Single-site studies suggest that a 2-week program of constraint-induced movement therapy (CIMT) for patients more than 1 year after stroke who maintain some hand and wrist movement can improve upper extremity function that persists for at least 1 year. To compare the effects of a 2-week multisite program of CIMT vs usual and customary care on improvement in upper extremity function among patients who had a first stroke within the previous 3 to 9 months. The Extremity Constraint Induced Therapy Evaluation (EXCITE) trial, a prospective, single-blind, randomized, multisite clinical trial conducted at 7 US academic institutions between January 2001 and January 2003. Two hundred twenty-two individuals with predominantly ischemic stroke. Participants were assigned to receive either CIMT (n = 106; wearing a restraining mitt on the less-affected hand while engaging in repetitive task practice and behavioral shaping with the hemiplegic hand) or usual and customary care (n = 116; ranging from no treatment after concluding formal rehabilitation to pharmacologic or physiotherapeutic interventions); patients were stratified by sex, prestroke dominant side, side of stroke, and level of paretic arm function. The Wolf Motor Function Test (WMFT), a measure of laboratory time and strength-based ability and quality of movement (functional ability), and the Motor Activity Log (MAL), a measure of how well and how often 30 common daily activities are performed. From baseline to 12 months, the CIMT group showed greater improvements than the control group in both the WMFT Performance Time (decrease in mean time from 19.3 seconds to 9.3 seconds [52% reduction] vs from 24.0 seconds to 17.7 seconds [26% reduction]; between-group difference, 34% [95% confidence interval {CI}, 12%-51%]; P<.001) and in the MAL Amount of Use (on a 0-5 scale, increase from 1.21 to 2.13 vs from 1.15 to 1.65; between-group difference, 0.43 [95% CI, 0.05-0.80]; P<.001) and MAL Quality of Movement (on a 0-5 scale, increase from 1.26 to 2.23 vs 1.18 to 1.66; between-group difference, 0.48 [95% CI, 0.13-0.84]; P<.001). The CIMT group achieved a decrease of 19.5 in self-perceived hand function difficulty (Stroke Impact Scale hand domain) vs a decrease of 10.1 for the control group (between-group difference, 9.42 [95% CI, 0.27-18.57]; P=.05). Among patients who had a stroke within the previous 3 to 9 months, CIMT produced statistically significant and clinically relevant improvements in arm motor function that persisted for at least 1 year. Trial Registration clinicaltrials.gov Identifier: NCT00057018.