|Back to Roundabouts|
Note: This document does not include graphics. The graphics in the original document are described where they appear in that document.
David Guth1, Daniel Ashmead2, Richard Long1, Paul Ponchillia1, & Robert Wall2
1Department of Blindness and Low Vision Studies
Western Michigan University
Kalamazoo, Michigan, USA 49008
2Department of Hearing and Speech Sciences
Vanderbilt University School of Medicine
21st Avenue South
Nashville, Tennessee, USA 37232
Abstract - We investigated the judgments of blind and sighted persons about the duration of gaps in traffic at four modern-roundabout intersections. Participants stood at roundabouts' crosswalks and indicated when they believed that gaps between vehicles were adequate for crossing to the pedestrian (splitter) island in the middle of the road before the arrival of the next vehicle at the crosswalk. We analyzed participants' latencies to detect gaps, their ability to detect gaps, and the frequency with which participants erroneously indicated that a gap was long enough to cross. The findings suggest that at some roundabouts, there are differences in pedestrians' abilities to determine whether it is safe to initiate a crossing, depending on whether they are using vision and hearing or hearing alone. Further, this effect appears to be related to the volume of vehicular traffic. Preliminary findings from three of the roundabouts were presented at IMC-10. This talk will review and expand upon those findings and then present new findings from an experiment at a single-lane urban roundabout at which pedestrians made judgments during both peak-traffic (rush) hours and off-peak hours.
This section provides general information about modern roundabout
intersections and their potential implications for blind pedestrians. An
expanded version of this introductory material may be found at
Roundabouts are unsignalized circular intersections and are common in many parts of the world. They are relatively new to the United States, but their numbers are increasing rapidly. The first U.S. roundabout was built in 1990 and by 2002 the number of U.S. roundabouts had grown to an estimated 700, with many more planned (see http://roundabout.kittelson.com). The design of U.S. roundabouts has not been standardized, with U.S. designers being influenced by differing practices employed in Europe (e.g., Brilon, 1991; Brown, 1995; Guichet, 1997; Sawers, 1996) and elsewhere (e.g., Troutbeck, 1993). Roundabouts present different challenges than traditional signalized intersections for individuals with blindness and visual impairments. The growing popularity of roundabouts in the U.S. has led to concerns about the accessibility of these intersections to pedestrians who are blind and visually impaired.
A typical modern roundabout (see Figure 1) is an unsignalized intersection with a circular central island and a circulatory roadway around the island. Vehicles entering the roundabout yield to vehicles already on the circulatory roadway. A dashed yield line for vehicles is painted at or near the outside edge of the circulatory roadway at each entering street. Roundabouts have raised or painted splitter islands at each approach to separate the entry and exit lanes of a street. These splitter islands are designed to deflect traffic (and thus reduce vehicle speed). Splitter islands also provide a pedestrian refuge midway through a street crossing.
Acknowledgements: This project is supported by the National Eye Institute of the U.S. National Institutes of Health (RO1EY12894-3). We thank Michael Niederhauser, Ed Myers, Ken Sides, John Gesink, and Ed Waddell for technical advice and assistance; and Duane Geruschat, Michael Glocken, April Shinholster, and Dae Shik Kim for assistance with data collection
Fig 1. Chief Okemos Roundabout (Okemos, Michigan, USA). Photograph courtesy of Dave Sonnenberg, Director of Traffic and Safety, Ingham County, Michigan Road Commission. This figure is an overhead photograph of a new roundabout near East Lansing, Michigan. The roundabout is similar to a T-intersections, with 3 legs. With the exception of one two-lane entry and one-two lane exit, all entry and exit lanes are one-lane wide. The photo shows well marked striped white crosswalks that cross splitter islands that are an estimated 5 feet wide.
Studies conducted in Western Europe, where roundabouts are common, and in the U.S., have generally found that roundabouts result in less severe vehicular crashes than more traditional intersections. This reduction in the rate of serious vehicular crashes is the most compelling reason cited by transportation engineers for the installation of roundabouts. Roundabouts increase vehicular safety for two main reasons (see Flannery, 2001; FHWA, 2000, Persaud, Retting, Gardner, & Lord, 2000; Retting, Persaud, Garder, & Lord, 2001). They reduce or eliminate the risk arising at signalized intersections when motorists misjudge gaps in oncoming traffic and turn across the path of an approaching vehicle. They also eliminate the often-serious crashes that occur at signalized intersections when vehicles are hit broadside by vehicles on the opposing street that have run a red light or stop/yield sign.
Concern has been raised about pedestrian safety at roundabout crosswalks for children, older adults (Alphand, Noelle, and Guichet, 1991) and people with disabilities (U.S. Access Board, 2001; American Council for the Blind, 2002). In the United States, the Americans with Disabilities Act (ADA) requires that public rights-of-way, including sidewalks and crosswalks, be accessible to pedestrians with disabilities. It appears that blind pedestrians in countries with many roundabouts avoid some of them (Guth, Long, & Ashmead, 2000) or are advised by O&M instructors (Orientation and Mobility Specialists' Association of Victoria, 2002) to avoid some of them. However, other than the present work, we know of no systematic efforts to determine whether such decisions and advice are valid.
In both studies, the participants included individuals who were totally blind and individuals with typical vision. All participants reported that they routinely traveled independently in urban areas and most participants had had previous experience negotiating roundabout intersections. Participants were asked to indicate when they believed it would be possible for them to cross from the curb where they stood to the splitter island before the arrival of the next vehicle at the crosswalk. They were instructed to assume that vehicular traffic would not yield to them, and to base their judgments on their normal walking speeds.
Data were collected using a portable computer linked to pushbuttons held by the participants. Each trial lasted 2 minutes, marked by the words "go" and "stop" said by an experimenter. Participants were instructed to press the button when they would have initiated a crossing and to hold the button down until they believed it would no longer be safe to initiate a crossing. After releasing the button, they were to press it again whenever another crossing opportunity occurred. Many participants said that it felt natural to press the button when they would first have started to cross, but that holding the button as long as it felt safe to start crossing seemed unnatural. Therefore, the data presented here are based only on initial button presses. In addition to the participants, an experimenter pressed a pushbutton whenever a vehicle arrived at the crosswalk. For example, on trials where participants were making judgments about exiting vehicles, all exiting vehicles were noted on the data record as they crossed the crosswalk. This record of vehicle presence at the crosswalk enabled us to relate the participants' judgments to the actual gaps in traffic. The computer sampled all pushbuttons every half-second. Thus, the raw data for a trial consisted of button status (pushed or not pushed) for 120 samples.
At the 3 Baltimore, Maryland roundabouts for which preliminary findings were reported at IMC-10, data collection occurred mostly from 9 a.m. to 4 p.m., between the morning and afternoon peak (rush) hours. At the Tampa, Florida roundabout that is the focus of this IMC-11 report, data collection occurred during both peak and non-peak hours. Before testing, participants received extensive familiarization to the roundabout and underwent a series of practice trials. Blind participants were positioned so that they had unimpeded acoustic access to the traffic.
During the test trials, participants stood on the sidewalk at one end of the crosswalk, about one step back from the curb line and aligned in the general direction of the crosswalk. At this distance, pedestrians did not noticeably affect the movements of vehicles. Blind participants completed the trials without their long canes and guide dogs. At least one blind and one sighted participant made judgments at the same time. This arrangement ensured that these paired participants were presented with identical traffic conditions.
The study was conducted at three modern roundabouts in metropolitan Baltimore, Maryland, U.S.A. These were a large, high-volume, urban, two-lane roundabout; an intermediate volume, urban, two-lane roundabout; and a low-volume, single-lane roundabout. A more detailed summary of the methods and findings of this study can be found at http://www.access-board.gov/research/roundabouts/bulletin.htm and a full research report is currently under review. The findings of the Baltimore study can be summarized as follows:
The remainder of this report is focused on data collected at one single-lane urban roundabout during the afternoon peak-traffic (rush) hour and off-peak hours. The roundabout had formerly been signalized and was along a commuter route so traffic volume was low to moderate during off-peak hours and heavy during peak hours. The experimental method was the same as used in the Baltimore study, with the addition of counting all vehicles using the roundabout during each trial. Each participant made judgments at one entry lane and one exit lane. A blind participant and a sighted participant were tested simultaneously in order to insure identical traffic conditions across the two groups. The blind participants included 5 dog-guide users and 5 long-cane users.
Figure 2 shows the distribution of gaps combined across entry and exit lanes at off-peak hours. The x-axis shows the gap duration, ranging from half a second to 110 seconds. The y-axis shows the percentage of all gaps that fell into each half-second "bin" of gap duration. For example, at the peak of the distribution, about 7 percent of gaps were 2 seconds long. Short gaps, up to 5 seconds long (shown in red), were too short to afford a safe crossing at typical walking speeds (the crosswalks were 16 to 17 feet long). Gaps from 5.5 to 9 seconds long (shown in green) were long enough for a safe crossing, but only if the pedestrian initiated the crossing as soon as the gap started. However, as in the Baltimore study, we found that that blind participants took several seconds longer than sighted participants to detect a crossable gap. The remaining gaps colored in blue are those longer than about 9 seconds, which we considered to be generally crossable by most pedestrians. Quite a few of these gaps were very long, a minute or more. Therefore, a pedestrian who wanted to be very conservative could wait for a very long gap, which in terms of listening means that there is little or no traffic sound in the area.
Fig 2. Mid-day gap distribution of gaps between successive vehicles, collapsed across entry and exit lane crosswalks. This graph is included for comparison of a similar graph given as Figure 3 for rush hour. In the mid-day graph, while there are many short gaps between one and 5 seconds that are labeled "not crossable" and a few gaps between 6 and 9 seconds labeled "crossable if prompt", there are many gaps of 10 seconds and longer. These are labeled "generally crossable".
Figure 3 shows the distribution of gaps at rush hour, when traffic volume was higher. Note how the distribution of gaps shifts to the left, toward smaller gaps between vehicles. Also, notice that there were very few long gaps. With more traffic the pedestrian is presented with fewer crossable gaps. At some point the distribution of gaps will shift enough so that a prudent pedestrian would have to wait very long for a safe crossing period, or might decide to avoid the crossing altogether. This could occur not only for individuals with visual impairments, but also for elderly persons, children on their way to school, or in the extreme, for anyone. We are continuing to explore the distribution of gaps as a critical variable in determining whether a roundabout is compatible with pedestrian travel in a given location.
Fig 3. Rush hour gap distribution of gaps between successive vehicles, collapsed across entry and exit lane crosswalks. This is similar to figure 2 except that most of the gaps fall in the 0-5 second range, a few in the 6-9 second range, and many fewer in the "generally crossable range" above 10 seconds than for mid-day.
This section switches the focus from data about vehicles to data about the decisions made by the participants. Recall that each time a participant pressed the button, he or she was indicating that it was possible to cross to the splitter island before the arrival of the next vehicle at the crosswalk. From these data and data about when the next vehicle arrived, we computed a measure we called the safety margin. Starting with the time at which the button was pressed (for example, 8 seconds before the arrival of the next vehicle), we factored in the time it would take the participant to make the crossing, which was 5 seconds. Thus, the pedestrian would have arrived at the splitter island 3 seconds before the vehicle (8-5=3) for a safety margin of +3. By this measure, a safety margin of zero would mean the vehicle reached the crosswalk just as the pedestrian finished the crossing, and a negative safety margin would be "risky": the driver may have been forced to slow down, the pedestrian may have had to make an evasive maneuver, or the vehicle may have hit the pedestrian.
Figure 4 shows the percentage of crossing decisions on which there was a negative safety margin. Data for blind participants are in yellow (left bar of pair) and data for sighted participants are in blue (right bar). The pair of bars on the left of the figure are for the mid-day trials, when traffic was lighter, and the pair on the right are for rush hour.
Fig 4. Percentage of negative safety margins across group and time of day. This bar graph shows 4 bars and the data that is shown is completely described in the next paragraph.
At mid-day, about 8% of the blind participants' crossing judgments had negative safety margins, compared to about 6% for sighted pedestrians, which indicates that both groups had generally safe crossing decisions. At rush hour, blind participants had negative safety margins on about 17% of their crossing judgments, compared to 7% for sighted participants. These findings suggest an important pattern. Under conditions of lighter traffic, a group of pedestrians with visual impairments was able to make crossing decisions with a margin of safety comparable to that of sighted pedestrians. When the traffic was heavier, however, there was a difference between groups, with blind pedestrians making more than twice as many decisions that could be considered risky. Indeed, after experiencing both conditions, several of the blind participants stated that whereas they would feel comfortable crossing at midday, they would have be very reluctant to cross at rush hour.
This section discusses differences in how quickly participants recognized the presence of crossable gaps. In our data analyses, we kept track of all gaps that were long enough for a pedestrian to cross before the arrival of the next vehicle. For each such gap identified by participants, we calculated the time lag or latency for them to detect the gap. This was simply the time from when the last vehicle passed the crosswalk until the pedestrian pressed the button, indicating that he or she would have started crossing. Figure 5 shows the mean latencies for each group. The sighted participants, shown in blue on the right, took an average of 0.8 seconds to indicate they would initiate a crossing. They reported that they often could see upcoming gaps that were crossable, waited for the car at the beginning of such a gap to pass by, and then immediately pressed the button. Blind participants, shown on the left in yellow, took an average of 5.5 seconds to report that they would start crossing. That is, once the gap began, it took an average of more than 5 seconds for the blind pedestrians to indicate that they would start crossing. This result, which was similar to but greater than that obtained at the 3 Baltimore roundabouts, appears to be due to the masking sounds of the last vehicle to travel through the crosswalk. The blind participants simply had to wait for the sound from that vehicle to fade away so that they could listen for other approaching vehicles. In practice, this implies that blind pedestrians require longer gaps between vehicles than sighted pedestrians. Consider again the Figures 2 and 3 distributions of traffic gaps, especially the gaps shown in green ("crossable if prompt"). For most sighted pedestrians, a gap of 5 or 6 seconds should be long enough for to cross a 16-foot crosswalk. But such a gap is functionally nonexistent for a blind pedestrian who loses the first 5 seconds to the masking sounds of vehicles that have recently traveled over the crosswalk. Thus, under conditions of heavy traffic, the accessibility of a crosswalk could be much different for these two groups of pedestrians.
Fig. 5 Latency to detect crossable gaps for blind and sighted participants. This bar graph shows two bars, one for the blind participants and one for the sighted participants. The data givne in the bar graph is completely described in the previous paragraph.
The overall pattern of findings across roundabouts suggests that not only does the accessibility of modern roundabouts to blind pedestrians vary across roundabouts (as found in the Baltimore study) but also under different traffic conditions at the same roundabout (as found in the Tampa study). In agreement with the empirical findings, the blind participants generally reported that they believed the Tampa roundabout to be "crossable" during mid-day and often "uncrossable" during rush hour. Traffic engineers design streets and intersections to be capable of safely and efficiently carrying vehicular traffic during peak hours. We believe that the same should be true for the design of crosswalks: an "accessible" crosswalk must be accessible throughout the day.
Alphand, F., Noelle, U., & Guichet, B. (1991). Roundabouts and road safety: State of the art in France. In W. Brilon (Ed.), Intersections without traffic signals II (pp. 126-140). Berlin: Springer-Verlag.
American Council for the Blind (April 11, 2002). News notes. (http://www.acb.org/newsnotes/nn020111.html)
Brilon, W. (Ed.). (1991). Intersections without traffic signals II. Berlin: Springer-Verlag.
Brown, M. (1995). The design of roundabouts. London: HMSO.
Federal Highway Administration. (2000). Roundabouts: An informational guide. [Publication FHWA-RD-00-067]. McLean,Virginia: U.S. Department of Transportation, Federal Highway Administration.
Flannery, A. (2001).Geometric design and safety aspects of roundabouts. Transportation Research Record No. 1751, 76-81.
Guichet, B. (1997). Roundabouts in France: Development, safety, design, and capacity. In M. Kyte (Ed.), Proceedings of the Third International Symposium on Intersections without Traffic Signals (pp. 100-105). Portland Oregon: University of Idaho.
Guth, D., Long, R., & Ashmead, D. (2000, August). Unpublished report to the American Council of the Blind: Interviews about roundabouts with blind pedestrians in England and France. (Available from D. Guth or R. Long at the contact address listed for this paper).
Orientation and Mobility Specialists' Association of Victoria - New South Wales Associates. (2002, May 14). Notes of meeting on pedestrian access to roundabouts. Unpublished.
Persaud, B., Retting, R., Garder, P., & Lord, D. (2000). Crash reductions following installation of roundabouts in the United States. Arlington, Virginia: Insurance Institute for Highway Safety.
Retting, R., Persaud, B., Garder, P., & Lord, D. (2001). Crash and injury reduction following installation of roundabouts in the United States. American Journal of Public Health, 91, 628-631.
Sawers, C. (1996). Mini-roundabouts: Getting them right. Wingham, United Kingdom: Euro-Marketing Communications.
Troutbeck, R.J., (1993). Capacity and Design of Roundabouts in Australia. Transportation Research Record No. 1398, 68-74.
U.S. Access Board (2001). Pedestrian Access to Modern Roundabouts: Design and Operational Issues for Pedestrians who are Blind. Washington, D.C.: Author