On the basis of the results of the previous experiments, the evoked spinal cord potentials may prove useful in evaluating clinical aspects of human spinal cord function. The purpose of this paper is to interpret and compare wave form characteristics which were recorded from electrodes placed on the cervical lamina of patients with cervical cord lesion. The investigation were performed on 53 subjects including 7 normal volunteers. The evoked spinal cord potentials to median or ulnar nerve stimulation were recorded from electrodes placed on the cervical lamina. Routinely, 100 responses were summated by a averaging computer. The normal response consisted of initially positive triphasic potentials (P1, N1,P2). The latency of P1 was 8.3 +/- 0.6 msec. In 13 patients with cervical spondylotic radiculopathy, responses were consistently identified as P1, N1, P2 in all cases although they decreased in amplitude, increased in duration and latency. The changes of response did not always concern with the severity of clinical symptoms. On the other hand, in 23 patients with cervical spondylotic myelopathy, it was possible to differentiate 4 types of potentials. They were remnant of P1, N1, P2(3 cases), decrease of P2 (2 cases), decrease of N1 (9 cases) and decrease or disappearance of N1, P2 (9 cases). In 9 cases showing decrease or disappearance of N1, P2 severe myelopathy was found clinically. These results suggest that the evoked spinal cord potentials may be helpful to assess the severity of cervical spondylotic myelopathy. However, wave form characteristics should be carefully analyzed because responses do not accurately reflect the extent of lesion. Recordings were also performed on 7 patients with cervical cord injury. In 1 case clinically judged to be complete transverse cord injury, responses in leads at the level of the lesion yielded a positive deflection followed by the first spike wave (P1). This would be expected with a killed end potential which arises in the level of cord lesion. However this potential could not be always detected in patients with complete cord injury but found in experimental animals. It is hoped that further refinement of these techniques may lead to methods that will be of value clinically for assessing the severity of a lesion and the prognosis for recovery.