The anesthetic effects and interactions of midazolam and fentanyl were determined in terms of their reduction of enflurane MAC in dogs, and the effects of their specific antagonists were also investigated. Control enflurane MAC was determined by the tail clamp method in 18 mongrel dogs. Each animal then received an iv loading dose of midazolam followed by a constant infusion at 9.6 .mu.g .cntdot. kg-1 .cntdot. min-1 designed to produce a stable enflurane MAC reduction of approximately 40%, and enflurane MAC was determined following a 60-min observation period during which time the midazolam concentration in plasma stabilized. Fentanyl was then administered in a series of three incremental loading doses (15, 30, and 225 .mu.g/kg) and infusions (0.05, 0.2 and 3.2 .mu.g .cntdot. kg-1 .cntdot. min-1) designed to produce enflurane MAC reductions of 30%, 50% and 65%, respectively. Enflurane MAC was again determined following a 60-min observation period for each new infusion. In nine dogs after the fourth determination of enflurane MAC, fentanyl was discontinued and 1 mg/kg naloxone was administered iv every 10 min until enflurane MAC was determined for the last time. In the other nine dogs, midazolam was discontinued and 1.5 mg/kg flumazenil (RO 15-1788) was administered and enflurane MAC determined for the last time. The midazolam concentration in plasma remained stable at 414 .+-. 134 ng/ml throughout the study, and in the absence of fentanyl reduced enflurane MAC by 40 .+-. 10% (mean .+-. SD). The addition of fentanyl produced significant further reductions in enflurane MAC. The additional enflurane MAC reduction attributable to fentanyl was slightly less but not significantly different from that predicted from the fentanyl concentration in plasma except for the highest infusion rate. After administration of naloxone, the degree of enflurane MAC reduction returned to that produced initially by midazolam alone. In contrast, the degree of enflurane MAC reduction after administration of flumazenil and in the presence of a continuous infusion of fentanyl was significantly less than that expected. Lower plasma concentrations of fentanyl interacted additively with midazolam. The development of acute tolerance to fentanyl may be an explanation for the less than additive interaction between midazolam and the highest fentanyl concentrations. This explanation is supported by the lower than expected MAC reduction attributable to the highest infusion of fentanyl after antagonism of midazolam by flumazenil.