Electromyographic feedback was compared with physical therapy for its relative effectiveness in training motor activity in patients with hemiplegia. A cross over design was used, so that each patient served as his own control. Strong reinforcement for attentional direction and for successful response to established criteria was used with both modalities. Electromyography through auditory and visual display was found effective in improving electromyographic activity, but more limited in training effective function as measured by active range of motion. In some cases it was counterproductive. Using an instrumental modality permits attention to only a limited number of motor elements involved in a motion, and is even more limited with respect to the interplay of activation and inhibition of those components. The activity of physical therapy, using tactile, proprioceptive, visual and communicative modalities, has the capacity for attending and drawing attention to many motor elements, as well as to the complexities of activation of some components concomitantly with the inhibition of others during the training process. Electromyographic feedback was felt to be especially useful in its ability to enhance figure-background discrimination and bypass possible areas of sensory obtundation.