To evaluate the clinical utility of somatosensory evoked potentials elicited by stimulation of the paraspinal region (PS-SEPs), waveform characteristics and scalp topography were studied in 23 normal controls. PS-SEPs were recorded in 22 patients with spinal lesions, and the relation between the abnormal PS-SEPs and clinico-radiological findings was investigated. 1) The normal control study showed clear waveforms on the scalp elicited by stimulation of the paraspinal region from C7 to the L2 spinous process level, with both bilateral and unilateral stimulation. The latency of PS-SEPs was gradually decreased when stimulation was moved in the caudo-rostral direction. 2) The normal range of spinal conduction time and conduction velocity was wide. The conduction velocity resulting from stimulation of the lumbar region tended to be slower than that from stimulation of the thoracic region, perhaps due to the difference in the length of the peripheral cutaneous nerves. 3) There was no correlation between the height and the latency of PS-SEPs. 4) Amplitudes of PS-SEPs after unilateral stimulation of the C7 and Th5 spinous levels were larger on the scalp contralateral to the side of stimulation. These findings were similar to SEPs elicited by median nerve stimulation. There was no significant amplitude laterality of PS-SEPs when the Th10 and L2 spinous levels were stimulated. Posterior tibial nerve SEPs showed a larger amplitude at the scalp ipsilateral to the side of stimulation than at the contralateral side. All these findings suggest that the cortical generator sites of PS-SEPs elicited by C7 and Th5 stimulation are located near the hand area, and those from Th10 and L2 stimulation are between the hand and the foot area. 5) Unilateral stimulation disclosed a clear laterality of sensory disturbance that was obscure when only bilateral stimulation was employed. PS-SEPs showed a high degree of abnormality when the patient had deep sensory disturbance. Generally, abnormal PS-SEPs were found caudal to the clinical sensory level, and a few cases showed abnormal PS-SEPs rostral to the sensory level. The latter might indicate that PS-SEPs detected subclinical sensory disturbance. It was therefore concluded that PS-SEPs are a useful tool for the objective evaluation of sensory disturbance, especially in cases of thoracic lesion, because conventional SEP studies, utilizing non-cephalic references do not provide clear identification of abnormal sensory levels along the thoracic spinal cord.