OBJECTIVES: To test the effects of cognitive training on subsequent motor vehicle collision (MVC) involvement of older drivers. DESIGN: Randomized, controlled, multisite, single-blind clinical trial. SETTING: Community-dwelling seniors at four U.S. sites: Birmingham, Alabama; Baltimore, Maryland; Indianapolis, Indiana; and State College, Pennsylvania. PARTICIPANTS: Nine hundred eight older drivers (mean age 73.1; 18.6% African American) who were randomized to one of three cognitive interventions or a control condition. INTERVENTIONS: Up to 10 sessions of cognitive training for memory, reasoning, or speed of processing. MEASUREMENTS: State-recorded MVC involvement up to 6 years after study enrollment. RESULTS: Speed-of-processing and reasoning training resulted in lower rates of at-fault collision involvement over the subsequent approximately 6-year period than controls. After adjusting for age, sex, race, education, mental status, health, vision, depressive symptoms, and testing site, participants randomized to the speed-of-processing and reasoning interventions had an approximately 50% lower rate (per person-mile) of at-fault MVCs than the control group (rate ratio (RR) = 0.57, 95% confidence interval (CI) = 0.34-0.96 for speed of processing), and (RR = 0.50, 95% CI = 0.27-0.92 for reasoning). There was no significant difference observed for the memory group. CONCLUSION: Cognitive speed-of-processing and reasoning training resulted in a lower at-fault MVC rate in older drivers than in controls. Considering the importance of driving mobility, the costs of crashes, and the benefits of cognitive training, these interventions have great potential to sustain independence and quality of life of older adults. More research is needed to understand the effects of different types and quantities of training.