The effects of 8-chloro-6-(2-fluorophenyl)-1-methyl-4H-imidazo[1,5-a] [1,4]benzodiazepine (midazolam, Ro 21-3981, Dormicum) in oral formulation of 15 and 30 mg on the sleep cycle of patients suffering from insomnia were assessed by means of polysomnographic recordings using a double-blind cross-over design. Both doses of midazolam were effective in improving sleep on short-term administration. In addition, significantly larger decrements of non-REM (NREM) sleep latency and of wake time through the 3rd third of night and nonsignificant trends toward smaller number of awakenings as well as shorter total wake time and longer NREM sleep time were induced by the 30 mg dose. Irrespective of the dosage sleep was almost exclusively increased at the expense of NREM sleep. Following 3 days treatment there was no rebound insomnia. These preliminary results suggest that the 15 mg dose could be appropriate in patients with difficulties in falling asleep, while the 30 mg dose would be more appropriate for patients who also experience difficulties in staying asleep.