This individual-level case control study analyzed the risk of occurrence of a pedestrian–motor vehicle collision at a given location on a state route in King County, Washington. With the full sample of collisions (1999-2004), binomial logit models estimated the odds of collision occurrence as related to the road and the neighborhood environments and adjusting for exposure. Separate models were run for SR-99, the principal transregional arterial with four-plus lanes ( n = 826), and for all the other state routes ( n = 1,602). The strongest significant correlates of the risk of collision occurrence were the presence of crosswalks with or without traffic signals, the facility's number of lanes, and the presence of nearby retail uses. Also positively significant were the number of traffic signals and the street-block size near the location and being located outside the city of Seattle. Exposure variables including road-level measures such as average daily traffic and posted speed and neighborhood-level measures such as the number of residential units and bus ridership were significant in at least one of the models. Employment density appeared to be an unreliable measure of exposure. Other pedestrian activity generators such as educational facilities were not significantly associated with the risk of collision occurrence. The strong significance of crosswalks, whether signalized or not, indicated that engineering approaches to safety should likely be complemented by education- and enforcement-based measures. Similarly, facilities in areas with concentrations of retail should become the targets of future safety programs.