Fifty patients who underwent diagnostic upper gastrointestinal endoscopy after midazolam sedation were randomized to receive (after completion of the examination) either the benzodiazepine receptor antagonist flumazenil or an identical-looking placebo. The speed of recovery from sedation was assessed by reaction time testing, measurement of critical flicker fusion frequency, and the semi-quantitative SOCA scoring system. Measurements were made up to 6 h post examination in all subjects, and at 12 and 24 h in all in-patients (n = 20). Flumazenil-treated patients were significantly more alert than those who received placebo at 10 min, 30 min, 1 h and 2 h (P less than 0.001 in all instances). Thereafter the two groups were similar. There was no evidence of recurrence of sedation in flumazenil-treated patients, nor did this drug adversely affect the period of anterograde amnesia between the administration of midazolam and flumazenil.