We studied 30 controls and 28 patients with multiple sclerosis (MS). M-waves, F-waves, and somatosensory responses evoked by median and peroneal nerve stimulation were used to estimate conduction times from the lumbar cord to the cortex; the cervical cord to the cortex; through the spinal cord; and in proximal segments of peripheral nerve. The central conduction times were significantly prolonged in the MS patients, giving an overall 79 percent incidence of abnormalities. Several patients who were asymptomatic in the tested arm (36 percent) or leg (16 percent) showed electrophysiologic abnormalities. By comparing conduction time from the lumbar cord to the cortex with the conduction time from the cervical cord to the cortex, the site of the lesion could be deduced, but not proven. The diagnostic yield from measuring the latencies of the somatosensory responses was 68 percent. We concluded that measuring central conduction times indirectly by use of M- and F-waves adds little to simple measurement of somatosensory responses.