Sleep quality and anxiety levels were examined using questionnaires and polysomnographic recordings in 22 chronic insomnia patients who regularly used benzodiazepines to treat their sleeping problems. After abruptly discontinuing their benzodiazepine medication, patients were randomly allocated to receive either a placebo or zolpidem 10 mg for 1 week, after which they entered an open extension phase, receiving zolpidem 10 mg for 3 weeks. Subjectively, sleep quality was considered mediocre during the use of a benzodiazepine hypnotic. One week after the discontinuation, an increase in sleep latency was observed in the placebo group, whereas zolpidem induced a significant decrease in sleep latency. Deterioration of other sleep variables (probably rebound) was not suppressed by zolpidem. An explanation for this could be the selective pharmacological profile of zolpidem. Polysomnographic differences between placebo and benzodiazepine and between placebo and zolpidem were not reflected by the subjective data on sleep and anxiety. Changes of sleep structure caused by hypnotics seem not always to be felt as such by patients. After 3-4 weeks of zolpidem treatment, the percentage of non-rapid eye movement-4 sleep increased significantly, corresponding with a significant subjective improvement of sleep quality. This indicates that zolpidem may restore physiological sleep.