Conditions that Influence Drivers' Yielding Behavior for Uncontrolled Crossings Eugene Bourquin, Robert Wall Emerson, and Dona Sauerburger
Abstract: Pedestrians with visual impairments need to cross streets where traffic signals and traffic signage are not present. This study examined the influences of several interventions, including a pedestrian's use of a mobility cane, on the behavior of drivers when they were expected to yield to a pedestrian crossing at an uncontrolled crossing.
This project was supported by Grant R01EY12894 from the National Eye Institute. The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the National Eye Institute or the National Institutes of Health.
Pedestrians must be able to analyze levels of risk in order to make decisions about street crossings, such as considering how much risk is acceptable and when it is preferable to look for an alternative to a given crossing. Assessing the level of risk can include determining the likelihood that drivers will yield to a pedestrian crossing a street. To some extent, all pedestrians depend on the yielding behavior of drivers, whether by intention or circumstance, and regardless of laws and regulations.
Research on yielding by drivers
Harrell (1993) found that dependency cues influenced drivers' behavior: "Motorists were significantly more likely to stop for a blind pedestrian than for a sighted pedestrian. This finding is consistent with the norm of social responsibility" (Harell, 1994, p. 529). Other studies also found that feelings of empathy (Batson, Chang, Orr, & Rowland, 2002; Taylor, 1998) and perceptions of dependence (Baker & Reitz, 1978) increased helping behavior toward people with physical and sensory disabilities. In a controlled experiment of 960 interactions between drivers and sighted pedestrians, researchers found that drivers slowed down or stopped when the speeds of their vehicles were low (Katz, Zaidel, & Elgrishi, 1975).
Ashmead, Guth, Wall, Long, and Ponchillia (2005) found that at roundabout entry lanes, drivers yielded to pedestrians who were standing at the edge of the crosswalk 36.4% of the time if the pedestrians were holding a white cane but only 20% if they were not. At exit lane crossings, where speeds were higher, none of the drivers yielded to pedestrians who were standing at the beginning of the crosswalk with no white cane and 9% of the time if the pedestrians had a white cane. Geruschat and Hassan (2005) reported that when pedestrians stood 1 foot from the curb at a roundabout, drivers yielded 45% of the time; when the pedestrian stood at the curb, drivers yielded 52% of the time; and when the pedestrian stood in the crosswalk, drivers yielded 60% of the time. When the long cane was present, drivers yielded 63% of the time, whereas when the long cane was not present, they yielded 52% of the time. Guth, Ashmead, Long, Wall, and Ponchillia (2005) found that at a roundabout, if the pedestrian had a clearly visible white cane, the likelihood that a driver would yield increased from 4% to 30% at the exit lane and from 44% to 60% at the entry lane.
No studies have observed how drivers behave if a pedestrian walks into the street. The purpose of the study presented here was to find out whether using a white cane, wearing an orange vest, or waving a red flag increased the likelihood of drivers yielding for pedestrians who walked toward their path in such a manner that a collision would occur if neither the driver nor the pedestrian stopped.
Method
Approval for the study was received from the Human Subjects Institutional Review Board of Western Michigan University. The data for the study were collected at four sites, two in Kalamazoo, Michigan, and two in the Maryland suburbs of Washington, DC. Each site was the residential intersection of a two-lane minor street with stop sign control and a two- or three-lane major arterial street with no traffic control. The drivers (who were not aware that a study was in progress) approached the intersection on the major street with a straight line of sight for at least 300 yards. The characteristics of each site are shown in Table 1. In Michigan, it was overcast and cool, and in Maryland, it was warm and sunny, with the sun high and toward the right of the drivers (slightly forward of the drivers at the Mount Tabor Road site, slightly behind them at the Old Chapel Road site).
At each site, the collaborating pedestrian walked into the first lane of the major street (empty of traffic) toward the second lane of the major street in such a way that he was on a collision path with a single approaching vehicle. Drivers in the second lane were used to ensure that they had more than three seconds of warning once the pedestrian had entered the street. Single approaching vehicles were chosen so that drivers would not be hit from behind if they came to a sudden stop. The pedestrian was 6 feet, 1 inch tall and about 185 pounds. He wore medium dark clothing and wraparound amber tinted eyeglasses for every trial. As he started to cross and continued walking, his head was facing forward or slightly away from the vehicle. When he was still in the first lane and 6 inches to 1 foot from the second lane, he stopped, rocked back, and forth, and turned his head slightly from side to side.
In this study the pedestrian crossed in one of the six following conditions:
Control:
no identifier or signal associated with the pedestrian.
Flag: held a bright red flag (16 inches × 15 inches) head high and angled forward and moved it from side to side (at about the same speed and arc that he had moved the white cane).
Vest: wore an orange vest with reflective strips over his clothing.
Cane: held a cane in his right hand, centered in front of his body, and moved it along the ground tapping from side to side in rhythm with his steps in an arc about 2 feet wide.
Cane waving: used a cane as in Condition 4, but, before he started to cross, he moved the tip of the cane over his head and then down to the ground twice.
Cane waving-vest: wore an orange vest while doing the same as in Condition 5.
Conditions 2-5 were selected because they had been suggested by mobility specialists and travel instructors as methods to keep travelers safer.
In preparation for the crossing trials, the collaborating pedestrian practiced walking at a steady, predictable speed across the street when using the cane or flag or nothing, timing his crossings to make sure he was walking at the same speed in all conditions. He then determined how many seconds were needed to walk across the first lane and step a few feet into the second lane. Collaborator 1 (see Figure 1) used the Timing Method for Assessing Speed and Distance of Vehicles (Sauerburger, n.d.) to train herself to predict when the vehicles approaching in the second lane were at such a speed and distance that they would be on a collision course with the pedestrian if he started to cross at that time.
After this preparation, Collaborator 2 positioned himself downstream from the approaching vehicles to measure the speed of the vehicles with a Class 1 laser speed gun (SpeedLaser, www.LaserAtlanta.com). The laser speed gun gave a real-time read-out of speed. The pedestrian stood near the edge of the major street, and another collaborator stood beside the pedestrian (away from the traffic approaching in the second lane) (see Figure 1). When a vehicle approached in the second lane without other vehicles close behind it and the vehicle was about 4.5 seconds away (5.5 seconds for the Winchell site), the pedestrian started to walk across the street (if he was doing one of the cane-waving conditions, he would already have waved the cane before starting to cross).
For each trial, we collected data on the type of vehicle, speed of approaching vehicles, and yielding status of each vehicle. Vehicles were classed as trucks (vans, pickup trucks, delivery vehicles, and semitrailers) or cars (essentially all vehicles not in the truck category). The speed of approaching vehicles was measured by the SpeedLaser gun at no less than 300 feet and no more than 400 feet from the crossing point. All the vehicles included in the data set demonstrated a fluctuation of less than 4 miles per hour (mph) as they approached and moved through the 300- to 400-foot zone. Speed was monitored as the pedestrian began crossing to categorize the vehicle's response. Vehicle responses were broadly categorized into yield or no yield categories. The yield category included vehicles that slowed to a full stop as well as "rolling yields." Rolling yields were coded when a vehicle had slowed to less than 10 mph (the lower limit of the speed gun) and were far enough away from the pedestrian that, if necessary, the driver could have stopped fully without affecting the pedestrian. The no yield category included vehicles that did not slow at all, sped up to the crossing point, or did not slow enough so that the pedestrian could continue to cross without a pedestrian-vehicle conflict. There were few instances in which a vehicle was not obviously in one category or another. Cases in which a vehicle hesitated or slowed minimally and then continued along the roadway were coded as nonyields. Vehicles that yielded generally did so obviously and at least 75 feet from the pedestrian, and vehicles that did not yield generally did not slow at all (and often honked or swerved as they passed the pedestrian).
For the purpose of the study, it was important that the pedestrian start to approach the second lane when the drivers perceived that they would hit him if they did not yield or slow down. If the pedestrian started too early and stopped before reaching the second lane while the vehicles were still too far back from the intersection, the drivers might realize that they did not need to yield because the pedestrian seemed to be yielding to them. If the pedestrian started too late, the drivers might have realized that he or she would not reach their lane before they passed, and they could therefore avoid hitting him if they just kept going or sped up. Also, if the pedestrian started too late, the drivers would have had less warning. Therefore, whenever drivers failed to yield and it was determined that the pedestrian had started crossing too early or too late, that trial was discarded. This determination was made by starting a timer whenever the pedestrian started to cross and stopping the timer when the vehicle passed the pedestrian or, if the vehicle slowed down but did not stop, when it was projected that the vehicle would have passed if it had not slowed down. If the time of passing (or projected time of passing) was fewer than 3.9 seconds or more than 6.5 seconds (4-7 seconds for the Winchell site), that trial was rejected. Rejection occurred 5 to 10 times throughout the data collection.
Results
Because of the disparities between the characteristics of the sites and the trials, analyses were performed both on the whole data set and on only the Kalamazoo data. Results from the Kalamazoo subset indicated general patterns of results that we then checked and extrapolated on by adding the Maryland data (for example, to obtain a larger data set with higher approach speeds).
The result of a driver yield or no yield was examined for each of the variables of crossing condition, type of vehicle, vehicle approach speed, and time of day. A chi square for yielding and crossing condition for the Kalamazoo data was significant--X2(5) = 61.3, p <.0001--which was supported in the larger data set--X2(5) = 70.9, p <.0001.
Table 2 shows the rate of yielding for each of the crossing conditions in the larger data set. The results in the table were evaluated using a Bonferroni correction, which resulted in a significance level of .05/5 or .01 being needed for statistical significance. There was a distinct break in the data when the cane was introduced; wearing a vest or waving a red flag created no difference in yielding than did using nothing at all.
A chi-square analysis showed that there was no significant difference in yielding rates by type of vehicle (truck versus car) when just the Kalamazoo data were analyzed--X2(1) = 0.76, p = .38--or when the larger data set was analyzed--X2(1) = 1.34, p = .25. A logistic regression showed that there was no significant difference in yielding rates for initial vehicle-approach speeds when the Kalamazoo data were analyzed, but there was a significant difference when the larger data set was analyzed (see Table 3). Note that in this analysis, an odds ratio of less than 1.00 indicates that the likelihood of a driver yielding decreases with an increase in approach speed.
There was a significant difference in yielding rates for the time of day of the data collection in the Kalamazoo data (X2(8) = 16.07, p = .04), which was not demonstrated in the larger data set (X2(8) = 12.72, p = .12). In general, it seems there may be a slight tendency for drivers to be less likely to yield later in the day, but the main differences seen in yielding were across the crossing conditions. A binary logistic regression was run with yielding (yes or no) as the dependent variable and crossing condition, vehicle approach speed, time of day, type of vehicle, and data site entered as possible predictors. For the entire data set, crossing condition was the primary predictor, followed by initial approach speed (see Table 4); the binary logistic regression was rerun using only the cane crossing conditions, and vehicle approach speed became the primary predictor, followed by the time of day (see Table 5).
Figure 2 shows the percentage of vehicles yielding at each approach speed, broken down by the conditions in which the pedestrian was crossing using a cane and the conditions in which he or she was not. Although vehicles always reduced their speed when yielding, the speeds in the figure denote the initial approach speeds of the vehicles before they began to yield. No trial had an approach speed of less than 20 mph.
Discussion and conclusions
Studies with waiting pedestrians showed that using a cane increased yielding from 52% to 63% (Geruschat & Hassan, 2005) and from 0 to 9% or 20% to 36.4% (Ashmead et al., 2005). When the pedestrian was actually walking in the street, using a cane increased the vehicles' yielding from 41% to about 90%. For pedestrians who are actually crossing the street instead of waiting to cross, the presence of a cane has a far greater impact on drivers. The effect of using a cane can be seen by virtue of there being a higher percentage of yielding for the cane-using conditions at every speed except 42 mph and 46 mph. The effect of approach speed can be seen by virtue of the fact that the yielding rate generally decreases as the speed increases. While the presence of a cane at lower speeds resulted in high rates of yielding, higher speeds substantially reduced the effect. This effect was not caused by less warning time for drivers who were approaching faster, since we controlled for this possibility and all the drivers had approximately the same warning time regardless of their speeds. This secondary effect of the speed was moderated by the time of day, which is most likely reflective of traffic patterns related to destination or the psychological state of drivers in the morning or afternoon.
The presence of a white cane reliably predicted significantly greater yielding by drivers. We hypothesized that on the basis of dependency cues, a pedestrian using a white cane may encounter greater yielding by drivers when crossing at uncontrolled crosswalks. The white cane is readily identified as a cue for dependence, causing drivers to respond more often within the norms of social responsibility. Conversely, an orange vest or waving flag, although visually salient, did not signal dependence.
Dependency cues alone may not fully explain the differences in response to a cane, vest, and flag, since the stimulus must be noticed (seen) before it can be processed for meaning. Research on visual cognition may offer insights. Psychologists have studied the sensory and cognitive processes that determine how people selectively notice or ignore visual information (Hughes, Vachon, & Jones, 2005; Most, Scholl, Clifford, & Simons, 2005; Ramachandran & Rogers-Ramachandran, 2005). Specifically, attentional set theory suggests that humans see what they are precued to notice, and Mack, Pappas, Silverman, and Gay (2002) found that objects that have relevance and meaning to the observer are more likely to be seen. In a particularly relevant laboratory experiment, Most and Astur (2007) found that drivers in a simulator collided less often when their attention set matched a motorcycle veering into their path. They concluded: "In sum, effects of attentional set extended to behaviour in a realistic, safety-relevant scenario, profoundly influencing the number of accidents that occurred. This is especially striking because the situation approximated an everyday task and the motorcycle was task-relevant" (p. 131). We infer from the results of our experiment that the white cane is noticed and understood because it is found within many drivers' precued attentional sets.
There are implications for practice. Those who encourage their students or family members to wear an orange vest or wave a flag to increase safety should be aware that it may have little or no effect during the day, other than perhaps raising false expectations of safety. Understanding these variables, pedestrians and orientation and mobility specialists can make better assessments about how to increase the likelihood that their crossing will be safer. For example, before starting to cross, pedestrians with visual impairments can display their canes prominently for drivers to see and should be aware that at crossings with faster traffic, they may experience less yielding. Pedestrians who do not use a cane because they do not require a tool to detect hazards may nonetheless carry a long white cane to use when crossing streets to increase the likelihood that drivers will yield as they cross. It is possible that since the drivers in this study were so greatly affected by the presence of a cane, drivers and other pedestrians may respond to the cane in other travel situations: when pedestrians are soliciting assistance from other pedestrians and drivers or expecting pedestrians and bicycle riders to avoid them on sidewalks and in congested areas.
Limitations of the study and the need for further research
The study was designed to elicit a real-life experience in which drivers were led to believe that they might hit a pedestrian if they did not stop. Even though everyday situations involve drivers approaching in platoons or clusters, and platooning can be dangerous (Appert-Roll& Santen, 2007), to ensure the safety of both the collaborators and the drivers, crossings were initiated only when the approaching vehicle from the right was not in a platoon and there was no approaching traffic to the left of the pedestrian.
There were no dog guide users in this study, so we do not know what effect the presence of a dog guide would have had on drivers who thought they were on a collision path with the pedestrian. Studying this effect would have been complicated by the fact that dog guides are trained to disobey any command to move forward into the path of an approaching vehicle. The only relevant study with guide dogs was Guth et al.'s (2005), in which the presence of a dog guide resulted in less yielding than a cane; in the lowest-yield condition (5%), the presence of the dog actually resulted in less yielding than no mobility device at all. While a general inference that the cane causes more yielding than a dog guide may be reasonable, conclusions about uncontrolled intersections will require further study.
Collaborators other than the crossing pedestrian were present within the drivers' general visual landscape. Both collaborators were in discrete locations and present across all the trials; however, we do not know the effects, if any, they had on yielding. Furthermore, we did not monitor the drivers' eye movements, and we do not have physical evidence of the drivers' attention during the trials.
Finally, we did not examine the conditions that influence drivers' yielding at night. There is some evidence that pedestrians' conspicuity with retroreflective materials improves drivers' recognition of pedestrians at night, but simply adding a vest over clothing was not effective (Wood, Tyrrell, & Carberry, 2005).
References
Appert-Rolland, C., & Santen, L. (2007). Accidents in platoons of vehicles. In T. P. Andreas Schadschneider, Reinhart Kühne, Michael Schreckenberg, and Dietrich E. Wolf (Eds.), Traffic and granular flow '05 (pp. 623-631). New York: Springer.
Ashmead, D. H., Guth, D., Wall, R. S., Long, R. G., & Ponchillia, P. E. (2005). Street crossing by sighted and blind pedestrians at a modern roundabout. Journal of Transportation Engineering, 131, 812-821.
Baker, L. D., & Reitz, H. J. (1978). Altruism toward the blind: Effects of sex of helper and dependency of victim. Journal of Social Psychology, 104(1), 19-28.
Batson, C. D., Chang, J., Orr, R., & Rowland, J. (2002). Empathy, attitudes, and action: Can feeling for a member of a stigmatized group motivate one to help the group? Personality and Social Psychology Bulletin, 28, 1656-1666.
Geruschat, D. R., & Hassan, S. E. (2005). Driver behavior in yielding to sighted and blind pedestrians at roundabouts. Journal of Visual Impairment & Blindness, 99, 286-02.
Guth, D., Ashmead, D., Long, R., Wall, R., & Ponchillia., P. (2005). Blind and sighted pedestrians' judgments of gaps in traffic at roundabouts. Human Factors, 47, 314-331.
Harrell, W. A. (1993). The impact of pedestrian visibility and assertiveness on motorist yielding. Journal of Social Psychology, 133, 353-60.
Harrell, W. A. (1994). Effects of blind pedestrians on motorists. Journal of Social Psychology, 134, 529-539.
Hughes, R. W., Vachon, F., & Jones, D. M. (2005). Auditory attentional capture during serial recall: Violations at encoding of an algorithm-based neural model? Journal of Experimental Psychology/Learning, Memory & Cognition, 31, 736-749.
Katz, A., Zaidel, D., & Elgrishi, A. (1975). An experimental study of driver and pedestrian interaction during the crossing conflict. Human Factors, 17(5), 514-527.
Mack, A., Pappas, Z., Silverman, M., & Gay, R. (2002). What we see: Inattention and the capture of attention by meaning. Consciousness and Cognition, 11, 488-506.
Most, S. B., & Astur, R. S. (2007). Feature-based attentional set as a cause of traffic accidents. Visual Cognition, 15, 125-132.
Most, S. B., Scholl, B. J., Clifford, E. R., & Simons, D. J. (2005). What you see is what you set: Sustained inattentional blindness and the capture of awareness. Psychological Review, 112, 217-242.
Ramachandran, V. S., & Rogers-Ramachandran, D. (2005). How blind are we? Scientific American Mind, 16, 96-95.
Taylor, C. J. (1998). Factors affecting behavior toward people with disabilities. Journal of Social Psychology, 138, 766-771.
Wood, J. M., Tyrrell, R. A., & Carberry, T. P. (2005). Limitations in drivers' ability to recognize pedestrians at night. Human Factors, 47, 644-653.
Eugene Bourquin, DHA, senior instructor, Helen Keller National Center, 315 Eighth Avenue, 10E, New York, NY 10001; e-mail: <oandmhk@msn.com>.
Robert Wall Emerson, Ph.D., professor, Department of Blindness and Low Vision Studies, Western Michigan, University, 1903 West Michigan Avenue, Kalamazoo, MI 49008-5218; e-mail: <robert.wall@wmich.edu>.
Dona Sauerburger, COMS, 1606 Huntcliff Way, Gambrills, MD 21054; e-mail: <dona@sauerburger.org>.