BACKGROUND Previous studies hypothesized that seasonal affective disorder (SAD) was caused by a circadian rhythm abnormality. The purpose of this study was to ascertain whether rest-activity rhythms were phase delayed, diminished in amplitude, or more poorly entrained to the 24-hour day. METHODS Twenty healthy adult controls and 25 outpatients meeting Rosenthal-National Institute of Mental Health criteria for SAD and DSM-III-R criteria for major or bipolar depression with seasonal pattern had their levels of activity recorded for 72 hours (weekdays) using wrist-worn actigraphs. RESULTS Subjects with SAD had activity levels that were 11% lower than controls (P = .03), and their levels of activity were most attenuated during the first 2 hours after arising (P = .004). The relative amplitude of the circadian rhythm did not differ between groups. Patients with SAD were phase delayed by 50 minutes for the entire period (P = .02). Analysis of each individual day indicated that patients were delayed by up to 70 minutes (P = .007). Interdaily stability, an index of coupling between the rhythm and its zeitigeber was reduced in SAD (P = .01). Compared with controls, patients with SAD had best-fit circadian periods that were 92% more deviated from 24 hours (P = .007) and daily acrophase (time of the peak of the fit circadian rhythm) times that were 110% more variable between days (P < .001). CONCLUSIONS Patients with SAD have circadian rest-activity rhythms that are significantly phase delayed and more poorly entrained to the 24-hour day.