Nineteen patients with radiologically confirmed stroke, and varying degrees of hemiparesis, were studied using somatosensory evoked potentials and the recently developed technique of transcutaneous motor cortex stimulation. The functional deficit caused by stroke was assessed at the time of evoked potential testing and again on follow-up 2 months after stroke. Stroke location and degree of recovery were compared with the evoked potentials elicited an average of 8 days after the acute event. The motor response was better (p less than 0.01) than somatosensory response at predicting an outcome in terms of functional recovery, both overall and when comparing patients with similar degrees of weakness. Normal somatosensory evoked potentials also predicted recovery but were not as sensitive as the motor evoked potentials (0.01 less than p less than 0.05). Motor evoked potentials in conjunction with clinical assessment may offer a means of more accurately predicting functional outcome following stroke than can be achieved with clinical assessment, with or without somatosensory evoked potentials. The additional finding that cortically evoked motor evoked potentials were delayed only by subcortical lesions is of interest and may extend the understanding of the pathophysiological mechanism of this response in humans.