Objective: To assess the feasibility of a trial to investigate the effectiveness of virtual reality-mediated therapy compared to conventional physiotherapy in the motor rehabilitation of the arm following stroke, and to provide data for a power analysis to determine numbers for a future main trial. Design: Pilot randomized controlled trial. Setting: Clinical research facility. Participants: Eighteen people with a first stroke, 10 males and 8 females, 7 right and 2 left side most affected. Mean time since stroke 10.8 months. Interventions: Participants were randomized to a virtual reality group or a conventional arm therapy group for nine sessions over three weeks. Main measures: The upper limb Motricity Index and the Action Research Arm Test were completed at baseline, post intervention and six weeks follow-up. Results: Outcome data were obtained from 95% of participants at the end of treatment and at follow-up: one participant withdrew. Compliance was high; only two people reported side-effects from virtual reality exposure. Both groups demonstrated small (78 points on upper limb Motricity Index and 4 points on the Action Research Arm Test), but non-significant, changes to their arm impairment and activity levels. Conclusion: A randomized controlled trial of virtual reality-mediated therapy comparable to conventional therapy would be feasible, with some suggested improvements in recruitment and outcome measures. Seventy-eight participants (39 per group) would be required for a main trial.