The dexamethasone suppression test (DST) and sleep EEG were compared in three different descriptive diagnostic schemes in 70 adult patients with nonpsychotic, unipolar major depressive disorder according to Research Diagnostic Criteria. Endogenous (E)/nonendogenous (NE), primary/secondary, and Winokur's family history subtypes were evaluated. The E/NE subclassification was best supported by both biological measures. The DST provided a highly specific (95%), relatively accurate measure (confidence interval of 87%), although its sensitivity was rather low (41%) in this largely outpatient sample. A REM latency of 62 min or less provided a more sensitive (66%) but less specific (79%) indicator of E depression. Stage 4 time and age may add to the information provided by REM latency alone in identifying E patients.