OBJECTIVE To investigate the effects of task-specific practice on hand function in stroke subjects who were given computer-assisted training and to look for associated changes in corticospinal connectivity. DESIGN Single case design experiments. SETTING Subjects' homes. SUBJECTS Seven chronic stroke subjects with impaired hand function. INTERVENTIONS Daily intensive practice of fast and accurate force changes in pinch grip and knob turning grip for four weeks. MAIN MEASURES Pinch and power grip strength and dexterity measured using the Ten-hole Peg Test and a timed nuts and bolts test were recorded twice a week throughout the study. Corticospinal connectivity was assessed by latency of electromyography (EMG) responses to transcranial magnetic stimulation (TMS) recorded during the baseline weeks before the training and again immediately after the training period. RESULTS Subjects completed between 5760 and 18560 force changes. Clinically relevant improvements in grip strength were evident in only two subjects. None of the subjects showed clinically significant improvement in dexterity measures. Statistical analysis showed that improved performance was related to the training phase in only two subjects. A reduction in latency of EMG responses to TMS was seen in two subjects but could not be attributed to the training phase. CONCLUSIONS The intensive training did not improve hand function in most subjects. Latencies of EMG responses to TMS were also unable to detect any change.