In previous work, we developed a lightweight wearable hand exoskeleton (HandSOME) that improves range of motion and function in laboratory testing. In this pilot study, we added the ability to log movement data for extended periods and recruited 10 chronic stroke subjects to use the device during reach and grasp task practice at home for 1.5 hours/day, 5 days per week, for 4 weeks. Seven subjects completed the study, performing 448±651 hand movements per training day. After training, impairment was reduced (Fugl-Meyer Test; gain=4.9±4.1; p=.039) and function was improved (Action Research Arm Test; gain=3.3±2.6; p=.032). There was a significant correlation between gains in the Action Research Arm Test and the number of movements during training (r=0.90; p=.005). Proximal arm control also improved, as evidenced by a significant reduction in the reach path ratio (p=0.038). Five subjects responded well to the treatment, having gains of 6 points or more on the Fugl-Meyer or Action Research Arm Test, and achieving significant gains in digit extension (gain=19.8±10.2 degrees; p=0.024). However, all of the gains that were significant immediately after training were no longer significant at the 3 month follow-up. This treatment approach appears promising, but longer periods of home training may be needed to achieve sustainable gains.