Background and Purpose-The increasing availability of robot-assisted therapy (RT), which provides quantifiable, reproducible, interactive, and intensive practice, holds promise for stroke rehabilitation, but data on its dose-response relation are scanty. This study used 2 different intensities of RT to examine the treatment effects of RT and the effect on outcomes of the severity of initial motor deficits. Methods-Fifty-four patients with stroke were randomized to a 4-week intervention of higher-intensity RT, lower-intensity RT, or control treatment. The primary outcome, the Fugl-Meyer Assessment, was administered at baseline, midterm, and posttreatment. Secondary outcomes included the Medical Research Council scale, the Motor Activity Log, and the physical domains of the Stroke Impact Scale. Results-The higher-intensity RT group showed significantly greater improvements on the Fugl-Meyer Assessment than the lower-intensity RT and control treatment groups at midterm (Pϭ0.003 and Pϭ0.02) and at posttreatment (Pϭ0.04 and Pϭ0.02). Within-group gains on the secondary outcomes were significant, but the differences among the 3 groups did not reach significance. Recovery rates of the higher-intensity RT group were higher than those of the lower-intensity RT group, particularly on the Fugl-Meyer Assessment. Scatterplots with curve fitting showed that patients with moderate motor deficits gained more improvements than those with severe or mild deficits after the higherintensity RT. Conclusions-This study demonstrated the higher treatment intensity provided by RT was associated with better motor outcome for patients with stroke, which may shape further stroke rehabilitation. Clinical Trial Registration-URL: http://clinicaltrials.gov. Unique identifier: NCT00917605.