INTRODUCTION
Table of Content

A central objective at the outset of this project was to identify and prioritize--in terms of likely consequences for safe driving performance at intersections--the age-related decrements in functional capability that can be linked through empirical evidence or logical inference to increased crash risk. This objective was addressed through a survey of selected research topics and consultations with experts in the field. The research topics of principal interest were: (1) age differences in functional capabilities of potential importance to the safe negotiation of intersections; (2) the relative overinvolvement of older drivers in specific crash types at intersections, and specific unsafe behaviors identified as causal or contributing factors; and (3) the critical driving task demands for identified maneuvers at designated intersection types, with assignments of relative (increased) risk for older drivers based upon mismatches between situational demands and drivers' (diminished) response capabilities. A table summarizing the expert consultations in this project activity is included at the end of this section.

To guide this effort, a conceptual framework was developed expressing the hypothesized relationship between intersection crash risk and age differences in functional capabilities. This framework, as diagrammed in Figure 1, assumes that the aging-crash risk relationship is mediated through unsafe driving behaviors which, by implication, can be reliably observed and quantified as a basis of comparison across individuals and across situations.

Figure 1. Hypothesized relationship between intersection crash risk and age differences in functional capabilities, mediated through unsafe driving behaviors.

Age-related functional changes of interest in the present review include those indicated both by "out-of-context" measures obtained in clinical/laboratory assessments, and by measures obtained using simulation or under closed-course or real-world driving conditions. Materials reviewed were selected from the burgeoning catalogue of (cross-sectional) studies on aging, on the basis of the presumed criticality of the functional deficits described therein to safe and effective driving performance at intersections. Such presumptions are underscored in this discussion through frequent references to specific task demands for intersection negotiation. The diagnosis, symptoms and consequences for safe driving performance of dementing illness, which afflicts a disproportionate number of older persons, are also addressed.

Prior analyses of intersection crashes were reviewed to document age differences in relative involvement rates for particular types of crashes. This information is important because it can lead directly to a specification of the vehicle maneuvers for which older drivers most often experience performance failures at intersections. For the various maneuvers, requirements for drivers' behavioral responses may then be identified, suggesting surrogates for crash risk which include, for example, (1) missed or slowed responses by drivers in the detection and/or avoidance of roadway hazards or conflicts with other road users, and (b) the violation of, or disregard for, traffic control devices or the expectations of other road users.

Building upon the review of literature and intersection crash analyses, a modified task analysis examined the critical driving task demands for specific intersection types under specific operating conditions, highlighting mismatches between situational demands and drivers' behavioral responses likely to result due to age-related diminished capabilities. Since the literature contains a dearth of controlled, prospective studies directly linking age-related diminished capabilities to differential crash involvements, and thus to increased crash risk at intersections, this task analysis inferred probable safety outcomes according to the following logic: where a specific functional decrement negatively influences the speed or accuracy of performance of a critical driving task, the safe negotiation of intersections whose features are associated with particular performance demands will be compromised, resulting in a higher incidence of unsafe driving behaviors and increased crash risk. As a result of this effort, problems in the negotiation of intersections by older drivers were prioritized according to their likelihood of occurrence and their consequences for traffic safety.

The consultations with experts contributing to this background synthesis solicited input in the form of published and unpublished research reports and personal communications. These were integrated, where appropriate, into each discussion topic addressed in the review. These contributions are acknowledged and summarized in Table 1.

Table 1. Status and content of expert consultations.

AGE DIFFERENCES IN FUNCTIONAL CAPABILITIES
Table of Content

Older persons disproportionately manifest a variety of measurable functional deficits with high construct validity as predictors of driving difficulty and therefore, presumably, of crash risk. While the empirical validity of such age differences as crash predictors remains at issue, a number of retrospective studies have established significant correlations between functional deficits and crash involvement. The following discussion explores the nature of deficits in sensory/perceptual, cognitive, and physical functions which appear with increasing frequency--whether the result of normative aging, trauma, disease, or dementia--among older persons, and the potential effects of such deficits on driving performance at intersections. In addition, a limited set of studies has been sampled to report older drivers' self-perceptions regarding the problems they experience as a result of diminished functional capabilities.

Briefly, the relationships between functional capabilities of older drivers and intersection negotiation likely to be of greatest operational significance can be summed up as follows. Age-related declines in spatial vision, including acuity (both static and dynamic) plus low- and mid-, as well as high-frequency spatial contrast sensitivity, will delay recognition of intersection features such as pavement width transitions, channelized turning lanes, island and median features across the intersection, and any nonreflectorized raised elements, and will delay comprehension of the information provided by pavement markings and traffic signs. This information loss in the early stages of the driver's vehicle control task will be compounded by attentional and decision-making deficits shown to increase with increasing age, with age differences in performance magnified as serial processing demands for conflict avoidance and compliance with traffic control messages increase during the intersection approach. Age-related decrements in the "useful field of view," selective attention, and divided attention/attention switching capabilities will slow the initiation of a driver's response when a lane change or other change of heading is required, either for hazard avoidance or to accomplish a desired intersection maneuver. In addition, less efficient working memory processes will translate into riskier operations for older drivers at intersections with increasing geometric complexity, and/or intersections in unfamiliar areas where concurrent search for and recognition of navigational cues disproportionately taxes "spare capacity" for lane-keeping and conflict avoidance. For turning drivers, an age-related diminished capability in judging the "least safe gap" ahead of oncoming vehicles may lead to inappropriate maneuvers. Finally, the execution of vehicle turning movements becomes more difficult for older drivers as bone and muscle mass decrease, joint flexibility is lost, and range of motion diminishes. Simple reaction time, while not significantly slower for older drivers responding to expected stimuli under nominal operating conditions, suffers operationally significant decrements with each additional response to an unexpected stimulus, i.e., as required in emergency situations.

To respond appropriately to all manner of stimuli in the roadway environment, a driver must first detect and recognize physical features of the roadway, traffic control devices, other vehicles, pedestrians, and a wide variety of other objects and potential hazards of a static and dynamic nature. On rare occasions, critical information concerning the presence or position of traffic is conveyed to a road user solely through an auditory signal; in the vast majority of cases, however, the visual system is preeminent at this (input) stage of processing.

It should be emphasized at this point that the classification of older individuals as visually impaired, from a human factors perspective, depends very much on the context of expected performance. For example, many individuals who are seriously affected by presbyopia (farsightedness), to the point of not being able to read without strong corrective lenses, may be relatively unaffected in viewing objects at a distance while driving. In another example, glare from ocular media scatter may pose serious problems at night, in rain or in bright sunlight but presents little difficulty on mildly overcast days. Nevertheless, a very high proportion of drivers age 60 and older will show a serious limitation in visual performance under at least some typical driving conditions. If even as small a group as the lowest-scoring-25 percent on critical functional tests are considered impaired, this number added to those experiencing clinical pathology gives an estimate of 1 in every 3 drivers over age 60 as potentially seriously impaired (Staplin, Breton, Haimo, Farber, and Byrnes, 1986).

The visual sensory input system is a complex biological composite of optical and neural components, including the cornea, aqueous humor, iris, lens, vitreous body, and retina. All of these elements change with age in ways that interact with each other and cause deterioration of visual performance. However, the largest single factor contributing to declining visual performance in the non-pathologic eye is increased light absorption and scattering in the crystalline lens. A distant second as a contributing factor is deterioration in the structures of the retina and neural pathway. All other factors, apart from pathology, may be grouped as minor in overall impact (Staplin, Breton, Haimo, Farber, and Byrnes, 1986). Physiological changes with aging in the lens and retina, and the most significant age-related visual pathologies, are summarized below before a detailed consideration of performance measures used to assess functional decline in the visual system.

The lens of the eye is a mechanically dynamic structure transparent to the visible spectrum, whose function is to focus an image clearly onto the neural retina where the process of seeing is initiated. Compared to the cornea and aqueous, the lens shows substantial changes with age that have serious consequences for visual performance (Spector and Sigelman, 1974; Spector, 1982). Its continued transparency during life is obviously critical to continued adequate visual performance. Two important changes in the lens occur as a function of normal aging. First, the lens grows constantly thicker and less able to contract in the act of accommodation for focusing on near objects. This causes the near focal point to continuously move out throughout life, eventually necessitating the use of reading glasses. The reduced elasticity that accompanies thickening also may slow down the act of accommodating to new targets nearer to the eye, although accommodation to view objects farther away appears less affected (Allen, 1956). Second, the lens becomes more yellow with age, indicating increased light absorption in a lens pigment that differentially absorbs short wavelength (blue) light.

Neither the reduced ability to accommodate to near objects, nor the slower speed of accommodation as an older individual's focus shifts to/from near versus far objects, can be reliably linked in the technical literature to impaired performance on specific driving tasks. Research on other effects of age-related increases in lens density deserve mention, however. First, studies have indicated a very large age difference in spectral absorption of blue light compared to that for wavelengths longer than 500 nanometers (nm). With aging, the entire absorption curve moves up, so that disproportionately less blue light gets through to the retina and the lens takes on a yellowish appearance. A documentable effect of the yellowing lens is a reduced ability to discriminate colors in the blue spectral region. This color defect shows up on a test such as the Farnsworth-Munsell 100-hue test as an increase in errors along a blue-yellow error axis (Verriest, 1963; Knoblauch, Podgor, Kusuda, Saunders, Hynes, Higgins, and DeMonasterio, 1986). The consequence for driving of a loss of color discrimination in the blue-yellow range is probably not great, even for very old and dense lenses. However, it could be expected that the more moderate increase in density for the longer wavelengths would lead eventually to significant reduction in the proportion of incident light penetrating to the retina, which could then be of serious consequence to driving performances in low light or low contrast situations.

The sudden formation of opacities in the form of cataracts in the lens is of much greater consequence than the slower physiological changes described above. A cataract results when protein particles in the lens increase in size to the point of producing significant light scatter. The cataract develops relatively quickly (1 to 5 years) compared to the slower lifelong changes in blue light absorption (Chylack, 1978). The major effect of the cataract on light is to back-reflect and scatter it. Back-reflection in a dense cataract may drastically reduce the proportion of incident light reaching the retina (Sigelman, Trokel, and Spector, 1974), but even less dense cataracts may reduce the contrast of the retinal image to a significant degree. A patient with a diffuse sclerotic cataract with daytime visual acuity of 20/25 will complain of glare in bright sunlight and may have difficulty seeing when looking into the headlights of an oncoming car at night. According to a recent review, there are no reliable data describing systematic effects of different levels of severity of cataracts on driving performance (Klein, 1991).

The retina is the multi-layered, innermost membrane of the eye that contains the initial neural substrate of vision, and is by far the most complex element of the visual system. Of all retinal changes with age, the most prevalent is the appearance of clinically-evident drusen (i.e., an accumulation of lipofuscin, a metabolic byproduct of outer segment renewal) in the retinas of 30 to 50 percent of individuals over the age of 60 (Feeney, Berman, and Rothman, 1980; Macular Phocoagulation Study Group, 1982). Between 1 and 5 percent of these persons go on to develop the pathological condition of senile macular degeneration (SMD), which is the leading cause of blindness in the over-60 age group (more people have glaucoma and cataracts, but fewer end up blind). Although people with age-related macular degeneration do not usually lose all of their sight, because of loss of central vision, they may be incapable of reading road signs or be unable to see cars (Klein, 1991).

Other important retinal pathologies include diabetic retinopathy and retinal artery and vein occlusions, all of which increase in frequency in old age. In diabetic retinopathy, chronic deterioration of retinal vascular support as a byproduct of the diabetic condition can lead to ischemia (insufficient blood flow), which in turn stimulates pathologic generation of new blood vessels. Color vision may be affected, with loss of the ability to discriminate yellow and blue. Additionally, contrast sensitivity may be affected, with losses across all spatial frequencies as retinopathy progresses (Klein, 1991). This process ultimately leads to vascular disorganization, hemorrhage and blindness. Similarly, blockage of venous or arterial flow can also stimulate vascular growth and the associated complications, which may seriously compromise vision. The diabetic situation is clearly in the chronic disease category, while vein and artery blockage appears more as an extension of the normal aging process.

Finally, a pathologic condition with relevance to driving found with increasing frequency among older persons is high interocular pressure (IOP), leading to glaucoma. Glaucoma eventually results in destruction of optic nerve fibers and is the second leading cause of blindness in older patients, affecting about 1 percent of those over age 60 (Greenberg and Branch, 1982; Viggosson, Bjornsson, and Ingvason, 1986). The condition is painless and patients are often unaware that they are suffering any deficits in visual field. There is a gradual constriction in the peripheral visual field, which can result in a total loss of vision. Drivers suffering from open-angle glaucoma and peripheral visual field loss may have difficulty seeing cars or pedestrians approaching from the side, and may show reduced contrast sensitivity (Klein, 1991).

Turning to performance effects of the aging eye, assessment techniques common to visual psychophysics provide a variety of tools that can be used to define the status of an intact visual system. Information provided by these functional assessments must then be evaluated in light of what is known of the relevant visual factors present in the driving situation. It is not always easy to make the connection between test performance and driving performance; the relative importance of performance factors such as image sharpness, glare, contrast and color can vary enormously depending upon such factors in the driving environment as the presence of rain, wet surfaces, frost, night, twilight or daylight conditions. These problems notwithstanding, functional tests in the following categories provide the best available information on performance of the aging visual system as it may relate to intersection driving: (1) spatial vision; (2) visual fields; (3) depth and motion perception; and (4) dark adaptation and glare recovery functions. A fifth category--(5) color vision--is deemed of lesser importance to the safe negotiation of intersections, but receives comment below. Accordingly, the following material will address each of the five categories of visual performance named above, presenting evidence of age differences in functional capability and citing studies of the effects of such differences on driving performance.

Spatial Vision

This category of vision assessment includes standard high contrast acuity testing, measurement of spatial contrast sensitivity or modulation transfer functions (MTFs), and the measurement of absolute and increment visual thresholds. The most systematic method for testing spatial vision is through determination of contrast sensitivity thresholds for a full range of sine wave gratings. A plot of contrast threshold against spatial or frequency then produces a modulation transfer function (MTF), which can be used to infer performance for any arbitrary stimulus configuration. Contrast sensitivity measurements can also be made at a specific spatial frequency of interest; the 4-minute gap size of a Landolt-C test stimulus as employed in many of the assessments reported in CIE Publication 19/2 (CIE, 1981) are an example of this test approach.

The characteristic MTF for spatial vision shows an inverted U-shaped function with peak sensitivity at an intermediate frequency. The cutoff of this function at the high frequency end depends on factors such as illumination and target size, and ranges from 15 to about 25 cycles per degree (c/deg) for foveal viewing at moderate illuminances (Campbell and Robson, 1968). When viewing conditions are kept constant and factors such as pupil size variation are taken into account, high and middle spatial frequency performance is found to decline with age, especially over age 40 (Derefeldt, Lennerstrand, and Lundh, 1979; Owsley, Sekuler, and Siemsen, 1983). Also, it has been shown that increased lens absorption with age alone cannot account for this decline in performance (Owsley, Gardner, Sekuler, and Lieberman, 1985).

Another, more widely examined aspect of spatial vision is acuity. Visual acuity is a test of high frequency spatial response at contrast levels far above threshold. Instead of measuring contrast threshold as a function of spatial frequency, acuity tasks measure the threshold spatial resolving power of the visual system--i.e., what separation is necessary in order to distinguish two high contrast features as being separate. In general, acuity performance can be predicted from high spatial frequency contrast sensitivity, but the converse is not true. Thus, it is not surprising that visual acuity, like high spatial frequency response, declines slowly at first, beginning at approximately age 40, then after the age of about 60 the decline accelerates (Weymouth, 1960; Richards, 1966; Richards, 1972). The Framingham study (Kahn, Leibowitz, Ganley, Kini, Colton, Nickerson, and Dawber, 1977) has provided evidence that about 10 percent of men and women between ages 65 and 74 have acuity worse than 20/30, compared to roughly 30 percent over the age of 75.

Dating at least back to Burg (1966), investigators have concentrated upon hypothesized relationships between acuity and driving competence. These studies have on the whole been quite disappointing, insofar as correlations between the driving record and static acuity scores are concerned. The ability of an observer to resolve moving targets--i.e., dynamic visual acuity (DVA)--has been found to have a stronger relationship with performance in many applied settings, however (Morrison, 1980; Long and Crambert, 1990). Shinar and Schieber (1991) point out that DVA may correlate more strongly with crash involvement, and especially among older drivers, because it combines multiple visual sensory and motor skills necessary for safe driving.

An investigation using the Pelli-Robson chart, which measures contrast sensitivity using letter stimuli that decrease in contrast but not in size, examined 1,475 ITT Hartford insurance policyholders' driving history data for differences in at-fault crash involvement (Pelli, Robson, and Wilkins, 1988). Contrast sensitivity was negatively correlated with at-fault crashes (r = -0.11); in addition, the researchers noted that since contrast sensitivity was negatively correlated with age itself (r = -0.40), the relationship between performance on the Pelli-Robson chart and crash involvement was probably understated (Brown, Greaney, Mitchel, and Lee, 1993).

In a study conducted to determine whether age-related differences in the ability to read highway signs could be measured by contrast sensitivity performance, Evans and Ginsburg (1985) used their own test chart--i.e., Vistech VCTS 6500--to obtain binocular contrast sensitivity measurements of 13 younger observers (ages 19 to 30) and seven older observers (ages 55 to 79) at spatial frequencies of 0.75, 1.5, 3.0, 6.0, 12.0, and 24 cycles per degree (cpd), while also measuring Snellen visual acuity. Observers then performed a highway sign discrimination task requiring each observer to view a movie film projection of an approaching road sign designating either a cross (+) or T intersection. The dependent variable was the discrimination distance for correct responses. Results showed that the difference in road sign discrimination distance was statistically significant; older drivers had to be significantly closer to the highway sign to determine whether it denoted a cross or T intersection, with a 25 percent average discrimination distance between the younger and older groups. The older group showed significantly lower contrast sensitivity than the younger group at 3.0, 6.0, and 12.0 cpd; significant correlations between highway sign discrimination distance and contrast sensitivity were shown at 1.5 and 12 cpd. There was no significant difference between Snellen acuities of each age group, and no significant correlation between Snellen acuity and discrimination distance.

A more recent study of the visibility distance of highway signs among young, middle-aged, and older observers by Kline, Ghali, Kline and Brown (1990) included the finding that icon signs provided superior visibility distances over text signs, particularly under dusk conditions. These authors suggested that older drivers may benefit disproportionately from the use of icon signs particularly at night, given their self-reported difficulties with signs and markings under conditions of low illumination. Advisory signs indicating proper lane position for specific maneuvers at intersections fall within this category. In an older driver survey by Yee (1985), 40 percent of the respondents reported that they never had difficulty reading traffic signs before they were too close to do any good; 33 percent seldom had difficulty reading them, and 24 percent sometimes did. Difficulty with traffic signs occurred most often on city streets (36 percent)--including signing at intersections--or on freeways through cities (31 percent).

More generally, a field investigation (Sivak, Olson, and Pastalan, 1981) of the effect of driver's age on nighttime legibility of highway signs indicated that older subjects perform substantially worse than younger subjects on a nighttime legibility task using a wide range of currently available sign materials. When subjects in two age groups (under age 25 and over age 61) were matched on high luminance visual acuity, the demonstrated legibility distances for the older subjects were only 65 to 75 percent of those for the younger subjects. These researchers concluded that age-related performance decrements on nighttime legibility tasks are primarily the result of sensory (visual acuity) deficits, rather than shortcomings in higher information-processing (e.g., reading/comprehension) skills (Sivak and Olson, 1982).

Aside from difficulties in the use of signing, problems for older drivers at intersections most likely to result from (age-related) deficits in spatial vision relate to the timely detection and recognition of pavement markings and delineation of curblines, medians, turning islands, and other intersection features. In a pertinent laboratory study, two groups of subjects (ages 19-49 and 65-80) viewing a series of ascending brightness and descending brightness delineation targets were asked to report when they could just detect a roadway heading (either left or right) from simulated distances of 30.5 and 61 meters (100 and 200 ft) (Staplin, Lococo, and Sim, 1990). Results showed that the older driver group required a contrast of 20 percent higher than the younger driver group to achieve the discrimination task in this study.

The comparative abilities of younger and older drivers to recognize downstream pavement markings has also been modeled extensively using the DETECT and PCDETECT programs developed by the Ford Motor Company (Bhise, McMahan and Farber, 1976). Analyses conducted for the Federal Highway Administration (Staplin et al., 1990) and for Transport Canada (ADI, 1991) using these computer models yield results consistent with related empirical studies: the age-related decline in spatial vision predicts delineation recognition ability for the best-performing quartile of the normative older (age 75+) driver population that is roughly equivalent to the poorest-performing quartile of the youngest (ages 18-35) driver group.

In summary, the attempt to relate studies of spatial vision functions to driving performance--with or without driver age as an independent variable--has been almost exclusively preoccupied with traffic control device (TCD) design elements, where high frequency cues predominate. Extrapolations of findings in the functional assessment literature to other intersection features must also give significant weight to the data describing drivers' response to mid- and low-frequency cues, however; age-related declines in contrast sensitivity also grow markedly for stimuli in the spatial frequency range below 12 cpd. Crucial issues in this regard include the timely and accurate perception of median and pavement edge boundaries which provide path guidance during the approach to and at intersections. Shifts in alignment, lane width transitions, and turning bays should be perceived at least at a 5-second preview distance. And, a particular need exists for left-turning drivers to pinpoint the exact location of islands, abutments, or other raised features across the wide intersections commonly encountered on suburban arterials in the United States, often while concurrently engaged in competing "effortful" working memory tasks.

Visual Fields

Age-related changes in visual fields can be measured either as a reduction in field area (contraction of the field limits) for different target sizes and intensities, or as an elevation in threshold values at distinct locations within the field limits. A kinetic testing method (Goldmann fields) has been used almost universally until the recent introduction of computer automated static techniques. Kinetic testing employs a movable spot of white light that is detected by the subject as it is brought slowly into the field of view from a starting point beyond the field limit. Isopters, or lines of equal detectability (i.e., lines that connect points in the visual field which are equally sensitive to the presence of the test stimulus), define the field limits for a given spot size and intensity. In general, field area declines as a function of decreasing target size and intensity. Automated static perimetry measures increment thresholds independently at many locations in the visual field using a small spot of light on a uniform background. The field area tested is usually restricted compared to Goldmann testing and thus the limits of the field are not recorded. Instead, threshold elevations at discrete locations are recorded and may be averaged to report a mean threshold elevation.

Decline in field area with age for both central and peripheral isopters using the kinetic method has been demonstrated (Drance, Berry, and Hughes, 1967). Similarly, a steady rise as a function of age in the mean threshold for static fields by a factor of 3 to 5 per decade has been reported (Bebie, Fankhauser, and Spar, 1967; Haas, Flammer and Schneider, 1986; Jaffe, Alvarado, and Juster, 1986). It is probable that some of this decline in sensitivity with age is attributable to non-neural factors such as progressive senile miosis, increased lens absorption, and increased light scatter in the ocular media. However, the relative importance of each of these factors is not well established.

Evidence describing the relationship between visual field loss and driving performance is mixed. In several large-scale crash analyses it has not been possible to find a significant correlation between the extent of drivers' visual fields and crash rates (Burg, 1967, 1968; Henderson and Burg, 1974; Cole, 1979). In a more recent study including over 17,000 volunteers, however, it was found that subjects with bilateral visual field defects had rates of crashes and convictions more than twice that of age- and gender-matched controls without bilateral defects (Johnson and Keltner, 1983). A study in Sweden using a driving simulator documented large individual differences among subjects with visual field defects, with most of them demonstrating impaired detection capability for test stimuli in the affected parts of the visual field (Lovsund, Hedin, and Tornros, 1991). In addition, it is important to note that the effect of a visual field loss on driving should be strongly related to its location: defects in the central field may be generally presumed to be more important than peripheral defects, and the horizontal meridian may be assumed to be most traffic relevant.

Logically, the impact of reduced visual field size in safe intersection use will be demonstrated by poorer performance in drivers' peripheral detection of vehicles and pedestrians during merging and turning maneuvers, respectively. When the visual field is restricted, increased eye movement may be invoked as a compensatory strategy. For a standard mounting of traffic signals to the right side of an intersection, it may be demonstrated that driver eye movement distances from the signal to a left-turning crosswalk must increase as the driver approaches the crosswalk. Older drivers who may have greater difficulty maintaining rapid eye movements and associated head movements are less likely to make correct judgments on the presence of pedestrians in a crosswalk, or on their walking speed (Habib, 1980). To the extent that a specific element of geometric design places exaggerated demands on the detection of peripheral objects, given older drivers' documented loss of range and flexibility of neck rotation, age-related decline in this visual function may increase the likelihood of maneuver errors at intersections.

Tarawneh, McCoy, Bishu, and Ballard (1993) included visual field measurements provided by a Keystone telebinocular testing device in a 2-year study of 105 drivers ages 65-88 at the University of Nebraska, where a variety of mental, physical, and functional status indicators were used to account for variance in subjects' on-road driving performance, emphasizing intersection turning maneuvers. Driving performance was evaluated using the on-road Driving Performance Measurement (DPM) protocol developed at Michigan State University (Vanosdall and Rudisill, 1979). In this study, a driver education expert trained in the use of the DPM technique evaluated subjects' speed control, directional control, and visual search, as they drove in their own cars. The DPM route was a 19-km (12-mi) circuit designed to evaluate the subjects in the situations that are most often involved in the crashes of older drivers. Therefore, their performance was evaluated at seven intersections, where they were required to make left turns at five intersections and right turns at the other two intersections. Four of the left turns were made from left-turn lanes onto four-lane divided arterial streets in suburban areas, and one was made from a left-turn lane onto a two-lane one-way street in an outlying business district. Two of the left turns were controlled by protected/permitted left-turn signal phases, two were controlled by permitted left-turn signal phases, and one was uncontrolled. One of the right turns was from a turning bay at a signalized intersection onto a four-lane divided arterial street in a suburban area. The other right turn was made from a stop-sign controlled approach at the intersection of two, two-lane two-way local streets in a residential area. The speed limits on the arterial streets were between 56-72 km/h (35-45 mi/h). The speed limit in the business district and residential areas was 40 km/h (25 mi/h). In all cases, maneuver performance was evaluated for: (1) the approach to the intersection; (2) the turning maneuver itself; and (3) the departure from the intersection. Correlational analysis of the study's results revealed a significant relationship between right visual field size and driving performance (r = 0.22).

Gianutsos (1991) distinguishes between measures of the "functional visual field"--i.e., the area of sensitivity for an individual under ordinary viewing conditions--and measures assessing the reaction time of subjects to evaluate and confirm the presence of a given stimulus at various locations in the periphery. She recommends use of such reaction time measures to obtain an index of visual performance more useful in predicting driving difficulty than can be obtained by traditional visual field testing procedures, and has developed PC-based protocols for this purpose.

Additional relevant findings may be cited from a simulator study of peripheral visual field loss and driving impairment which also examined the actual driving records of the study participants, and used multiple regression analyses to predict both simulator crashes and real-world crashes (Szlyk, Severing, and Fishman, 1991). It was found that visual function factors, including acuity as well as visual field measures, could account for 26 percent and 6 percent of the variance in real-world and simulator crashes, respectively. When these factors were combined with simulator response indices, including deviation in lateral lane position, out-of-lane events, brake pedal pressure, and reaction distance, 71 percent of the variance in real-world crashes and 80 percent of the variance in simulator crashes could be accounted for in the study sample. Also, greater visual field loss was associated in the simulator data with greater distance traveled ("reaction distance") before responding to a peripheral stimulus (e.g., a stop sign). While age was one variable according to which experimental and control groups were matched in this research, with both groups including participants ranging in age from their late 20s to late 60s, there was no attempt in this research to account for study outcomes in terms of age per se. This study is noteworthy for two reasons, however. First, subjects with peripheral field loss attempted to compensate for those losses through increased lateral eye movement, a strategy that is likely to be applied less effectively as a person advances in age beyond the range included in this study. A subsequent simulator study by Szlyk, Brigell, and Seiple (1993) which did incorporate age as an independent variable similarly indicated that lateral and vertical head movements, but not eye movements, increased for patients with hemianopic visual field loss relative to an older, normally sighted group. Even more important, the authors conclude that predicting an individual's ability with regard to complex driving performance depends upon the interaction of visual and what they term "visuocognitive" variables.

In this context, it is crucial to distinguish reduced visual field size or sensitivity as a sensory function from the related component of visual attention commonly termed "useful field of view" (UFOV), for which reliable age differences have also been demonstrated. This distinction will be elaborated upon in a later discussion of cognitive performance effects, where a body of evidence linking driver difficulties with intersection use, specifically, to UFOV impairments is reviewed.

Depth and Motion Perception

Tests of stereo depth perception examine a person's ability to judge relative distances without the aid of monocular cues. Arguably, a person suffering a substantial decrement in this ability may evidence difficulty in gap acceptance judgments at intersections. One recent driver performance study addressing this topic presented test slides to subjects of different age groups, which consisted of six, yellow, diamond-shaped targets, each containing four black circles. One of the four circles on each target (either the top, the bottom, the left, or the right) was designed to appear to be "floating" toward the subject while viewing the slide through the vision tester. The angles of stereopsis (seconds of arc) tested were 400, 200, 100, 70, 50, and 40. The smaller the number, the more effectively an individual can discriminate the depth cues present in these stimuli. Reading the first five targets correctly (i.e., identifying the location of the floating circle) was scored as acceptable depth perception. If a subject missed two consecutive targets, the angle of stereopsis of the last correctly read target was recorded as his/her depth perception score. The mean results for three age groups (ages 18-55, ages 56-74, and age 75+) on this measure were 112, 117, and 217 seconds of arc, respectively, with standard deviations of 106, 103, and 140 (Staplin, Lococo, and Sim, 1992).

Another study utilizing a vision tester to measure depth perception (Tarawneh et al., 1993) demonstrated a significant correlation (r = 0.35) between this variable and intersection negotiation performance, using the DPM on-road evaluation protocol (Vanosdall and Rudisill, 1979). [NOTE: The evaluation protocol was summarized above under the discussion of age differences in peripheral visual fields.]

While accurate perception of the distance to intersection features such as islands, pedestals, and other raised features is important for the safe use of these facilities, researchers have placed a relatively greater emphasis on motion perception, where dynamic stimuli--usually other vehicles--are the primary targets of interest. Motion perception is related to dynamic visual acuity, but unlike DVA, the perception of angular motion appears to be primarily limited by age-related deficits in neural mechanisms, rather than oculomotor ones (Shinar and Schieber, 1991). Prior investigations have addressed motion perception abilities pertinent to driving, including time-to-collision and gap-acceptance judgments, though only a subset has compared older and younger subjects.

In time-to-collision (TTC) estimates drivers estimate how long it takes, moving at a constant speed, to reach specified points in their paths (Purdy, 1958). They are hypothesized to be based either on an "optic-flow" process, in which the driver's analysis of the relative expansion rate of an image (such as an oncoming vehicle) over time provides the estimate of TTC directly, (Gibson, 1966; Lee, 1974, 1976) or on a cognitive process in which TTC is estimated using speed and distance information. In the first case, the driver relies on two-dimensional information--that is, angular separation cues (the image gets larger)--to estimate TTC; in the second, the driver calculates TTC on the basis of three-dimensional information. Several studies (Schiff and Detwiler, 1979; Cavallo, Laya, and Laurent, 1986) have supported the optic-flow model and the idea that two-dimensional, angular separation cues, separate from background information suffice to allow drivers to estimate TTC.

Relative to younger subjects, a decline (possibly exponential) for older subjects in the ability to detect angular movement has been reported. Using a simulated change in the separation of taillights, indicating the overtaking of a vehicle, a threshold elevation greater than 100 percent was shown for drivers ages 70 to 75 versus those ages 20 to 29 for brief (0.3 second) exposures at night. In this study, older subjects required 3.1 min of arc for detection of motion, compared to younger subjects who required only 1.43 min of arc (Hills, 1975). Older persons may in fact require twice the rate of movement to perceive that an object's motion-in-depth is approaching, given a brief (2.0 seconds) duration of exposure. In related experiments, older persons required significantly longer to perceive that a vehicle was moving closer at constant speed: at 31 km/h (19 mi/h), decision times increased 0.5 second between ages 20 and 75 (Hills, 1975). The age effect was not significant when the vehicle was moving away from the subject.

Next, research has indicated that relative to younger subjects, older subjects underestimate approaching vehicle speeds (Hills and Johnson 1980). Specifically, Scialfa, Guzy, Liebowitz, Garvey, and Tyrrell (1991) showed that older adults tend to overestimate approaching vehicle velocities at lower speeds and underestimate at higher speeds, relative to younger adults. Furthermore, analysis of judgments of the "last possible safe moment" to cross in front of an oncoming vehicle has shown that older persons (especially men) allowed the shortest time margins at 96 km/h (60 mi/h) approach speeds--older persons accepted a gap to cross at an average constant distance of slightly less than 152 m (500 ft), whereas younger men allowed a constant time gap and, thus, increased distance at higher speeds.

Hills (1980) measured actual crossing times for 10 subjects in each age and gender group, with each driver using his/her own vehicle on a test track. Young male drivers demonstrated a much shorter mean crossing time (2.5 seconds) than any of the other classes, and younger drivers (of both genders) showed a much smaller within-subject variance than older drivers of the same gender. Darzentas, McDowell, and Cooper (1980) used the results of Hills' data in a simulation model to estimate conflict involvement for each class of subject as a function of main-road flow and speed. In the model, a conflict occurs when a poor gap acceptance decision is made by a driver, causing an oncoming vehicle to decelerate to avoid collision. The model was run for main-road flows from 500 to 900 vehicles per hour for each class of driver. In each case, the number of conflicts increased linearly with flow. Older drivers were involved in more conflicts than young drivers of the same gender, and male drivers were involved in more conflicts than females in the same age class at all flows. Crossing drivers made more judgment errors in front of faster main-road vehicles. Additionally, older drivers were more likely than younger drivers of the same gender to cause a conflict in the crossing maneuver, for a wide range of vehicle speeds.

Dark Adaptation and Glare Sensitivity

Although intersection lighting installations are common in many suburban and most urban locations, rural and/or residential settings may contain unlit intersections, and drivers' dark adaption capabilities may be tested in transitions between lit and unlit areas as well. Tests of dark adaptation of the rod and cone photoreceptors in the retina of the eye measure the time course of the improvement in threshold sensitivity with cumulative time in the dark. The normal function for a young adult shows a rapid fall in the threshold for the first few minutes followed by a brief leveling out to the cone plateau, and is then followed by a second rapid drop over 10 or 15 minutes to effectively reach the rod plateau after about 30 minutes in the dark. However, many studies have shown a progressive elevation of both rod and cone thresholds with age (McFarland, Domey, Warren and Ward, 1960; Pitts, 1982), with an accelerated loss above the age of 60 which appears to parallel the increase in lens density documented earlier in this review.

One study found that the elevation in dark-adapted thresholds with age was greatest for shorter (blue) compared to longer wavelengths, and was able to account for most of this difference in terms of increased lens density (McFarland et al., 1960). That lens density contributed strongly to the elevated thresholds in this study was demonstrated by a control group of aphakic subjects⁽¹⁾ who showed approximately 1 log unit more sensitivity than their natural-lens age mates (out of a 1.3 log unit difference); the remaining 0.3 log unit difference could be accounted for by pupillary and neural changes. Similarly, an earlier review concluded that about 1.5 log units out of a total threshold elevation of 2.0 from 20 to 70 years of age can be accounted for primarily by changes in the lens (1.2 log units), and somewhat less by pupillary changes (Kahn et al., 1977).

The impact for the older driver of lost sensitivity under nighttime conditions should be assessed against the nature of the night driving task. Even at night, most visual information is processed by the cone or daylight system in the foveal region; artificial lighting raises the illumination level to the photopic range so that reading and tracking functions can occur. The peripheral rod system participates primarily by alerting the driver to a weaker signal away from the foveal line of sight that may then be oriented to, with the foveal cones. The implication of a loss in rod sensitivity is that a much brighter peripheral signal will be needed to elicit proper visual attention from the driver, and that signals now falling below threshold will be ignored. In fact, the signal may need to be 10 to as much as 100 times brighter, depending on driver age and object color. Since both rod and cone thresholds increase with age, it is also true that more light will be needed to bring important tasks such as reading and tracking (path maintenance) above the cone limit. Indeed, for steadily-increasing numbers of normatively aged drivers, objects depending on reflected light for driver detection may fall close to the elevated cone threshold.

This disadvantage for the older motorist can be further compounded by environmental and/or operational conditions, and age differences in glare sensitivity and glare recovery which penalize this group. First, the stray light introduced into a driver's eyes from roadway glare sources--most notably oncoming vehicles--can create special problems for older individuals. At intersections, additional light from roadside sources and even traffic signals can create glare problems for older drivers. At relatively low pavement luminance levels, glare--or, more specifically, veiling luminance--can be treated as a contrast sensitivity reduction factor, and its effect can be compared with the direct effect of age on contrast sensitivity noted earlier.

In summary, between ages 20 and 70, aging directly reduces contrast sensitivity by a factor of about 3; older drivers are thus at a greater relative disadvantage at lower luminance levels than younger drivers. At the same time, the magnitude of the "glare factor" with respect to its detrimental effect on a 20-year-old versus a 70-year-old driver increases by a factor of about two. Assuming that the effects of age and glare on contrast sensitivity are independent, older drivers are very much at a disadvantage in (night) driving situations in which glare is prevalent (Farber and Matle, 1989). A study of age and the brightness of pavement edge lines referenced earlier reported that an older driver test group required a contrast of 20 percent higher than a younger group to correctly discriminate roadway heading (Staplin et al., 1990); adding glare to the identical test protocol magnified the difference in performance between the two groups, and it was observed that glare limited the ability of the older group to discriminate direction-of-curve as a function of distance to the point of curvature, but not the younger group.

Color Vision

Mediation of color vision occurs at the retinal level in a two-stage process, initiated by photon catches in the three cone types, and transformed through the middle retinal layers into two opponently coded color signals. One of the most commonly applied tests of color vision, and one of the few for which data on normal aging are available, is the Farnsworth-Munsell 100-hue test, which measures color discrimination by requiring the observer to arrange 85 very closely adjacent color samples. A number of studies document the increase in 100-hue error scores as a function of age (Verriest, 1963, Verriest, van Laetham, and Uvijls, 1982; Knoblauch et al., 1986). In these studies a differential increase in blue-yellow errors is reported for subjects with no observable ocular pathology. The mean error of this type for naive subjects over 70 years of age is greater than 100, compared to a mean error score of 37 for the 20- to 30-age decade. While the precise locus of this effect is unclear, studies of changes in color matching and wavelength discrimination performance suggest a minimal role for retinal factors in the age-related loss of blue-yellow discriminability (Ruddock, 1965; Moreland, 1978).

Driver performance studies that have examined the effects of age and color vision have keyed on motorists' responses to sign and signal elements. As one example, a laboratory study has shown that increasing driver age (in conjunction with greater numbers of signs and higher background complexity in a roadway scene) leads to increased error rates in the recognition and identification of traffic signs for the particular color combinations of white-on-green and white-on-black (Woltman, Stanton, and Stearns, 1984). Another study conducted to determine the impact of dimming traffic signals at intersections at night found that older persons have reduced levels of sensitivity to intensity and contrast, but not to color (Freedman, Davit, Staplin and Breton, 1985). Tests of color vision have been included in assessment batteries administered by Tarawneh et al. (1993), Brown et al. (1993), and Temple (1989), among others. Correlations of deficiencies in color vision with on-road driving performance were not significant and, where significant correlations with simulator performance could be demonstrated, findings have not suggested any practical consequence for performance of critical driving tasks at intersections.

In sum, age-related deficits in color sensitivity arguably may account for a statistically significant portion of the variance in the conspicuity of selected traffic sign elements, but there is no compelling reason to believe that older drivers will experience operationally significant differences in the overall ability to safely negotiate intersections--at least for individuals who were not anomalous during their younger and middle-aged years--specifically as the result of deficits in color vision.

Older Drivers' Self-Perceptions of Declining Visual Skills

As a complement to the empirical evidence cited above for various aspects of diminished sensory capability, the self-perceptions of older drivers themselves of problems experienced due to declining vision may be noted. Questionnaires and focus group studies have provided information about older drivers' perceptions of their visual abilities as they relate to driving, and what driving difficulties they may experience as a result of their visual impairments (Gutman and Milstein, 1988; Milstein and Gutman, 1988; Benekohal, Resende, Shim, Michaels, and Weeks, 1992; Kosnik, Sekuler, and Kline, 1990; and Klein, Klein, Fozard, Kosnik, Schieber and Sekuler, 1992). Many visual impairments occur gradually over time, and go unnoticed because the nervous system is good at "filling in the gaps" that may be missing in the visual fields. It has been shown during driving assessment and counseling, for example, that older drivers with visual field impairments think they can see and are aware, when in fact they are not, resulting in overconfidence in their self-appraised driving ability. When driving, such a person may be able to see the roadway, but may not perceive a cyclist on the right, an oncoming vehicle on the left, or a person in an intersection⁽²⁾.

In a focus group study conducted by Gutman and Milstein (1988), the most frequently cited impairments that made driving difficult for the 162 participants across the three age groups studied (56-65, 66-75, and 76+) were poor vision (by 19 percent), poor night vision (by 13 percent), and glare from the sun or headlights (11.7 percent). Looking specifically at the responses of the participants age 76 and older, a greater percentage of drivers in this cohort reported difficulty with seeing at night (18.5 percent) and glare (22.2 percent) compared to drivers ages 56-65 and those ages 66-75 (Gutman and Milstein, 1988; Milstein and Gutman, 1988). Difficulty seeing/reading signs and or signals and poor vision were given as reasons for older driver crash involvement by 25 percent of the Gutman and Milstein (1988) focus group participants across the three age groups, and by 40.7 percent of those age 76 and older.

In a study to investigate whether there were differences in visual functioning between older individuals who are current drivers and older persons who have given up driving, it was found that ex-drivers had more trouble with glare when watching TV, reading small print, reading an advertisement on a passing bus, seeing clearly at dusk, and rated their vision as less satisfactory than their driving counterparts, regardless of age; these five questions probing these visual difficulties were significantly correlated with driving status (Kosnik, Sekuler, and Kline, 1990). Sixty-seven percent of the ex-drivers in the Kosnik et al. study gave up driving because of their visual problems, although there was no significant difference between the percentages of drivers and ex-drivers reporting glaucoma or age-related maculopathy. This study points to the fact that no single visual problem was responsible for drivers deciding to stop driving; instead former drivers exhibited declining visual abilities in several areas. Visual difficulties were comprised of loss in the overall quality of vision, performing visual tasks at a slower rate, problems locating and reading signs embedded in the cluttered surround of other signs, difficulty reading small print, trouble reading a sign on a passing bus, difficulty seeing at night, and difficulty seeing in dim light. In fact, the question with the highest correlation--reading an advertisement on a passing bus--illustrates a deficit in dynamic visual acuity. This ability, which requires the coordination of visual and motor skills, was identified in the section of this report addressing age differences in visual functions as one of the more significant performance effects of aging.

A related study surveyed adults ranging in age from 22 to 92 to gain a greater understanding of the visual difficulties they encounter while driving, as well as in the performance of everyday tasks (Kline, Kline, Fozard, Kosnik, Schieber, and Sekuler, 1992). Participants consisted of 397 volunteers from the Baltimore Longitudinal Study of Aging, divided into four age groups: 20-39, 40-59, 60-79, and 80 and older. The survey instrument contained 8 questions regarding the respondents' motor vehicle driving experience, followed by 18 items assessing the level of visual difficulty with various driving tasks, such as problems with oncoming headlights, seeing the instrument panel at night, judging speed, etc.

Analysis of the driving tasks component showed that age was strongly related to number and type of miles driven annually. The older drivers drove fewer miles annually, during rush hour, and at night. There was no relationship between age and driving environment (rural vs urban). A factor analysis conducted on the 18 vision-driving items revealed the following five factors: general vision/driving problems; illumination driving problems; age-related driving problems; gender; and health and education. Age loaded only with age-related driving problems. This included an age-related decline in reported visual quality, and increased level of difficulty with increasing age on eight of the visual/driving items (reading a street sign in time, seeing past dirt/rain on windshield, dim instrument panel, judging own speed, surprise when merging, other vehicles move too quickly, unexpected vehicles in the periphery, and windshield glare).

The results of this study may help to explain the types of driving problems older drivers encounter as a result of their diminishing visual capabilities. The high frequency with which older drivers report unexpected vehicles in their periphery and when they are merging is consistent with laboratory research demonstrating age-related declines in visual search for peripherally presented targets, shrinking of the visual fields, and binocular field losses, as described earlier in this section.

Finally, studies utilizing longitudinal health information on large numbers of community-living older participants have found that declining visual function is a significant factor associated with voluntary driving cessation (Marottoli, Ostfeld, Merrill, Perlman, Foley, and Cooney, 1993; Stewart, Moore, Marks, May, and Hale, 1993).

Compounding the varied deficits in visual capabilities associated with increasing age, an overall slowing of mental processes has been postulated beginning as early as the fifth decade and accelerating for most individuals as they 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 driving difficulties at intersections. The cognitive functions included in this processing stage perform attentional, decisional, and response selection functions crucial to safe intersection negotiation given the "tactical" and "operational" task (cf. Michon, 1979) demand levels associated with everyday operating conditions on current system facilities.

It is useful to distinguish between the generalized functional decline resulting from one of various pathological conditions which occur more frequently in the aged--most importantly the dementias--and which may predispose individuals to respond less effectively across the full range of driving tasks, versus the age-related decrements in particular cognitive functions among the normatively aging population, which can be logically or empirically related to driver behavior at intersections. The emphasis below is on the latter category of diminished capabilities, although the consequences of dementia for driving and an examination of current issues and controversy surrounding this topic are addressed at the conclusion of this discussion.

The vast literature on cognitive functions and their assessment makes a fundamental distinction that must be taken into account in the effort to focus this review on aspects of performance most relevant to intersection use by older versus younger persons. Two major categories of cognition have been defined--variously termed crystallized and fluid, or product and process--which refer, respectively, to measures reflecting the accumulated knowledge from earlier processing, and measures reflecting the efficiency of acquiring, transforming, retaining, and applying new information. In the former category, results from many studies using a wide variety of psychometric tests (e.g., vocabulary, general knowledge) show that age effects are very small, and that sometimes older persons score higher than younger persons. By comparison, measures reflecting the efficiency of current processing often show older adults to be at a disadvantage in relation to younger adults (Salthouse, 1990). The degree of age-related decline in fluid, or process cognitive functioning varies a great deal from one older individual to another, and is strongly affected by task variables.

The range of tasks performed by an older driver in the approach to and negotiation of an intersection will be addressed in detail in a subsequent section of this review. For present purposes, it may confidently be asserted that the cognitive aspects of safe intersection negotiation depend upon a host of specific functional capabilities. Most prominent among these are: (1) the access and retrieval of previously learned information to recognize and comprehend all manner of stimuli in the roadway environment, as well as the organization and integration of such information in "working memory" as required for vehicle control and navigational decisions; (2) efficient search and scanning operations, in which the most salient stimuli are discriminated from the ones that are less relevant at that instant; and (3) divided attention to process more deeply and respond as needed to the most salient stimuli, while allocating resources (i.e., serial, or "effortful" processing capacity) among multiple (shared) tasks.

Age Differences in Memory Functions

Memory functions are constantly coming into play as drivers must remember the route they wish to follow, the information acquired from traffic control devices as an action is initiated, and the rules for expected behavior in specific situations, at a minimum. For effective maneuvering at intersections, certain aspects of spatial (non-verbal) learning and memory may also be of particular importance. Memory is also important as a factor in other cognitive tasks, particularly "working memory," which allows the integration of continuous sensory information over time, the manipulation of information in memory for problem solving and decision making, and the division of attention between multiple, relevant sources of information such as an intersection control display and oncoming traffic. For the commonly-distinguished categories of sensory (iconic), short-term (primary), and long-term (secondary) storage, however, the operational significance of demonstrated age differences in these functions is less apparent. Pending successful registration of incoming sensory information, only gross deficits in short-term memory processes are likely to disrupt vehicle control in familiar situations where drivers can rely on crystallized knowledge to perform overlearned responses. Accordingly, this literature is briefly summarized below. The one exception in this area is "working memory," a topic addressed both in this discussion and again in the following section devoted to attentional processes, with which it is inextricably linked.

The "earliest" memory function engaged in the ongoing processing of roadway information is sensory memory, termed "iconic" memory for the visual sensory register. Research has shown that: (1) older persons do not require more time to establish a legible icon, but their icons are more susceptible to interference from distracting visual information perceived just before or just after a target stimulus (Walsh, Till, and Williams, 1978; Cerella, Poon, and Fozard, 1982); (2) the persistence or duration of icons differs with age, in that a light source can flash at a slower rate for older than for younger persons and still be perceived as a continuous (steady) signal (Kline and Schieber, 1980); and (3) the capacity of iconic memory is very large for young and old alike (Sperling, 1963).

Primary memory stores information that has been processed beyond the level of the sensory registers but which, nevertheless, is short-lived if it does not receive further attention and processing. Age-related studies of digit and letter spans suggest that there are, at best, marginal differences in the primary memory capacity of older and younger adults (Drachman and Leavitt, 1972; Parkinson, Lindholm, and Inman, 1982). Neither have significant differences between older and younger adults been observed in their respective recency effects (Craik, 1968; Smith, 1975). At least two studies investigating interference effects have measured more rapid loss of information from primary memory in older persons (Talland, 1965; Inman and Parkinson, 1983), but two earlier ones reporting no age-related differences (Kriauciunas, 1968; Keevil-Rogers and Schnore, 1969) may also be cited.

More importantly, the construct of working memory signifies that the primary store is a place where information is operated upon. In other words, primary memory cannot simply be construed as a repository for information; it is also the unit that performs higher level cognitive functions. Clearly, an important factor influencing the performance of these functions is the speed with which information is processed. There is now a general consensus among investigators that older adults tend to process information more slowly than younger adults, and that this slowing not only transcends the slower reaction times often observed in older adults but may, in part, explain them (Anders, Fozard, and Lillyquist, 1972; Eriksen, Hamlin, and Daye, 1973; Waugh, Thomas, and Fozard, 1978; Salthouse and Somberg, 1982; Byrd, 1984). Part of this general cognitive slowing seems to be attributed to an increase in the time it takes older adults to retrieve information from primary memory (Waugh et al., 1978; Hunt, 1978). Insofar as information in primary memory has a limited "lifespan," one would expect older adults to perform poorly on short-term memory tasks that require substantial attentional resources or on tasks that require the reorganization of to-be-remembered information. Thus, while compelling evidence does not exist to suggest that older adults differ from younger adults in either the capacity of, or the rate with which information is lost from, primary memory, older drivers will still be at greater risk in situations such as intersections that require rapid mental operations for appropriate vehicle control, especially when they are simultaneously required to perform such operations and retain other (e.g., navigational) information for future use.

Secondary memory, often labeled "long-term" memory, is generally considered to be a permanent store of unlimited capacity. Investigators of age differences in this memory function have focused on the efficiency both of encoding of information in secondary memory and of retrieval processes. Age-related decrements in the efficiency of retrieval of information from secondary memory--e.g., a deficit in recall but not in recognition memory performance--is the most common finding in the literature (Schonfield and Robertson, 1966; Hultsch, 1975; Craik, 1977; Rankin and Hyland, 1983). It should be noted, however, that the largest age-related differences in recall are found in intentional recall tasks, where subjects are allowed to process information in the manner of their choice (Thomas and Ruben, 1973; Eysenck, 1974; Till and Walsh, 1980; Poon, Walsh-Sweeny, and Fozard, 1980). These studies suggest that while older adults are often capable of employing strategies that promote efficient encoding, they do so far less spontaneously than do younger adults. In other words, it would appear that older subjects are relatively less able to organize material spontaneously in ways that render the material more easily remembered. Finally, three factors have been shown to minimize age-related deficits: practice (Treat, Poon, and Fozard, 1981; Howard, 1986), self-paced learning conditions (Canestrari, 1963; Hulicka and Wheeler, 1976), and familiarity (Poon and Fozard, 1978; Barrett and Wright, 1981; Hultsch and Dixon, 1983).

The relationship of secondary memory deficits among older drivers to intersection negotiation difficulties seems tenuous, at least with respect to semantic information. It is interesting to note, however, that nonverbal memory studies testing spatial and configurational variables also show older persons to be at a disadvantage relative to younger persons. It has been reported that older adults exhibit relatively slower performance and increased error rates on tasks requiring mental rotation of block drawings (Gaylord and Marsh, 1975), while other studies have found age-related deficits in accuracy but not speed of mental rotation (Herman and Bruce, 1983), or deficits in speed of mental rotation of geometric figure drawings without significant age differences in error rates (Berg, Hertzog and Hunt, 1982). Also, research using standardized tests of memory for designs (e.g., the Benton Visual Retention Test and the Wechsler Memory Scale) has shown consistent age-related declines in performance (Hulicka, 1966; Arenberg, 1978). Deficits in spatial memory are more likely to impact driving performance at intersections because navigational decisions must be acted upon at these locations. Navigational uncertainty will logically increase the likelihood of erratic maneuvers during an intersection approach; also excessive slowing--to the point where other traffic is disrupted--may characterize the behavior of a driver who is lost, as he/she processes additional cues from the environment in an attempt to retrieve sufficient spatial knowledge for a maneuver decision at the intersection. Walsh, Krauss, and Regnier (1981), in a study of the relationships between spatial ability, environmental knowledge, and environmental use by older individuals, reported that the use of services and facilities and the confidence with which trips are initiated from home are directly linked to spatial ability and spatial knowledge. There is also clinical evidence for demented populations that a disruption in spatial skills is the most common reason cited by older drivers in self-acknowledgments of diminished functional capacity (Odenheimer, 1989).

Age Differences in Attentional and Decisional Processes

The following material addresses two complementary functions that are essential to the safe and effective use of intersections, and that have been associated with significant age differences. The first involves the earliest stage of visual attention used to quickly capture and direct attention to the most salient events in a driving scene. The second involves the division of attention between targets of recognized importance to a driver, prior to a vehicle maneuver during the approach to or during the negotiation of an intersection. As commonly referenced in the technical literature, these cognitive processes are considered under the headings "selective attention" and "divided attention."

The most promising work addressing issues of selective attention and traffic safety arose, interestingly, from the general failure of earlier studies to find a reliable relationship between visual field sensitivity and motor vehicle crash experience (cf. Burg, 1968; Henderson and Burg, 1974; Waller, Gilbert, and Li, 1980). At the same time, investigators of age-related diminished capabilities, following reports of disproportionately high crash and violation rates for older drivers indicating specific problems with turning and merging maneuvers and failure-to-yield, especially at intersections (Campbell, 1966; Moore, Sedgley, and Sabey, 1982; Kline, 1986; Staplin and Lyles, 1991), noted that all these activities involve the processing of information from the peripheral visual field. Driving, however, unlike conventional visual field sensitivity tests, involves complex scenes with moving and/or distracting stimuli, plus the necessity of constantly dividing one's attention between central and peripheral vision. Thus, a preferred paradigm for conducting research in this area has emerged--the "functional" or "useful" field of view (UFOV). Measures of this field involve the detection, localization and identification of targets against complex visual backgrounds (Sanders, 1970; Verriest, Barca, Dubois-Poulsen, Houtmans, Inditsky, Johnson, Overington, Ronchi, and Villani, 1983; Verriest, Barca, Calbria, Crick, Enoch, Esterman, Friedman, Hill, Ikeda, Johnson, Overington, Ronchi, Saida, Serra, Villani, Weale, Wolbarsht, and Zinirian, 1985). UFOV is also influenced by the presence of distractors or multiple stimuli in the field of view (Drury and Clement, 1978; Sekuler and Ball, 1986; Scialfa, Kline, Lyman, and Kosnik, 1987; Ball, Beard, Roenker, Miller, and Griggs, 1988), as well as the time available to process the display (Bergen and Julesz, 1983; Ball, Roenker, and Bruni, 1990).

Most importantly, tests assessing the useful field of view appear to be better predictors of problems in driving than are standard field tests. One study examining state crash records for 53 (older) drivers who had been tested for visual/cognitive capabilities accounted for 20 percent of the variance in crash frequency with a composite predictor variable that included mental status and the size of the useful field of view; this model was much stronger than predictions based only upon visual sensory function which excluded measures of information processing at higher levels (Owsley, Ball, Sloane, Roenker, and Bruni, 1991). In this study, drivers with restrictions in UFOV had 15 times more intersection crashes than those with normal visual attention. A following study by the same researchers examining the driving records of over 300 drivers confirmed the predictive power of UFOV. In this study, the correlation between crash frequency and useful field of view exceeded r = 0.55; in other words, the UFOV measure alone accounted for over 30 percent of the variance in crash experience among this study sample (Ball, Owsley, Sloane, Roenker, and Bruni, 1994).

It must be reiterated that UFOV research incorporates measures of selective attention and speed of visual information processing to arrive at an overall measure of performance. Since the UFOV measure depends upon information coming through a driver's visual sensory channel, people with serious visual loss are also likely to evidence serious impairment in UFOV. The converse is not true, however--many adults who evidence impairments in UFOV have normal visual fields. UFOV is therefore a more comprehensive measure of information processing ability than visual sensory status alone.

The relationship between UFOV and older driver performance was explored further in a simulator study conducted by Walker, Sedney, and Mast (1992). The age-related narrowing of the UFOV was examined in this research using dynamic vehicle targets presented in realistic contexts on large-screen video systems, as opposed to the smaller CRT test monitor used by the applied vision researchers most active in developing this paradigm. A central tracking task of varying difficulty simulated the control tasks of driving, while vehicle images were introduced on the left and right periphery, and on a screen to the rear of the subject. Results indicated significant slowing of older (ages 65-70) drivers' responses to peripheral targets as the effort required to perform the forward tracking task was increased, while no effect of central task loading was obtained for young (ages 20-25) and middle-aged (ages 40-45) drivers. Age differences in simple reaction time (RT) as an explanation of these results was subsequently ruled out, supporting the interpretation of significant narrowing of the UFOV with age in a test protocol more closely representing the visual cues present during actual driving. The study authors, noting the strong predictive relationships between UFOV and intersection crash involvement found in the literature (see Ball and Owsley, 1991), suggest that reduced UFOV may contribute to the "looked, but didn't see" crash category. It is interesting to note that older drivers experience roughly the same proportion of lane-changing crashes as drivers in other age groups, but individuals over the age of 70 are twice as likely to be cited as at fault in this crash type (Monforton, Dumala, Yanik, and Richter, 1988).

As the UFOV paradigm has attracted increasing attention as a potential predictor of traffic safety outcomes, other investigations of age-related differences in this functional capability have yielded more equivocal results. Perry, Koppa, Huchingson, Ellis, and Pendleton (1993) have reported a study using performance on a "brief field of view" (BFOV) measure--a closely related technique for measuring a subject's ability to obtain information from the center of a briefly presented array while simultaneously detecting a peripheral target--to predict performance in controlled field studies of traffic signal detection at systematically varying degrees of eccentricity from the driver's forward line of sight. The controlled field study conditions simulated an intersection with a three-lane approach; drivers reported briefly presented traffic signal configurations (a different color in each lane) while they were driving toward the signal array. In the laboratory, the older drivers in this study did not differ from younger subjects in their ability to process information in the central 5 area of focus, but had greater difficulty acquiring data from portions of the visual field distal to the center. The results of Perry et al. (1993) parallel those of others in this respect (cf. Owsley et al., 1991). However, Perry et al. (1993) did not find a significant correlation between the laboratory BFOV measure and signal identification performance under the controlled field conditions for older drivers in this study. The authors suggest that inclusion of a skilled motor task (driving) in the field test could be an important issue, such that older drivers' relative differences in maneuvering skill/motor performance could have differentially reduced their processing capacity available for the perceptual (signal identification) task.

Brown, Greaney, Mitchel, and Lee (1993) employed the UFOV testing protocol of Ball, Roenker, Bruni, Owsley, Sloane, Ball, and O'Connor (1991) to measure visual attention capabilities of a group of 1,475 ITT Hartford insurance policyholders ranging in age from 50 to 80 and above. These individuals were divided into two groups, according to the presence or absence of recent at-fault crashes on their driving records, and the researchers tested for significant correlations between crash status and each of a number of measures from a psychophysical test battery (including UFOV). The obtained correlation for UFOV and at-fault crashes was 0.05, characterized by the authors as "unexpectedly low," though statistically significant. It should be noted that the older drivers participating in this study were volunteers, raising questions about selection bias toward the most capable members of this cohort. Also, a noisy, crowded test environment was described which may have yielded unrepresentative visual attention measures.

The importance of selective attention and attention switching to the safe performance of older drivers has also been argued by Parasuraman and Nestor (1991). These researchers cite the application of dichotic listening measures to demonstrate impairments in the ability of mild and moderate Alzheimer's patients to disengage or reorient attention, while their ability to initially adopt a focused attention state remained unaffected (Greenwood, Parasuraman, and Haxby, 1989, 1991). To the extent that a driver's approach to and negotiation of an intersection is an effortful, capacity-demanding processing task involving the continuous monitoring of competing external stimuli as well as internal vehicle controls/displays, a loss of efficiency in attention switching has at least a high construct validity as a predictor of crash likelihood. Further, a recent meta-analytic study of predictors of driving crash involvement (Arthur, Barrett, and Alexander, 1991) found (auditory) selective attention to be the most valid predictor.

Drivers' difficulties in the negotiation of intersections also should reflect the divided attention demands they face in such situations. Given the concurrent demands for lane selection, and vehicle control for path maintenance, plus vigilance for potential conflicts with other vehicles and pedestrians, it is important to highlight recent efforts to measure age differences in this critical cognitive activity.

Of particular interest is the research program underway at the Traffic Research Centre in The Netherlands, including driving simulation studies with two continuous performance tasks--a compensatory lane-tracking task and a (self-paced) visual choice reaction time task. Researchers in this laboratory took the important step of controlling for impairments already present at the single-task level. That is, single task difficulty was adjusted to stable and equivalent levels for younger and older subjects before initiating experiments on allocation-of-resources effects under divided-attention conditions. One notable experiment was conducted by Ponds, Brouwer, and van Wolffelaar (1988). For their tracking task, in which subjects used the steering wheel to compensate for "sidewinds" that pushed the vehicle away from a straight-ahead heading, a time-on-target (TOT) score was the dependent measure, defined as the time the subject's car was wholly within its lane boundaries. For the visual reaction time task, subjects had to count the number of dots in a randomly generated array superimposed within a predefined rectangular area in their forward field of view; either 9 dots out of 40 possible locations were filled in (50 percent of cases) or 8 or 10 dots (25 percent of cases each) on a trial, with counting accuracy as the dependent variable on this self-paced task. A new dot array was presented as soon as the subject had performed a "9" versus "not 9" choice to the previous array, separated by a 500 milliseconds visual masking stimulus. Subjects' resource allocation between the two tasks was governed by verbal instructions within separate blocks of test trials, according to five different strategies: concentrate solely on tracking, emphasize tracking, give equal attention to tracking and dot counting, emphasize dot counting, and concentrate solely on dot counting.

The Ponds et al. (1988) study constructed performance-operating-characteristic (POC) curves based on the results obtained under each resource allocation strategy, where performance in one task is plotted as a function of performance on the other task, for each block of test trials. Differences between POCs thus represent differences in divided attention ability. This analysis revealed a clear decline in dual task performance for older (mean age = 68.6) versus younger (mean age = 27.5) and middle-aged (mean age = 46.7) subjects, manifested principally through larger performance decrements on the tracking task. This outcome is explained in part as reflecting the self-paced nature of the visual choice task. Assuming that both of these adaptive, continuous tasks compete for the same attentional resources (cf. Wickens, 1984), these results are important in establishing an empirical basis for the reports of exaggerated difficulties for older drivers in divided attention conditions, and particularly at intersections.

A follow-on study by Brouwer, Waterink, van Wolffelaar, and Rothengatter (1991) sought the locus of the divided attentional impairment reported above. Since the response mode for the visual discrimination task was a button-push, and since it has been documented that aging particularly affects the integration of motor skills (Korteling, 1991), Brouwer et al. (1991) replicated the earlier work using an additional, vocal response mode for this task. The tracking task remained as described above. Using correct dot counts and time on target measures, as noted earlier, these researchers again plotted POC curves. Divided attention deficits for older subjects due principally to tracking task impairments were again indicated, and differences between young and old subjects were larger when they responded manually (button-push) on the dot-counting task than when they responded vocally (though vocal responding led to more errors). This finding supported the hypothesis that response integration may play a significant role in age-related divided attention deficits associated with performance of (simulated) driving tasks.

Finally, Brouwer, Ickenroth, Ponds, and van Wolffelaar (1990) varied this research methodology such that the dot array in the self-paced visual choice task was presented peripherally as well as in the driver's central field of view. This study was prompted by the observation that their initial effort left out one key component of divided attention demands under actual driving conditions: active visual search for information at unpredictable locations. According to other investigators (Plude and Hoyer, 1985), and as documented earlier in this review, the efficiency of visual search processes is especially age-sensitive. Brouwer et al. (1990) found that varying the resource allocation strategy (via instructions) did not influence older drivers on the dot-counting task for centrally-presented patterns, but had a significant effect for peripheral stimuli. The shifting allocation strategies presumably affected the extent of active visual search (i.e., involving eye movements). This finding suggests that if drivers must increase their attention to--for example--an unfamiliar roadway feature downstream to make appropriate maneuver decisions during an intersection approach, an impairment in the discrimination of peripheral targets is likely.

Another perspective on this problem is provided by attempts to measure the mental workload imposed upon vehicle operators under varying traffic conditions. As attentional demands for varying driving tasks shift according to situation, increase in task loading may produce few or no measurable increases in error rates as the operator allocates more resources to the task in question. At some point, when all available resources are allocated, a sharp increase in errors results from further task loading. At low levels of load, an individual's resources not committed to a task represent "spare capacity." In describing age differences in attentional ability related to safe performance at intersections, it would clearly be useful to establish the level of demand that can be met before the "break point" in error rate occurs. By requiring the operator to perform a subsidiary task that utilizes unallocated attentional resources, estimates of both spare capacity and primary task load can be derived. A field study of subsidiary task measures in driver loading is noteworthy in this regard (Zeitlin, 1993).

The Zeitlin (1993) study built upon earlier work (Zeitlin and Finkelman, 1975) indicating delayed digit recall and random digit generation, using oral responses, to be appropriate subsidiary tasks for driver workload measurement, according to these criteria: (1) minimal interaction with the primary task; (2) greater performance degradation as a function of decreased capacity than the primary task; and (3) monotonic or predictable changes in performance as a function of spare capacity. Data reported by Zeitlin (1993) were collected over a 4-year period for van pool drivers commuting from upstate New York to New York City while traversing a mix of rural secondary roads, limited access highways and expressways, urban arterials, and city streets. During a 2-minute test period on both inbound and outbound commutes, subjects performed each subsidiary task under conditions of varying primary task difficulty. The primary driving task difficulty was gauged in terms of speed and traffic density, the frequency of brake applications by the driver, and subjective ratings of driving difficulty. A workload index calculated by dividing the number of brake actuations by the square root of speed--a composite measure of steady state and transient driving conditions--was correlated highly (r = 0.834) with errors on the digit recall task. In addition, the digit recall task correlated highly (r = 0.615) with rated driving difficulty. This convergence supported the author's assertion that the digit recall subsidiary task can provide a good measure of spare capacity and can be used to infer primary task workload.

While the study reported by Zeitlin (1993) did not examine age differences, specifically, it deserves mention in this review because of the indicated sensitivity of the working memory measure (delayed digit recall) to concurrent attentional demand under actual driving conditions. The construct of working memory (cf. Salthouse, 1990), which incorporates the allocation and control of attentional capacities (Baddeley, 1986), is a dominant theoretical framework guiding research on age differences in cognition. Baddeley (1986) has characterized working memory as "a system for the temporary holding and manipulation of information during the performance of a range of cognitive tasks." An age-related limitation in the information processing capability of working memory is the central tenet of much of the work which demonstrates a decline by older subjects on cognitive tasks, including the selective attention and divided attention processes cited above as crucial to safe intersection negotiation.

Finally, decisional processes drawing upon working memory crucial to safe performance at intersections may be illustrated through a study of alternative strategies for presentation of left-turn traffic control messages (Staplin and Fisk, 1991). This study evaluated the effect of providing advance left-turn information to drivers who must decide whether or not they have the right-of-way to proceed with a protected turn at an intersection. Younger (mean age 37) and older (mean age 71) drivers were tested using slide animation to simulate dynamic approaches to intersection traffic control displays, with and without advanced cueing of the "decision rule" (e.g., LEFT TURN MUST YIELD ON GREEN ) during the intersection approach. Without advanced cueing, the decision rule was presented only on a sign mounted on the signal arm across the intersection as per standard practice, and thus was not legible until the driver actually reached the decision point for the turning maneuver. Cueing drivers with advanced notice of the decision rule through a redundant upstream posting of sign elements significantly improved both the accuracy and latency of all drivers' decisions for a "go/no go" response upon reaching the intersection, and was of particular benefit to the older test subjects. Presumably, the benefit of upstream "priming" is derived from a reduction in the requirements for serial processing of concurrent information sources (sign message and signal condition) at the instant a maneuver decision must be completed and an action performed. The differences in maneuver decision responses demonstrated in this study illustrate both the potential problems older drivers may experience at intersections due to working memory deficits, and the possibility that such consequences of normal aging can to some extent be ameliorated through improved engineering design practices.

Older Drivers' Self-Perceptions of Declining Cognitive Skills

Drivers participating in the Gutman and Milstein (1988) focus group study were asked to report what they were most concerned about with respect to their driving ability. Across all age groups, the biggest fear was loss of attention/concentration. This fear was reported by 30 percent of the drivers age 76 and older, 26 percent of those ages 66 to 75, and 24 percent of those ages 56 to 65. The majority of these drivers (41 percent across all age groups) believed that older driver inattention was the primary reason for crashes involving older drivers. Over one-half (55.6 percent) of the drivers age 76 and older stated that older drivers' crashes are a result of reduced attentiveness, compared to 37 percent of drivers ages 66 to 75, and 39.5 percent of drivers ages 56 to 65. Older drivers' misjudgment of other vehicles and drivers was cited as causing crashes by 7 percent of the participants, across all age groups.

In the Kline et al. (1992) survey described earlier, the older participants reported greater difficulty judging both the speed of their vehicle as well as that of other vehicles, and expressed a concern over other vehicles "moving too quickly." This perceived difficulty is consistent with increased proportions of turning crashes and right-of-way violations for this group. The older drivers' perceptual/cognitive difficulties in judging vehicle speeds is in agreement with research describing older drivers' deficits in motion detection and "least safe gap" judgments, as discussed earlier.

Contemporary views on the effects of age on body movements as they relate to driving behavior clearly distinguish between the complementary processes of response initiation and movement time (see Stelmach and Nahom, 1992). Response initiation, in turn, is presumed to reflect capabilities in the areas of response preparation, response selection, and response programming, and is sensitive to changes in task complexity and requirements for speed versus accuracy of response. Measures of movement execution, by comparison, address individual and group differences in movement trajectories and kinematics, kinetics (muscular force), coordination, joint flexibility, and sensory-motor integration. Prior studies have been consistently characterized by cross-sectional research designs, and, as elaborated below, have consistently indicated that there is age-associated slowing in motor performance, across all component processes. Questions remain, however, as to the aspect(s) of psychomotor capability which account for the greatest variance in the overall accuracy or latency measures of driver response.

It has been hypothesized by Salthouse (1985) that nearly all cognitive motor processes are slowed by approximately the same proportional amount with increasing age. The stages of response selection and response programming may be particularly sensitive to age-related decline, however. Response selection represents the output stage of sensory/perceptual and cognitive processing of information from the roadway environment, up to and including decision-making. Given the dynamic nature of the driving task, individuals are continuously engaged in the discrimination of "most relevant" stimuli, and subsequent initiation of a best--or at least an adequate--vehicle control response.

Investigations of response selection compare reaction times (RTs) as the number of response alternatives, or level of uncertainty, increases. In simple RT tasks, only slight age differences are commonly obtained (Gottsdanker, 1982). This type of study may generally be characterized as one in which a substantially suprathreshold and unambiguous signal and a well-learned response are both known in advance of a test trial. In contrast, using a choice RT paradigm, many researchers have determined that older adults are significantly slower than younger adults when response uncertainty is increased, indicating a disproportionately heightened degree of risk for older drivers when faced with two or more choices of action (Simon and Pouraghabagher, 1978; Vegega, 1989).

Tarawneh (1991) examined findings published by proponents of both "parallel" and "sequential" (serial) models of driver information processing, seeking to determine the best estimator for older individuals of a perception-reaction time (PRT) encompassing six different component processing operations for determining traffic signal change intervals: (1) latency time (onset of stimulus to beginning of eye movement toward signal); (2) eye/head movement time to fixate on the signal; (3) fixation time to get enough information to identify the stimulus; (4) recognition time (interpret signal display in terms of possible courses of action); (5) decision time to select the best response in the situation; and (6) limb movement time to accomplish the appropriate steering and brake/accelerator movements.

Tarawneh's (1991) review produced several conclusions. First, the situation of a signal change at an intersection is among the most extreme, in terms of both the information-processing demand and subjective feelings of stress that will be experienced by many older drivers. Second, the most reasonable interpretation of research to date indicates that the best "mental model" to describe and predict how drivers respond in this context includes a mix of concurrent and serial-and-contingent information-processing operations. In this approach, the most valid PRT estimator will fall between the bounds of values derived from the competing models thus far, also taking age-related response slowing for recognition, decision making, and limb movement into account. Tarawneh indicated the need to increase design values--relative to those derived from studies of young drivers--by 5 to 45 percent for the stimulus recognition phase, 20 to 100 percent for response decision, and 20 to 90 percent for limb movement, to accommodate 95 percent of the older driver population. (The net increase in the design standard for signal change time recommended by this author was from 1.0 to 1.5 seconds.)

A contrasting set of results was obtained in a recent FHWA-sponsored study of traffic operations control for older drivers (Knoblauch, Nitzburg, Reinfurt, Council, Zegeer, and Popkin, 1995). This study compared the decision/response times and deceleration characteristics of older drivers (ages 60-71+) with those of younger drivers (under age 60) at the onset of the amber signal phase.

Results of the Knoblauch et al. (1995) study showed no significant differences in 85th percentile decision/response times between younger and older drivers when subjects were close to the signal. When subjects were further from the signal at amber onset, older drivers had significantly longer decision/response times than the younger drivers. The authors suggested that the significant differences between older and younger drivers occurred when the subjects were relatively far from the signal, and that some older subjects will take longer to react and respond when additional time is available for them to do so. Thus, they concluded that the older drivers were not necessarily reacting inappropriately to the signal. In terms of deceleration rates, there were no significant differences, either in the mean or 15th percentile values, between the older and younger subjects. Together, these findings led the authors to conclude that no changes in amber signal phase timing are required to accommodate older drivers.

A study by Stelmach, Goggin, and Garcia-Colera (1987) examined response preparation where the independent variables were pretrial information level (complete, partial, or none) concerning which arm to move, the direction of movement, and the extent of movement. They demonstrated a significant and disproportionate slowing of response for older (ages 60-65) versus both young (ages 18-25) and middle-aged (ages 40-47) adults as uncertainty level increased. Based on related work, Goggin, Stelmach, and Amrhein (1989) concluded that preparatory intervals and length of precue viewing times appear to be crucial determinants of age-related differences in movement preparation and planning. When older adults are permitted to have longer stimulus exposures and longer interstimulus intervals, they exhibit less slowing of movement (Eisdorfer, 1975; Goggin et al., 1989). 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. In this regard, intersections that require weaving or successive lane changes within a restricted timeframe--as often found, for example, in dual left-turn lane geometries, or where a lane change to merge with traffic from an acceleration lane is required after negotiating an auxiliary right-turn lane with a channelizing island--should be the most difficult for this user group.

The "programming" of driver responses is a closely related issue. Stelmach, Goggin, and Amrhein (1988) predicted that older adults would have greater difficulty in situations in which anticipated driving actions must be altered. Subjects received pretrial information about the type of movement which was to occur following a cue. Accurate pretrial information (80 percent probability) defined a "planning" condition, and inaccurate information (20 percent probability) defined a "restructuring" condition in this experiment. As expected, older subjects were slower to initiate a response than younger subjects, and particularly when performing under the restructuring condition. These researchers conclude that older drivers will have greater difficulty in situations in which anticipated driving actions must be rapidly altered. The previously noted facilitation by Staplin and Fisk (1991) of maneuver decisions and hence, response selection during intersection left turn approaches by "priming" the driver with redundant upstream signing further underscores this age difference.

A related measurement of physical response capability was undertaken by Staplin, Lococo, and Sim (1990) in an experiment examining cumulative latencies for brake, accelerator, and steering wheel responses in a driving simulator. Three conditions were tested: (1) a baseline condition, where only a single control response was required; (2) a two-movement response sequence; (3) and a three-movement response sequence. The various permutations of response types within each sequence were tested (e.g., accelerate-brake-steer), with right- and left-steering responses equally distributed across trials. Slides with simple icons (red ball, green ball, and right- and left-pointing blue arrows) cued the subjects to make specific control movement sequences on a given trial. The slides were presented for a 400 millisecond (ms) duration with a 50 ms interstimulus interval, at a common fixation point. Results showed an advantage for younger subjects in performing a single control response that was very small, while the relative decrement for older subjects in speed of response widened progressively as the required control movement sequences included two and three reactions. These data were interpreted as an indication that older drivers will be at relatively greater risk than younger or middle-aged drivers when they must override a just-initiated vehicle control movement with one or more successive movements. Again, the need to avoid geometric designs which increase the likelihood that older drivers will be called upon to execute multiple responses in rapid succession is emphasized.

The movement execution factors contributing to response slowing in older adults, apart from the issues of response selection, programming, and preparedness so important to movement initiation, is relatively more straightforward. A review by Welford (1984) indicates that movement time--the interval between the initiation of movement and its completion--is significantly slower among the older population than among the young. Age-related motor impairments have been linked to decreases in muscle mass and elasticity, decreases in bone mass, and a reduction of central and peripheral nerve fibers (Welford, 1982). Muscular atrophy and related neural losses during aging are known to disproportionately affect the ability to control movement rapidly and accurately (Larsson, Grimby, and Karlson, 1979). Goggin and Stelmach (1990) reported findings which show that muscular force control may be impaired in older adults, with the result that movement corrections during movement execution are slower and much less efficient. In addition, the synchronous activation of muscles on one side of the body versus the other, as well as the inhibition of inappropriate postural responses, has been shown to be more difficult for older than for younger adults (Stelmach, Phillips, DiFabio and Teasdale, 1989). These findings suggest that older individuals may have a diminished capability to perform coordinated voluntary movements as required on a continual basis for safe and effective vehicle control.

Finally, the slowing of psychomotor responses of older drivers reflects a decline in head and neck mobility, which accompanies advancing age. Joint flexibility, which is an essential component of driving skill, has been estimated to decline by approximately 25 percent in older adults (Smith and Sethi, 1975), 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 similarly may be expected to hinder the timely recognition of conflicts during turning and merging maneuvers at intersections (see Ostrow, Shaffron, and McPherson, 1992). Of respondents to an older driver survey conducted by Yee (1985), 35 percent reported problems with arthritis and 21 percent indicated difficulty in turning their heads to scan rearward while driving.

The practical consequences of restricted head and neck movement on driving performance at T-intersections were investigated by Hunter-Zaworski (1990), using a simulator to present videorecorded scenes of intersections with various levels of traffic volume and sight distance in a 180 field of view from the driver's perspective. Two subject groups, drivers ages 30-50 and drivers ages 60-80, depressed a brake pedal to watch a video presentation (on three screens), then released the pedal when they judged that it was safe to m