Background and Purpose. Individuals with stroke can improve upper-extremity function in the chronic phase, providing rehabilitation is intensive, attended, and of sufficient duration. Virtual reality has been used in motor retraining; however, off-the-shelf game consoles may not be appropriate for those with marked motor impairment and high finger or arm spasticity. The objective of this study was to investigate the feasibility and effects of training in individuals with chronic stroke who are either high-functioning or low-functioning, and also spastic. Case Descriptions. Four volunteers, 3 men and 1 woman, were recruited from a local aphasia support group. All individuals were chronic post stroke with right-side hemiplegia. Training took place at the Tele-Rehabilitation Institute at Rutgers University. The intervention was performed on the Rutgers Arm II, a prototype training table that senses supported arm movement and grasp strength and tilts to resist or assist reach. Participants played games that adapted automatically to each individual's motor abilities. The games were practiced over 6 weeks, 3 sessions every week, with sessions lasting up to 1 hour. The 4 participants were evaluated by a senior physical therapist before, immediately following, and 6 weeks after the intervention. No occupational or physical therapist was present during the training sessions. Outcomes. The primary outcomes were changes in the affected upper-extremity subset of the Fugl-Meyer test and self-reported changes in the participants' activities of daily living. Improvements in active range of motion and grasp strength were secondary outcomes. All individuals improved in Fugl-Meyer scores and retained these gains (participant 1, 45 to 50; participant 2, 16 to 22; participant 3, 12 to 20; participant 4, 42 to 51). Participants 2 and 3, who presented with severe motor impairment, began using their affected arms in daily activities subsequent to training. All participants improved in their shoulder, elbow, and finger flexion active range of motion. Remarkably, participants 2 and 3, who were unable to exert force in grasping or pinching pre-training, could now do so and retained these gains at 6 weeks following the intervention. Well-being and mood seemed to improve in all participants. Discussion and Conclusion. Results show that motor retraining in virtual reality is feasible, well-tolerated by participants, and benefitting them. The Rutgers Arm II system was able to train participants, who varied greatly in their degree of motor impairment, but without a clinician being present. The present study contributes to the body of knowledge on novel virtual rehabilitation interventions for the upper extremity. Key Words: Upper extremity, Virtual reality, Sustained grasp, Rutgers Arm, Stroke. BACKGROUND AND PURPOSE The direct and indirect costs associated with head injuries from stroke or trauma approach $200 billion a year in the United States alone.1'2 At the completion of standard-of-care motor retraining (6 to 9 months post stroke), individuals present with varying levels of upper-extremity (UE) functional recovery. Their recovery depends not only on the type and severity of the initial injury to the central nervous system, but also on the intensity and length of rehabilitation subsequently provided. Diminished upper-extremity function in the chronic phase post stroke negatively affects the individual's degree of independence in activities of daily living (ADLs), mental health, and social life. The person's motivation to improve and the important role knowledge of results feedback plays in motor relearning3 both point to the need to investigate virtual reality as an element of therapy. Virtual worlds4 have been shown to provide rich knowledge of results and to increase motivation if tailored to the individual's abilities.5'6 Virtual reality previously has been investigated as a way to conduct motor rehabilitation following a stroke. …