In motor skill learning, the greater the dose of training, the greater the efficacy of training, the lower the efficiency of training, and the better the long-term retention.Whether such principles apply to motor training after stroke is unclear. Here, we developed novel mixed-effects models of the change in the quality of arm movements during and following training fitted to data from a recent randomized controlled trial of the effect of the dose of training in chronic stroke. Analysis of the model's learning and retention rates demonstrated an increase in efficacy of training with greater doses, a decrease in efficiency of training with both additional doses and additional bouts of training, and fast initial decay following training. Two additional effects modulated retention: a positive "self-training" effect, and an un-expected negative effect of dose.Our results suggest that for patients with sufficient arm use post-training, self-training will further improve use, but additional therapy may be in vain.