Psychotropic drug-free hospitalized veterans with nonseasonal major depressive disorders or depressed forms of bipolar disorder were treated with light for 1 week. Twenty-five patients were randomly assigned to bright white light treatment (2000-3000 lux), and 26 patients were randomized to dim red light placebo control treatment. Unlike those treated with dim red light, those treated with bright white light showed declines in three measures of depression during treatment. Partial relapse appeared within 2 days. A global depression score showed a statistically significant (p = 0.02) difference favoring bright white light treatment. Two bright-light-treated patients became mildly hypomanic, but side effects were mild. Improvement was not correlated with patient expectations; indeed, patients expected somewhat greater benefit from the placebo. Patients treated in summer responded as well as those treated in winter. Baseline electroencephalogram (EEG) sleep stage data (e.g., rapid eye movement; REM latency) did not predict treatment responses. These 1-week treatment results suggest that bright light might produce benefits for patients with nonseasonal depression. Bright light should not be recommended for routine clinical application before additional assessments with longer treatment durations are done.