Background: Recovering useful hand function after stroke is a major scientific challenge for patients with limited motor recovery. We hypothesized that sequential training beginning with proximal bilateral followed by unilateral task oriented training is superior to time-matched unilateral training alone. Proximal bilateral training could optimally prepare the motor system to respond to the more challenging task-oriented training. Methods: Participants: Twenty-six participants with moderate severity hemiparesis Intervention: Participants received either 6-weeks of bilateral proximal training followed seq uentially by 6-weeks unilater al task-oriented training (COMBO) or 12-weeks of unilateral task-oriented tr aining alone (SAEBO). A subset of 8 COMB0 and 9 SAEBO participants underwent three functional magnetic resonance imaging (fMRI) scans of hand and elbow movement every 6 weeks. Main Outcome Measures: Fugl-Meyer Upper extremity scale, Modified Wolf Motor Function Test, University of Maryland Arm Questionnaire for Stroke, Motor cortex activation (fMRI). Results: The COMBO group demonstrated significantly grea ter gains between baseline and 12-weeks over all outcome measures (p = .018 based on a MANOVA test) and specifically in the Modified Wolf Motor Function test (time). Both groups demonstrated within-group gains on the Fugl-Meyer Upper Extremity test (impairment) and University of Maryland Arm Questionnaire for Stroke (functional use). fMRI subset analyses showed motor cortex (primary and premotor) activation during hand movement was significantly increased by sequential combination training but not by task-oriented training alone. Conclusions: Sequentially combining a proximal bilateral before a unilateral task-oriented training may be an effective way to facilitate gains in arm and hand function in those with moderate to severe paresis post-stroke compared to unilateral task oriented training alone.