Postoperative performance following total intravenous anaesthesia (TIVA) using midazolam and alfentanil was studied with and without the administration of a single dose of a benzodiazepine antagonist, flumazenil (Ro 15-1788). Performance was compared with a reference group anaesthetized with thiopentone, alfentanil and nitrous oxide. All patients were assessed by use of a rating scale which took into account the degree of sedation, amnesia, comprehension and cooperation as well as temporal and spatial orientation. There was a slow recovery following TIVA with somnolence and amnesia lasting several hours. Administration of flumazenil 1.0 mg i.v. at extubation caused a significant reduction of sedation (P less than 0.001) during the first postoperative hour, with patients fully awake or only lightly sedated, but was later followed by resedation. The patients of the reference group were moderately sedated during the observation period. Five and six hours postoperatively there was no difference between the groups. Amnesia was more profound in the groups that received midazolam; the effect of the antagonist could only be seen for 15 min after its administration. Comprehension and cooperation, as well as orientation, were equally good in the antagonist and in the reference group during the immediate postoperative period, whereas in the TIVA group a gradual improvement over the first hours was seen. In the antagonist group there was no increase in the number of analgesic requirements, no anxiety attacks or other adverse effects. It is concluded that flumazenil offers an improvement in postoperative performance following TIVA induced by midazolam and alfentanil, but the effects are of short duration.