We studied in a sleep laboratory the effects of bedtime 15 mg midazolam and 20 mg midazolam on the sleep and morning performance of healthy subjects with polygraphically verified sleep onset insomnia. Six subjects received 15 mg midazolam and six subjects received 20 mg midazolam for 14 consecutive nights which were preceded by a three-night placebo baseline and followed by a three-night placebo period. The medications were administered in a double-blind manner. The results were that both doses increased total sleep time, reduced sleep latency, reduced wake time after sleep onset and reduced the number of awakenings. There was no difference between the doses. Midazolam had its main effect by decreasing wake time in the first third of the night. We found no evidence of tolerance, drug withdrawal rebound insomnia, or drug-induced morning performance decrements. We did find evidence of an anterograde amnesia produced by the drug.