Studies of motor imagery and motor learning have thus far been concerned only with its effects on healthy subjects. Therefore, in order to investigate the possible involvement of the basal ganglia, the effectiveness of motor imagery in the acquisition of motor constants in a graphomotor trajectorial learning task was examined in 11 non-demented mildly affected Huntington's disease (HD) patients and 12 non-demented Parkinson's disease (PD) patients. The patients received, after baseline, 10 min of motor imagery training, followed by a motor practice phase. Additionally, a test battery for visual imagery abilities was administered in order to investigate possible relations between visual and motor imagery. The results showed that imagery training alone enabled the HD patients to achieve a significant approach to movement isochrony, whereas the PD patients showed no marked improvements, either with motor imagery or with motor practice. Furthermore, the PD patients had more difficulties than the HD patients in solving the visual imagery tasks. Subsequent correlational analysis revealed significant relationships between the degree of caudate atrophy in the HD patients and their performance in the visual imagery tasks. However, there were no substantial correlations between the performance on the visual imagery tasks and the improvement of motor performance through motor imagery, which indicates that visual and motor imagery are independent processes. It is suggested that the dopaminergic input to the basal ganglia plays an important role in the translation of motor representations into motor performance, whereas the caudate nucleus atrophy of the HD patients does not seem to affect motor imagery, but only the visual imagery process. Furthermore, the deficits found in PD patients might also be related to their limited attentional resources and difficulties in employing predictive motor strategies.