A group of 27 patients with definite (n = 20) or probable (n = 7) RDC major depressive disorder underwent 2 sleep EEGs and 1 TRH test while in a drug-free depressive phase. A short mean REM latency (< 60 min) identified 55.5% of major depressives while added use of blunted TSH responses (delta max. < 5 µU/ml) increased that percentage by 11 %. When patients were subdivided into RDC endogenous and nonendogenous, mean REM latency and global depression scores distinguished the 2 groups, while delta TSH did not. A short mean REM latency identified endogenous depression with 80% specificity and 76% sensitivity. The combination of REM latency and delta TSH reduced the specificity to 60%, and therefore cannot be recommended for differentiating endogenous from nonendogenous depression