Cortical somatosensory evoked potentials (CSEP) allow monitoring of spinal cord function during surgery. Ketamine has been shown to enhance CSEP amplitude, but there is no previous study comparing its effects with those of other anaesthetic regimens. Therefore, we have compared the effects of ketamine with those of fentanyl, both combined with midazolam, on CSEP monitoring during major spine surgery. Twenty patients with normal preoperative CSEP were allocated randomly to a ketamine or fentanyl group. Anaesthesia was induced with ketamine 3 mg kg-1 or fentanyl 6 micrograms kg-1 i.v., and midazolam 0.3 mg kg-1 i.v in both groups, and maintained with continuous i.v infusion of ketamine 2 mg kg-1 h-1 or fentanyl 3 micrograms kg-1 h-1, combined in both groups with midazolam 0.15 mg kg-1 h-1 and 60% nitrous oxide in oxygen. CSEP were elicited by tibial posterior nerve stimulation and measured P1 and N1 latencies, and P1-N1 amplitude, CSEP were recorded before and after induction, at 15 min, 1 and 2 h after induction, during skin closure and after removal of nitrous oxide. Both groups were comparable in characteristics, duration of surgery, mean arterial pressure and temperature. CSEP latencies were not significantly affected in either group. CSEP amplitude decreased significantly over time in the fentanyl group (from mean 2.02 (SEM 0.41) to 0.95 (0.17) microV, P < 0.05), but not in the ketamine group (from 1.33 (0.36) to 1.05 (0.31) microV, ns). Nevertheless, we did not observe any significant differences in amplitudes or latencies between the two groups. The delay in obtaining the first voluntary postoperative motor response was significantly greater in the ketamine group (170 (54) vs 55 (17) min, P < 0.01). Both ketamine and fentanyl allowed us to obtain reliable CSEP during major spine surgery, and there were no significant difference between these two anaesthetic regimens for CSEP monitoring, but a longer delay for voluntary postoperative motor assessment was observed in the ketamine group.