Auditory (AER) and somatosensory evoked responses (SSER) were recorded simultaneously in eight patients under anaesthesia before surgery. We studied the effects of equi-MAC end-expiratory concentrations of isoflurane (0.65-0.75%) and nitrous oxide (60-65%). The anaesthetics were changed at random in three consecutive 10-min periods so that each patient received both drugs. From the AER recorded from the vertex and inion, Pa and Nb latency and amplitude were measured. N13, P20 latency and N13 amplitude were measured from SSER recordings from the neck and P15, N20, P25, N35, P45 latency and P15-N20, N20-P25, P25-N35 and N35-P45 amplitude from the scalp over the hand area of the sensory cortex. Compared with nitrous oxide, isoflurane significantly increased the latencies of the AER waves Pa (P = 0.02) and Nb (P = 0.02), and the SSER waves N20 (P = 0.001) and P25 (P = 0.04). We were unable to demonstrate significant differences in Pa and Nb amplitude between isoflurane and nitrous oxide that we had seen previously. However, the amplitude of the SSER wave N20 was reduced significantly by nitrous oxide compared with isoflurane (P = 0.0004). This wave (N20) is thought to emanate from the thalamo-cortical radiations, and our findings may be explained by an analgesic effect of nitrous oxide mediated by endogenous opioids.