Peformance in a prelearned choice reaction time task was studied 6 months after surgery in rats with ibotenate-induced lesions of the striatum either with or without striatal grafts, and in sham-operated controls. The long postoperative interval allowed full transplant maturation and the establishment of appropriate connections by the transplants. The animals were trained prior to surgery on a visual choice reaction time task which requires that a movement is made away from stimuli signalling reward. The use of multiple measures allowed a thorough analysis of several aspects of the animals' performance. Whilst sham-operated control animals recovered normal (preoperative) performance rapidly, the lesioned animals had a severe performance deficit. Although the transplanted animals were initially at least as deficient in performance as the lesioned group, repeated testing led to an amelioration of the lesion-induced deficit according to two distinct measures of spatial bias and reaction time. On a third measure, latency to complete the lateralized movement, the grafted group were initially worse than the lesioned group but repeated testing resulted in significant recovery. These results suggest that postoperative training may help to optimize the efficacy of graft-induced recovery, and that animals may need to learn to use a transplant in order for it to confer functional benefit in complex prelearned tasks.