HIGHWAY OPERATIONS AND DRIVER PERFORMANCE
The safety of the highway system is determined by the outcomes of interactions between the driver, the vehicle, and the highway environment (system) itself. If drivers can successfully operate their vehicles on the highway system without physically impinging on others or losing control, then it is generally considered to be safe. To the extent that operations break down, then the system is either less safe (e.g., an accident occurs), less efficient (congestion occurs), or both. In this context, the classic traffic engineering definition of highway or traffic operations centers around the flow of traffic through the system—i.e., the fundamental relationship between the flow (q), density (k), and speed (v) of traffic, all of which are interrelated (see Khisty, 1990; ITE, 1992). Common measures of effectiveness related to highway operations include level of service, delay, and average travel time.
There is a fairly obvious interaction between highway operations (q-k-v interactions) and system design insofar as design considerations such as narrower or wider lanes, or more steep or gradual grades, can hinder or enhance operations (Highway Capacity Manual, 1985). For example, more permissive geometries will, all else being equal, generally tend to increase average travel speeds, decrease density, and increase capacity. Likewise, there is an interaction between operations and vehicle characteristics as they interact with the highway system and/or the user. For example, trucks will slow considerably on a steep grade and affect overall operations on that segment of the system. As another example, knowledge of a vehicle's actual braking distance or acceleration potential will presumably affect the distance at which a driver will follow another vehicle and will affect gap acceptance behavior. Finally, operations are also related to the characteristics of the motorists. As an example, the following distances that are acceptable to motorists will define the density (k) and, indirectly, speed (v) and flow (q) that will be observed; assumed and actual perception-reaction times will likewise impact not only the safety of operations but also levels of service. The primary concern here is with the interaction between highway operations and driver characteristics, and especially those characteristics of the driver that tend to vary with the aging process.
There is a large and growing body of literature regarding the capabilities and limitations of older persons that affect their ability to negotiate the highway safely, from the standpoint of highway operations. The decline of drivers' capabilities as they age is well-established, and has been the subject of investigation for some years, dating back to the 1960's. The results of these investigations, as documented in the companion volume (volume I) to this synthesis, has caused the list of problems associated with aging to virtually pass into the realm of "common knowledge" for traffic engineers, human factors engineers, and health care professionals, among others. These problems include degradation of visual acuity, lowered contrast sensitivity, diminished information processing capabilities, increased complex reaction times, and so on. This litany of performance deficits encountered (on average) as the population ages is thought to be at the root of older drivers' problems evidenced in highway operations.
Documentation of the operational problems of older drivers has been quite thorough (see McCoy, Ashman, and Foster, 1991; Lerner, Morrison, and Ratte, 1990; Lerner, Huey, McGee, and Sullivan, 1995; Knoblauch, Nitzburg, Reinfurt, Council, Zegeer, and Popkin, 1995; McKelvey and Stamatiadis, 1988; Iowa DOT, 1987; NHTSA, 1993; Staplin, Lococo, and Sim, 1993; Staplin, Harkey, Lococo, and Tarawneh, 1996). The fact that older drivers are generally over-represented in traffic accidents (relative to their exposure) as well as in specific kinds of accidents (e.g., left turns at intersections) is well known. Likewise, the types of driving errors for which older drivers are often cited in accident situations are easily recited, and include failure to yield the right of way, improper lane usage, and other problems where diminished visual and decision-making skills would appear to have explanatory value. These problems contrast significantly with those of the other high-risk group for traffic accidents—the youngest drivers—who are typically cited for such stereotypically aggressive characteristics such as speeding and following too closely. Both groups, however, tend to share a fairly accurate assessment of the problems of the other group as well as those that their own age group exhibits. More often than not, though, they are less likely to attribute such problems to themselves (i.e., "other people" in my age group have these problems; I don't") (Nelson et al., 1993).
Although the types of accidents in which older drivers are involved intuitively result from their diminished (on average) physiological and psychological capabilities, there have been considerable problems in successfully relating these "causes" with the ultimate "effects"—traffic accidents (e.g., NHTSA, 1993). Perhaps even more importantly, there has been little evaluation of the effectiveness of the remedies suggested to address the "older driver problem" in highway safety. This is due, in part, to the time required to undertake accident-based analyses—i.e., a sufficient number of sites must be treated with some countermeasure and sufficient time must pass in order to accumulate enough data (accidents) to determine whether the countermeasure is successful. It is also due to the fact that the evaluations simply aren't being attempted. Thus, with respect to highway safety, the current situation can be summed up as follows: (1) there is a fair amount of knowledge about the performance of older persons on key driving-related dimensions (although verification in actual driving situations is somewhat limited); (2) there is better and less ambiguous knowledge about highway-safety-related problems (i.e., relevant types of accidents have been fairly well identified); and (3) there is little that proves the intuitively-appealing linkage between presumed cause and effect. While this all too briefly sums up the state of the knowledge relating older drivers and highway safety, less attention has been paid to the relationships between highway operations and the impact of an aging population of drivers.
At issue is the impact of older drivers on highway operations, and the need for translating their characteristics into operational concerns, and, in turn, into design considerations. The sparse research that has been undertaken in this area has generally been done with small groups of subjects in field experiments with differing levels of control (e.g., from closed-course tests to in-vehicle observation and measurement of driver behavior). This is due, in part, to the problems with differentiating driver age in field observations, e.g., there is no convenient and inexpensive way to collect driver age information as vehicle speeds are being collected with an automatic device.
Further, the gap between what is known about human factors characteristics of older drivers and how these characteristics relate to traffic operations in real-world situations is quite large. In general, researchers have not often dealt with the relationship between older driver characteristics and operations-related issues per se. One of the purposes of the current review is to fill in that gap to the extent possible. More likely, the next step will be to identify those driver characteristics identified from human factors studies and interpret them from an operational perspective—for example, how do laboratory tests of motion detection translate into car-following behavior which, from a traffic operations standpoint, impacts traffic density.
Already, many practitioners seem quite aware of the conflicts between accommodating the stereotypical older person and maintaining a high level of service. This is especially true for city traffic engineers who deal daily with tasks such as the appropriate timing and phasing of traffic signals and public pressure to accommodate older pedestrians at selected intersections. Maintaining high levels of service for vehicles is diametrically opposed to providing increased crossing times for slower-moving pedestrians.
In the sections that follow, material is organized first according to the classic dichotomy of intersection versus non-intersection (link) operations. Within each of these categories of highway operations, material is further organized around specific operational issues that have been identified and explored by researchers. Preceding these discussions, a review of key assumptions about driver behavior in operations-related guidelines is presented, while an overview of the driving task and ITS issues concludes this section of the synthesis.
Finally, it should be noted that in the discussion that follows, the distinction between older persons and the degradation of performance characteristics which are often associated with aging becomes blurred. It must be remembered that age per se is not generally acceptable as a proxy for the diminished capabilities that are often associated with aging. Some older persons have performance characteristics that are better than those of much younger persons, and some perform far worse; while on average, performance degrades in old age, the variation among elderly driver cohorts also is considerably higher than among younger drivers.
Assumptions in Operations-Related Calculations
The characteristics of drivers are central in defining the operational state of the system. It is therefore useful to include a brief review of the assumptions that are made in the Highway Capacity Manual (TRB 1985). In the first chapter of the Highway Capacity Manual (HCM), the basic tenets of the calculations for level of service (LOS) for uninterrupted and interrupted flow are set forth. At the most basic level, this is where driver characteristics are defined as having an impact. These fundamental concepts are elaborated upon in later chapters of the HCM and detailed calculation procedures are presented.
Uninterrupted Flow. In uninterrupted flow, the fundamental q-k-v relationships are such that as q (flow or volume) increases, k (density of vehicles) increases, and v (speed of vehicles) decreases. Both the speed and density of the vehicles in the system are driven by characteristics of the vehicle operators (although, in some instances, speed can result from vehicle characteristics such as when heavy vehicles are going up a grade). It is the operators who determine how close they are willing to follow a leading vehicle (density) and how fast they are willing to go (notwithstanding their relative ability to attain that speed given interactions between vehicles). In words from the HCM, "As capacity is approached, flow becomes more unstable because available gaps in the traffic are fewer. At capacity, there are no usable gaps in the traffic stream..." Different LOS are then defined in terms of relatively arbitrary points along the way from no flow to capacity. The point is that what constitutes an acceptable or usable gap is defined by the driver and this is fundamental to determining the expected or actual LOS of a facility. To the extent that the definition of what constitutes an acceptable gap or vehicle spacing changes over time with changes in the driver population, both theoretical and observed LOS and capacity will change.
Interrupted Flow. Four basic issues are identified: (1) the concept of green time at signalized intersections; (2) saturation flow rate and lost times at intersections; (3) flow at stop and yield signs; and (4) delay.
Green time at intersections refers simply to the allocation of green time between the different movements. The primary impact of driver or pedestrian characteristics would be through the allocation of green time for a movement which was based on the time required for pedestrians to clear an intersection. For example, to the extent that older pedestrians walk more slowly than others, minimum green allocations to the minor movement could be increased thereby reducing green time for the major movement. Likewise, allocation of an all-pedestrian scramble phase also reduces the green available to vehicular movement.
According to the HCM, "the saturation flow rate for an approach is defined as the flow rate per lane at which vehicles can pass through a signalized intersection in a stable moving queue." Time is lost from the available green time through the change interval and start-up and clearance of vehicles. Start-up lost time refers to the delay that occurs between the signal turning green and the vehicle beginning to move. This time is greatest for the first vehicle and then incrementally decreases to the saturation headway (which is defined as a sort of steady-state, car-following distance). Clearance lost time is the portion of the change interval that is not used by motorists (e.g., motorists near the signal will continue through the intersection after yellow onset).
Driver characteristics will affect the recognition of and reaction to green onset, the acceptable following distance in a moving queue (i.e., acceptable car-following distances), and the use of the yellow signal. In general, it would be expected that older drivers would have slightly longer reaction times, require longer car-following distances, and make less use of the yellow time. Thus, it is likely that a higher proportion of older drivers in the traffic stream might have a significant effect on LOS and the actual capacity of an intersection.
Operations at stop- and yield-controlled intersections are defined in the HCM as dependent on "judgmental tasks." Two critical factors are identified: (1) the gaps that are available in the flow of traffic on the primary roadway, and (2) the acceptability of those gaps to motorists on the minor (controlled) approach. Motorist characteristics will impact both of these factors. In terms of traffic flow on the primary roadway, car-following behavior will define the size and number of gaps. To the extent that older persons drive more slowly than others, they will cause at least marginally more and larger gaps to be available in the traffic stream. Older persons are also likely to have different gap acceptance behavior than others while waiting on the minor approach—this will not only effect their own behavior (more delay) but also the behavior of those who are queued behind them. The effect of the latter, as elaborated below is that the longer a motorist waits for a gap or has waited in the queue, the shorter the gap that will be accepted. Thus, the HCM assumption of a constant critical gap is questionable, which directly impacts capacity calculations.
Thus, there are several conflicting results of changing driver characteristics that will effect the operation of stop- and yield-controlled intersections: the potential availability of larger if not more gaps due to different car-following behavior of older drivers in the traffic stream; the needs of older drivers for larger gaps to cross or turn into the traffic stream; and the effects of prolonged waits in the queue on the length of acceptable gaps.
Other Driver-Oriented Considerations in the HCM. There are several other implicit or explicit considerations (assumptions) of driver characteristics in the HCM procedures. One of these relates to the manner in which level of service is measured. "Average stopped-time delay is the principal measure of effectiveness used in evaluating level of service at signalized intersections" (HCM, 1985). Given that the definition of stopped-time delay is straightforward (the time a vehicle is stopped in queue), it is related to driver characteristics indirectly through other mechanisms; for example the start-up time may be longer for older drivers. To the extent that delay on the link between intersections is considered when making delay calculations for arterials, it also is effected by speed selection by different types of drivers.
Next, it is important to recognize that the HCM is largely empirically based—for example, the impact of wider or narrower lanes on driver behavior (chosen speed and following distance) finds its way into the HCM calculations as an adjustment factor in the uninterrupted flow calculations. These adjustments are based on how "average" drivers seem to behave on roadway sections with wider or narrower lanes, i.e., not as many vehicles will, on average, traverse a section with narrower lanes in a given unit of time, all other factors being equal. Of course, this in turn depends on who the drivers were on a given day when the data were collected—one of things taken as "equal" by the HCM is the driver. To the extent that few or no older drivers were observed in the empirical base of data, "all other factors" clearly are not equal. What may be required is an adjustment for the percentage of older drivers in the traffic stream—similar to heavy trucks, terrain, and so on. Interestingly, in the calculations for basic freeway segments, there are adjustments for the "driver population," although the HCM does not indicate that one of the parameters of interest of the population might be consideration of the age distribution.
Intersection (Interrupted Flow) Operations
There are several aspects to intersection operations that are potentially affected by the actual behavior of older persons and allowances that should be made for them. In the context of an intersection, design tends to be defined more broadly than just the physical design of the roadway, and is meant to include, for example, signal timing. From an operational perspective, there is concern with not only the safety of the intersection operation, but also the traditional measures of effectiveness such as delay and level of service.
Intersection level of service (LOS) is basically a function of delay, regardless of whether the intersection is uncontrolled, stop- or yield-controlled, or signalized. The manner in which intersection operations are potentially affected by older persons can best be illustrated by the following tasks required of a driver as he/she approaches and passes through the intersection:
· Detection/Recognition: The driver must first recognize that an intersection is ahead. The initial recognition may be as a result of physically seeing and recognizing some characteristics of the intersection (e.g., a signal, crossing traffic, or actually seeing the intersecting roadway, or slowing traffic ahead) or by virtue of a traffic control device (TCD). Moreover, the encounter with the intersection can range from completely expected (the area is well known to the driver) to completely unexpected (the area is unknown).
· Control, Guidance, and Navigation: The general action that is required at the intersection is dependent on the navigational needs of the driver. For example, is a turn necessary, or is it appropriate to travel straight through? More specific aspects of vehicle control and guidance also are important.
· Identification of Specific Actions: Given the need to turn or not, the driver must then determine what more specific type of action is required at the intersection. The range of actions includes simply proceeding with some degree of caution (the driver has the right-of-way); proceeding with caution and making a judgment about surrendering the right-of-way (e.g., an uncontrolled intersection); and responding to a TCD (sign and/or signal) in an appropriate way, given the presence or absence of other vehicles.
· Execution of Actions: Finally, the driver executes the appropriate maneuver(s) (e.g., stop and turn, proceed straight ahead without stopping) and passes on through the intersection.
In this scenario, there are numerous places where the characteristics of the driver can have an effect on operations. To the extent that this has been identified in the operations-related literature, it is addressed below.
Detection/Recognition of the Intersection. Depending on the environment in which the driver is operating, initial recognition of the intersection can occur in several ways. The first is when the intersection is unexpected and recognition occurs in terms of recognition/conspicuity and then legibility of TCD’s (e.g., a flashing beacon over an intersection, or the standard intersection-ahead symbol sign). If a turn is necessary or if the intersection serves as an advance notice of another potential turn further downstream (e.g., "turn at the first street past the intersection of Street X and Avenue Y), differences in recognition and legibility distances for older and younger drivers could have an effect on operations parameters. This is perhaps the easiest type of intersection detection/recognition problem to examine.
On the approach to the intersection, the driving "style" of older versus younger drivers can be characterized by differences in the speed at which each group drives. There is considerable anecdotal information that older persons drive more slowly than younger persons (Lerner, 1991; Lerner, Morrison, and Ratte, 1990; Nelson, Evelyn, and Taylor, 1993); this is supported, in turn, by the study of violation data related to accidents (Lyles, 1993; Stamatiadis, Taylor, and McKelvey, 1991) where older drivers were less likely than other age groups to be cited for speed-related offenses when they were involved in accidents. There were no comprehensive field studies that were found wherein the speeds of older drivers were explicitly compared with those of others, although Walker, Alicandri, Sedney, and Roberts (1990) found, in a simulator-based experiment, that older persons drove more slowly and were more likely to make navigational errors.
Another aspect of driver performance as the intersection is approached is car-following behavior. Hoffman and Mortimer (1994) indicate that in field tests, there is general underestimation of time-to-collision estimates in a car-following mode. Estimated time-to-collision is about 0.8 of the actual value in a car-following mode, versus 0.6 reported elsewhere in laboratory results without both vehicles moving. No differentiation by age group was provided. However, Korteling (1990) suggests that older drivers do not differ from control groups in car-following tasks in the field. Interestingly, the difference between lab and field-oriented testing here is consistent with the same sort of comparison between lab and field for perception-reaction time testing—i.e., field-testing results generally show smaller differences than would be expected based on laboratory results. This is somewhat unexpected and perhaps points to unknown compensatory strategies about which one can hypothesize. For example, one such strategy might be to purposefully follow at greater and more cautious distances. This is supported by Lerner (1991) who provides an extensive review of the literature. Moreover, the reduced speed at which older drivers apparently operate would act to offset car-following problems as the older driver, because of the speed differential, would more likely be the lead vehicle in a platoon.
Thus, in terms of approaching an intersection, it would be expected that older drivers would be driving more slowly, which would have the general impact of slowing the overall traffic stream and possibly increasing platoon formation. At the same time, the slower speeds would seem to have a generally ameliorating effect on sudden turns and maneuvers which might result from unexpectedly coming upon an intersection.
While the present literature search identified no studies of how drivers specifically identify intersections, one way that an intersection can be recognized is by virtue of the TCD’s that are deployed to warn the approaching driver. In one study of stop-controlled intersections, driver behavior with both STOP AHEAD and STOP signs was examined, versus behavior with only the latter. In this study, Zwahlen (1988) found that drivers who saw the two different situations did not significantly vary their behavior in terms of speeds, placement within the lane, and eye-scanning behavior. At night, there was some limited evidence that the STOP AHEAD sign resulted in lower speeds. Unfortunately, the subjects were almost all college-aged and no differentiation by age was possible. Still, this result is interesting given the significant age differences likely to be observed in sign visibility.
In work on construction zone signs, Kuemmel (1992) reported that older drivers had, on average, legibility distances of about 30.5 m (100 ft) less when compared with younger drivers for both day and night conditions. In a study on supplemental interchange signs using a simulator, Hummer (1989) found that older persons had differential speed responses and recognition distances that were 15 to 17 percent lower than other age groups (which equates to a 30.5-m [100-ft] difference at about 183 m [600 ft] for the control group). More recently, Mace (1994) tentatively reported that while older drivers generally have more problems with low contrast signs, their visibility distances do not degrade at night, compared to younger drivers, whose night visibility distances are 15 to 20 percent shorter than during the day for positive contrast signs.
In a study that was concerned with the adequacy of low-beam headlights, Olson and Sivak (1983) studied the detection of pedestrians at night with low-beam illumination (no other lighting). Results of this study showed that 83 percent of the older drivers (age 65 and older) could not stop in time for pedestrians wearing dark clothes, and 23 percent could not stop in time to avoid hitting pedestrians wearing a light-colored shirt. The performance of older drivers was appreciably worse than that of younger drivers (under the age of 30), who "scored" 10 to 45 percent and 3 percent, respectively. In a summary of research conducted at the University of Michigan's Transportation Research Institute, Sivak (1987) reported that nighttime legibility of signs is age related with older drivers' legibility distances being only 65 to 77 percent of their younger counterparts. This is not inconsistent with the separate results reported by Kuemmel and Hummer noted above.
Thus, if identification of an intersection is based on the legibility of TCD’s which alert the driver, older drivers would apparently recognize it later than would a younger driver. Although the studies cited are for different types of signs and/or in different situations, the overall "message" contained in the results is much the same.
However, it should be noted that there were no studies identified that focused on the identification of specific kinds of roadway situations (e.g., such as an intersection), and the conclusions are somewhat indirect. The older driver, who would typically be approaching an intersection at a relatively slower speed than other motorists, would be more likely to experience problems identifying both the intersection itself and the activities within it; but, the lower speed helps to compensate for any problems that might result from lower visibility and/or recognition distances, with the result that operations (i.e., LOS) may be only marginally degraded in some instances.
Control, Guidance, and Navigation Requirements at the Intersection. As noted above, the general action that is required at the intersection is dependent on the navigational needs of the driver—e.g., is a turn necessary or is it appropriate to travel straight through? Unfortunately, there has been little work regarding the operational aspects of selecting lanes and turning strategies. However, anecdotal evidence was reported by Staplin et al. (1996) that older drivers quite often suddenly find themselves in the wrong lane, either because they have certain expectations about lane use derived from intersections encountered on the same roadway (e.g., the right lane is for both right-turn and through maneuvers in one location, but in a downstream location, it becomes dedicated for right turns only, and the left lane is for through- and left-turn maneuvers); or, the advanced signing is inadequate or lacking, and pavement markings are covered by cars at the intersection. The biggest problem with turn-only lanes reported by older drivers who participated in focus groups conducted by these researchers was that there is not enough warning for this feature. The appropriate amount of advance notice specified by drivers ranged widely, from 5 car lengths to 1.6 km (1 mi). Still 64 percent of the participants said that multiple warning signs are necessary when the right lane becomes a turn-only lane.
In the same focus group, 79 percent of the group reported that overhead lane use signs are far more effective than roadside-mounted signs for this type of warning. Several participants suggested a combination of roadside and overhead signs in combination with painted roadway markings. Although painted roadway markings were deemed helpful, 84 percent of all participants stated that they are useless in isolation from signs, because they are usually at the intersection and are obscured by traffic, and they are frequently worn and faded. The result is that these drivers end up in the wrong lane and must go in a direction they had not planned for, or they try to change lanes at a point where it is not safe to do so. Thus, a general conclusion from this study is that overhead signing posted in advance of, as well as at, an intersection provides the most useful information to drivers about movement regulations which may be difficult to obtain from painted arrows when traffic density is high or when pavement markings are obscured by snow or become faded, or where sight distance is limited.
In one study, the installation of overhead lane-use control signs in advance of six intersections in Michigan contributed to a reduction in the total number of accidents by 44 percent in a 1-year period, and a reduction in the incidence of accidents caused by turning from the wrong lane by 58 percent (Hoffman, 1969). Older drivers (as well as their younger counterparts) have been shown to benefit from redundant signing (Staplin and Fisk, 1991). In addition to redundant information about right-of-way movements at intersections, drivers should be forewarned about lane drops, shifts, and merges through advance warning signs, and ideally these conditions should not occur close to an intersection. Advance route or street signing as well as confirmatory signing/route assurance assemblies across the intersection will aid drivers of all ages in deciding which lane will lead them to their destination, prior to reaching the intersection.
Identification and Execution of Specific Actions. Negotiation of the intersection itself has been identified as being a high risk maneuver to all drivers, and especially to older drivers in terms of violations and accident involvement. In addition to problems related to safety per se, hesitancy on the part of the driver will tend to increase delay for other users with whom they interact. The problems include vehicle-vehicle interactions (e.g., is it safe/appropriate to make a left turn in front of an oncoming vehicle) and those between vehicles and pedestrians. However, while it is clear that these situations will affect operations, very little work has been done on quantifying that effect either in the field or theoretically. Staplin et al. (1996) recently showed that older drivers require longer gaps in the opposing stream of through traffic when turning left, and because they are less likely to position themselves within an intersection while waiting to turn when compared to younger drivers, their intersection clearance times are longer.
Operationally, once the driver actually enters the general area of the intersection, it is necessary to execute the required maneuvers while keeping track of numerous other potentially conflicting activities. The number and complexity of these activities depend on whether the intersection is signalized or not.
At a signalized intersection, other vehicles must be monitored, as well as pedestrians and the status of the signal. The largest impact on traffic operations (flow) is the timing and phasing of the signal. In turn, change interval duration, left-turn maneuver durations, and pedestrian signals are three aspects which are potentially most significantly affected by driver characteristics.
As the driver approaches a signalized intersection with the intention to go straight through, one of the key dilemmas occurs if a yellow signal is encountered—that is, to stop or continue through. Stimson, Zador, and Tarnoff (1980) found that conflicts (with a crossing vehicle using part of the red phase) were dependent on geometry, travel speed, and, possibly, pavement condition. In this context, they asserted that conflicts could be eliminated with small increases in the yellow phase, although the results were site-specific. Butler (1980) noted the variation in parameters used to determine the change interval. This theme was also remarked upon by Zador, Stein, Shapiro, and Tarnoff (1985) who showed that "less adequate average clearance intervals had higher average crash rates than those with more adequate average clearance intervals;" that is, as clearance intervals decreased, potential conflicts with cross streets increased. The implication being that if "design" or "average" drivers are exposed to conflicts by inadequate clearance intervals, it is even more likely that drivers who are "less than average" in terms of functional capability would have even higher risks.
Chang, Messer, and Santiago (1985), based on vehicle observations, have argued for a simpler approach to the timing of change intervals. They studied yellow response time (time from yellow onset to brake light illumination), deceleration rates, and driver behavior from different distances from the intersection at yellow onset. They note that 95 percent of the vehicles going through the intersection at yellow onset continued through when they were less that 4.5 s from the intersection and that mean, median, and 85th percentile start delays (at green onset) were 1.8, 1.7 and 1.0 s. They suggest that 4.5 s of yellow might be an appropriate "constant" time, and that the start delay times should be used in developing "all-red" intervals. Lin et al. (1987) addressed this same issue and suggested a 4.0 s (plus/minus) correction for "vehicle supply patterns." For example, with congested conditions (non-clearing queues) the yellow time would be increased by 0.5 to 1.0 s (higher likelihood of "going" vehicles throughout the yellow) while for low flow situations, yellow times would be reduced 0.5 to 1.0 s. Gordon and Robertson (1989) provide some indirect support for longer yellow times during high-flow periods in terms of the higher number of violations of "running the red."
None of the studies noted in the last two paragraphs had any differential results by driver age. In a separate study, Ranney and Pulling (1990) reported that, for a relatively small number of drivers (23 drivers ages 30 to 51 and 21 drivers ages 74 to 83), using a closed course experiment, there were no significant differences between younger and older drivers in running red signals.
It should be noted that the most recent edition of the ITE Traffic Engineering Handbook (1992) recommends that the length of the yellow interval be defined by:
y = t + v / (2a + 2 Gg)
where: y = length of yellow interval.
v = velocity of approaching vehicle.
a = deceleration (recommended as 10 ft/s).
t = perception/reaction time of 1.0 s.
G= gravitational acceleration (32 ft/s2).
g = grade in percent divided by 100.
For level terrain, the formula above results in yellow intervals that range from 3.6 s at 56 km/h (35 mi/h) to 5.0 s at 88 km/h (55 mi/h). This formulation is from an ITE committee's recommendation in 1985 (which was also the genesis of the work by Chang and others reported above). Other determinations of the appropriate yellow time are based on the elimination of the so-called dilemma zone so that motorists will be able to either stop or accelerate and clear the intersection. In the formula shown above, it is easy to see that the assumption of a perception/reaction time is key in the determination of the yellow signal. In another paper, Chang, Messer, and Santiago (1984) also point out the fact that observation of drivers in the field indicated that the perception and brake reaction times as well as the deceleration rate will vary with speed. They argue that different brake reaction times and deceleration rates should be used for different approach speeds.
Potential problems with this formula notwithstanding, it is widely referenced when detailing signal operations. For example, the Michigan Manual of Uniform Traffic Control Devices (1994) indicates that the change interval be determined using "accepted engineering principles" and makes specific reference to the Transportation Engineering Handbook.
In this context, intersection operations are affected by older drivers only insofar as their general behavior at yellow onset is substantially different than that of the "design driver" (e.g., they are more likely to stop at yellow onset regardless of their time/distance from the intersection) or that the clearance interval is modified to accommodate decidedly different reaction times (as per the ITE formulation). Staplin et al. (1996) found that when drivers were asked what their usual reaction was to a yellow traffic signal—continue through or stop—almost half of the young/middle-aged drivers indicated they continue through the intersection, compared to only 20 percent of drivers ages 65 to 74 and 10 percent of the drivers age 75 or older. As an example of the relative sensitivity of the ITE formula, at 72 km/h (45 mi/h) the suggested yellow interval is 4.3 s. A 20 percent increase in the assumed reaction time would add 0.2 s to the yellow time, yielding 4.5 s, while a 20 percent increase in speed (from 72 to 87 km/h [45 to 54 mi/h]) yields about a 1 s increase, to approximately 5.0 s.
With respect to the adequacy of the 1.0 s assumption, Taoka (1989) examined driver response time data from several sources in order to test the fit of different distributions, and showed that the appropriate distribution is the log-normal distribution and the median, mean, and 85th percentile times were 1.2, 1.4, and 1.9 s, respectively. This implies that the assumption for the clearance time calculation should be more on the order of 1.2 s to 1.8 s rather than the 1.0 s currently assumed in the ITE formulation but that the AASHTO assumption of 2.5 s may correspond with the 95th percentile driver. After reviewing the literature, Neuman (1989), allowed for perception-reaction times ranging between 1.5 to 3.0 s, depending upon the situation.
Research studies evaluating the clearance interval length to accommodate older drivers' diminished perception-reaction time have shown mixed findings. Citing documented age differences in varying components of response speed, a recent analysis by Tarawneh (1991) of the perception-reaction time of elderly drivers to traffic signal changes at intersections 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 elderly population. The net increase in the design standard for signal change time recommended by this author was from 1.0 to 1.5 s. Similarly, Wortman and Matthias (1983) found that the 85th percentile perception-reaction time value at most intersections approached 2 s. Gordon, McGee, and Hooper (1984) estimated a median perception-brake reaction time of 1.23 s and recommended using the 85th percentile value of 1.77 s. Garber and Srinivasan (1991) conducting an accident analysis using a Virginia DOT data base of 7,000 intersection accidents involving drivers age 50 or older, found that an increase in amber time for a given speed reduces the involvement ratio for the elderly, which implies that the amber period of 3 to 5 s is probably not sufficient for the elderly because of their longer reaction times.
However, Knoblauch, Nitzburg, Reinfurt, Council, Zegeer, and Popkin (1995), comparing the responses and braking ability of older and younger drivers to amber onset in a controlled field study where subjects drove their own vehicles, found no significant differences between driver age groups in on-brake reaction times when the subjects were 3.0 to 3.9 s from the signal, nor in deceleration rates. They did, however, find differences in driver behavior between the two age groups in trials where the light turned amber when drivers were farther away from the signal. When the subjects were 4.0 to 4.9 s away from the light, there were significant differences in 85th percentile on-brake reaction times and decision/response times. On a 32 km/h (20 mi/h) approach, the older subjects' 85th on-brake reaction time percentile value was 1.26 s, while the younger subjects' was 0.82 s. There was no difference between drivers when the speed was 48 km/h (30 mi/h). When drivers were 4.0 to 4.9 s away from the light at amber onset, the decision/response time of the older drivers was 1.38 s at 32 km/h (20 mi/h) and 0.88 s at 48 km/h (30 mi/h), compared to that of the younger subjects' which was 0.50 s at 32 km/h and 0.46 s at 48 km/h. These differences were significant. The older drivers apparently take longer to react and respond when additional time is available for them to do so, but the authors state that these differences do not indicate that older drivers are necessarily reacting inappropriately to the signal. They concluded that current standards for amber signal duration and red signal onset do not need to be modified. Mixed responses were provided by older drivers regarding the length of the amber signal in focus groups conducted by these researchers earlier in the project; some older drivers reported that the yellow phase was too short, while others felt it was too long. Many favored a uniform time for the yellow phase, however. Older drivers participating in earlier focus groups indicated a need for longer yellow clearance intervals (Staplin, Lococo, and Sim, 1990).
Hauer (1988) states that although the perception-reaction time of 1 s may be too short, and this shortfall may be particularly difficult for older drivers, the issue is more complicated than the "mechanistic application of a formula." He states that if a longer perception-reaction time were used in the calculation, the duration of the clearance interval would also be extended with two consequences. First, what is added to the clearance interval is taken away from the length of time for the green signal, increasing the frequency of stopping and vehicular delay and thus adversely affects safety. Second, it is widely believed that if long clearance intervals were provided, drivers might gradually begin to encroach on these. Thus, the 1-s perception-reaction time appears to be used more as a device to get reasonable results than as a serious reflection of actual reaction times. It has less to do with what is assumed to be a proper reaction time and more with how drivers behave at the onset of the yellow signal and how this behavior depends on its duration. The Traffic Control Devices Handbook states that since excessively long yellow intervals may encourage driver disrespect, a maximum of about 5 s is usually used for the yellow interval if a long interval is required. If a longer phase-change interval is needed, than the additional time should be provided by an all-red interval.
With regard to the safety of a flashing "end of green" warning phase, it may be noted that Mahalel and Zaidel (1985) reported that flashing the green light for 3 s just prior to the amber light increases the "indecision zone" and consequently increases the probability for rear-end collisions. An accident analysis of 319 urban intersections in Israel over a 3-year period showed that the mean number of rear-end collisions was higher at intersections equipped with flashing green programs than those with an amber-only program. The number of right-angle collisions increased as well with the flashing green, but the effect was not statistically significant. More recently, Mussa, Newton, Matthias, Sadalla and Burns (1996) conducted a driving simulator study to evaluate the effect of flashing the amber light at the end of the green phase (prior to the onset of solid amber) on improving driver anticipation. The indication sequence was green, green/flashing amber, amber, then red (referred to as a four-phase program). The amber was set to flash for a duration depending on the approach speed to the intersection, and the frequency of flashing was set at 2 Hz. At 40.3 km/h (25 mi/h) the amber flashed for 2.5 s, and was followed by 4.0 s of solid amber. At 72.5 km/h (45 mi/h), the amber flashed for 4.5 s, and was followed by 4.3 s of solid amber. Forty-one subjects aged 18 to 58 participated. Although the four-phase program (compared against the standard three-phase program) reduced red light violations, reduced the severity of maximum accelerations and decelerations, and reduced kinematically-defined inappropriate stop or cross decisions, study results also indicated that the four-phase program increased the size of the indecision zone by a factor of two, increasing the potential for rear-end collisions at intersections. The larger the indecision zone, the more vehicles are likely to be in the zone, and the higher the likelihood of successive drivers making conflicting decisions. In addition, first response times for the four-phase program showed substantially larger variability than those for the three-phase program, and the analysis of first response time variability in relation to stopping decision variability indicated that unsafe driver behavior is likely with the four-phase program.
Turning to a discussion of left-turn maneuvers at signalized intersections, the yellow time is not as clearly calculated for vehicles that turn left, since left-turning vehicles often use the yellow-red transition as the opportunity to make the turn, especially during congested periods. Conversely, during lower flow conditions, the left-turn movement is much different from the straight-through movement. The nature and timing of the former is considerably less easily modeled and the variance in behavior is presumed to be high. The more critical left-turn maneuver occurs when the turning motorist must judge the gap available to make the left turn across the path of an oncoming vehicle.
Part of the left-turning task complexity is whether drivers understand the rules under which the turns will be made—for example, do the drivers understand the message that is being conveyed by the signal and any ancillary signs. If the signals and markings are not understood, at a minimum there may be delay in making a turn or, in a worse-case scenario, may be an accident. Kettelson and Vandehey (1991) argue that the Highway Capacity Manual and AASHTO guidelines are based on gap acceptance logic that is incorrect. For example, experiencing increasing delays (e.g., while waiting to turn) will cause motorists to accept smaller gaps over time than they might normally.
It is also necessary that drivers understand the meaning of signals and accompanying signs to promote efficient use of green time. An accident analysis conducted by Curtis, Opiela, and Guell (1988) indicated that approaches with any form of supplemental signing for left-turn control experience more accidents than those without such signs. Additionally, a driver comprehension analysis conducted in a laboratory setting with drivers ages 30 to 60 and older, showed that green displays (those with the green ball alone, green arrow alone, or combinations of green ball and green arrow on the left-turn signal) were correctly interpreted with widely varying frequency, depending on the signals shown for the turning and through movements (Curtis et al., 1988). In most cases, performance declined as age increased; older drivers were correct approximately half as often as the youngest drivers. Most driver errors, and especially older driver errors, indicated signal display interpretations that would result in conservative behavior, such as stopping and/or waiting. A summary of the results of the Curtis et al. (1988) study follows.
Curtis et al. (1988) found that the simple green ball under permissive control was correctly interpreted by approximately 60 percent of the subjects. For protected/prohibited operation, the green arrow (with red ball for through movement) was correctly answered by approximately 75 percent of drivers. For protected/permitted operation, the green ball alone was correctly answered by only 50 percent of the respondents, while the green arrow in combination with the green ball had approximately 70 percent correct responses. When the green ball with the green arrow were supplemented by the R10-12 sign "left turn yield on green (ball)," only 34 percent of drivers answered correctly. This test result suggests that the MUTCD recommended practice may result in some driver confusion, as test subjects answered correctly more often when the sign was not present, even when the effects of regional differences in familiarity with the sign were considered. Green arrows were better understood than green balls. Conversely, red and yellow arrows were less comprehensible than red and yellow balls. Potentially unsafe interpretations were found for red arrow displays in protected/prohibited operations. The yellow arrow display was more often treated as a last chance to complete a turn when compared to a yellow ball. Driver errors were most frequent in displays that involved: (1) flashing operations; (2) multiple faces with different colors illuminated on the left-turn signal head; and in particular, (3) different colors on the turn and through signals.
When Hummer, Montgomery, and Sinha (1990) evaluated motorists' understanding of left-turn signal alternatives, they found that the protected signal was by far the best understood, permissive signals were less understood, and the protected/permitted (p/p) the least understood. When a green ball for through traffic and a green arrow for left turns were displayed, the protected signal was clearly preferred over the permissive and p/p signals, and the leading signal sequence was preferred more often than the lagging sequence. Respondents stated that the protective signal caused less confusion, was safer, and caused less delay than the permissive and p/p signals. It should be noted, however, that while older persons were in the sample of drivers studied, they made up a very small percentage (8 of 402) and differences were hard to substantiate. A similar study by Staplin and Fisk (1991) showed that the "left turn yield on green (ball)" sign was the most problematic. This study included older and younger drivers and showed that older drivers had higher error rates and increased decision latencies for situations were the left turn was not protected.
More recently, Knoblauch, Nitzburg, Reinfurt, Council, Zegeer, and Popkin (1995) examined the lack of understanding associated with a variety of protected and permitted left-turn signal displays. They found that many drivers, both young and older, do not understand the protected/permitted signal phasing, and suggested that efforts to improve motorist comprehension of left-turn signal phasing should be targeted at the entire driving population. In focus group discussions conducted as part of the same project, many older drivers reported that they avoid intersections that do not have a protective left-turn arrow or those where the time allowance for left turns was too short. In addition, the situation where the green arrow eventually turns to a solid green ball was generally confusing and not appreciated by the older participants. Among the recommendations made by the older drivers were: (1) provide as many protected left-turn opportunities as possible; (2) standardize the sequence for the left-turn green arrow so that it precedes solid green or red; (3) lengthen the protected left-turn signal; (4) lengthen the left turn storage lanes so that turning traffic does not block through traffic; (5) make traffic signal displays more uniform across the U.S., including the warning or amber phase; (6) standardize the position and size of signals; (7) provide traffic lights overhead and to the side at major intersections; (8) provide a warning to the driver when the light is about to turn yellow, such as a pulsing green phase of one to two seconds; (9) paint a yellow line in the pavement upstream of the signal such that, if the driver has not reached the line before the light has turned yellow, he/she can not make it through before the red light; (10) provide backplates around lights to minimize glare from the sun; and (11) eliminate decorations on signal heads as they are often green and red and may be confusing near signal faces.
Bonneson and McCoy (1994) also found a decreased understanding of protected and permitted left-turn (p/p) designs with increased age, in a survey conducted in Nebraska with 1,610 drivers. In this study, the overlap phase (left-turn green arrow and through green ball illuminated) was the least understood by drivers wishing to turn left, with only one-half of the respondents answering correctly; most of the respondents who erred chose the safer course of action, which was to wait for a gap in oncoming traffic. In terms of signal head location, 4 to 5 percent more drivers were able to understand the p/p display when it was centered in the left-turn lane (exclusive) as opposed to having the head located over the lane line (shared). Although the difference was statistically significant, Bonneson and McCoy point out that the difference may be too small to be of practical significance. In terms of lens arrangement, significantly more drivers understood both the permitted indication and the protected/MUTCD indication (left-turn green arrow and through red ball) in vertical and horizontal arrangements than in the cluster arrangement. An analysis of sign use compared the exclusive cluster lens arrangement over the left-turn lane and exclusive vertical lens arrangement over the through lanes with and without the use of an auxiliary sign ("left turn yield on green [ball]"). Overall, the results indicated a significantly higher correct response rate when there was no sign. Designs with a sign decreased driver understanding by about 6 percent. However, for the permitted indication the sign appeared to help driver understanding, whereas during the overlap and protected indications it appeared to confuse drivers. Comparisons between the protected/MUTCD indication and a modified protected indication (green arrow with no red ball), showed that for the horizontal p/p designs, 25 percent more drivers were able to understand the protected indication when the red ball is not shown with the green arrow, and for the vertical and cluster p/p designs, 12 percent more drivers understood the modified protected indication. The point is that from an operational perspective, hesitancy as a result of misunderstanding will decrease the level of service and possibly result in accident situations.
Numerous studies have found that: (1) protected left-turn control is the safest, with protected/permitted being less safe than protected, but safer than permitted (Curtis et al., 1988; Matthais and Upchurch, 1985; Fambro and Woods, 1981); and (2) transitions from protected operations to protected/permitted operations experience accident increases (Warren, 1985; Cottrell and Allen, 1982, Cottrell, 1985; Florida ITE, 1982; Agent, 1987). According to Fambro and Woods (1981) for every left-turn accident during a protected phase, 10 would have occurred without protection. Before-after studies where intersections were changed from protected to permitted control have shown 4- to 7-fold increases in left turn accidents (Agent, 1987; Florida ITE, 1982).
Williams, Ardekani, and Asante (1992) conducted a mail survey of 894 drivers in Texas to assess motorists' understanding of left-turn signal indications and accompanying auxiliary signs. Drivers over age 65 had the highest percentage of incorrect responses (35 percent). Results of the various analyses are as follows: (1) the use of a green arrow for protected-only left turns produces better comprehension than the use of a circular green indication, even when the circular green indication is accompanied by an auxiliary sign; (2) for a five-section signal head configuration, the display of a green left turn arrow in isolation produces better driver understanding than the simultaneous display of a circular red indication and a green left turn arrow; (3) the "left turn yield on green (ball)" auxiliary sign was associated with the smallest percentage of incorrect responses, compared to the "left turn on green after yield" sign, the "protected left on green" sign, and the "left turn signal" sign; and (4) the percentage of incorrect responses was 50 percent lower in the presence of a circular red indication compared to a red arrow; the red arrow was often perceived to indicate that a driver may proceed with caution to make a permissive left turn.
In another study conducted by Curtis et al. (1988), it was found that the Delaware flashing red arrow was not correctly answered by any subject. The incorrect responses indicated conservative interpretations of the signal displays which would probably be associated with delay and may also be related to rear-end collisions. Drivers interpreted the Delaware signal as requiring a full stop before turning, because a red indication usually means "stop," even though the signal is meant to remind motorists to exercise caution but not necessarily to stop unless opposing through traffic is present. Hulbert, Beers, and Fowler (1979) found a significant difference in the percentage of drivers under age 49 and those over 49 who chose the correct meaning of the red arrow display. Sixty-one percent of the drivers over age 49 chose "no turning left" compared to 76 percent of the young and middle-aged drivers. Although other research (Noel, Gerbig, and Lakew, 1982) has concluded that the left turn arrow is more effective than the red ball in some left-turn situations where special turn signals and exclusive turn lanes are provided, drivers of all ages will be better served if signal indications are consistent. Therefore, it is recommended that the use of the arrow be reserved for protected turning movements and the color red be reserved for circular indications to mean "stop."
Hawkins, Womak and Mounce (1993) surveyed 1,745 drivers in Texas to evaluate driver comprehension of selected traffic control devices. The sample contained 88 drivers age 65 and older. Three alternative signs describing the left-turn decision rule were evaluated: (1) R10-9, "protected left on green arrow "(in the Texas MUTCD but not the National MUTCD); (2) R10-9a, "protected left on green" (in the Texas MUTCD but not the National MUTCD); and (3) R10-12, "left turn yield on green (ball)." The R10-12 sign did the best job of the signs in the survey informing the driver of a permissive left turn condition, with 74.5 percent choosing the desirable response. Of those who responded incorrectly, 13.6 percent responded that they would wait for the green arrow, and 4.3 percent made the dangerous interpretation that the left turn was protected when the green ball was illuminated. Drivers age 65 years and older were among the respondents who were most likely to choose an incorrect response.
The role of "working memory" in decisional processes crucial to safe performance at intersections may be illustrated through a human factors 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 [ball]") 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.
Stelmach, Goggin, and Garcia-Colera (1987) found that older adults were particularly impaired when preparation was not possible, showing disproportionate response slowing when compared with younger subjects. When subjects obtained full information about an upcoming response, reaction time (RT) was faster in all age groups. Stelmach et al. (1987) concluded that older drivers may be particularly disadvantaged when they are required to initiate a movement in which there is no opportunity to prepare a response. Preparatory intervals and length of precue viewing times are determining factors in age-related differences in movement preparation and planning (Goggin, Stelmach, and Amrhein, 1989). When preparatory intervals are manipulated such that older adults have longer stimulus exposure and longer intervals between stimuli, they profit from the longer inspection times by performing better and exhibiting less slowness of movement (Eisdorfer, 1975; Goggin et al., 1989). Since older drivers benefit from longer exposure to stimuli, Winter (1985) proposed that signs should be spaced farther apart to allow drivers enough time to view information and decide what action to take. Increased viewing time will reduce response uncertainty and decrease older drivers' RT.
The differences in maneuver decision responses demonstrated in the Staplin and Fisk (1991) 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. Staplin and Fisk (1991) also showed that older drivers had higher error rates and increased decision latencies for situations were the left turn was not protected. In particular, the most problematic displays were those with only one steady illuminated signal face (green ball) accompanied by a sign that indicated that it was not safe to proceed into the intersection with the assumption of right-of-way ( "left turn yield on green [ball]"). A correct response to this combination depends on the inhibition of previously-learned "automatic" responses; a signal element with one behavior (go) was incorporated into a traffic control display requiring another conflicting behavior. When comparing these data to other studies showing superior comprehension for the "left turn yield on green (ball)" sign, it is important to keep in mind that demands on subjects varied across the different efforts, with an emphasis on the dynamic intersection approach simulated by Staplin and Fisk (1991).
Hummer, Montgomery, and Sinah (1991) evaluated leading and lagging signal sequences using a survey of licensed drivers in Indiana; an examination of traffic conflicts; an analysis of accident records; and a simulation model of traffic flow, to evaluate motorists' understanding and preference for leading and lagging schemes as well as determining the safety and delay associated with each scheme. Combinations of permissive and protective schemes included: (1) protected-leading, in which the protected signal is given to vehicles turning left from a particular street before the green ball is given to the through movement on the same street; (2) protected-lagging, in which the green arrow is given to left-turning vehicles after the through movements have been serviced; (3) protected-permissive, in which protected left turns are made in the first cycle and a green-ball signal allows permissive turns later in the cycle; and (4) permissive-protected, in which permissive turns are allowed first in the cycle and protected left turns are accommodated later in the cycle. The protected-leading and protected-permissive schemes are known as "leading," and the protected-lagging and permissive-protective are known as "lagging" schemes. Of the 402 valid responses received, 248 respondents preferred the leading, 59 preferred the lagging sequence, and 95 expressed no preference. The most frequent reasons given for preference of the leading sequence were: it is more like normal; it results in less delay; and it is safer. There are apparent trade-off’s here, however, i.e., the leading sequence was associated with a higher conflict rate with pedestrians and a higher rate of run-the-red conflicts (drivers turning left during the clearance interval for opposing traffic), while the intersections with a lagging sequence were associated with a significantly higher rate of indecision conflicts than the leading intersections due to violations in driver expectancy. Overall, it is judged that consistency in signal phasing across intersections within a jurisdiction, as well as across jurisdictions, should be a priority, and that use of a leading protected left-turn phase offers the most benefits.
Upchurch (1991) compared the relative safety of five types of left-turn phasing using Arizona Department of Transportation (ADOT) accident statistics for 523 intersection approaches, where all approaches had a separate left-turn lane, 329 approaches had two opposing lanes of traffic, and 194 approaches had three opposing lanes. The five types of left-turn phasing included: (1) permissive; (2) leading exclusive/permissive; (3) lagging exclusive/permissive; (4) leading exclusive; and (5) lagging exclusive. For the 495 signalized intersections on the State highway system, most samples represented a 4-year accident history (1983 to 1986). For 132 signalized intersections in six local jurisdictions, samples ranged from 4 months to 4 years, all between 1981 to 1989. When the accident statistics were stratified by left-turn volume and opposing traffic volume (vehicles per day), the following observations and conclusions were made for sample sizes greater than 5, eliminating any conclusions about lagging exclusive phasing:
• Leading exclusive phasing had the lowest left-turn accident rate in almost every case. This was true in every left-turn volume range and every opposing volume range except one (19 out of 20 cases). Lagging exclusive/permissive was the exception for 3 opposing lanes and left turn volumes of 0 to 1000.
• When there were two lanes of opposing traffic, lagging exclusive/permissive tended to have the worst accident rate.
• When there were three lanes of opposing traffic, leading exclusive/permissive tended to have the worst accident rate.
• When there were two lanes of opposing traffic, the order of safety (accident rate from best to worst) was leading exclusive, permissive, leading exclusive/permissive, and lagging exclusive/permissive. However, there was a small difference in the accident rate among the last three types of phasing.
• When there were three lanes of opposing traffic, the order of safety (accident rate from best to worst) was leading exclusive, lagging exclusive/permissive, permissive, and leading exclusive/permissive.
Upchurch (1991) compared the accident experience of 194 intersections that had been converted from one type of phasing to another in a simple before-after design. For each conversion, 4 years of before accident data and 4 years of after accident data were used, where available. At approaches having two opposing lanes of traffic, the statistics for conversions from permissive to leading exclusive/permissive and vice-versa reinforced each other, suggesting that leading exclusive/permissive is safer than permissive. At approaches having three opposing lanes of traffic, the statistics for conversions from leading exclusive to leading exclusive/permissive and vice-versa reinforce each other, suggesting that leading exclusive is safer than leading exclusive/permissive.
Parsonson (1992) states that a lagging left-turn phase should be used only if the bay provides sufficient storage, as any overflow of the bay during the preceding through movement will spill into the adjacent through lane, blocking it. A lag should also be reserved for those situations in which opposing left-turn movements (or U turns) are safe from the left-turn trap (or are prohibited). Locations where the left-turn trap is not a hazard include T-intersections, those where the left-turn (or U turn) opposing the green arrow is prohibited or is allowed only on a green arrow (protected-only phasing). In addition, driver expectancy weighs heavily in favor of leading left turns, and driver confusion over lagging left turns results in start-up losses. In a survey of 34 engineers from 19 states, Parsonson (1992) concluded that lagging left turns were not popular with many respondents and were used only when necessary and safe.
Thus, from an operational perspective, left turns by older drivers at signalized intersections can likely impact operations in the sense that they are somewhat less likely to use the allowed turning time appropriately--e.g., they may not turn when it is permitted (and safe) to do so, causing unneeded delays to themselves and other drivers that may be queued behind them.
Next, traffic signals at many signalized intersections must also accommodate pedestrians. Traffic operations are affected by pedestrians in several ways: (1) if an all-way pedestrian signal is provided, all traffic stops during the pedestrian "scramble" phase; (2) green and amber phases may be timed to accommodate the actual crossing; and (3) right-turn-on-red (RTOR) movements may be prohibited to make pedestrian movements safer. All of these pedestrian signal allocations tend to decrease green time (or turning opportunities in the case of an RTOR) for vehicular movements. If the pedestrians of concern are older persons, it has been hypothesized that even more green time will be lost for vehicles, since the walking speeds will be slower, and longer pedestrian crossing times will have to be accommodated.
Hauer (1988) discusses the fundamental conflict between vehicles and pedestrians, which are manifested in the scenarios above, in the context of design of curb radii. In one instance, curb radii may be lengthened to better accommodate turning vehicles; in all likelihood, turning speed will be increased, as well as the distance that a pedestrian will have to cross. Hauer also notes wide variations in walking (crossing) speeds, although 1.2 m/s (4.0 ft/s) is typically assumed as the standard. He cites several sources where recommended speeds are in the 0.76 to 1.13 m/s (2.5 to 3.7 ft/s) range, which purport to represent an 85th percentile speed accommodating older persons. Other studies reported by Hauer showed that speeds varied from an 85th percentile speed of about 0.67 m/s (2.2 ft/s) but a maximum of about 2.1 m/s (6.9 ft/s). By way of sample calculations, he also notes that for a 18-m (60-ft) wide street, a pedestrian moving at 0.91 m/s (3.0 ft/s) takes 20 s versus 15 s at 1.2 m/s (4.0 ft/s). Fruin, Ketchame, and Hecht (1988) studied four intersections in Manhattan (New York City) and recommended that a median walking speed of 1 m/s (3.3 ft/s) should be assumed (rather than the "current" 1.37 m/s [4.5 ft/s]) but that a 3.0 s start-up time at the curb (in response to a "go" signal) should be allowed. For an 18-m (60-ft) wide street, the differences in the two methods would yield: 13.3 s crossing time at 1.37 m/s (4.5 ft/s), and 21.2 s for 1 m/s (3.3 ft/s) plus 3.0 s for start-up. This would result in a 7.9 s increase in the allocation of green time to the minor street if the pedestrian crossing time of the major street was the limiting criterion.
More recently, Hoxie and Rubenstein (1994) measured the crossing times of older and younger pedestrians at a 21.85-m (71.7-ft) wide intersection in Los Angeles, CA and found that older pedestrians (age 65+) took significantly longer than younger pedestrians to cross the street. In this study, the average walking speed of the older pedestrians was 0.86 m/s (2.82 ft/s), with a standard deviation of 0.17 m/s (0.56 ft/s), versus 1.27 m/s (4.17 ft/s), with a standard deviation of 0.17 m/s (0.56 ft/s) for the younger pedestrians. Of the 592 older pedestrians observed, 27 percent were unable to reach the curb before the light changed to allow cross traffic to enter the intersection, and one-fourth of this group were stranded by at least a full traffic lane away from safety.
Knoblauch, Nitzburg, Dewar, Templer, and Pietrucha (1995) conducted a series of field studies to quantify the walking speed, startup time, and stride length of young/middle-aged pedestrians and pedestrians age 65 and older under varying environmental conditions. Analysis of the walking speeds of 3,458 pedestrians under age 65 and 3,665 pedestrians age 65 or older crossing at intersections showed that the mean walking speed for younger pedestrians was 1.51 m/s (4.95 ft/s) and for older pedestrians was 1.25 m/s (4.11 ft/s). The 15th percentile speeds were 1.25 m/s and 0.97 m/s (4.09 ft/s and 3.19 ft/s) for younger and older pedestrians, respectively. These differences were significant at the 0.05 level. Among the many additional findings with regard to walking speed were the following: pedestrians who start on the WALK signal walk slower than those who cross on either the flashing DON'T WALK or steady DON'T WALK; the slowest walking speeds were found on local streets while the faster walking speeds were found on collector-distributors; sites with symbolic pedestrian signals had slower speeds than sites with word messages; pedestrians walk faster where RTOR is not permitted, where there is a median, and where there are curb cuts; faster crossing speeds were found at sites with moderate traffic volumes than at sites with low or high vehicle volumes. For design purposes, a separate analysis was conducted for pedestrians who complied with the signal, as they tended to walk slower than those who crossed illegally. The mean crossing speed for the young compliers was 1.46 m/s (4.79 ft/s) and for the older compliers was 1.20 m/s (3.94 ft/s). The 15th percentile speed for the young compliers was 1.21 m/s (3.97 ft/s), and was 0.94 m/s (3.08 ft/s) for the older compliers. Older female compliers showed the slowest walking speeds, with a mean speed of 1.14 m/s (3.74 ft/s) and a 15th percentile of 0.91 m/s (2.97 ft/s). One of the slowest 15th percentile values [0.89 m/s (2.94 ft/s)] was observed for older pedestrians crossing snow-covered roadways. It was concluded from this research that a mean design speed of 1.22 m/s (4.0 ft/s) is appropriate, and where a 15th percentile is appropriate, a walking speed of 0.91 m/s (3.0 ft/s) is reasonable. It was also determined by Knoblauch et al. (1995) that the slower walking speed of older pedestrians is due largely to their shorter stride lengths. The stride lengths of all older pedestrians are approximately 86 percent of younger pedestrians.
Knoblauch et al. (1995) also measured start-up times for younger and older pedestrians who stopped at the curb and waited for the signal to change before starting to cross. The mean value for younger pedestrians was 1.93 s compared to 2.48 s for older pedestrians. The 85th percentile value of 3.06 s was obtained for younger pedestrians, compared to 3.76 s for older pedestrians. For design purposes, the authors concluded that a mean value of 2.5 s and an 85th percentile value of 3.75 s would be appropriate. These data specifically did not include pedestrians using a tripod cane, a walker, or two canes; people in wheelchairs; or people walking bikes or dogs. The MUTCD (1988) states that under normal conditions, the WALK interval should be at least 4 to 7 s in length so that pedestrians will have adequate opportunity to leave the curb before the clearance interval is shown. Parsonson (1992) noted that the reason this much time is needed is because many pedestrians waiting at the curb watch the traffic, and not the signals. When they see conflicting traffic coming to a stop, they will then look at the signal to check that it has changed in their favor. If they are waiting at a right hand curb, they will often take time to glance to their left rear to see if an entering vehicle is about to make a right turn across their path. He reports that a pedestrian reasonably close to the curb and alert to a normal degree, was observed to require up to 4 or 5 s for this reaction, timed from when the signal changes to indicate that it is safe to cross, to stepping off the curb. It may be remembered that older pedestrians stand farther away from the curb, and may or may not be alert. In addition, there are many drivers who run the amber and red signals, and it is prudent for pedestrians to "double check" that traffic has indeed obeyed the traffic signal, and that there are no vehicles turning right on red or (permissive) left on green before proceeding into the crosswalk. Because older persons have difficulty dividing attention, this scanning and decision-making process requires more time than it would for a younger pedestrian.
Parsonson (1992) reports that the state of Delaware has found that pedestrians do not react well to the short WALK and long flashing DON'T WALK timing pattern. They equate the flashing phase with a vehicle yellow (clearance) period. The Florida DOT and the city of Durham, Ontario, provide sufficient WALK time for the pedestrian to reach the middle of the street, under the assumption that the pedestrian will not turn around when the flashing DON'T WALK phase begins.
Somewhat surprisingly, several pedestrian-oriented studies make no overt mention of differential problems that may relate to older persons, or handle them in either an indirect or very general fashion. These include a review of an FHWA safety program (Vallette and McDivitt, 1981) and the presentation of a methodology for evaluating the feasibility of grade- separated pedestrian crossings (Lindley, 1985). Two other studies were directed to examination of pedestrian signals (Khasnabis, Zegeer, and Cynecki, 1982) and RTOR accidents involving pedestrians (Zegeer and Cynecki, 1985). The work by Khasnabis et al. was a literature review of other work and noted that few pedestrian-oriented studies actually related accident savings to signalization and were more likely to be concerned with compliance with signals: pedestrian signals may result in delay to both pedestrians and vehicles, and in many instances compliance is quite low. Low compliance and understanding was encountered for both specific groups (students) and for certain signal indications, (e.g., flashing WALK). The latter result was also noted by Zegeer, Opeila, and Cynecki (1982). Of note is that none of the numerous suggestions for further study relate to older persons. In the latter study, several different sign and marking configurations were tested (e.g., "no turn on red," "no turn on red (ball)," "no turn on red when pedestrians are present," and, for pedestrians a marking on the street "look for turning vehicles"). Virtually all signs and markings were effective in reducing conflicts and were recommended in various situations. However, no age-related data were presented and there were no comments made about relating pedestrian crossings to reduced operational efficiency.
More recently, however, there have been several surveys, accident analyses, and focus groups conducted with older pedestrians to document their difficulties. In a survey of older pedestrians in the Orlando, FL area, 25 percent of the participants reported difficulty seeing the crosswalk signal from the opposite side of the street (Bailey, Jones, Stout, Bailey, Kass, and Morgan, 1992). Older pedestrians have been observed to wait for longer gaps between vehicles before attempting to cross the road. In one study, approximately 85 percent of the pedestrians age 60 years or older required a minimum gap of 9 s before crossing the road, while only 63 percent of all pedestrians required this minimum duration (Tobey et al., 1983). The decline in depth perception may contribute to older persons' reduced ability to judge gaps in oncoming traffic. It may be concluded from these studies that older pedestrians do not process information (presence, speed, and distance of other vehicles) as efficiently as younger pedestrians, and therefore require more time to reach a decision. Other researchers have observed that older pedestrians do not plan their traffic behavior, are too trusting about traffic rules, fail to check for oncoming traffic before crossing at intersections, underestimate the speed of approaching vehicles, and follow other pedestrians without first checking for conflicts before crossing (Mathey, 1983; Jonah and Engel, 1983). Knoblauch, Nitzburg, Dewar, Templer, and Pietrucha (1995) report that locating the curb accurately and placing the foot is a matter of some care, particularly for the elderly, the very young, and the physically handicapped.
Harrell (1990) used distance stood from the curb as a measure of pedestrian risk for intersection crossing. Observations of 696 pedestrians divided among three age groups (less than 30 years old, age 31 to 50, and age 51+) showed that the oldest group stood the farthest from the curb, that they stood even further back under nighttime conditions, and that older females stood the farthest distance from the curb. The author uses this data to dispel the findings in the older pedestrian literature that they are not cognizant of the risks of exposure to injury from passing vehicles. Similarly, it may be argued that this behavior keeps them from detecting potential conflict vehicles and makes speed and distance judgments more difficult for them, while at the same time limiting their conspicuity to approaching drivers who might otherwise slow down if pedestrians were detected standing at the curbside at a crosswalk.
A study of pedestrian accidents conducted at 31 high pedestrian accident sections in Maryland between 1974 and 1976, showed that pedestrians 60 years old or older were involved in 53 (9.6 percent) of the accidents, and children under 12 showed the same proportions. The pedestrians age 60 and older accounted for 25.6 percent of the fatal accidents. Compliance with traffic control devices was found to be poor for all pedestrians at all study locations; it was also found that most pedestrians keyed on the moving vehicle rather than on the traffic and pedestrian control devices. Only when the traffic volumes were so high that it was impossible to cross did pedestrians rely on traffic control devices (Bush, 1986).
Garber and Srinivasan (1991) conducted a study of 2,550 accidents involving pedestrians which occurred in the rural and urban areas of Virginia to identify intersection geometric characteristics and intersection traffic control devices which were predominant in crashes involving elderly pedestrians. Accident frequency by location and age for the accidents within the cities showed that while the highest percentage of accidents involving pedestrians aged 59 years or less occurred within 45.7 m (150 ft) from the intersection stop line, the highest percentage of accidents for pedestrians age 60 years or older (51.8 percent) occurred within the intersection.
More recently, Knoblauch, Nitzburg, Reinfurt, Council, Zegeer, and Popkin (1995) reported that older adults are over-involved in crashes while crossing streets at intersections compared to younger pedestrians. In their earlier analysis of the national Fatal Accident Reporting System (FARS) data for the period of 1980 through 1989, 32.2 and 35.3 percent of the deaths for pedestrians ages 65 to 74 and age 75 or older, respectively, occurred at intersections (Reinfurt, Council, Zegeer, and Popkin, 1992). This compared to 22 percent or less for the younger age groups. Analysis of the North Carolina motor vehicle crash file for 1980 through 1990 displayed somewhat smaller percentages, but showed the trend of increasing pedestrian accidents at intersections as age increased. Further analysis of the North Carolina data base showed that pedestrians age 65 or older as well as those ages 45 to 64 experienced 37 percent of their accidents on roadways with four or more lanes. This compares to 23.7 percent for pedestrians ages 10 to 44 and 13.6 percent for those age 9 years or younger. Older pedestrians were over-represented in fatal crashes involving a left turning vehicle; these crashes accounted for 5.4 percent of the U.S. fatal crashes with pedestrians age 65 or older and 1.0 percent of cases involving younger pedestrians. Older pedestrians were also slightly over-represented in right-turn crashes, including right-turn-on-red. The North Carolina file showed that right-turn crashes in general accounted for 5.7 percent of crashes for pedestrians age 65 or older, compared to 2.7 percent for the other age groups. Older pedestrians were also over-represented in crashes involving a backing vehicle.
Habib (1980) reported a 2 to 1 ratio of left-turn to right-turn pedestrian accidents in a study conducted on Manhattan Island. The major contributing causes of left-turn pedestrian accidents in this study were poor visibility from within the vehicle (blockage by the driver side A-pillar) and bad driving habits. Possible causes provided for older pedestrians' over-representation in left turning crashes included: (1) their slower walking speed; (2) their lack of awareness that vehicles may be turning left during their WALK signal or on the green light; (3) they may inadequately search for left-or-right turning vehicles prior to stepping off the curb; (4) their inability to react quickly when a left-turning vehicle enters their path; (5) their misunderstanding of the flashing WALK signal; and (6) their over-reliance on signals to protect them (Knoblauch et al., 1995). Van Houten and Malenfant (1995) reported that near-hit conflicts associated with vehicles turning right on green and right on red combined were only 88 percent of the number associated with vehicles turning left on green. Prohibition of left turns is a practice used when pedestrian volumes are high, opposing traffic is heavy, and/or when separate turning lanes are not available. This prohibition reduces the conflicts and potential accidents between left-turning vehicles and pedestrians, although it disrupts traffic patterns, and often simply migrates the problem to the next intersection (Reinfurt, Council, Zegeer, and Popkin, 1992). Another countermeasure is to provide exclusive left-turn phasing. Exclusive left-turn phasing, however, will only reduce conflicts with pedestrians when pedestrians are provided with a DON'T WALK signal during the interval when the green arrow is given to left-turning motorists, and pedestrians comply with the signal (Zegeer and Zegeer, 1988). Van Houten and Malenfant (1995) found that signs erected next to the pedestrian signal, and prompts painted in the crosswalk instructing pedestrians to look for turning vehicles increased the percentage of pedestrians looking for all threats and almost eliminated conflicts between motorists and turning vehicles.
With respect to signalization for pedestrians, Zegeer, Opiela, and Cynecki (1982) conducted an accident analysis to determine whether pedestrian accidents are significantly affected by the presence of pedestrian signals and by different signal timing strategies. They found no significant differences in pedestrian accidents between intersections that had standard-timed (concurrent walk) pedestrian signals compared with intersections that had no pedestrian signals. Concurrent or standard timing provides for pedestrians to walk concurrently (parallel) with traffic flow on the WALK signal. Vehicles are generally permitted to turn right (or left) on a green light while pedestrians are crossing on the WALK interval. Other timing strategies include early release timing, late release timing and exclusive timing. In early release timing, the pedestrian WALK indication is given before the parallel traffic is given a green light, allowing pedestrians to get a head start into the crosswalk before vehicles are permitted to turn. In late release timing, the pedestrians are held until a portion of the parallel traffic has turned. Exclusive pedestrian phasing is a countermeasure where traffic signals are used to stop motor vehicle traffic in all directions simultaneously for a phase each cycle, while pedestrians are allowed to cross the street. Barnes Dance or scramble timing is a type of exclusive timing where pedestrians may also cross diagonally in addition to crossing the street. Exclusive timing is intended to virtually eliminate turning traffic or other movements which conflict with pedestrians while they cross the street. This timing strategy causes excessive delays to both motorists and pedestrians, however; in the Zegeer et al. (1982) analysis, exclusive-timed locations were associated with a 50 percent decrease in pedestrian accidents for intersections with moderate-to-high pedestrian volumes when compared to both standard-timed intersections and intersections that had no pedestrian signals. Elderly road users (age 65 or over) recommended the following pedestrian-related countermeasures for pedestrian signs and signals, during focus group sessions held as a part of the research conducted by Knoblauch, Nitzburg, Reinfurt, Council, Zegeer, and Popkin (1995): (1) reevaluate the length of pedestrian walk signals due to increasingly wider highways; (2) implement more Barnes Dance signals at major intersections; and (3) provide more "yield to pedestrians" signs in the vicinity of heavy pedestrian traffic.
Zegeer and Cynecki (1986) tested a "look for turning vehicles" pavement marking in a crosswalk, as a low-cost countermeasure to remind pedestrians to be alert for turning vehicles, including RTOR vehicles. Results showed an overall reduction in conflicts and interactions for RTOR vehicles and also for total turning vehicles (RTOR and RTOG). Even with a RTOR prohibition, researchers have found that approximately 20 percent of motorists commit a RTOR violation when given the opportunity (Zegeer and Cynecki, 1986). Of those violators, about 23.4 percent result in conflicts with pedestrians or vehicles on the side street.
Zegeer, Stutts, Huang, Zhou, and Rodgman (1993) report that the use of audible signals may be helpful for visually impaired pedestrians to indicate when the WALK signal is actuated. However, the signal only indicates when the walk interval is activated, and does not guarantee that motorists will not run red lights or turn across a pedestrian's path from a side street (RTOR). Although audible signals have been installed in many U.S. cities (St. Paul, MN; San Diego, CA; and Harrisburg, PA, to name a few), some organizations providing services to the blind are opposed to audible signals. In response to Carlson and Weiss's (1983) recommendations for senior citizen and handicapped pedestrians, the National Federation for the Blind issued a resolution in which they "condemn and deplore the use of buzzers and bird calls as a travel aid for the blind" because a blind person's awareness of approaching vehicles and ability to judge the movement of traffic assures his or her safety. The use of buzzers provides no additional information to the pedestrian, but sends a message to the public reinforcing negative societal attitudes surrounding the competence of blind people. The Minnesota Chapter of the National Federation for the blind states that audible signals pose a serious hazard, because the sounds emitted are a distraction and obscure the vehicular sounds a visually-impaired individual must hear in order to determine whether it is safe to cross. Oliver (1989) notes that audible pedestrian signals are only feasible at certain complex and confusing intersections frequented by visually impaired individuals. These include public transit stops, shopping centers, and medical and educational facilities. Among the criteria he found necessary for their operation were that they: must not be annoying to the average pedestrian or resident; must have sound levels from 10 to 120 dB; must have upper and lower sound levels; must sound only when the walk signal is displayed; must have a distinct sound for each direction; and should be actuated by either pedestrian, timer, or both. He emphasized that audible devices are not meant to be a substitute for a visually impaired person's orientation and mobility skills, but rather an aid to them.
Wilson (1980) conducted a before and after study on the effects of installing an audible pedestrian signal at a signalized intersection in the United Kingdom. The following behaviors were noted after installation of the signal: pedestrian crossing time decreased by 5 percent during the "green man" phase; pedestrian delay after the onset of the "green man" phase decreased by 20 percent; and there was a significant reduction in the proportion of pedestrians failing to complete their crossing before the vehicle green phase began. There was no separate analysis of blind pedestrian crossing behavior in this study. A limitation of audible signals noted by Uslan, Peck and Waddell (1988) is that blind pedestrians have difficulty locating the proper pedestrian signal pole and push button. They conducted a study of 27 blind pedestrians crossing intersections with and without audible signals, and found that the audible signals did in fact, aid the blind pedestrians at complex intersections, however, at complex intersections, audible signals require intensified listening, necessitating instruction and practice in their use.
More recently, Van Houten, Malenfant, and Van Houten (1996) evaluated the effectiveness of an experimental auditory pedestrian signal designed to prompt pedestrians to look for turning vehicles. A horn speaker was erected at each crosswalk at an intersection in the downtown area of Clearwater, FL, which played the following message only on the corner where the crosswalk button was pushed: "Please wait for walk signal." Then, 200 ms before the walk signals were illuminated for traffic crossing Cleveland Street, the following message was played at all four corners: "Look for turning vehicles while crossing Cleveland Street," and 200 ms before the walk signals were illuminated for traffic crossing Garden Avenue, the following message was played on all four corners: "Look for turning vehicles while crossing Garden Avenue." Messages were digitally recorded in English using an adult female's voice in one test condition, and a male child's voice in another condition. Pedestrians were scored for checking the three potential threat turning vehicle paths (right on red, left on green, and right on green) as they crossed the roadway, and the number of motor vehicle-pedestrian conflicts was also recorded. A reversal design was employed where baseline data were collected, followed by the adult auditory signal, then a second baseline period followed by a re-introduction of the adult auditory signal, then the child auditory signal, and finally the adult auditory signal. During the initial baseline phase, 16.3 percent of the pedestrians did not look for any threats and the number of conflicts per 48 pedestrians per session averaged 1. Results showed that the introduction of the female adult auditory signal was associated with a 75 percent decrease in the percentage of pedestrians who did not look for threats (4.2 percent did not look compared to 16.3 percent during the baseline), and a 75 percent reduction in conflicts, compared to the baseline condition. The reintroduction of the female adult auditory signal was associated with a decrease of 76 percent in the percentage of pedestrians who did not look for threats and a decrease of 100 percent in the average number of conflicts per session, compared to the initial baseline. The introduction of the child's voice was associated with a lower percentage of pedestrians who did not look for threats (only 2.5 percent did not look) and the elimination of vehicle-pedestrian conflicts. The reintroduction of the adult female's voice was associated with a slightly higher percentage who did not look for threats (3.2 percent), but there were no vehicle-pedestrian conflicts. Subjective comments provided by a sample of 100 pedestrians indicated that 71 percent thought the signal was a good idea and was helpful. The sample included four visually-impaired pedestrians (no age was provided in the report) who each said that it was very helpful to them because they could tell when the signal changed.
Pedestrian volumes can be used to warrant a signal and, in turn, pedestrian characteristics can be used to modify signals when they are placed. Again, as an example, the Michigan Manual of Uniform Traffic Control Devices (1994) provides a warrant for a signal based on pedestrian volumes but states that for the major street crossing, they "may be reduced as much as 50 percent ... when the predominant pedestrian crossing speed is below 1 m/s (3.5 ft/s)." An additional warrant is provided for school children and based on gaps in the traffic stream that would allow groups of children to cross safely. This is of interest primarily because of the implicit assumption in the formula of the pedestrian walking speed of 1.2 m/s (4.0 ft/s). Interestingly, the manual indicates that the WALK indication interval does not need to "equal or exceed the total crossing time calculated for the street width, as many pedestrians will complete their crossing during the flashing DON'T WALK clearance interval." It has been anecdotally noted that some pedestrians, and especially older persons, will turn back when the flashing DON'T WALK indication begins. The manual also notes that crossing times and signal duration be calculated so that the pedestrian can "leave the curb and travel to the center of the farthest traveled lane before opposing vehicles received a green indication (normal walking time is assumed to be 1.2 m/s (4 ft/s)." For the older person, who might be walking at speeds substantially less than 1.2 m/s (4.0 ft/s), the signal sequence may well be very confusing as part way across the intersection they are "told" DON'T WALK in a flashing mode, and then, before they reach the opposite curb, in a steady mode—even for those who immediately begin crossing on the WALK signal.
As another point of reference, the Highway Capacity Manual (1985) equates the following pedestrian walk speeds to different levels of service: A = 1.3 m/s (4.3 ft/s); B = 1.28 m/s (4.2 ft/s); C = 1.2 m/s (4.0 ft/s); D = 1.16 m/s (3.8 ft/s); E = 0.76 m/s (2.5 ft/s); and F < 0.76 m/s (2.5 ft/s).
While it is not known how widespread the practice is of accommodating older pedestrians (and presumed lower walking speeds), anecdotal comments received from city traffic engineers in Michigan and Ohio (especially those from jurisdictions in major metropolitan areas such as Detroit and Cleveland) indicate that there is considerable pressure to provide some sort of special treatment in, at least, spot locations (e.g., near "elderly housing" projects or retirement developments). It is part of the Policy for the Older Road-User Program in Florida (Florida DOT, 1991) to consider "slower walk speeds for signal timing and increasing the use of refuge islands at spot locations" in 22 Florida counties. Indeed, in Dade County (Miami) the characteristics of the population using specific intersections are considered and changes are made to accommodate older pedestrians notwithstanding the likely degradation of other operational characteristics (e.g., intersection LOS) (Pivnik 1994). Zegeer and Zegeer (1988) stress the importance of "tailoring" the most appropriate traffic control measures to suit the conditions at a given site. This is because the effect of any traffic control measure is highly dependent on specific locational characteristics, such as traffic conditions (e.g., volumes, speeds, turning movements), pedestrian volumes and pedestrian mix (e.g., young children, college students, older adults, handicapped), street width, existing traffic controls, area type (e.g., rural, urban, suburban), site distance, accident patterns, presence of enforcement, and numerous other factors.
In summary, in spite of some relatively aggressive attempts at accommodation of different pedestrian populations, the literature does not indicate that the full extent of the operational trade-off’s between accommodating older pedestrians and degradation of level of service for vehicular traffic has been resolved or, in many cases, has even been considered. Similarly, the impacts of the introduction of "special" operational characteristics at a given intersection in a system of intersections (progressive signals) have apparently not been investigated in any systematic manner.
At an unsignalized intersection, the situation is somewhat different. While other vehicles and any pedestrians must be monitored, there is (obviously) no signal. Thus, the driver must be more alert to other vehicles and more closely monitor his/her position and speed. Unsignalized intersections would also typically require more active driver movements (e.g., head-turns) in the monitoring process. To the extent that an older driver has difficulties in, for example, judging gap times or even in keeping track of oncoming vehicles, there will be some degradation of LOS for the intersection. This issue is explored below.
One of the problems considered is gap acceptance of drivers at T-intersections. The attention afforded this situation in studies is presumably derived, in part, from the lack of confounding which would be present if oncoming traffic also had to be accounted for by the driver. In this sense, the T-intersection is experimentally cleaner, and results would apply to four-way intersections when opposing traffic was not present.
Darzentas, McDowell, and Cooper (1980) undertook a study of gap acceptance for the simple turning maneuver with subjects classified by age (31 to 40 versus 61 to 70) and gender (although data had been previously reported by others). They found that the minimum acceptable gap is negatively correlated with speed of the approaching vehicle for each class of drivers, that females are more conservative than males, and that the judgement of older females is "appreciably worse than that of other classes of drivers." While the original data were not provided, results of their regressions of minimum gap acceptance with approach speed (ft/s) showed the following estimates: at about 48 km/h (30 mi/h), older females would accept a minimum gap of 11+ s, young females about 10.4 s, older males about 7.8 s, and younger males about 6.8 s. At about 88 km/h (55 mi/h), the comparable figures were 8+ s for all females (younger females were slightly higher—the lines had crossed) and about 6 s for all males (the lines crossed at about this speed). Older drivers showed more variation than younger drivers, males less than females, and there were not statistically significant differences between young and old males. As an interesting aside, the authors do not discuss any possible cohort effect for the older women (e.g., older women in 1980 may not be as experienced or comfortable with driving as women the same age today because of changing roles in society and long-term driving habits). In other work, Darzentas and McDowell (1981) looked at variations between day and night behavior (although without age differentiation) and found that drivers behave more conservatively regarding far-side gaps at night (gaps are 1.0 s longer), whereas there are no differences for near-side gap acceptance between day and night conditions.
In a more recent study, Hunter-Zaworski (1990) examined both younger and older drivers as well as those with and without restricted range-of-neck movement in a simulator-based experiment. Decision time increases with age. Only younger persons were found to be able to overcome their impairment. The overall decision times were: 14.4 s for impaired older persons, 12.1 s for unimpaired older persons, 11.4 s for impaired younger persons, and 11.3 s for unimpaired younger persons. There were no differences in standard deviation by age, although they were somewhat higher for all impaired persons.
In one of the few papers reviewed that explicitly relates changes in driver behavior to operational considerations, Kettelson and Vandehey (1991) argue that use of median gap acceptance times in calculating capacity and LOS of minor movements at unsignalized intersections is biased, primarily because virtually all gaps greater than 14 to 15 s are accepted. Moreover, drivers were observed to accept shorter gaps as a function of how long they had already waited—e.g., a driver who waits long enough will likely eventually accept a gap which is shorter than one he/she already rejected. That is, the critical gap to be accepted is not constant (as is assumed in the Highway Capacity Manual). Thus, HCM calculations will be in error until situational influences, as noted above, can be taken into account.
Fitzpatrick (1991) examined the gap acceptance behavior of auto and truck drivers and found that a gap of 8.25 s was accepted by 85 percent of the automobile drivers for both right and left turns at a moderate-to-high volume intersection. For low volume situations (with slightly different geometry), a gap of 10.5 s was accepted by 85 percent. For truck drivers, the respective gaps were 10.0 and 15.0 s. Fitzpatrick also compares these results with those of several other gap acceptance studies and concludes, in general, that while these results are comparable, the HCM appears to be on the low side. The general difficulties of comparing different studies because of differing geometric characteristics are also noted.
Overall, the work on gap acceptance that was reviewed directly or indirectly (e.g., earlier work was summarized in Fitzpatrick's work) has indicated that there are most likely differences in the behavior of older and younger drivers at unsignalized intersections and that, notwithstanding these differences, calculations of operations-related parameters (e.g., capacity) are probably already subject to some error because of the underestimation of gap acceptance embodied in standard reference materials such as the Highway Capacity Manual. Not only will increasing numbers of older drivers in the population have an effect, but the assumption of a constant critical gap appears unwarranted as the result of observations that it may decrease with increasing wait time, which would lead to an underestimation of capacity.
Non-Intersection (Uninterrupted Flow) Operations
Non-intersection (link) operations also have the potential to be affected by varying characteristics of driving behavior, particularly those that are commonly associated with older drivers. Operating a vehicle on the highway system at non-intersection locations is generally a less complex task (than at an intersection) and the judgments that drivers must make are more straightforward. The logical list of concerns includes the following: operating speed; interaction with other vehicles, including car-following behavior on simple tangent sections; operations on horizontal curves; stopping, passing, intersection, and decision sight distances as they relate to operations; and merging and weaving at freeway ramps.
Speed and Density. The speed at which a vehicle operates is a function of the driver's perceived capabilities of the vehicle, the characteristics of the environment (roadway system), and perceptions about his/her own driving capabilities and skills. Based on reviews of accident and violation statistics and the self-reports of the drivers themselves, younger drivers tend to drive more aggressively and faster while older drivers tend to drive more slowly.
The question that remains is how lower speeds affect overall operation of the roadway. The answer depends on the kind of system that is being considered. For a freeway-level facility, slower drivers will typically cause more passing maneuvers in relatively free-flowing traffic, although it seems unlikely that the LOS would be affected in a significant way. As the traffic volumes increase, there will be more platoon formation and there may be more problematic behavior (e.g., drivers passing on the right, more drivers weaving through slower-moving traffic) and some delays might even result although this, of course, depends on the numbers of drivers in various age groups that are present. Similar behavior would occur on two-lane roads, although the slower driver would have a more immediate effect as the passing opportunities would, generally, be more limited and platoons would form more immediately and last longer (depending, of course, on the volume).
Work done in support of the HCM on passenger car equivalents and, to a degree, passing lanes indicates that insertion of a slow-moving vehicle into the traffic stream has a multiplicative impact (it has the same effect as introducing more than one vehicle into the stream)—whether that vehicle is moving slowly as a result of performance characteristics (e.g., a heavy truck that loses speed on an upgrade) or because of human factors (i.e., its driver does not want to, or is not competent to drive at higher speeds. Unfortunately, most of the vehicle equivalency work is empirical in nature, based on observation in the field; nothing similar has been attempted simply using slower-moving passenger vehicles, although it would not be difficult to construct a mathematical model which could be used to predict such equivalencies.
As mentioned, work on passing lanes and the passing maneuvers can also provide some insight into the effect of slower moving vehicles (in this instance hypothesized to be driven by older persons) on the traffic stream. Although the literature in these areas was not comprehensively reviewed, two relatively recent papers are of interest. Harwood, St. John, and Warren (1985) studied passing lanes on two-way, two-lane highways—both in terms of adding short four-lane sections and standard passing lanes (i.e., a three-lane section). They found that implementation of these measures led both to increased passing rates (more than twice) and reduced platooning (less than one-half) relative to the two-lane sections. In addition, they found that operational effects persisted for several miles downstream from the treatment. Speeds were found to increase somewhat (3.5 km/h [2.2 mi/h]) in the treated area, but it was not possible to assess upstream and downstream speed effects per se since there was considerable variance due to different geometry and other roadway characteristics. An investigation by Kaub (1990) implies that the "need" to pass becomes greater as volumes increase (more platoons form) and shows that the passing maneuvers themselves become more problematic as the perceived and real safety declines, as measured by the number of aborted passing attempts. An abort was identified as an overtaking vehicle going into the oncoming lane and then pulling back into line without passing. Interestingly, work by Lyles (1981) on passing-zone TCD’s showed that aborted passes could be reduced by more judicious use of passing-zone signs. However, the overall point is that, not unexpectedly, slower-moving vehicles (regardless of why they are slower moving) tend to create platoons of vehicles in traffic, and the effect increases with increasing same-direction and opposing volumes. Geometry which allows for more passing zones and passing lanes tends to reduce the platooning effect and, in turn, increases real and perceived levels of operation. The installation of passing lanes is typically studied in terms of delay and accident reduction and related to vehicle mix on the presumption that differential vehicle performance causes the speed differential (which, in turn, causes platooning and so forth). If the speed differentials that are encountered by virtue of the mix of drivers in the field can be predicted, that too could become a parameter which could warrant passing lane installation.
Related to speed is vehicle density, which can be operationalized as the distance between vehicles. Much of the laboratory work that deals with time-to-collision is done in the context of a moving vehicle overtaking a vehicle ahead, or estimating this parameter relative to a vehicle that is approaching (e.g., in the context of making a left turn across the path of an oncoming vehicle). Hoffmann and Mortimer (1994) have done one of the few studies where time to collision is measured between two moving vehicles. They note that estimates of time to collision were about 0.8 of actual (versus 0.6 reported elsewhere). This would explain the observed propensity of many motorists to follow too closely. Their results were not, however, differentiated by age group. Moreover, it would seem that older drivers, by virtue of their slower speeds are less likely to be in a following mode—rather they would be the followed.
While there is both analytical and anecdotal evidence that older drivers drive more slowly than the "average" driver, and it is logical to expect that if there were enough slower drivers that operations would degrade (i.e., a lower level of service), the operational effects of such behavior have not been widely discussed in the literature nor has the magnitude of the effect been estimated.
Control, Guidance, and Navigation. Control, guidance, and navigation are the three principal components of the driving task as defined by Alexander and Lunenfeld (1986). Using an upcoming freeway interchange as an example, from an operational perspective the driver needs to monitor the vehicle status (e.g., speed, turn signals), control the vehicle within a lane, speed up and/or slow down to merge or weave, and select the appropriate lane to achieve the purposes of the trip while not impinging on other drivers. The literature that was identified in the search provided some insight on operations-related issues and illustrates the lack of work that has been done that explicitly relates such issues to operational concerns per se.
It should be noted that at the grossest level of trip planning (e.g., overall route selection), King (1986, 1986a) found that there were no differences by age or driving experience in route selection behavior, and Wenger, Spyridakis, Haselkorn, Barfield, and Conquest (1990) found that differing behavior in terms of route diversion could not be explained by driver age. However, there is at least anecdotal evidence that older persons may plan their trips to avoid certain kinds of roadways, unfamiliar routes, and certain conditions (e.g., night driving, poor weather).
Driving Task and ITS Issues
Of recent concern at the level of vehicle control is the interaction between the vehicle and the driver as a result of the renewed interest in automated vehicles and highways. Hancock and Parasuraman (1992) and Sheridan (1991) provide an overview of the issues involved. Hancock and Parasuraman argue that the debate over intelligent transportation systems (ITS) is not dissimilar from that which occurred with respect to aircraft control some time ago. The issues (from both papers) include: regulation of driver mental workload; keeping the driver in the loop; design of in-vehicle navigation aids; in-car display conflicts; matching ITS to individual and group differences in driver behavior; training and licensure; evolution and integration of technology; and traffic management and information trust. The paper essentially sets forth an agenda that should be reviewed. Of primary interest to the effort here is that it is not currently known how older drivers of either this or the next generation will react to the significantly different in-vehicle driving environment that ITS technologies will offer. The extent to which older persons choose (or are permitted to choose) to interact with ITS features may have significant impact on the operations-related outcomes. It is important to note, however, when comparisons are made between the current debate and that which occurred over aircraft-related control issues, that while the tasks for an aircraft may be much more complex, the training for the "average" driver is disproportionately less.
Related to the general issues just noted is recent work by Walker, Alicandri, Sedney, and Roberts (1990) which was directed to a simulator-based study of driver response in route selection and navigation, with the navigation aids ranging from strip maps to more complex audio and visual devices. Notwithstanding the fact that the older drivers in the test sample were probably better than normal because of self-selection biases, older persons in this study drove more slowly; had larger variability in lateral placement; had longer reaction times to vehicle gauge changes; drove in the "other" lane after turns; and were more likely to make navigational errors. It was also noted that there was less within-group variation in results than had been reported elsewhere (which may have been due in part to the self-selection bias and the fact that several subjects withdrew because of "simulator sickness"). Overall, subjects were observed to not degrade in terms of lateral placement with increasing complexity, which may be due to trading off lateral placement for other factors (e.g., monitoring gauges). The relatively clear implication of this work is that the widely-discussed operations improvements which are hypothesized as a result of ITS technology may not be achieved by older persons unless it is indirectly through its use by other motorists (i.e., market penetration of ITS does not have to be anywhere near 100 percent for benefits to accrue to all users).
Related to the implementation of ITS, there is also literature that relates to the driving task directly—for example, visual searching and vehicle control. Staplin, Lococo, Sim, and Gish (1996) measured the speed of subjects' lane change decisions, in a simulated freeway driving study, and found predictable slowing in decision latency by older versus younger drivers. The effect was most pronounced for the oldest (75+) subjects, however, with the 65 to 74 age group typically closer to younger (ages 25 to 45) drivers in speed of response than to the 75+ age group. In a fairly recent paper, Taoka (1991) discusses "spare glance" duration in terms of how drivers allocate their visual search time among different tasks/stimuli. The tasks ranged from mirror glances during turning to reading roadway name signs. While specific results were not differentiated by age, Taoka asserted that 85th percentile glance times at signs (about 2.4 s) were likely too long as 2.0 s is the maximum that a driver should divert from the basic driving task. Taken together, these results imply that the older driver would more likely be represented in the tail of the distribution and, therefore, should have problems dividing attention between searching for/reading signs and the basic driving task. These results can be compared to much earlier work by, for example, Robinson, Erickson, Thurston, and Clark (1972) who, while estimating that 90 percent of the information used in driving is visual, made covert observations of random drivers at stop signs and of known subjects in freeway lane-changing situations. For the stop and turn maneuvers, drivers had search times of 1.2 s to 2.6 s and for the merging experiment used 0.8 to 1.6 s for mirror glances and 0.8 to 1.0 s for direct glances. The mirror glance durations were relatively consistent with some of Taoka's data.
Results such as Taoka's and those of Robinson et al. (1972) combined with those of Walker et al. (1990) and Staplin et al. (1996) imply that drivers, and especially older drivers, may have some difficulty appropriately allocating time and attention to different aspects of the driving task. To the extent that this is realized by the driver, there will be some compensation—e.g., drivers may slow down to be able to successfully search for the appropriate information or they may not check other aspects of vehicle operation (e.g., gauges). The results of the compensation will either merely impede operations (e.g., slowing which potentially will cause platoon formation) or may reduce safety (e.g., erratic lane use or sudden speed changes). To some degree ITS technologies may be capable of aiding the driver in both simple execution of different aspects of the driving task and in compensation although, as was noted above, the technology itself has pros and cons and it is not clear whether the overall driving task is simplified or made more complex. For example, an audible signal which tells the driver to exit at the next off-ramp moves the task of determining whether the next exit is the appropriate one from a visual to an audio-oriented task—one realm of information processing becomes more complex (audio) and one becomes simpler (visual). Whether this will be a net benefit to the driver, as well as the consequences for highway operations, is unknown at this time.