Somatosensory evoked cortical potentials (SSEP's) were recorded in 27 healthy subjects using tibial and peroneal nerve stimulation with cephalic and non-cephalic references. Four major peaks were present in all recordings. Analysis of these components showed that SSEP's collected after tibial nerve stimulation with non-cephalic reference (linked earlobes) produced the most consistent clearly defined component peaks. Average latency, amplitude, and interpeak latency differences are presented for these SSEP's. Significant correlations were obtained between the height of the individual and the P1, N2, P2, and N3 latencies, and the N3-P1 interpeak latency. These results suggest that reproducible SSEP's can be obtained from tibial nerve stimulation in normal subjects using minimal numbers of stimulus presentations (28 to 64). The SSEP's from 34 patients with varying degrees of spinal cord trauma were compared with the SSEP's from normal subjects. These comparisons involved the P1, N2, P2, and N3 latencies and the interpeak latency values, as well as the amplitude values. Patients with normal sensory and motor neurological examinations could be distinguished from patients showing decreased sensory and motor findings or clinically complete lesions on the basis of peak latency and interpeak latency values. The latter two groups could not be distinguished from one another. In general, all patient groups had SSEP's of lower amplitude than did normal individuals, but the groups could not be distinguished from one another. These results indicate that SSEP's can be a useful clinical tool for differentiation of complete from incomplete spinal cord lesions, but do not invariably predict recovery of function.