The aim of this study was to reassess the efficacy of flumazenil for reversal of sedation with midazolam. Twenty-four ASA I or II patients undergoing elective surgery under epidural anaesthesia participated. Following epidural block, midazolam was administered to keep the patient sleepy but still responsive to verbal commands. At the end of surgery the patients were randomly allocated to receive, in a double-blind manner, either flumazenil (0.1 mg.ml-1) or placebo. The study drug (maximum dose: 10 ml) was titrated until the patient became fully awake. Sedation was assessed with the Modified Steward Coma Scale (MSCS), the Trieger test (TT) and Critical Flicker Frequency (CFF). The assessments were done before anaesthesia (baseline), at the end of surgery immediately before administration of study drug, and serially afterwards, at 10, 30, 60, 90, 120, 150 and 180 min. Analyses of variance for repeated measures and pooled t tests were used. The duration of surgery was (mean +/- SD) 0.72 +/- 0.25 hr in the flumazenil group and 0.74 +/- 0.28 hr in the placebo group. The total dose of midazolam was 7.2 +/- 2.2 mg for the flumazenil group and 8.9 +/- 2.7 mg for the placebo group. The volume of study drug administered was 5.5 ml +/- 1.9, equivalent to 0.55 mg, for the flumazenil group and 6.7 +/- 2.2 ml for the placebo group. Critical Flicker Frequency is the only measure which revealed a difference (P < 0.005) between the flumazenil and placebo groups and this occurred only at the ten-minute assessment.(ABSTRACT TRUNCATED AT 250 WORDS)