Nine patients with rapid cycling bipolar disorder were treated with a total of 13 trials of bright light therapy in the morning (n = 5), evening (n = 3), or midday (n = 5). In each instance, the patient's mood ratings during 3 months of light therapy (added to a stable medication regimen) were compared to his or her mood ratings during 3 months on the same medication but without light treatment. Of the 3 light therapy schedules, only midday lights appeared to have beneficial clinical effects, improving mood ratings in 3 patients. In contrast, the morning light therapy trial was terminated prematurely in 3 patients because of clinical instability. Light treatment was better tolerated if patients discontinued it on days when they were hypomanic. The clinical and theoretical implications of these preliminary findings are discussed.