In order to assess the relationship of the risk of injury requiring hospital attendance in children riding bicycles to sociodemographic factors and to measures of exposure, a population-based case-control study is being undertaken in a large area of suburban Melbourne, Australia. Particular attention is given to the measurement of individual exposure in several dimensions. Analysis of interim data from 109 cases and 118 controls shows that 51% of injuries occurred while the child was playing rather than making a trip on the bicycle and only 22% involved another vehicle. Boys used bicycles more commonly than girls but there was minimal evidence of an increased risk of injury in boys, adjusting for exposure. There was no evidence for an age trend in injury risk, but children from families in the lowest income category were at significantly increased risk. Exposure measures showed complex patterns of association with injury risk. Estimated time spent riding was more closely associated with risk than distance travelled, with an odds ratio of 2.2 (95% confidence interval 1.1-4.2) for children riding for more than 3 hours per week compared to children riding less than 1 hour. Riding more than 5 km on the sidewalk was also associated with increased risk (odds ratio 3.1, 95% CI 1.1-8.5). The elevated risk associated with sidewalk riding may be due to difficulties in negotiating uneven surfaces. The case-control study provides an ideal design for this type of investigation but valid and reliable measurement of exposure is difficult.