This study evaluated laboratory sleep and performance after placebo, after 5, 10 and 20 mg of midazolam and after 30 mg of flurazepam. EEG recordings showed that 20 mg of midazolam significantly decreased sleep latency and stage 1 sleep, increased stage 2 sleep, and delayed the onset of the first REM period when compared with placebo. Subjective reports from the patients showed that the dose decreased the frequency of awakenings. In the morning, 9 h after drug ingestion, performance on most tests was affected very little by 20 mg of midazolam. However, performance on two psychomotor tests was slightly impaired after 20 mg of midazolam, while a test of free recall was slightly improved. On most variables, 5 mg of midazolam acted like placebo, while 10 mg of midazolam was intermediate between placebo and 20 mg of midazolam. Preliminary data suggest that 20 mg of midazolam may be as efficacious in inducing sleep as 30 mg of flurazepam, but may show fewer performance decrements in the morning.