To evaluate the effectiveness of flumazenil in reversing midazolam-induced conscious sedation and general anesthesia, we gave either flumazenil or placebo to 55 patients in a double-blind manner after surgery. Whether surgery was done under conscious sedation (CS group) or general anesthesia (GA group) depended on the procedure. Recovery was assessed by an Observer Assessment of Alertness and Sedation (OAAS) Scale, Finger-Nose (F-N) test, and picture recall and recognition. OAAS and F-N tests were done at baseline, 0 minutes (before the test drug administration), and at 5, 15, 30, 60, 120, and 180 minutes. Picture recall and recognition were tested at 180 minutes and 24 hours, respectively, after test drug administration. Patients who were sedated during surgery (CS group) received a mean of 10.3 ± 5.3 mg midazolam. The patients receiving general anesthesia (GA group) were given 21.0 ± 8.2 mg midazolam. In the CS group, improvement in scores on the OAAS scale and F-N test was similar after administration of both flumazenil and placebo. In the GA group, flumazenil produced significantly greater improvement in the OAAS scores at 5 and 15 minutes, and in the F-N test scores at 15 minutes. In both groups, picture recall and recognition improved significantly immediately after flumazenil administration, but this improvement was generally not sustained for pictures shown at later times. These results imply that flumazenil is beneficial for reversing amnesia briefly after midazolam-induced sedation. However, flumazenil hastens recovery only when larger doses of midazolam are used for general anesthesia.