Background. Several driving retraining programs have been developed to improve driving skills after stroke. Those programs rely on different rehabilitation concepts. Objectives. The current study sought to examine the specific carryover effect of driving skills of a comprehensive training program in a driving simulator when compared with a cognitive training program. Methods. Further analysis from a previous randomized controlled trial that investigated the effect of simulator training on driving after stroke. Forty-two participants received simulator-based driving training, whereas 41 participants received cognitive training for 15 hours. Overall performance in the on-road test and each of its 13 items were compared between groups immediately posttraining and at 6 months poststroke. Results. Generalized estimating equation analysis showed that the total score on the on-road test and each item score improved significantly over time for both groups. Those who received driving simulator training achieved better results when compared with the cognitive training group in the overall on-road score and the items of anticipation and perception of signs, visual behavior and communication, quality of traffic participation, and turning left. Most of the differences in improvement between the 2 interventions were observed at 6 months poststroke. Conclusions . Contextual training in a driving simulator appeared to be superior to cognitive training to treat impaired on-road driving skills after stroke. The effects were primarily seen in visuointegrative driving skills. Our results favor the implementation of driving simulator therapy in the conventional rehabilitation program of subacute stroke patients with mild deficits.