Bright light is the recommended treatment for winter seasonal affective disorder (SAD). Previously we showed that the antidepressant effect of morning (but not evening) light was greater than placebo after 3 weeks of treatment. Here, we determined if the magnitude and direction of circadian rhythm phase shifts produced by the bright light in the previous study were related to the antidepressant effects. Twenty-six SAD patients from the original sample of 96 had their rectal temperature continuously monitored while they participated in a placebo-controlled parallel design conducted over six winters. After a baseline week, there were three treatments for 4 weeks-morning light, evening light, or morning placebo. Bright light was produced by light boxes (approximately 6000 lux). Placebos were sham negative ion generators. All treatments were 1.5 h in duration. Depression ratings were made weekly by blind raters. Circadian phase shifts were determined from changes in the timing of the core body temperature minimum (Tmin). Morning light advanced and evening light delayed the Tmin by about 1 h. The placebo treatment did not alter circadian phase. As the sleep schedule was held constant, morning light increased and evening light decreased the Tmin to wake interval, or phase angle between circadian rhythms and sleep. Phase advance shifts and increases in the phase angle were only weakly associated with antidepressant response. However, there was an inverted U-shaped function showing that regardless of treatment assignment the greatest antidepressant effects occurred when the phase angle was about 3h, and that patients who moved closer to this phase angle benefited more than those who moved farther from it. However 46% of our sample had a phase angle within 30 min of this 3 h interval at baseline. So it does not appear that an abnormal phase angle can entirely account for the etiology of SAD. A majority (75%) of the responders by strict joint criteria had a phase angle within this range after treatment, so it appears that obtaining the ideal phase relationship may account for some, but not all of the antidepressant response. In any case, regardless of the mechanism for the antidepressant effect of morning light, it can be enhanced when patients sleep at the ideal circadian phase and reduced when they sleep at a more abnormal circadian phase.