Midazolam (15 mg p.o.) was compared with placebo and oxazepam (15 mg) in 12 healthy volunteers and in seven patients suffering from sleep disorders in a single-blind cross-over study. Each treatment period lasted for seven days. The last two nights were spent in a sleep laboratory to evaluate the efficacy of the three compounds. The drugs were given to the patients every day and to the volunteers only on the recorded nights immediately before going to bed. The subjects rated their quality of sleep every morning after administration. Midazolam shortened (P = 0.025) the sleep latency to first stage 2 (t2 = 29 min) compared with placebo (t2 = 58.7) min and oxazepam (t2 = 55.4 min) in the group of patients; in the group of volunteers t2 was shortened (P = 0.05) only by midazolam (t2 = 17.2 min) compared with placebo (t2 = 24.6 min). REM suppression was not found in the group of patients, while sleep stages 3 + 4 were slightly reduced. However, a suppression of REM by midazolam (P = 0.025) and oxazepam (P = 0.01) was observed in the volunteers compared with placebo. The effects of midazolam seemed to be related to its pharmacokinetics. The drug increased the amount of stage 3 + 4 (P = 0.01) and suppressed REM (P = 0.005) compared with oxazepam and placebo, only during the first 3 h, when it was measurable in plasma. Midazolam was rated by the patients more favourably than oxazepam (P = 0.025) and placebo (P = 0.05). The volunteers noted no difference amongst the three treatments, but reported hangover effects after oxazepam.