The traditional means of localizing sensorimotor cortex during surgery is Penfield's procedure of mapping sensory and motor responses elicited by electrical stimulation of the cortical surface. This procedure can accurately localize sensorimotor cortex but is time-consuming and best carried out in awake, cooperative patients. An alternative localization procedure is presented that involves cortical surface recordings of somatosensory evoked potentials (SEP's), providing accurate and rapid localization in patients under either local or general anesthesia. The morphology and amplitude of median nerve SEP's recorded from the cortical surface varied systematically as a function of spatial location relative to the sensorimotor hand representation area. These results were validated in 18 patients operated on under local anesthesia in whom the sensorimotor cortex was independently localized by electrical stimulation mapping; the two procedures were in agreement in all cases. Similar SEP results were demonstrated in an additional 27 patients operated on under general anesthesia without electrical stimulation mapping. The following three spatial relationships between SEP's and the anatomy of the sensorimotor cortex permit rapid and accurate localization of the sensorimotor hand area: 1) SEP's with approximately mirror-image waveforms are recorded at electrode sites in the hand area on opposite sides of the central sulcus (P20-N30 precentrally and N20-P30 postcentrally); 2) the P25-N35 is recorded from the postcentral gyrus as well as a small region of the precentral gyrus in the immediate vicinity of the central sulcus: this waveform is largest on the postcentral gyrus about 1 cm medial to the focus of the 20- and 30-msec potentials; and 3) regardless of component identification, maximum SEP amplitudes are recorded from the hand representation area on the precentral and postcentral gyri.