The goal of this study was to determine whether transcranial magnetic stimulation (TMS) measures of the corticomotor pathways might reflect and predict functional improvements during rehabilitation of a hemiparetic upper extremity. Ten patients with variable levels of functional impairment in upper extremity use were enrolled at least 3 months after a stroke. TMS was used to obtain serial measures of motor thresholds and motor-evoked potential (MEP) size for a muscle from both hands before and after each session of an intervention aimed at improving functional motor control. Functional ability and cortical map area and volume were measured before each therapy session. At intake, all TMS measures from the affected side were impaired compared with the unaffected side but they did not reliably predict the level of functional improvement. Motor thresholds decreased, whereas MEP amplitude and map size increased with treatment. The amount of change in affected side measures was correlated with the amount of improvement in hand/arm function. Normalization of MEP amplitude asymmetry in response to the 1st 2 therapy sessions predicted long-term improvement in Fugl-Meyer motor score. Within limits, TMS appears to be useful in both moderate and more impaired patients as a physiological assay of treatment-induced plasticity and behavioral gains. Methodological differences in the literature, however, currently obscure a full understanding of the potential contributions of TMS to rehabilitation research.