventional vision test protocols, the lowest resolution targets (20/200) were always presented first, when subjects' lack of familiarity with the novel MultiCAD procedures would be most likely to impact their performance. This interpretation is supported by the latency data for correct responses, which show a consistent decrease in response time for the lower resolution versus the higher resolution stimuli. There were no clear trends in the data presented in Figures A-1 and A-2 as a function of subject age group, except for an increase in response time at all acuity levels for the old-old subjects.


Dynamic Acuity. The sample's dynamic acuity capabilities are documented in Figures A-3 and A-4. For each of the three levels of test stimuli that were presented--20/40, 20/80, and 20/200--the following results are summarized: percent of sample which meets the indicated performance level, for young-old vs. old-old drivers (Figure A-3); and mean response time for correct responses for these age groups (Figure A-4). A sharp drop in performance was noted for the 20/40 versus both the 20/80 and 20/200 stimuli, for both age groups. Slower response times were noted for the old-old subjects, for two of the three levels of target resolution (20/80 and 20/200).


Static Contrast Sensitivity. The sample's static contrast sensitivity capabilities are presented in Figures A-5 and A-6. For each of the four levels of test stimuli that were presented--two spatial frequency levels (15 and 7.5 cycles/degree), each at low and high contrast--the following results are summarized: percent of sample which meets the indicated performance level, for young-old vs. old-old drivers (Figure A-5); and mean response time for correct responses, for these age groups (Figure A-6). These data show clearly superior contrast sensitivity performance by the study sample as a function of decreasing target resolution and increasing target contrast, as expected. The effect of age was most apparent in the performance deficits manifested by the old-old subjects, specifically for the low contrast, low resolution targets; at a high contrast level, performance deteriorated somewhat with increasing age for the high resolution targets. In terms of response time, subjects required an average of 1.5 seconds longer to respond to the low contrast, high resolution targets than to the high contrast, low resolution targets. The old-old subjects were slower to respond to all targets than the young-old subjects, with the greatest disparity in performance for the high resolution, high contrast targets.


Dynamic Contrast Sensitivity. The sample's dynamic contrast sensitivity capabilities are presented in Figures A-7 and A-8. For each of the four levels of test stimuli that were presented--two spatial frequency levels (15 and 7.5 cycles/degree), each at low and high contrast--the following results are summarized: percent of sample which meets the indicated performance level, for young-old vs. old-old drivers (Figure A-7); and mean response time for correct responses for these age groups (Figure A-8). These data show trends quite similar to the static target presentations, with the poorest performance at high spatial frequency and low contrast levels. Generally, a lower percentage of subjects in the old-old group were able to resolve the targets compared to subjects in the young-old group, and when they could, they were longer to respond.


Neck Flexibility. The sample's neck flexibility, expressed in terms of degrees rotation of the head to the left and right, is presented in Figure A-9. Slightly greater flexibility was demonstrated for a head turn to the right than for a turn to the left, and for young-old subjects compared to old-old subjects.


Angular Motion Sensitivity and Useful (Functional) Field of View. Figures A-10 and A-11 document the error rates and brake reaction times, respectively, for the study sample in response to central and peripheral visual targets representing potential threats or conflicts. As a reminder, four stimulus types were presented: (1) lead vehicles slowing with brake lights activated; (2) lead vehicles slowing without brake lights activated; (3) vehicles at 15-degree offset moving on intersecting (90) path; and (4) pedestrian at 30-degree offset moving on intersecting (90) path. For approximately 20 to 40 percent of the events where a brake response would have been appropriate in the MultiCAD driving video, subjects failed to respond. Generally, the old-old subjects failed to respond to potential threats more frequently than the young-old subjects, especially in the case of a lead vehicle braking without brake lights activated. However, young-old subjects demonstrated almost twice the error rate of the old-old subjects, when the target was a pedestrian at 30 degrees of eccentricity. Looking at response times for correct responses, there was very little difference as a function of age group. The longest latencies were shown for a lead vehicle ahead that stopped or slowed without its brake lights activated. Subjects responded close to 5 seconds after the lead vehicle began to decelerate, for this particular set of driving scenarios.

continuation