To determine whether commonly used psychoactive drugs increase the risk of involvement in motor vehicle crashes for drivers > or = 65 years of age, the authors conducted a retrospective cohort study. Data were obtained from computerized files from the Tennessee Medicaid program, driver's license files, and police reports of injurious crashes. Cohort members were Medicaid enrollees 65-84 years of age who had a valid driver's license during the study period 1984-1988 and who met other criteria designed to exclude persons unlikely to be drivers and to ensure availability of necessary study data. There were 16,262 persons in the study cohort with 38,701 person-years of follow-up and involvement in 495 injurious crashes. For four groups of psychoactive drugs (benzodiazepines, cyclic antidepressants, oral opioid analgesics, and antihistamines), the risk of crash involvement was calculated with Poisson regression models that controlled for demographic characteristics and use of medical care as an indicator of health status. The relative risk of injurious crash involvement for current users of any psychoactive drug was 1.5 (95% confidence interval (CI) 1.2-1.9). This increased risk was confined to benzodiazepines (relative risk = 1.5; 95% CI 1.2-1.9) and cyclic antidepressants (relative risk = 2.2; 95% CI 1.3-3.5). For these drugs, the relative risk increased with dose and was substantial for high doses: 2.4 (95% CI 1.3-4.4) for > or = 20 mg of diazepam and 5.5 (95% CI 2.6-11.6) for > or = 125 mg of amitriptyline. Analysis of data for the crash-involved drivers suggested that these findings were not due to confounding by alcohol use or driving frequency.