This study was performed in order to evaluate the accuracy with which pedestrians estimate the speed of nearby automobiles. A total of 87 subjects were involved in this experiment, with 487 useful speed estimates being obtained from them. The vehicle speeds were measured using radar guns, and the moving cars were located upstream, downstream, and adjacent to the subjects. A survey form was used to obtain attributes about each subject. A multivariate regression was used to allocate portions of the variation in the percent error among the various factors that describe the subject and the moving vehicle. Statistical significance was discovered for the main effects and some interactions among: gender, driving experience, a self-assessment of ability to estimate such speeds, the use of the posted speed limit, the speed of the target vehicle, and the location of the vehicle on the roadway. The implications for speed estimation accuracy are presented, both for categories of subjects and for individual subjects. The standard deviation in the accuracy of individuals was remarkable for its large magnitude and its persistence over all the categories of subjects and also the different vehicle speeds and locations.