Additional therapy may improve poststroke outcomes. Self-rehabilitation is a useful means to increase rehabilitation time. Mechanized systems are usual means to extend time for motor training. The primary aim was to compare the effects of self-rehabilitation using a mechanized device with control self-exercises on upper extremity impairment in patients with stroke. Phase III, parallel, concealed allocation, randomized controlled, multicenter trial, with 12-month follow-up. Patients aged 18 to 80 years, 3 weeks to 3 months poststroke with a Fugl-Meyer Assessment score of 10 to 40 points, were randomized to the Exo or control groups. All undertook two 30-minute self-rehabilitation sessions/day, 5 days/wk for 4 weeks in addition to usual rehabilitation. The Exo group performed games-based exercises using a gravity-supported mechanical exoskeleton (Armeo Spring). The control group performed stretching plus basic active exercises. Primary outcome was change in upper extremity Fugl-Meyer Assessment score at 4 weeks. Two hundred fifteen participants were randomly allocated to the Exo group (107) or the control group (108). Mean age (SD), 58.3 (13.6) years; mean time poststroke, 54.8 (22.1) days; and mean baseline Fugl-Meyer Assessment score, 26.1 (9.5). There was no between-group difference in mean change in Fugl-Meyer Assessment score following the intervention: 13.3 (9.0) in the Exo group and 11.8 (8.8) in the control group (P=0.22). There were no significant between-group differences in changes for any of the other outcomes at any time point (except for perception of the self-rehabilitation). There was no between-group difference in cost utility at 12 months. In patients with moderate-to-severe impairment in the subacute phase of stroke, the purchase and use of complex devices to provide additional upper limb training may not be necessary: simply educating patients to regularly move and stretch their limbs appears sufficient. URL: https://www.clinicaltrials.gov; Unique identifier: NCT01383512.