The paper describes a method of collecting exposure data with the objective of measuring pedestrian accident risk by comparing the exposure data with pedestrian accidents. Over 97% of pedestrian accidents in New South Wales, Australia, during 1971 occurred in urban areas. For comparison with these accidents, pedestrian exposure was measured at four sites in Sydney during July and August 1973. The numbers of pedestrians and vehicles passing through small road sections (average 40 m) during 5-min intervals were observed. During each hour all sections at each site were observed consecutively. Based on a theoretical model, exposure was defined as the product of the numbers of pedestrians and vehicles observed in the same section in the same 5-min interval. This exposure was described by identical pedestrian, vehicle and environment characteristics as were the accidents, although there were probably differences in application of definitions. Accident rates (accidents divided by exposure) were analysed and indicated, with some exceptions, good agreement with expectation and past research on pedestrian accident risk and factors affecting it.