U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
|Publication Number: FHWA-HRT-13-044 Date: August 2013|
Publication Number: FHWA-HRT-13-044
Date: August 2013
The objective of this project was to conduct studies on how driver behavior and performance are affected by messaging features within the right-of-way (ROW) in a variety of contexts and to use the new information in the development of evidence-based guidance that supports more effective communication to roadway users.
A review of literature on messaging in the ROW indicated that little is known about the interaction of traffic control device (TCD) messaging and the roadway environment. The review suggested several potential approaches for advancing understanding of this interaction and provided a basis for the series of experiments that followed.
An initial study used a multidimensional scaling (MDS) procedure to identify the factors that drivers consider when viewing roadway environments. Two factors were identified. The dominant factor was aesthetic pleasantness or degree of built structures and clutter. That is, areas with few built structures were judged aesthetically pleasing, and those with a lot of built structures were judged cluttered and less pleasant. A less dominant factor was the degree of organization in the scene. However, organization was correlated with aesthetic pleasantness-organized environments were judged more pleasing, even if built up. It is suggested that roadway classifications based on user judgments may be useful in organizing guidance on TCD design and placement. However, the present study, which examined only major arterials in one community, provides just a hint of the potential utility of such a system.
To communicate effectively, TCDs must be detected. Detection relies on both visibility and conspicuity. The factors that affect visibility are reasonably well understood. Conspicuity, the properties of a sign that make it likely to capture visual attention, is less well understood. Two common measures of conspicuity are sign recall and eye glances to signs. The present project used an eye tracker to record eye gaze location and queried participants on sign recall immediately after selected signs were passed. The most striking finding was that speed limit identification was good regardless of whether the speed limit signs received a gaze, whereas warning sign message identification was more dependent on the sign receiving a gaze. Speed limits were correctly identified about 80 percent of the time whether or not gazes to speed limit signs were detected. Warning sign messages that did not receive a gaze were recalled only 37 percent of the time. When gazed at, warning sign messages were correctly identified with about the same frequency as speed limits. Drivers who were unfamiliar with the roads were more likely to recall warning signs than drivers who were familiar with the roads. Drivers who were not asked to recall TCDs were somewhat less likely to glance at the signs than those who had been asked to recall TCDs. However, regardless of whether drivers were asked to recall signs or not, fewer than half of warning or speed limit signs received glances. The classification dimensions identified in the MDS study did not significantly correlate with either glance or recall performance.
It was concluded that neither glance nor recall data alone are sufficient to characterize the conspicuity of TCDs.
The ability to detect and identify signs with peripheral or near peripheral vision was also examined. In the laboratory, speed limit and warning signs were presented in various roadway scenes. Outdoors, the same signs were presented with two different background environments.
These experiments showed that drivers can detect the presence of signs in their far peripheral vision (20-70° from the point of gaze, depending on background conditions). Sign detection can be followed by a glance to the sign to read its message. However, the field experiment results suggested that drivers read some signs without looking directly at them. The final study in this project measured the critical angle at which warning and speed limit signs can be read in the absence of a direct look.
Five speed limit signs and five text-based warning signs served as target stimuli in a laboratory experiment to determine the peripheral angle at which the signs can be read. Each sign was presented in each of six roadway contexts. Each of the roadway contexts was a panoramic view that was included in the MDS study. A common psychometric technique, the method of limits, was used to measure the angle between the direction of gaze and the content on signs that could be identified.
Speed limit identification performance was superior to warning sign identification performance. Identification performance for both sign types decreased as the angle away from the fixation point increased. Speed limit signs were identified with over 80 percent accuracy at angles out to ±9° from gaze location. Warning signs were identified with above 80 percent accuracy with a 3° gaze offset, and identification remained above chance accuracy out to 15°.
The sign identification findings confirm that traffic sign messages can be recognized in the absence of foveal glances to them. Thus, failures to gaze at signs cannot be taken as evidence that drivers are unaware of sign content. Awareness of sign content, then, is a better measure of sign conspicuity than eye glances to a sign. This does not mean that glance data are not useful. In the laboratory, where the participant's only task is to identify signs, the probability of correctly reading a familiar sign without a direct glance is high. In an actual roadway environment, where the driver must attend to more than just signs, the probability of correctly reading a sign without a glance is probably lower. Furthermore, if gaze direction does not fall within 9° of a sign, it is unlikely that the content of the sign will be registered.
There is no generally accepted measure of the conspicuity of TCDs. Eye tracking has been used to assess whether TCDs receive glances, but the research reported here indicates that some TCD content was noticed and remembered even though the TCD did not receive a direct eye glance. In addition, about 20 percent of signs that received glances could not be recalled 2 s after they were passed. A field test that used an eye-tracking device and asked drivers to recall selected signs did not show a predictable effect of sign environment on glance probability or recall. However, two other psychophysical testing methods indicated that background environment influences sign detectability, especially regulatory sign detectability. With light colored or cluttered surrounds, speed limit signs were about half as detectable as they were with dark, uncluttered surrounds. Although further research is needed to develop empirically based guidance, the results of these studies suggest that guidelines for practitioners should be developed to suggest when TCD conspicuity enhancements are needed and which available enhancements are likely to be most effective in specific situations.
With the surrounds explored in these studies, warning sign detectability did not vary by predictable or practically significant amounts as a function of background. However, this should not be taken to mean that warning signs detectability would not be markedly degraded by more complex surrounds than were explored in these studies. In particular, the present results should not be used as support for posting more than one sign on a roadside support, as the literature suggests that crowding would reduce readability.
The final section of this report provides guidelines for ensuring the effectiveness of traffic control signs. These guidelines are based on an integration of the present findings with guidelines and recommendations from the Manual on Uniform Traffic Control Devices (MUTCD) and human factors principles from other resources.
The immediate objective of this project was to conduct studies on how driver behavior and performance are affected by messaging features within the ROW in a variety of contexts. The ultimate objective was to develop evidence-based guidance that would support roadway designers and planners in more effectively communicating to roadway users.
In the first phase of this project, a literature review was conducted to identify what is known about how the roadway environment interacts with messaging features. A small study was conducted to identify environmental features that are important to motorists' perceptions. This study was followed by an on-road observational study in which drivers' glances to TCDs and recall of those TCDs were recorded. Because the literature on TCDs emphasizes the importance of visibility and conspicuity, a series of experiments were conducted to assess conspicuity of warning and speed limit signs as a function of the environment in which they are viewed. Based on the literature and the studies conducted as part of this project, recommendations for practitioners were developed and future research needs were identified.
A review of the literature on driver behavior suggested that little is known about the effects of the interaction of the roadway environment with roadway messaging features. Lay categorized sign use into four stages: (1) detection, (2) reading, (3) understanding, and (4) action. The present studies address factors that affect the first two stages, sign detection and reading.(1)
To ensure detection, a sign or marking must be visible and conspicuous. For visibility, a sign must be located in the driver's field of view, in a clear line of sight, and in adequate lighting. Conspicuity is generally defined in terms of the probability of an object being noticed. However, whether an object is noticed is not easily defined. For example, one measure of conspicuity used by Cole and Jenkins was whether participants fixated on an object.(2) However, Luoma found that despite a pedestrian warning sign being fixated on by 62 percent of drivers, only 8 percent of those drivers were aware of the warning when asked to identify the sign seconds after it was passed.(3) This suggests that eye fixations alone do not ensure that a sign has been attended to. Alternatively, there are cases in which stimuli can be perceived without a driver fixating on them. For instance, Luoma found that after drivers turned at an intersection, more than half reported being aware that the intersection lacked lane markings even though an eye tracker showed that they had not fixated on the areas of pavement where the lane markings would have been.(3) Eye tracking in dynamic environments is complex, and basic research in this area continues to evolve.(4-6) Therefore, terms such as fixation, glance, gaze, and look should be interpreted with caution. It should not be assumed that field data are measured with the same level of accuracy as laboratory data obtained when the observer's head is fixed, the lighting is optimized for measurement, the stimuli are static, and the sampling rate for eye position is equal to or greater than 240 Hz. For driving studies in which eye glances are reported, it can only be assumed that the evidence that an object was (or was not) brought into or near the center of gaze was adequate. As discussed in this report, observers can perceive objects without looking at or fixating on them. In a three-dimensional, dynamic world, it can be difficult to determine whether the 2° of foveal vision are focused on a large, near object or a small, far object that is juxtaposed with the near object in a two-dimensional plane. The approximately 2° cone of fine vision provided by the fovea of the eye describes a 17-ft (5.2-m) cone at a distance of 500 ft (153 m) from the observer. Many objects can fall within that cone of vision, and attention can be focused on any object within that cone as well as objects outside that cone if fine detail discrimination is not required.
Because recall and eye glance data do not appear to provide definitive measures of conspicuity, the inclusion of additional measures is desirable. Wertheim recently suggested a critical conspicuity distance, which is the lateral distance away from an object that a person can fixate on and still detect or identify the object.(7) This measure may be important when observers are primed to attend to certain targets. For example, drivers might be primed to attend to traffic signs. When signs are outside of this critical conspicuity distance, they are likely to go unnoticed. Variations on Wertheim's technique were used in two of the studies reported here.
Most investigators are in agreement that the environment around a sign is an important factor in determining its conspicuity. (See references 2 and 8-10.) However, there are few studies that investigate how to systematically characterize the roadway environment in ways that are relevant to TCD conspicuity. Typical engineering classifications are based on land use (e.g., rural or urban) or roadway function (e.g., feeder, minor arterial, or major arterial) rather than on drivers' perceptions or the appearance of the immediate surrounds of a TCD.
The first study reported here was conducted to assess one possible method of roadway classification that would be more sensitive to drivers' perceptions. The second study was a field study in which an instrumented vehicle was used to record drivers' eye glances to TCDs. After passing each selected TCD, the drivers were asked to identify the TCD so that eye glances could be related to TCD recall. The third study assessed TCD detection conspicuity, the largest angle away from a sign that a driver could gaze and still detect sign presence. The fourth study assessed TCD identification conspicuity, the largest angle away from a sign that a driver could gaze and still read a sign.