BACKGROUND Disturbances of serotonergic neurotransmission appear to be particularly important for the pathophysiology of winter depression. This study investigated whether fluoxetine has antidepressant effects comparable to bright light in the treatment of seasonal affective disorder (winter type). METHODS A randomized, parallel design was used with rater and patients blind to treatment conditions. One week of placebo (phase I) was followed by 5 weeks of treatment (phase II) with fluoxetine (20 mg per day) and a placebo light condition versus bright light (3000 lux, 2 h per day) and a placebo drug. There were 40 patients (20 in each treatment condition) suffering from seasonal affective disorder (SAD) according to DSM-III-R who had a total score on the Hamilton Depression Scale of at least 16. RESULTS Forty patients entered phase II and 35 completed it (one drop-out in the fluoxetine group and four in the bright light group). Fourteen (70%) of the patients treated with bright light and 13 (65%) of those treated with fluoxetine were responders (NS). The remission rate in the bright light group tended to be superior (bright light 50%, fluoxetine 25%; P = 0.10). Light therapy improved HDRS scores significantly faster, while fluoxetine had a faster effect on atypical symptoms. Light treatment in the morning produced a significantly faster onset of improvement, but at the end of treatment the time of light application seemed not to be crucial. CONCLUSIONS Both treatments produced a good antidepressant effect and were well tolerated. An apparently better response to bright light requires confirmation in a larger sample.