This paper describes the preliminary results of an ongoing study of the effects of two training approaches on motor function and learning in persons with hemi paresis due to cerebrovascular accidents. Eighteen subjects with chronic stroke performed eight, three-hour sessions of sensorimotor training in haptically renedered environments. Eleven subjects performed training activities that integrated hand and arm movement while another seven subjects performed activities that trained the hand and arm with separately. As a whole, the eighteen subjects made statistically significant improvements in motor function as evidenced by robust improvements in Wolf Motor Function Test times and corresponding improvements in Jebsen Test of Hand Function times. There were no significant between group effects for these tests. However, the two training approaches elicited different patterns and magnitudes of performance improvement that suggest that they may elicit different types of change in motor learning and or control.