NCHRP 3-62
Guidelines for Accessible Pedestrian Signals
Prepared for
National Cooperative Highway Research Program
Transportation Research Board
National Research Council
TRANSPORTATION RESEARCH BOARDNAS-NRC PRIVILEGED DOCUMENT This Report, not released for publication, is furnished only for review to members of or participants in the work of the National Cooperative Highway Research Program (NCHRP). It is to be regarded as fully privileged, and dissemination of the information included herein must be approved by the NCHRP. |
Accessible Design for the Blind
March 2004
The research could have not been done without the cooperation and major assistance of the city of Portland Oregon. In particular, Bill Kloos provided the full cooperation of the Signals section of the Portland Department of Transportation and Jason McRobbie and Dave Grilley very creatively designed and installed the temporary pushbutton integrated accessible pedestrian signals. The authors also wish to acknowledge those individuals and organizations that provided assistance in recruiting and interviewing research participants, scheduling data collection and providing hospitality to participants: Michael Yamada, Ben Horner, Rebecca Mock, Bill West and the ARC of Multnomah County. Their able assistance facilitated a smooth and enjoyable research experience for both researchers and participants.
ACKNOWLEDGEMENT OF SPONSORSHIPThis work was sponsored by the American Association of State Highway and Transportation Officials, in cooperation with the Federal Highway Administration, and was conducted in the National Cooperative Highway Research Program, which is administered by the Transportation Research Board of the National Research Council. |
An Accessible Pedestrian Signal (APS) is a device that communicates information about pedestrian timing in non-visual format such as audible tones, verbal messages and/or vibrating surfaces.1 It is critical that any APS system provide clear information as to which crosswalk has the walk interval. Presently, APS in the
Many cities in the U.S install pedestrian pushbuttons for crosswalks in both directions on a corner on a single pole. There are various reasons for this practice. Many installations use a single pole on a corner for span wires or mast arms to support traffic signals; money is saved by placing both pushbutton units on the same pole rather than installing additional poles with associated wiring and conduit. Poles are sometimes installed at various locations on the corner due to limited right of way and AASHTO design guidelines discourage poles near the curb for vehicular safety reasons and to prevent damage to poles from turning trucks.
The Manual on Uniform Traffic Control Devices (MUTCD) 4E.08 says “pedestrian pushbutton detectors should be capable of easy activation and conveniently located near each end of the crosswalks”2. However, poles where pushbuttons are located vary greatly in distance from the crosswalk. MUTCD 4.E.09 provides that “At corners of signalized locations with accessible pedestrian signals where two pedestrian pushbuttons are provided, the pushbuttons should be separated by a distance of at least 3 m (10 ft)”, and pushbuttons for APS should be located “Adjacent to a level all-weather surface to provide access from a wheelchair, and where there is an all-weather surface, wheelchair accessible route to the ramp; within 1.5 m (5 ft) of the crosswalk extended; within 3 m (10 ft) of the edge of the curb, shoulder, or pavement; and parallel to the crosswalk to be used.”3 The
This research is part of the National Cooperative Highway Research Program Project 3-62, Guidelines for Accessible Pedestrian Signals, which is developing future recommendations, guidelines, and standards for pushbutton placement and for APS installation. This experiment was undertaken to build on the knowledge base of why APS are needed and understand the importance of appropriate installation techniques. The panel overseeing this project requested additional research on the question of whether the installation of APS on two poles separated by at least ten feet was necessary, or if there were options for installation of both devices on a single pole that would provide unambiguous WALK information. This study is the first part of that investigation. Data were also collected on the behaviors of the general pedestrian population with the 1- and 2-pole configurations. The results of those observations are still being analyzed and will be provided in a subsequent report.
The primary purpose of this study was to determine the effect of three factors on the ability of pedestrians with visual or cognitive impairments to determine which crosswalk at a corner had the WALK signal. The factors under investigation were: 1.) whether there were significant advantages to installing two poles on a corner, each with its own pushbutton-integrated APS, over installing two pushbutton-integrated APS on a single pole, 2.) whether proximity of these poles to the curb influenced the ability of participants who had visual or cognitive impairments to push the correct pushbutton and to know when the WALK signal came on for the street in front of them, and 3) whether there were significant advantages to using the same sound vs. two different sounds vs. two speech messages to indicate the walk interval.
The participants were ninety adults who were visually or cognitively impaired and who, by self-report, independently traveled outdoors on at least one route and crossed signalized intersections.. Half of the participants had some degree of visual impairment and half of the participants were cognitively impaired.
Local recruiters, having extensive knowledge of, and contacts within agencies or organizations relevant to each group of participants, were hired and the participants were recruited by word of mouth, flyers, and distributed e-mails. Many participants were associated with one of the following agencies: The Oregon Commission for the Blind, The ARC (formerly Association for Retarded Citizens) of Multnomah and Clackamas Counties, Independent Living Resources, The Brain Injury Association of Oregon, Traumatic Brain Injury Club, Legacy Good Samaritan Hospital and Medical Center Young Adult Support Group.
The following information was gathered during personal interviews with persons with visual impairments who had expressed an interest in participating: date of birth, highest level of education, amount of vision (including ability to see walk/don’t walk signs, poles, and crosswalk lines), etiology, date of onset of disability, additional disabilities, type of travel aid used, self rating of travel ability, frequency of independent travel, and frequency of crossing at unfamiliar intersections. Interviews with participants who were cognitively impaired gathered similar information: date of birth, highest level of education, amount of vision, date of onset of disability, additional disabilities including seizures, living situation, street crossing practices, frequency of independent travel, and frequency of crossing at unfamiliar intersections.
Participants were selected for participation with the goal of comprising groups that included persons with a wide age range, balanced according to frequency of independent travel, and frequency of crossing at unfamiliar intersections. They were paid $20.for their participation.
Forty-five participants with visual impairments were divided into three subgroups, of 15 participants each, based on their ability to see pedestrian signals, crosswalk lines and poles. The group that had the least vision (referred to as totally blind) reported that they were totally blind or had only light perception and could not see pedestrian signals, crosswalk lines, or poles. This group ranged in age from 21 to 78 with a mean age of 47.6 years. Another group (referred to as legally blind) reported they were legally blind; some of them could occasionally see pedestrian signals, and either occasionally or usually could see crosswalk lines and poles. This group ranged in age from 22 to 85 with a mean age of 51.6 years. A third group (referred to as low visual acuity) had visual impairments but they were not necessarily legally blind and could usually see pedestrian signals, crosswalk lines and poles. This group ranged in age from 19 to 74 with a mean age of 48.6 years. Participants from all the subgroups who normally used a long cane as a travel aid used it during the research. No participant used a dog guide as a travel aid.
Forty-five adults with cognitive disabilities were also divided into three subgroups of 15 participants. One group was comprised of people who had had head injuries. This group ranged in age from 22 to 57 with a mean age of 45.9 years. Another group was comprised of people who had had strokes and whose vision and hearing might have been affected, but who were not legally blind. This group ranged in age from 35 to78 with a mean age of 52.5 years. A third group was comprised of people who were developmentally delayed. This group ranged in age from 19 to 85 with a mean age of 49.1 years. All participants were cognitively capable of personally consenting to participation and all said that they independently crossed streets at signalized intersections.The intersection of NE 7th Avenue and NE Multnomah Street in Portland, Oregon was equipped with eight pushbutton-integrated APS units that were mounted on temporary poles, with pushbuttons at 42 inches high. In pushbutton-integrated APS, audible signals come from the pushbutton housing, unlike pedhead-mounted APS, in which the audible signals come from a speaker mounted in or on the pedhead. The pushbutton-integrated APS units used in this study included a pushbutton locator tone, automatic volume adjustment in response to ambient noise level, a tactile arrow that vibrated during walk interval (aligned in the direction of travel on the associated crosswalk), an audible “click” sound and red LED to confirm that the button had been pressed, and an audible WALK indication which varied during the experimental conditions.
The pushbutton locator tone is intended to inform approaching pedestrians that they need to push a button to actuate a pedestrian timing, and aid them in locating the pushbutton. It repeats once per second and is adjusted to be audible only six to twelve feet from the pushbutton. The audible WALK indications varied during the experimental conditions and were changed by the researchers with the use of a handheld programming unit supplied by the manufacturer.*
Two APS mounted on a single pole were used on two corners and APS mounted on two separated poles were used on the other two corners. On corners A & D (the northwest and northeast corners), the two APS for crossing both streets at each corner were mounted on a single pole; on corners B & C (the southwest and southeast corners) the two APS on each corner were mounted on two poles separated by at least 10 feet (see Figures 1 and 2), meeting the guidelines of the MUTCD as previously described.
Distance of the poles from the street was also varied. On corner A, the pole was installed approximately 3 feet from each street, and on corner D, the pole was installed approximately 10 feet from each street. The same was true for the corners with two pushbutton poles: on corner C, each pole was installed approximately 3 feet from the street, and on corner B, each pole was installed approximately10 feet from the street.
There were two possible sound conditions on each corner. The audible WALK indication was set by the researcher for either: two different tones (cuckoo and fast tick) or two same tones (fast tick) on the corners having two poles. On the corners with just one pole, the audible WALK indication was either two tones (cuckoo and fast tick) or two speech messages (“7th. Walk sign is on to cross 7th,” or “Multnomah. Walk sign is on to cross Multnomah.”). (See Table 1.)
The total duration of all audible and visual WALK indications was seven seconds. There was a 0.5 second pause between repetitions for all sound conditions. Each speech message repeated twice during each walk interval, and the cuckoo and fast tick messages repeated three times during each walk interval. This variability was due to the length of the messages. Each speech message was approximately 3 seconds and each repetition of the cuckoo or fast tick was approximately 2 seconds.
The choice of tones and speech messages was based on current practice and on recent laboratory research. 5, 6 The single tone used was a rapid tick, as this tone has been found to be more highly detectable than other tones in use or proposed for APS. When two tones were used, they were the rapid tick and cuckoo. The cuckoo sound was selected over the chirp sound, which it is often used in association with, because it is not the same as any birdcall in the
To reiterate, two sound conditions (two tones or two speech messages) were assigned to corners having a single pole (corners A and D), and two sound conditions (two different tones or same tone) were assigned to corners having two poles (corners B and C). Testing every sound condition at every corner would have resulted in an experiment that required two sessions. Additionally, testing the same tone from each APS at corners having a single pole would have been non-informative. If both APS have the same tone coming from the same location, there is no cue to indicate which crosswalk is being signaled. However, on corners having two pushbutton poles separated by at least 10 feet, it may be possible to hear which pushbutton is sounding. Researchers recognized that either two different tones or speech messages could have been used in APS at corners having two poles; both would have provided information that was redundant to the cue that was provided by the location of the APS. A decision was made to use the most commonly used strategy in the
All ninety participants were individually tested in approximately one-hour sessions during the summer of 2003. First the participants were familiarized with a non-functioning demonstration unit of the APS that included the tactile arrow indicating the direction of the associated crosswalk. Participants then listened to a tape recording of the various sounds that would come from the APS. The sounds included a locator tone and the three possible WALK indications. The participants were told that they were to identify the walk interval on the street in front of them by raising their hand. Participants were told that they could use the visual WALK signal if they could see it, the pushbutton sounds or messages, or the traffic. They were encouraged to use any information that they would normally use to determine when the WALK signal comes on. Then the participants were guided to the starting location for the first trial.
Before the participant arrived at the corner, the APS devices were configured to one set of the possible WALK indication sound conditions. The APS devices on corners having two poles were randomly set to either the same tone condition (fast tick) or two tones condition (cuckoo and fast tick). While the starting settings were determined randomly, the two corners with two poles were always set up for the same condition (#1 or #2) for each block of trials. The two APS units on each corner having a single pole were randomly set to either the two tone, or the speech message condition. Both APS on the single pole corners were always set for the same condition within a block of trials. (See Table 1.)
Each participant completed 16 trials, run in two blocks of 8 trials with a short rest period (usually five to ten minutes) in between. For an intersection such as was used in this experiment, a pedestrian can approach each of the four corners from two different directions. Therefore, as shown in the chart, there are eight possible unique approaches/crossings at the intersection. Each participant completed all eight approaches in a randomly determined order both prior to, and after, the rest period. During the rest period, the APS devices at each corner were reconfigured to use the alternate sound condition.
For all trials, participants were positioned to start 30 feet from the pushbutton. For each trial, participants were instructed: “When I say Go, first, go push the button to cross the street in front of you. Then, stand where you would wait to cross the street. Raise your hand when the WALK comes on for the street in front of you, which is [street name].” When participants were 20 feet from the pushbutton, a stopwatch was started; it was stopped when participants pushed the correct pushbutton. If a participant pressed the incorrect pushbutton, but continued to search for the other APS, the time continued to run and was stopped if and when the participant pressed the correct pushbutton. On the other hand, when participants pressed the incorrect pushbutton and went and stood at the street demonstrating their belief that they had pressed the correct button, no time was recorded. In this instance, the experimenter would then press the correct pushbutton and the remainder of the trial would be recorded. A stopwatch was also used to time the delay between the onset of the correct WALK signal and when participants raised a hand indicating their judgment that the WALK signal had come on for the street in front of them.
Once participants had pressed a pushbutton and taken up a position where they would normally wait to cross the street, one of the researchers always pressed the other pushbutton. By doing so, and in combination with beginning participants at certain times during the intersection phasing, the researchers attempted to ensure that on half of all trials the first WALK signal to come on was that for the street in front of the participants while on the other half of the trials the first WALK signal was for the street beside participants. However, it was not entirely possible to achieve this balance given testing constraints.
For every trial, as soon as participants raised their hands to indicate their belief that the WALK signal for the street in front of them had come on, the trial ended and the experimenter directed or guided them to the next starting point. A trial was also ended if, after pressing the pushbutton, a participant waited through four consecutive walk intervals without responding.
Other information recorded on each trial included which pushbuttons the participants investigated and in which order, and which walk signal was the first to come on after the participants pressed the correct pushbutton. After the participants had completed all 16 trials, a short survey was administered to learn the attitudes and preferences of the participants towards the various pushbutton arrangements and sounds. The questionnaire addressed: which corner was the easiest and hardest, how easy it was to find and use the pushbuttons when the poles were near to or away from the curb, and when the APS were on one pole or two poles, whether the participant used the arrow to determine which crosswalk the button controlled, which cues the participant used (i.e., visual signal, audible signal, or traffic) to decide when the WALK had come on, and which walk indication was the participant’s favorite and least favorite. Also included was the researcher’s observation of what cue the participant seemed to be using during the trials.
The vast majority of significant findings, both practically and statistically, were found in the data collected from the two sub-groups having the least vision, referred to as totally blind and legally blind. Most of the results presented here will therefore focus upon these 30 participants. Relevant findings from the low visual acuity subgroup and the cognitively impaired group follow. Mean substitution was used to deal with all missing walk signal response delays and pushbutton location times (for example, if a given totally blind subject was missing pushbutton location data on corner A, southbound approach, speech condition, the mean for the totally blind subjects on corner A, southbound approaches, speech condition was substituted).
For the purposes of our analysis and discussion, a correct response was one in which the hand raise occurred during one of the 7-second periods in which the visual pedestrian signal for the street in front of them was in WALK. It was possible for a correct response to occur following either a correct or an incorrect pushbutton activation. This resulted from the fact that an experimenter always pushed whichever pushbutton the participant had not. This allowed response data to be collected on trials in which an incorrect button push had occurred. An incorrect response was defined a bit more narrowly than a correct response. An incorrect response was recorded when the hand raise occurred during the first walk interval for the side street. That is, after the participants pressed a pushbutton and took up a position to wait to cross the street, the visual and audible WALK signal to cross the street beside them came on first, and participants mistook this information to mean the WALK signal had come on for the street in front of them.
As explained in the procedure section, it was not possible to ensure that for each trial there would be an equal probability that the first WALK signal to come on would be for the street in front of the participant vs. the street beside the participant. Therefore, the number of trials in which an “error” was possible varies between corners and individual pushbutton position and sound conditions. The error rates presented in this section reflect the number of incorrect responses divided by the number of trials in which an error was possible (i.e. trials in which the first WALK signal after a pushbutton was pushed was for the street beside the participants).*
One of the most striking findings of this study was the very low error rate at the two-pole near the curb configuration of corner C (4/53, 7.55%); see table 2. No other pushbutton configuration resulted in less than a 26.9% error rate. Also, the other two pole configuration (corner B) resulted in a lower error rate than both of the single pole corners. However, due to a lack of independence of the measures (i.e. repeated measures design), no statistical test (i.e. Chi-Square) can confirm whether or not the two pole configurations resulted in overall lower error rates than the corners with a single pole. Nevertheless, the actual differences between error rates in most of the situations discussed here are quite substantial.
For the corners with two poles, there were two possible audible signal conditions, same tone (sound condition 1) or two different tones (sound condition 2). In the same tone condition, a fast tick indicated the onset of the WALK signal for both street crossings, while in the two different tone condition, a cuckoo signaled the northbound or southbound crossing and the fast tick signaled the eastbound or westbound crossing. For the corners with a single pole, there were also two possible audible signal conditions, two different tones (condition 2) or two speech messages (condition 1). The two tone sound condition produced higher error rates on each corner than the corresponding same tone condition or speech message condition. (see Table 3). Overall, the two different tone sound condition resulted in a 36.27% error rate (37/102). For the two pole APS arrangements at corners B and C, the two different tone sound condition resulted in 12 errors on 52 possible trials (23.08%), while the same tone sound condition resulted in half as many errors (6/53, 11.32%). On corner C (two pole near the curb configuration), the error rate was a mere 3.57% (1/28) for same tone condition. For the single pole APS arrangements at corners A and D, the two tone sound condition was especially troublesome, causing errors on 50% of the possible trials (25/50). The speech message condition resulted in much reduced error rate of 18.97% (11/58).
For pedestrians to safely cross at signalized intersections, they should begin crossing during a walk interval for the street they wish to cross. It is also important that they begin crossing as soon after the onset of the WALK signal as possible, so as to allow themselves adequate time to cross the street before the onset of perpendicular traffic.
WALK signal response delays (measured from the onset of the correct walk interval) for totally blind and legally blind participants were both low and consistent for the two-pole, near the curb pushbutton arrangement at corner C. Regardless of the sound condition (same tone vs. two different tones) or approach direction, it took participants approximately 2.0 seconds to correctly respond to the onset of the walk signal at corner C. Planned comparisons revealed significantly faster responses to the signals at corner C than those at corner A (F(1,29)=8.598, p<0.01), corner B (F(1,29)=6.156, p<0.05), and corner D (F(1,29)=31.459, p<0.001), (see Figure 3).
It was discussed earlier that for corners with a single pole, the two speech messages condition resulted in a much reduced error rate compared to the two different tone condition. The effect of the speech messages on the walk signal response delay is not as clearly defined or, in most instances, as positive. At corner A, the speech message WALK indication resulted in significantly faster responses than did the fast tick WALK signal (F(1,29)=8.333, p < 0.01), however, the mean response delay for the speech condition (2.3 sec) was slower than that for the cuckoo WALK signal (1.9 sec, see figure 4). This difference did not however achieve significance (F(1,29)=1.694, p > 0.05). At corner D, the mean response delay for the speech message WALK signal was quite high (3.1 sec) and was slower than the mean response delays for the fast tick WALK signal (2.7 sec) and for the cuckoo WALK signal (2.4 sec). The speech message WALK signal resulted in significantly slower responses than the cuckoo (F(1,29)=6.976, p < 0.05), while the difference between response delays for the speech message and the fast tick failed to reach significance (F(1,29)=2.407, p > 0.05).
Another factor of particular interest is how quickly blind participants were able to locate and press the appropriate pushbutton given the different APS arrangements at the four corners of the intersection. The pushbutton location times were measured from the time the participant crossed a line that was 20 feet away from the appropriate pushbutton to the time that they pressed the correct pushbutton. Again, when participants were unable to find the correct pushbutton or pressed the incorrect pushbutton and stopped searching for the correct button, no pushbutton location time was recorded. On average, pushbutton location times were fastest at corners D (13.6 seconds) and A (14.95 seconds, see Figure 5). While the difference between mean pushbutton location time at corners D and A was not statistically significant (F(1,29)=2.497, p>0.05), the mean pushbutton location time at corner D was significantly faster than that at corners C (F(1,29)=4.838, p<0.05) and B (F(1,29)=20.156, p<0.001). The mean pushbutton location time at corner A was significantly faster that that at corner B (F(1,29)=13.049, p<.01), but was not significantly faster than that at corner C(F(1,29)=1.395, p>0.05).
One result of interest was the very large discrepancy in pushbutton location times between northbound approaches to corner C (12.0 seconds), and westbound approaches to the same corner (21.6 seconds). Figure 5 clearly shows that the mean pushbutton location time for the northbound approaches to corner C was amongst the fastest across all approaches and corners. Also, the mean pushbutton location time for northbound approaches to corner C was significantly faster than the mean pushbutton location time for westbound approaches to the same corner (F(1,29)=81.968, p<0.001). Possible reasons for this discrepancy will be addressed in the discussion.
The analysis of the data collected from participants with low visual acuity (the subgroup with the most vision) demonstrated very few statistically significant differences; however, the majority of the trends in the data matched those for the totally blind and legally blind sub-groups. While the overall error rate for the low acuity group was only 5.36% (6/112), five of the six errors occurred on trials with the two different tone sound condition on corners A and D, the two most difficult conditions for the totally and legally blind participant sub-groups. Also similar to the analysis described above, mean WALK signal response delay at corner C (1.1 sec) was faster than those at all of the other corners (corner D, 1.2 sec; corner B, 1.3 sec; and corner A, 1.4 sec). The differences were moderately significant between mean WALK signal response delays at corners C and A (p=.066) and those at corners C and B (p=.077). It is of particular interest that although most members of this group of participants could see the visual walk signal, the same pattern of results as those obtained from participants with less vision emerges. This point will be addressed further in the discussion.
Although the speech condition resulted in no errors with this subgroup, it did result in the slowest response times at both corners A and D. At corner A, the speech message resulted in significantly slower responses than did the cuckoo (F(1,29)=16.230, p < 0.01), but the mean WALK signal response delay for the speech message (1.6 sec) was not significantly slower than that for the fast tick (1.4 sec). At corner D, no comparisons reached significance; however the mean WALK signal response delay for the speech message (1.3 sec) was slower than the delays for the cuckoo (1.1 sec) and the fast tick (1.1 sec).
One difference from the participants who are totally blind or legally blind is in the ability of the participants with low visual acuity to efficiently find the correct pushbutton. The low visual acuity participants had little difficulty finding the correct pushbutton for the westbound approach to corner C. As a result, the fastest pushbutton location times for this subgroup were obtained at corner C (6.6 sec), although all pairwise comparisons failed to reach significance. Mean pushbutton location times at corners A and D were again similar to one another (both were 7.2 seconds due to rounding), and the slowest pushbutton location times were again recorded at corner B (7.7 sec).
The analysis of the data collected from the 45 participants with cognitive impairments similarly has few statistically significant differences; however the majority of trends in the data again matched those of the 30 totally blind or legally blind participants. The participants with cognitive impairments made very few errors (10/362) resulting in an overall error rate of 2.76%. The few errors committed did follow a pattern similar to those of the visually impaired participants. Only two of the ten errors occurred on corners with two poles (corners B and C), and more errors were committed in the two different tone conditions (6/10) than in the same tone or speech message conditions (4/10). Once again, the mean WALK signal response delay at corner C (0.9 sec) was faster than those at all of the other corners; however, there was no significant main effect for corner (F < 1.0) and all mean WALK signal response delays were less than 1.05 sec. Once more, all of these participants had sufficient vision to be able to see the visual walk signal, and yet the pattern of results is again similar to the visually impaired participants.
Mean pushbutton location times at corners D (6.19 sec) and C (6.22 sec) were both significantly faster than those at corners A (6.59 sec) and B (6.57 sec). The very small difference between the most extreme mean pushbutton location times (0.40 sec) does, however, qualify the importance of the significant differences. Once more, it is noted that for these participants the westbound approaches to corner C did not cause greater difficulty in finding the correct pushbutton than northbound approaches to the same corner.
The self-report questionnaire data is similar to the behavioral results is some ways and dissimilar in others. While corner C produced the most favorable behavioral results, the 45 participants with visual impairments gave the single-pole-close-to-the-curb configuration (Corner A) the most votes as the overall easiest corner (27/45, 60.0%), while the two-poles-near-the-curb configuration (Corner C) received the second most votes (13/45, 28.9%). Corner B (the two-poles-further-from-the-curb configuration), which produced some of the worst pushbutton location times and walk signal response delays, received the most votes as the most difficult pushbutton arrangement (34/45, 75.6%). The survey questions that most conflicted with behavioral results assessed participants favorite and least favorite sounds. Even though the fast tick resulted in the best response accuracies and consistently fast walk signal response delays, it only received 17.8% of the vote as the favorite sound (8/45), and received the most votes as the least favorite sound (25/45, 56.6%). The speech message was the most popular of the sound messages (29/45, 64.4%), and also received the least votes as the least favorite sound (6/45, 13.3%).
Safe and legal pedestrian crossings at signalized intersections require that pedestrians correctly identify which crosswalk has the WALK indication. Positive information is provided by visual pedestrian signals, and by APS (audible and vibrotactile) where they are available. Where pedestrian phasing is concurrent with vehicular phasing, the onset of the WALK signal can also be inferred by a surge of vehicular traffic going parallel to the crosswalk, from the lane/s closest to the pedestrian. In the absence of APS, pedestrians who are unable to see visual pedestrian signals typically attempt to begin crossing with this vehicular surge.
At complex, unfamiliar, signalized intersections without APS, pedestrians who are blind often fail to begin crossing within the walk interval for the desired crosswalk.7 Even when APS are present, pedestrians who are blind sometimes find audible signal information ambiguous in terms of which crosswalk is being signaled.8 It is critical that any APS system provide unambiguous information. The results indicate that pushbutton location affects the ambiguity of WALK signal information.
On the corner having pushbutton-integrated APS on two poles near to the street, participants in the totally blind and legally blind sub-groups indicated that the street in front of them had the WALK signal, when, in fact, the WALK signal was to cross the street beside them, on only 7.55% of possible trials. This pole arrangement for APS devices corresponds to the arrangement that is required in some other countries, including Sweden, Denmark, and Australia.9 For all other pushbutton-integrated APS arrangements tested in this study, participants in the totally blind and legally blind sub-groups indicated that the street in front of them had the WALK signal, when it was actually the WALK signal to cross the street beside them, on at least 26.9% of trials. On those trials, if they had been crossing the street, they were likely to have started crossing at the onset of traffic on the street perpendicular to their direction of travel.
Use of the configuration with two pushbutton poles with APS near the curb, both having the same tone, resulted in excellent accuracy by participants who were unable to use visual pedestrian signals (totally and legally blind sub-groups). This pole configuration and use of same tone sound was also beneficial to participants having more vision (low visual acuity sub-group and cognitively impaired group), in determining which crosswalk had the WALK signal. This corresponds to the pushbutton pole arrangement plus WALK signal system that is common in countries using pushbutton-integrated APS.9
Those participants who had vision sufficient to see visual pedestrian signals (i.e. the cognitively impaired participants and the low visual acuity participants) were recorded by experimenters as having typically used the visual pedestrian signals to determine the onset of the correct walk interval (and they also typically reported reliance on the visual pedestrian signals). It is therefore of interest that their results are in the same direction as those for the two sub-groups who were always or usually unable to see the visual pedestrian signals. It appears that, whether or not it was on a conscious level, or observable to experimenters, the participants who could see the visual pedestrian signal information were nonetheless influenced by both the pushbutton pole arrangement and the WALK signal sound. If they were not influenced by pole arrangement and the nature of the walk sound, accuracy would be expected to be the same under all pole arrangements and sounds.
Pedestrians who do not begin crossing promptly following the onset of the walk interval may not complete their crossings before the onset of perpendicular traffic, at which time they are particularly at risk for crashes. Recent research on pedestrians who are blind making crossings at complex, unfamiliar, signalized intersections, found that on 51.4 % of trials, crossings were not initiated during the walk interval, and on 26.9% of trials, crossings were completed after the onset of perpendicular traffic.7 Therefore, it was of interest to measure the delay in response to the onset of the WALK signal for the correct crosswalk.
In this study, for participants who were always or usually unable to see visual pedestrian signals (totally and legally blind sub-groups), responses to the onset of the WALK signal for the correct crosswalk were significantly fastest on the corner having two pushbutton poles with APS near the curb (corner C), and responses were uniformly fast at this corner regardless of sound condition. Although the over-all mean walk signal response delay for this corner was approximately two seconds, well within the normal walk interval, delay would likely be longer if actually making crossings, as pedestrians who are blind typically require a little time to determine whether there are vehicles turning across the crosswalk. Mean walk signal response delay for blind pedestrians crossing at complex, unfamiliar signalized intersections without APS has been found to be 6.4 seconds. 7
Responses to the correct WALK signal were also fastest at the corner having two pushbutton poles with APS near the curb for the participant groups who always or usually could see the visual pedestrian signal. Mean walk signal response delays were lower for these groups than for the groups who were always or usually unable to see visual pedestrian signals.
It appears that, whether or not it was on a conscious level, or observable to experimenters, the response delay of participants who could see the visual pedestrian signal information was influenced by the pushbutton pole arrangement. Installing pushbuttons on two poles near the curb seems to promote fast reaction time to the onset of the WALK signal, and is thus likely to promote faster initiation of crossings for pedestrians with visual and cognitive disabilities.
By comparing the time to press the correct pushbutton at corners having pushbuttons in different pole configurations, we are able to infer ease of location of the correct pushbutton. Nearly all participants were observed to visually or tactually use the tactile and/or visual arrows on APS to confirm which crosswalk the APS was associated with, and 82 of 90 participants (91%) reported that they used the arrows. The results of this research, therefore, would not necessarily be applicable to pushbuttons that did not have tactile and visual arrows that were carefully aligned with the direction of travel on the associated crosswalk.
For the participant sub-groups in this study who had the least amount of vision and who were rarely or never able to see visual pedestrian signals, time to press the correct pushbutton was marginally faster at the two corners having two pushbuttons on a single pole. The mean pushbutton location time for the corner having two pushbutton poles far from the curb was longest. However, any advantage in speed of pressing the correct push button is far out-weighed in importance by the greater accuracy in identifying which crosswalk had the WALK signal. Errors in determining which crosswalk has the WALK signal may result in crashes, while delay in locating the correct pushbutton does not typically have negative consequence for life safety.
For the low visual acuity sub-group, unlike the sub-groups having even less vision, mean pushbutton location time was fastest for the corner having two pushbutton poles near the curb. However, as for the sub-groups having less vision, mean pushbutton location time was slowest for the corner having two pushbutton poles far from the curb. For the group having cognitive impairments, the corner having two pushbuttons near the curb resulted in one of the two fastest times, which were not statistically different from one another.
Therefore, although the results for time needed to find and press the correct pushbutton are not as clear-cut or definitive as for speed and accuracy in identifying the crosswalk being signaled, having two pushbutton poles near the curb generally appeared to promote speed (and by inference, ease) of locating the correct pushbutton.
Speed in locating the pushbutton appears to have been significantly influenced by the relationship between pushbutton location as designed to test the hypotheses of this experiment, and pushbutton location with relation to geometric idiosyncrasies at each corner. There may have been some difference attributable to slight differences in perceived loudness of each APS device as well. It was extremely difficult to make the perceived loudness of the APS signal the same from each device. Simply setting all devices the same for volume and ambient noise response is not sufficient because differences in the nature and distance of reflective surfaces at each corner also influence perceived loudness. At each corner, there was a significant difference in time to press the correct pushbutton depending on the direction from which the corner was approached.
For example, at the corner having two pushbutton poles near the curb, corner C (see Figure 2), mean time for pushing the correct pushbutton on westbound approaches was much longer than northbound approaches to the same corner. When approaching westbound, participants passed very close to the pushbutton to cross the street beside them (the parallel street) and the locator tone of that APS was usually very audible to participants. They may have been able to hear the locator tone for the pushbutton to cross the street in front of them (correct pushbutton) when they approached the incorrect pushbutton, but the loudness of the sound would have been much less. (Recall that locator tones are only intended to be heard 6 to 12 feet from the pushbutton, and the pushbuttons were approximately 10 feet apart.) Therefore it commonly occurred that participants would check the arrow on the APS for crossing the parallel street before finding and pressing the pushbutton for the street in front of them. When approaching the corner northbound, participants did not pass quite as close to the pushbutton for the parallel street because the pushbutton pole was on a bulb-out in the curb line, so they did not delay to check the arrow for the parallel street before finding and pushing the correct button.
The two different tone condition may have produced more errors than the same tone condition at all corners because some participants with visual impairments had previous experience with cuckoo signals at other locations in the city and may thus have been primed to respond to the cuckoo, regardless of whether it was the correct signal. This could also explain why the cuckoo resulted in the fastest response rates at the single pole corners. Nonetheless, when pushbuttons were on two poles near the curb, accuracy was equally good with two tone and same tone conditions.
The two different tone condition may have had particularly high error rates on the single-pole corners because there was no redundant APS location information to indicate which crosswalk was associated with which sound. The speech message condition resulted in lower error rates than the two tone condition, but mean WALK signal response delays for the speech message WALK condition were slower. This may reflect greater cognitive demand for processing the speech messages than the tones, or may reflect the different lengths of the sounds and messages themselves. A single complete speech message lasted approximately three seconds, whereas a single cuckoo lasted somewhat over one second.
Subjective judgments of participants, recorded on the survey, were variance with objective measures in three important ways. First, corner A (one pole close to the curb) had the most preferred pole arrangement with corner C (two poles far from the curb) a somewhat distance second choice. This is in contrast to objective measures in which corner C produced the fastest and most accurate responses to the walk indication. Because accuracy in responding to the walk indication is essential to safe street crossing at signalized intersections, despite the subjective preference, the arrangement at corner C is recommended.
Second, speech messages were used at the locations with two pushbuttons on a pole and resulted in better accuracy than two tones sound condition, but less accuracy than the fast tick. Speech messages also resulted in slower response times than the fast tick indication. When it is possible to separate the poles, the fast tick WALK indication is recommended.
Third, the fast tick was the least preferred walk indicator, but it produced the fastest and most accurate responses to the walk indication. Because inaccuracy in responding to the walk signal, that is, beginning the crossing when the wrong walk signal comes on, has direct consequences for safety in crossing, the fast tick is preferred over other walk indications used in this research. Verbal messages may have been rated highly simply because participants were primarily native English speakers, because ambient sound was never excessively loud, and because it provided additional information, that is, the name of the street to be crossed. The cuckoo may have been rated more highly than the fast tick simply because it was a more familiar walk indication to some participants. Therefore despite some dislike of the fast tick WALK indication, it is the recommended sound because it results in fast and more accurate decisions regarding which crosswalk has the WALK.
The recommended location for two pushbutton-integrated APS on a single corner is consistent with Guidance in MUTCD 4E.09.
· Place each APS device on a separate pole, located as close as possible to the curb line.
· Place each APS as close as possible to the line of the associated crosswalk that is furthest from the center of the intersection.
· Place the two APS at least 10 feet apart.
The recommended WALK indication for APS that are located according to these recommendations is a rapid tick, or percussive sound, at 10 repetitions per second. Both APS should have the same WALK indication. Where it is technically infeasible to install two pushbuttons on a corner on two separate poles, it is recommended that verbal WALK messages following the model “Multnomah. Walk sign is on to cross Multnomah” be used. Two different tones to indicate which crosswalk has the walk interval at an intersection where there are two pushbutton-integrated APS on a single pole provides ambiguous information and may result in pedestrians crossing with the wrong signal. Speech WALK indications should be used sparingly, however, as it is not possible to make them understandable under all ambient sound conditions, and to all pedestrians. In this experiment, speech messages were evaluated in a situation where all participants were told the name of the street that they were approaching. If the pedestrian is not familiar with the area or is confused about the street names, the speech message, which begins with the street name, will not clarify which crossing has the WALK. If speech messages are used, additional features may be needed on the device to provide street name information to the pedestrian who is unfamiliar with the intersection. These features would include a high-contrast tactile arrow oriented parallel to the direction of travel on the associated crosswalk, and a pushbutton information message that provides the name of the street controlled by the pushbutton. Braille street name information on the APS is also desirable.
Jurisdictions desiring to standardize the WALK indication should also standardize the location of APS so that they will provide unambiguous information regarding which crosswalk has the walk interval. Where APS are installed in a variety of locations, engineering judgment is required to determine, for each intersection, whether the rapid tick or a speech WALK indication will provide the most unambiguous information to pedestrians.
Research planned as part of NCHRP Project 3-62 will develop additional guidance.
Figure 1. Intersection diagram for NE Multnomah St. and NE 7th Ave., Portland, OR.
Figure 2. Diagram of Corner C (southeast corner of NE Multnomah St. and NE 7th Ave., Portland, OR)
Figure 3. Mean WALK signal response delay by corner.
Figure 4. Effects of sound condition on WALK signal response delay at corners with two APS on one pole.
Figure 5. Mean pushbutton location time by corner.
Figure 6. Mean pushbutton location time by corner and approach direction.
Table 1. Sound conditions and crossing direction
Table 2. Error in recognizing correct WALK signal, by corner.
Table 3. Error in recognizing correct WALK signal, by corner and sound condition.
Corner A – one pole close to the curb |
Sound condition A1: speech messages |
Sound condition A2: two tones |
APS for crossing southbound |
Multnomah, walk sign is on to cross Multnomah |
Cuckoo |
APS for crossing eastbound |
Seventh, Walk sign is on to cross seventh |
Fast tick |
Corner B – two poles 10 feet from the curb |
Sound condition B1: same tone |
Sound condition B2: two tones |
APS for crossing northbound |
Fast tick |
Cuckoo |
APS for crossing eastbound |
Fast tick |
Fast tick |
Corner C – two poles close to the curb |
Sound condition C1: same tones |
Sound condition C2: two tones |
APS for crossing northbound |
Fast tick |
Cuckoo |
APS for crossing westbound |
Fast tick |
Fast tick |
Corner D – one pole 10 feet from the curb |
Sound condition D1: speech messages |
Sound condition D2: two tones |
APS for crossing westbound |
Seventh, WALK sign is on to cross Seventh |
Fast tick |
APS for Crossing southbound |
Multnomah, WALK sign is on to cross Multnomah |
Cuckoo |
Corner |
# of incorrect responses/total possible instances |
error rate |
A |
17/56 |
30.36% |
D |
19/52 |
36.54% |
C |
4/53 |
7.55% |
B |
14/52 |
26.92% |
Corner/sound condition |
# of incorrect responses/ total possible instances |
Error Rate (%) |
A1 / Speech |
7/33 |
21.21 |
A2 / 2 Tones |
10/23 |
43.48 |
D1 / Speech |
4/25 |
16.00 |
D2 / 2 Tones |
15/27 |
55.56 |
C1 / Same Tones |
1/28 |
3.57 |
C2 / 2 Tones |
3/25 |
12.00 |
B1 / Same Tones |
5/25 |
20.00 |
B2 / 2 Tones |
9/27 |
33.33 |
1. Manual on uniform traffic control devices Washington, D.C.: U.S. Department of Transportation, Federal Highway Administration. 2003, Part 4A.02
2 Manual on uniform traffic control devices Washington, D.C.: U.S. Department of Transportation, Federal Highway Administration. 2003, Part 4.E08
3. Manual on uniform traffic control devices Washington, D.C.: U.S. Department of Transportation, Federal Highway Administration. 2003, Part 4E.09
4. Draft Guidelines for Accessible Public Rights-of-Way. Washington, D.C.:
5. Wall, R.S., Ashmead, D.H., Bentzen, B.L. and Barlow, J.M. a. Audible pedestrian signals: Characteristics of perception in noise. (Submitted for publication.)
6. Wall, R.S., Ashmead, D.H., Bentzen, B.L. and Barlow, J.M., b. Audible pedestrian signals for street crossing. (Submitted for publication.)
7. Bentzen, B.L., Barlow, J.M. and Bond, T. Pedestrians who are Blind at Unfamiliar Signalized Intersections: Research on Safety. Transportation Research Board Annual Meeting Compendium of Papers, January 2004.
8 Bentzen, B.L., Barlow, J.M. and Franck, L.. Addressing barriers to blind pedestrians at signalized intersections. ITE Journal, September 2000, 70-9, 32-35.
9 Barlow, J.M., Bentzen, B.L. and Tabor, L. Accessible pedestrian signals: Synthesis and Guide to Best Practice. Berlin, MA: Accessible Design for the Blind, Prepared for the National Cooperative Highway Research Program, Project 3-62. (2003)
* APS units were manufactured by Polara Engineering, 4115 Artesia Ave., Fullerton, CA 92933
* Error rates were calculated using a few different operational definitions of what constituted an error, and regardless of the method chosen the rank order of the corners remained the same. Therefore, this method for calculating error rates was chosen because the researchers felt that it best addressed the issue of whether or not the pole placements and sound conditions provided either ambiguous or unambiguous cues for the onset of the WALK signal.