Background. Mirror therapy is a priming technique to improve motor function of the affected arm after stroke. Objective. To investigate whether a mirror therapy-based action observation (AO) protocol contributes to motor learning of the affected arm after stroke. Methods. A total of 37 participants in the chronic stage after stroke were randomly allocated to the AO or control observation (CO) group. Participants were instructed to perform an upper-arm reaching task as fast and as fluently as possible. All participants trained the upper-arm reaching task with their affected arm alternated with either AO or CO. Participants in the AO group observed mirrored video tapes of reaching movements performed by their unaffected arm, whereas participants in the CO group observed static photographs of landscapes. The experimental condition effect was investigated by evaluating the primary outcome measure: movement time (in seconds) of the reaching movement, measured by accelerometry. Results. Movement time decreased significantly in both groups: 18.3% in the AO and 9.1% in the CO group. Decrease in movement time was significantly more in the AO compared with the CO group (mean difference = 0.14 s; 95% confidence interval = 0.02, 0.26; P = .026). Conclusion. The present study showed that a mirror therapy-based AO protocol contributes to motor learning after stroke.