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 summary report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-RD-02-003
Driver Error Report
Depending on the source, driver error is cited as the cause of 45 to 75 percent of roadway crashes and as a contributing factor in the majority of crashes (Hankey, et al, 1999). However, which driver errors lead to a crash, and the degree to which these errors contribute to the crash, often cannot be specified. Instead, driver error is used as a general catchall category that is invoked when environmental and mechanical causes are no longer considered.
The objectives of the Identification and Evaluation of Driver Errors project included:
The remainder of this executive summary highlights the work that was conducted and the major project results.
LITERATURE REVIEW AND PRELIMINARY DATABASE EXAMINATION
The first project task was to gather and review relevant information. Three areas were reviewed.
The first area reviewed was standards documents that are typically used by roadway and traffic engineers: the Manual on Uniform Traffic Control Devices (MUTCD: U.S. Department of Transportation 1988) and A Policy on Geometric Design of Highways and Streets [American Association of State and Highway Transportation Officials (AASHTO), 1994]. (The 2000 version of the MUTCD was not yet available.) This review showed traffic engineers that use relatively straightforward design principles to integrate driver performance and driver–vehicle system performance into the design of roadway delineation and traffic control devices. The principles are intended to ensure that drivers have the information they need to in time to take the correct action: Violation of these principles increases crash risk. The review suggested that one countermeasure to crashes is to comply with the design principles, and that driver errors are often caused by failures to follow the principles.
The second area of reviewed was how error has been defined and characterized in previous research and how researchers have used taxonomies to classify errors. This review was to: (1) provide insight into driver error from experts in the field, (2) evaluate driver error classification methods used in the past, and (3) investigate how human error is classified outside of the driving domain and how they are related to causal or contributing factors. Three primary conclusions were drawn from this literature review: (1) complex diagrams of cognitively developed modeling are of limited value; (2) the underlying causes of a crash may be complex and involve multiple contributing factors or interactions between factors; (3) several models of human/driver error classification could be merged to form a suitable taxonomy to facilitate the current project.
The third area was an examination of several accident databases, with emphasis on driver error extraction and categorization, to support later project phases.
DRIVER ERROR TAXONOMY DEVELOPMENT
Tasks that supported development of driver error taxonomies were: (1) analyses of national and State accident databases, (2) conduct of focus groups with investigating officers, and (3) interviews with drivers about critical incidents in which they were involved.
The Fatality Analysis and Reporting System (FARS), the Highway Safety Information System (HSIS), the State Data Program (SDP), and the North Carolina Narrative Accident Database were analyzed. Searches of narrative fields in the North Carolina database proved to be more enlightening as to the causes of crashes, i.e., why crashes happened, whereas the non–narrative fields in the other databases were primarily useful for determining what happened, but not why. A number of taxonomies were developed. For example, taxonomies were developed that detailed characteristics of the roadway (e.g., infrastructure elements) and the characteristics of the driver (e.g., driver age). A taxonomy format that was useful in detailing why the crash occurred was the tree diagram. An example of a tree diagram taxonomy is shown in figure 9B in the main body of this report. Each successive branch of a tree provides more detailed information than the previous branch.
Five investigating officer focus groups were conducted. Police departments from Blacksburg, Roanoke, and Arlington (Virginia) represented small, medium, and large municipalities, respectively. Two divisions of the Virginia State Police were also included. The officers indicated which driver errors they believe are most frequent, what they thought were the underlying causes of the errors, and what infrastructure changes might reduce the frequency of those errors.
Fifty–four drivers from three age groups and from three driving environments (rural/small town, medium–sized city, and large city/metropolitan area) were interviewed about crashes and "close–calls" they had experienced. The drivers provided candid responses that resulted in a multitude of crash contributing factors. These factors were organized into a collision taxonomy and a close–call taxonomy that emphasized the multi–contributing factor nature of crashes. A model of contributing factors that affect driving performance was developed. That model reflects the multiple contributing factors and factor interactions that characterize most crashes.
CRITICAL INCIDENT INVESTIGATION
A critical incident is a traffic event in which a conflict occurs between two or more vehicles or between a vehicle and a pedestrian. A conflict requires at least one of the participants involved take evasive action to avoid a collision. Site surveillance was carried out at 31 roadway sites. The sites represented a variety of geometric features, and traffic control devices. Over 200 hours of video recordings were made. Data were collected at the peak traffic periods. Trained video analysts reviewed the recordings. Over 1,200 critical incidents were captured and analyzed. Each critical incident was characterized with regard to: site geometry, traffic count, location, and time of day.
Four analyses were conducted on the critical incident data: (1) generalized infrastructure analysis, (1) specific site analysis, (3) time–of–day/day–of–week analysis, and (4) most serious incident evaluation.
The generalized infrastructure analysis resulted in "Event Occurrence Taxonomies." These taxonomies were hierarchically structured. The highest taxonomy level was the the roadway geometry of the roadway site. The second level described the event (e.g., right turn). The third level described the interaction that occurred between the vehicles, or the vehicle and the pedestrian. The fourth level defined the precipitating element that affected the path of the primary vehicle. Probability models that might be incorporated into traffic simulation models were also developed.
The specific site analysis produced a list of incident clusters. This analysis was conducted only two sites: one of moderate complexity, and one of high complexity. The specific site analysis proved useful for identifying contributing factors, identifying clusters of incidents with common causal factors, and exploring countermeasures to mitigate the identified causal factors.
The time–of–day/day–of–week analysis indicated that: (1) the presence of traffic control officers has a dramatic effect on reducing the number of critical incidents, (2) critical incidents were more prevalent at the town site, (3) and critical incidents were more prevalent earlier in the week, and higher severity incidents occurred more frequently later in the day.
The most serious incident evaluation revealed that there is a similarity between critical incidents and the crash taxonomies developed earlier in the study. "Willful inappropriate behavior" was the principal contributor in 57 percent of all the critical incidents. "Inadequate knowledge" and "infrastructure" were principal contributing factors in 23 and 20 percent of the incidents, respectively.
Three primary products were delivered: (1) the probability models that predict the likelihood of a critical incident, given a certain set of characteristics; (2) a recommended methodology for analyzing sites for high frequency or severity of critical incidents; (3) and validated taxonomies that categorize critical incidents into a structured, hierarchical format.
DEVELOPMENT OF INFRASTRUCTURE–RELATED COUNTERMEASURES
To further characterize the infrastructure contribution, this analysis focused on those critical incidents for which infrastructure was a contributing factor.
Forty–three instances of infrastructure contributions were identified and described. The descriptions included diagrams that accurately and concisely the relation between the incident and the infrastructure. Candidate countermeasures were developed for each of the forty–three instances. These countermeasures provide the logical concepts for solving the problems. Many of the solutions were to modify the sites to conform to standard engineering practice, whereas others involved introduction of novel traffic control devices. Alternatives to the recommended countermeasures were also suggested. The cost of each recommendation was approximated.
The forty–three critical incidents were characterized by eight commonly recurring factors: (1) visibility blockage caused by large vehicles; (2) pedestrian right–of–way violations; (3) left turns at signalized intersections; (4) right of way confusion at two–way stop–controlled intersections; (5) entrance and exit lane inadequacies; (6) driveways near or in an intersections; (7) intersections in close proximity to one another; and (8) time–of–day dependent lane control.
Each commonly recurring factor is described, the probable cause of the driver errors is identified, potential countermeasures are proposed, and additional research recommendations are made.
DEVELOPMENT OF RECOMMENDATIONS
Three specific sets of recommendations are proposed: (1) guidelines for coding of crash data, (2) guidelines for including critical incident risk in traffic simulation models, and (3) revisions to highway design guidelines.
To improve the utility of crash databases for the analysis of the causes of driver errors it is recommended that: (1) a national uniform coding scheme be developed that meets the needs of all stakeholders be developed; (2) human factors principles be applied to the design of paper based and computerized reporting forms; (3) a section listing principal contributing factors and driving performance be included on police accident reporting forms; (4) that re–transcription of police reports be eliminated; (5) that in the development of reporting forms, police officers be included in the requirements analysis, test, and evaluation; and (6 ) safety evaluations of roadways include interviews with the police responsible for those roadways.
The probability models developed in this project are for critical incidents, not crashes. A linear relationship between critical incident probability and crash probability was assumed.
Because most of the critical incidents that were observed were related to deviations from current FHWA approved guidelines, it was concluded additional emphasis on the consequences of deviating from design guidelines be added and eight guideline statements were proposed.
For further information, or for a full copy of the report, please contact:
Topics: research, safety, human factors, Manual on Uniform Traffic Control Devices (MUTCD)
Keywords: research, safety, driver error
TRT Terms: driver errors, accident causes, infrastructure