Previous studies have shown that the motor training in a virtual-environment with the augmented feedback promotes motor learning in normal subjects and in long-term post-stroke patients. We evaluated whether this approach could be useful also for treating patients with arm motor deficits due to a recent stroke. Thirty-eight patients were included in the study within 3 months from an ischemic stroke in the territory of the middle cerebral artery. Twenty-five subjects received training with the Reinforced Feedback in Virtual Environment (RFVE) therapy for the arm, and thirteen patients received an equal amount of a conventional rehabilitation (CR) therapy focused to the upper limb. Before and after therapy, the autonomy of daily living activities were assessed with the Functional Independence Measure (FIM) and the degree of motor impairment was measured with the Fugl-Meyer scale for the upper extremity (FM-UE). The RFVE therapy group showed significant improvements in the FM-UE and the FIM scale mean scores. The conventional therapy determined smaller and not statistically significant scores improvements. These data indicate that the recovery of arm motor function in patients after a recent stroke appear to be speeded up by an augmented feedback provided in a virtual-environment.