BACKGROUND: Neurologically impaired persons seem to benefit from driving-training programs, but there is no convincing evidence to support this notion. The authors therefore investigated the effect of simulator-based training on driving after stroke. METHODS: Eighty-three first-ever subacute stroke patients entered a 5-week 15-hour training program in which they were randomly allocated to either an experimental (simulator-based training) or control (driving-related cognitive tasks) group. Performance in off-road evaluations and an on-road test were used to assess the driving ability of subjects pre- and post-training. Outcome of an official predriving assessment administered 6 to 9 months poststroke was also considered. RESULTS: Both groups significantly improved in a visual and many neuropsychological evaluations and in the on-road test after training. There were no significant differences between both groups in improvements from pre- to post-training except in the "road sign recognition test" in which the experimental subjects improved more. Significant improvements in the three-class decision ("fit to drive," "temporarily unfit to drive," and "unfit to drive") were found in favor of the experimental group post-training. Academic qualification and overall disability together determined subjects that benefited most from the simulator-based driving training. Significantly more experimental subjects (73%) than control subjects (42%) passed the follow-up official predriving assessment and were legally allowed to resume driving. CONCLUSIONS: Simulator-based driving training improved driving ability, especially for well educated and less disabled stroke patients. However, the findings of the study may have been modified as a result of the large number of dropouts and the possibility of some neurologic recovery unrelated to training.