Hemiparesis resulting in functional limitation of an upper extremity is common among stroke survivors. Although existing evidence suggests that increasing intensity of stroke rehabilitation therapy results in better motor recovery, limited evidence is available on the efficacy of virtual reality for stroke rehabilitation. In this pilot, randomized, single-blinded clinical trial with 2 parallel groups involving stroke patients within 2 months, we compared the feasibility, safety, and efficacy of virtual reality using the Nintendo Wii gaming system (VRWii) versus recreational therapy (playing cards, bingo, or "Jenga") among those receiving standard rehabilitation to evaluate arm motor improvement. The primary feasibility outcome was the total time receiving the intervention. The primary safety outcome was the proportion of patients experiencing intervention-related adverse events during the study period. Efficacy, a secondary outcome measure, was evaluated with the Wolf Motor Function Test, Box and Block Test, and Stroke Impact Scale at 4 weeks after intervention. Overall, 22 of 110 (20%) of screened patients were randomized. The mean age (range) was 61.3 (41 to 83) years. Two participants dropped out after a training session. The interventions were successfully delivered in 9 of 10 participants in the VRWii and 8 of 10 in the recreational therapy arm. The mean total session time was 388 minutes in the recreational therapy group compared with 364 minutes in the VRWii group (P=0.75). There were no serious adverse events in any group. Relative to the recreational therapy group, participants in the VRWii arm had a significant improvement in mean motor function of 7 seconds (Wolf Motor Function Test, 7.4 seconds; 95% CI, -14.5, -0.2) after adjustment for age, baseline functional status (Wolf Motor Function Test), and stroke severity. VRWii gaming technology represents a safe, feasible, and potentially effective alternative to facilitate rehabilitation therapy and promote motor recovery after stroke.