Purpose: To evaluate the efficacy of an expanded form of Constraint-Induced Movement Therapy (eCIMT) that renders CIMT, originally designed for treating mild-to-moderate upper-extremity hemiparesis, suitable for treating severe hemiparesis. Methods: Twenty-one adults ≥1 year after stroke with severe upper-extremity hemiparesis (with little or no capacity to make movements with the more-affected hand) were randomly assigned to eCIMT (n = 10), a placebo-control procedure (n = 4), or usual care (n = 7). The participants who received usual care were crossed over to eCIMT four months after enrollment. The CIMT protocol was altered to include fitting of orthotics and adaptive equipment, selected neurodevelopmental techniques, and electromyography-triggered functional electrical stimulation. Treatment was given for 15 consecutive weekdays with 6 hours of therapy scheduled daily for the immediate eCIMT group and 3.5 hours daily for the cross-over eCIMT group. Results: At post-treatment, the immediate eCIMT group showed significant gains relative to the combination of the control groups on the Grade-4/5 Motor Activity Log (MAL; mean = 1.5 points, P < 0.001, f = 4.2) and a convergent measure, the Canadian Occupational Performance Measure (COPM; mean = 2.3, P = 0.014, f = 1.1; f values ≥0.4 are considered large, on the COPM changes ≥2 are considered clinically meaningful). At 1-year follow-up, the MAL gains in the immediate eCIMT group were only 13% less than at post-treatment. The short and long-term outcomes of the crossover eCIMT group were similar to those of the immediate eCIMT group. Conclusions: This small, randomized controlled trial (RCT) suggests that eCIMT produces a large, meaningful, and persistent improvement in everyday use of the more-affected arm in adults with severe upper-extremity hemiparesis long after stroke. These promising findings warrant confirmation by a large RCT.