Spinal cord evoked potentials (SCEPs) after spinal cord stimulation were used as a method of spinal cord monitoring during surgery of 6 extramedullary and 14 intramedullary spinal cord tumors. SCEPs were recorded from an epidural electrode placed rostral to the level of the tumor. Electrical stimulation was applied on the dorsal spinal cord from a caudally placed epidural electrode. The wave forms of SCEPs consisted of a sharp negative peak (N1) in 15 cases and two negative peaks (N1 and N2) in 5 cases. The N2 wave was markedly attenuated by posterior midline myelotomy, whereas the N1 activity showed less-remarkable changes by myelotomy. An increase in N1 amplitude was observed after the removal of the tumor in four extramedullary and three intramedullary cases. Of six patients that showed decreased N1 amplitude after the removal of the tumor, five patients developed postoperative motor deficits. However, there were four false-negative cases and one false-positive case in regard to changes of N1 amplitude and postoperative motor deficits. Four false results occurred in intramedullary cases. In two of them, postoperative symptoms indicated intraoperative unilateral damage to the spinal cord. The position of the stimulating electrode, the difference in thresholds of the axons for electrical stimulation between the right and left side of the spinal cord, or the change of the distance between the electrode and the spinal cord surface may account for these false results. Thus, our analysis of the changes of SCEP wave forms and early postoperative symptoms indicates that the sensitivity of this monitoring method to detect intraoperative insults to the spinal cord is unsatisfactory in spite of the reproducible wave forms. We conclude that SCEP monitoring can be used as an alternative method or in combination with other types of evoked potentials in patients with severe spinal cord lesions who show abnormal somatosensory evoked potentials preoperatively.