Because of the shortcomings in accident data, such as incomplete reporting and inaccurate information on accident reports, a need was found for indicators of accident potential. Nearly half of the 209 locations identified in Kentucky as hazardous by accident criteria were found to be falsely identified due to random accident occurrences. Accident repeatability from one year to the next was found to be poor at 60 intersections (r = 0.64) and 170 spot locations (r = 0.59). Up to two years of accident data were found to be necessary to obtain a reliable base of accident data.

Conflict counts were conducted at five intersections in Central Kentucky to determine characteristics of conflict data. Good reliability was found between observers in simultaneous counts of conflicts and weaves with r-values as high as 0.93. Traffic volumes accounted for only about 30 percent of the variation in numbers of conflicts. Conflict numbers, types, and rates were found to be very repeatable at one intersection. A revised procedure for collecting data in Kentucky, which increased the sample size and reduced the required manpower, was described. Modified data sheets were also developed for signalized and unsignalized intersections.

Results from evaluations of safety improvements in Kentucky using conflicts, erratic meanuevers, and accidents were summarized. Reductions in accidents (85 percent) and conflicts (81 percent) were found at intersections where left-turn signal phasing was added. Installation of green-phase extension resulted in conflict and accident reductions of 62 and 54 percent, respectively, at several high-speed intersections. Erratic maneuvers were reduced by 27 percent after installing raised pavement markers at five freeway lane-drop locations. Procedures for intersection analysis using conflict diagrams were described. Conflict counts were recommended during routine inspections of suspected hazardous locations.

Report Date


Report Number

No. 482

Digital Object Identifier



Offered for publication to the Transportation Research Board.