Fifty‐nine hospitalized patients participated in a double‐blind study: 19 received 15 mg midazolam, 20 received 50 mg oxazepam, and 20 placebo. The three groups were comparable with regard to age, sex, height, weight, and degree and type of insomnia. The sleep‐onset latency was shorter with midazolam than with placebo or oxazepam (Mann‐ Whitney test, alpha less than 0.05). With regard to total sleep duration and the number of nocturnal awakenings, there was no difference between the midazolam and oxazepam groups, whereas there was a difference between these two groups and placebo. More subjects of the midazolam group felt calm and refreshed on awakening. Safety, assessed by clinical examination and laboratory tests, was excellent. This study confirms the usefulness of midazolam in treating ‘early’ insomnia, i.e. difficulty in falling asleep.