Although the endogenous subtype in depression has long been thought to have prognostic significance, to date no long-term maintenance treatment trial has examined the relative risk of recurrence in patients meeting criteria for this subtype. Following our analysis of the primary hypotheses regarding the relationship between treatment assignment and outcome [Frank et al. (1990) Arch. Gen. Psychiatry 47, 1093-1099], we now examine psychobiologic and maintenance treatment correlates for these recurrent unipolar patients grouped according to melancholic, endogenous but not melancholic, and non-endogenous subtype at index presentation. No differences were observed among the three groups in overall survival time; however, in the 52 patients who received psychotherapy without active medication during the maintenance phase, length of survival was inversely related to endogeneity. Interestingly, no differences were found among the three groups in EEG sleep parameters when studied either at baseline or following recovery.