We examined 24-hour melatonin rhythms from 20 patients with seasonal affective disorder (SAD) and 20 healthy volunteers. Patients and controls were individually matched for age, sex, and month of study. Plasma samples were taken at hourly intervals, and were assayed for melatonin by radio-immunoassay. The 24-hourly melatonin estimations for each individual were fitted to a cosine curve, and the significance of the curve fits was calculated. Two analyses were performed. In analysis 1 the following were calculated: (a) cosine fit, (b) significance of fits, (c) mean amplitude and acrophase (peak) and (d) mean melatonin levels. The curve fits were highly significant for all but three subjects (two patients, one control), but there were no significant differences in any measure between the two groups. In analysis 2 the comparisons were repeated and restricted to the 18 patients and 19 controls in whom there was a statistically significant melatonin rhythm. Again there were no significant differences between groups. These results suggest that the circadian rhythm of melatonin is not abnormal in SAD, and that the therapeutic effect of light in SAD is not mediated by phase shifts in melatonin secretion.