Recently, it was hypothesized that acute or cumulative suppression of non-REM sleep intensity might be related to the therapeutic effects of antidepressants. This intensity has been proposed to be expressed in the EEG power density in non-REM sleep. In the present study, the relationship was examined between the changes of EEG power density in non-REM sleep and the changes in clinical state in 8 depressed patients during treatment with trazodone. A 1-week wash-out period was followed by 1 week of placebo administration, a medication period of 5 weeks and a 1-week placebo period. To minimize systematic influences of sleep duration and non-REM-REM sleep alterations, EEG power was measured over the longest common amount of non-REM sleep stages 2-4 (168.5 min), accumulated from sleep onset onwards. During trazodone treatment, the 13- and 14-Hz bins showed a significant reduction in EEG power. No clear-cut change, however, was observed in the EEG power of the delta frequency range (1-4 Hz) which is considered to be the principle manifestation of non-REM sleep intensity. Furthermore, no overall significant relationship between EEG power suppression and clinical improvement could be demonstrated.