RÉSUMÉ

Understanding the human−machine interface has led to novel technologies that have provided new tools for therapists who practice the labor-intensive restorative treatments in neurological rehabilitation settings. These devices are safe and effective and are now providing the means to understanding some aspects of neurological recovery as a motor learning phenomenon. In this article, we review the results of two meta-analyses on the impact of rehabilitation robotics for the upper extremity in stroke recovery. Results demonstrate that motor performance can be improved in patients with subacute and chronic stroke, and that so-called motor plateaus may indicate something other than a biological optimum. We review some of our results with both subacute and chronic stroke patients. We then focus on the specificity of training and review some of our results on generalization (1) across different limb segments but same workspace, (2) same limb segments and workspace but untrained movements, and (3) same limb segments but different workspace.