Five aphasics and five controls were compared in their response to three perceptual discrimination programs presented in automated fashion. The programs, composed of shapes based upon those necessary to form English capital letters, were concerned with the variables of shape discrimination, orientation of form, and transition of solid shape to outline figure. Aphasics and controls differed significantly in response latencies and error rates to sets of pre-test items representing each program. Aphasics were given automated training with those programs on whose pre-test they had an error rate greater than 10%. On follow-up testing one week after training, response latency decreased and differed significantly from pre-test latency, and the error rate became comparable to that of the normal controls.