The results of this research synthesis were directed to: (1) provide the rationale and justification for the recommendations subsequently developed for the Older Driver Highway Design Handbook, and (2) identify gaps in knowledge and priority issues for future research investigating the relationship between various aspects of highway design and operations, and safe driving performance. While the former objective is addressed in a separate project deliverable, recommendations for future research priorities are discussed below. Because of the larger context of this project, an emphasis on understanding age differences in driver performance will be preserved when discussing research needs with respect to elements of highway design.

The following paragraphs describe gaps in the present knowledge that impede progress in the areas of highway geometry, highway operations, and highway lighting. In addition, priority areas for future research to understand and explain the performance of drivers, particularly older drivers, in response to these elements of highway design are identified.

Initially, this discussion will focus on research needs in the area of highway geometry. A broad set of issues can be defined for improving our understanding of the relationship between the design of specific geometric elements and highway safety. Foremost among these in the present contract are:

(1) Where are particular driver groups, such as older drivers, having problems?

(2) How does the performance of older drivers compare to their younger counterparts?

(3) What changes to specific elements of geometric design can be introduced that will enhance older road-user performance in particular, and thus system safety overall?

Because of the interrelated nature of the set of issues above, suggested research objectives and study methodologies are considered together in the discussion that follows.

First, it will be useful to conduct accident analyses to determine if older drivers are overrepresented when compared to younger drivers. This approach has been used in the past and continues to be a relatively good means of determining where older road users are involved in accidents. However, more detailed analyses can and need to be conducted using the FHWA Highway Safety Information System (HSIS). Several of the states included in the HSIS data base include detailed geometric design data that can be linked to the accident data in such a way that would allow for examining accidents for different age groups associated with specific geometric design elements. Research areas that may be appropriate for such accident analyses include horizontal curvature, vertical curvature, interchanges/ramps, cross-section elements, and of course, at-grade intersections.

Also, comparing the performance of older and younger drivers can be accomplished through observational studies. If it is determined that older drivers are indeed overrepresented in certain classes of accidents involving specific geometric design elements, then observational studies may be justified to help determine why older road users are having more problems than younger road users. Driver expectancy is one of the most important areas for future research in this regard. The geometry of the roadway should transmit the proper messages to drivers. Changes in geometry should take place in a transitional fashion giving the driver adequate time to feel the change. For example, the addition of new lanes and lane drops at intersections should be consistent with driver expectancy at such locations. Such studies could be performed in a simulated environment, on a test track, or in actual real-world conditions, depending upon the selected measures of performance and their relevance to actual driving situations.

Performance-based observational studies should key on the same geometric elements noted above. In the case of horizontal curvature, a study could be set up on a test track or in the real world to evaluate speed and tracking characteristics of older and younger drivers on various degrees, radii, and lengths of curve. Measures of effectiveness (MOE's) may include approach speed, curve entrance speed, curve exit speed, lateral placement at various points on the approach and within the curve, encroachments into the adjacent lane or onto the shoulder, and erratic maneuvers. Observational studies targeted on vertical curvature would examine the design elements of percent upgrade, percent downgrade, total change in grade, length of curve, and location of curve in relation to other geometric design features such as horizontal curves. As with horizontal curves, it may also be feasible to examine the traffic control devices associated with the vertical curves. Both sag and crest curves could be included with a range of grades, change in grades, and curve lengths. MOE's may include approach speed, speed on either side of the curve, speed at the point of change in grade, lateral placement at various points within the curve, and encroachments into the adjacent lane or onto the shoulder. With respect to interchanges/ramps, including lane drop locations and weaving sections, performance-based observational studies could examine age differences in driver behaviors of operational significance. These should include at least the speed at the point of lane change, relative location of the lane change, conflicts with other vehicles, use of mirrors or turning of head, and time to complete the maneuver.

Finally, future research focusing on driver behavior at intersections should also include performance-based studies. More data are needed to reliably identify differences in older driver performance compared to younger driver performance in terms of tracking behavior through an intersection, gaps accepted and rejected, conflicts with other vehicles entering the intersection, and time to maneuver through the intersection. Once it has been determined that older drivers do indeed have a significantly greater problem with a particular element of geometric design relative to younger drivers, and the differences in performance relative to this design element between the two age groups are well understood, changes in the geometric design element to accommodate older drivers can be implemented. Determining what changes will and will not work will require both laboratory and field research. Simulation and other techniques can be used to model the traditional design elements and the more accommodating design elements. Drivers can then be exposed to these different approaches in the field to determine if they hold any promise for further testing, and, if successful to be recommended for implementation. Field evaluations, including long-term site studies with traffic observations and accident histories, should be conducted to evaluate design changes after implementation.

The next set of critical gaps in knowledge lead to the identification of future research priorities addressing the relationship between highway operations and driver performance. If operations degrade to such an extent that safety becomes an issue, then design alterations or other significant changes may be warranted. Otherwise, the operational changes may only need to be calculated, which may require new or revised analytical methods. A number of highway operations-related research priorities identified through this synthesis are given below.

(1) Investigate the relationship between driver age and speed.

In general, there is only secondary or anecdotal data available regarding the relationship between driver age and speed. Considerably more data should be collected on the speed distributions of different driver age groups, controlling for highway types, ambient light, and other potential characteristics of the highway environment (e.g., curve versus tangent sections, wet and dry conditions) so that the age-speed relationship can be clearly defined. At the same time, theoretical (perhaps through simulation) work should be done to assess the impacts of different distributions of drivers' speeds on platoon formation, gap distributions, and lane-changing behavior. For example, what is the effect of having 10 percent older drivers with distribution x in the traffic stream, 20 percent, and so on. This work will have applicability in assessing the extent to which increasing the percentage of older drivers in the traffic stream will impact capacity and level-of-service calculations.

(2) Investigate the car-following behavior of older drivers to assess their impact on the density of the traffic stream.

It is expected that older drivers will follow leading vehicles at greater distances than average (although this needs to be tested). If this is true, the percentage of older drivers in the traffic stream would have the impact of reducing the definitions of the values for the critical and maximum density of the stream. The extent to which the expected lower traffic speeds of older drivers offset their expected tendency to follow at greater distances also needs to be examined.

(3) Update the Highway Capacity Manual (HCM) with speed and gap distributions obtained from field study data or computer simulations.

Knowledge of the speed and gap distributions should be used to modify HCM assumptions and procedures as appropriate. For example, it may be that the speed effect is significant enough to require a "correction factor" in terms of passenger-car equivalents to be considered in capacity/level-of-service calculations.

(4) Investigate the relationship between increased pedestrian walk times and level-of-service calculations for signalized intersections.

The relationship between allowances that need to be made for increased pedestrian walk times and level-of-service/capacity calculations for signalized intersections needs to be investigated and refined further. This would seem to be a relatively straightforward investigation that could be done through simulation and/or use of existing HCM-based software. Practicing traffic engineers need to have some guidance about the relative trade-offs between accommodating pedestrians at isolated sites and in the context of more sophisticated signal networks.

(5) Determine thresholds for the number of older drivers in the traffic stream to warrant the implementation of protected signal phasing.

Based on the literature, it does not seem likely that the by-now-classic left-turn problems that older drivers have at uncontrolled intersections or at signalized intersections with permissive turning will significantly impact general operations. However, a sufficient population of older drivers in the traffic stream may have a significant effect and theoretical work could be done to determine what the population size would have to be to cause an operational problem. While the operations effect is not hypothesized to be large (i.e., level of service is not expected to be adversely affected), the safety problems should not be ignored.

(6) Investigate older drivers’ understanding of and response to ITS technologies.

Many important program and policy decisions rest upon the expected benefits of ITS in terms of improved highway operations. While the research agenda for ITS demonstration projects places attention on human factors issues, the response of drivers, particularly older drivers, to full-scale implementation of ITS technologies is unknown. In the area of vehicle design, older drivers’ preferences for simple, uncluttered displays and a resistance to novel control interfaces are well documented; unanticipated and potentially negative impacts of ITS traffic control devices on highway operations due to misunderstanding, information overload, or driver motivational factors should be carefully studied. Laboratory/focus group work coupled with naturalistic field studies will both contribute valuable data in this area.

Turning to a consideration of future research priorities in the area of highway lighting and driver performance, a general concern based on the studies reviewed here is that existing standards have not been applied to meet the needs of older drivers, even though current standards in highway lighting incorporate the needs of older drivers in terms of increased contrast thresholds and increased glare susceptibility. None of the studies reviewed use older observers as the "design driver." In some cases, older observers are used in a study (e.g., the age range in the Menard and Cariou [1994] study is 20 to 53); however, age was not used as an independent variable in the report of their results. In most of the studies reviewed, VL’s are typically calculated assuming observer ages from 20 to 23. It may be the case that lighting engineers are using older "design drivers" in the design of lighting systems and that these practices are not reported in the literature. In any event, this review has primarily focused upon current lighting standards rather than current lighting practices.

The current visibility model emphasis on Small Target Visibility (STV) retains average pavement luminance and pavement luminance uniformity as secondary criteria, but with a major change in emphasis. A minimum average pavement luminance is recommended to mitigate oncoming headlight glare, and non-uniformity of pavement luminance is encouraged up to the point where it becomes visually unacceptable with respect to comfort. Under these new concepts it becomes theoretically possible to design a lighting system with low levels of pavement luminance (low power costs) if the system has low visability glare and enhanced contrast between the target and its background.

Research needs for continuing improvements to lighting systems, that account for age differences in driver performance, include the following.

(1) Conduct accident analyses in which the effects of visibility and age are isolated.

There are many problems inherent to accident studies that cannot be easily controlled. Unfortunately, they are currently the primary means of quantifying benefits in terms of dollars saved and comparing this to the capital equipment, energy and maintenance costs. Based on such cost-benefit analyses, it is clear that highway lighting increases the safety of night driving. What cannot be determined from the existing accident analyses is the extent to which visibility is the primary contributing factor to accidents of older drivers at night. There is a need for accident analyses in which the effects of visibility and age are isolated. The extent to which improved lighting standards will have an added benefit for older drivers cannot be precisely determined based on current data.

(2) Assess dynamic visual performance where the complete range of spatial, temporal, average luminance and age are manipulated factorially.

Age-related glare and contrast thresholds effects are perhaps the two largest contributions to visibility deficits experienced by older observers. One shortcoming of existing standards is the lack of agreement on how to assess transient luminance and/or contrast effects. When targets are temporally modulated, older observers show pronounced deficits in their ability to process these targets. Based on the psychophysical data (reported in the companion volume to this report), one would expect that the ability to detect targets in the presence of luminance and contrast transients caused by roadway lighting is much worse among older drivers. What is needed is agreement on a psychophysical methodology, visual performance data, and an explicit methodology for applying the data to VL calculations. Although data exists on the spatio-temporal contrast sensitivity of older observers (Owsley et al., 1983), the results may not generalize to roadway lighting conditions. One approach to obtaining the appropriate data set would be to perform a parametric study similar to Blackwell's in which contrast thresholds for a large range of spatial frequencies, temporal frequencies and adaptation levels are assessed for different age groups. Although parametric studies of the spatio-temporal contrast sensitivity as a function of light level exist (van Nes, Koenderink, Nas and Bouman, 1967; Wright and Johnston, 1983; Kelly, 1984), there is no single study in which the complete range of spatial, temporal, average luminance and age are manipulated factorially.

(3) Fourier analyze modulation transfer and contrast sensitivity functions.

An alternative approach that significantly simplifies the characterization of the visual system to various combinations of different parameters involves the use of linear systems theory. The primary advantages of this approach are that: (1) the impulse response function of the visual system to spatial and temporal modulations can be used to predict visual response for arbitrary combinations of spatial and temporal parameters; (2) the effects of various intervening optics can be readily combined; and (3) the impulse response function is a functional representation of underlying physiological mechanisms such as spatial and temporal receptive field filtering. The accuracy and generalizability of a vision model based on visual physiology is expected to be higher when predicting visibility of complex targets. Such a model may actually simplify VL calculations because the effects of age-related changes in glare and contrast threshold can be calculated using one function; namely, the modulation transfer function (MTF) of the visual system. One way to obtain the MTF is by comparing the contrast sensitivity function (CSF) of young observers with that of older observers. As a starting point, predictions based on linear systems analyses should be applied to existing psychophysical data, particularly Blackwell's data.

(4) Determine interactions among lighting and visibility parameters.

Another shortcoming of existing lighting practices is the inability to adequately assess the effects of various lighting geometries and distributions, and how these changes interact. Current techniques involve measuring VI or VL for static targets. This makes it difficult to evaluate the relative effectiveness of changing luminaire height versus luminaire spacing for actual driving conditions. For example, does moment-to-moment detection performance while driving correlate more with the maximum contrast, minimum contrast, the rate of change of luminance non-uniformities, the temporal frequency of contrast polarity changes, and/or the average contrast? No standards currently exist to address these issues.

(5) Update criteria for visibility distance requirements.

More effort should be given to predicting visibility requirements for specific driving tasks and driver ages. Clear and Berman (1983) have criticized the CIE approach on the basis that most of the parameters involved in calculating RVP and RTP are not measurable quantities and therefore cannot be determined a priori without conducting an experiment. An alternative approach was suggested that involves first calculating the visibility distance provided by a given lighting system for a specific target and then calculating the age-dependent minimum safe detection distance. To aid these calculations, image processing techniques were proposed that would simplify calculations of target contrast for various lighting scenarios. This is a much more defensible approach than establishing a minimum required VL since this value depends on the particular task that drivers will be performing. Other approaches that assess speed-accuracy as well as primary-secondary task tradeoffs should also be considered.

Finally, this review has addressed a number of areas in which current standards for lighting design can be improved. One tool that lighting engineers need is a convenient means of assessing complicated tradeoffs among various lighting geometry and distribution parameters. Not only do lighting parameters interact for physical measurements such as contrast (i.e., Keck's analysis of the effects of light overlap on contrast polarity), they also interact when assessing visibility requirements as a function of driver age (i.e., visibility versus glare for unidirectional lighting). Put simply, more light is not always better. Explicit, convenient techniques need to be developed that aid the engineer in making complicated lighting parameter tradeoffs given specific site characteristics.