In 72 consecutive depressed hospitalized patients afternoon plasma cortisol was measured in three ways before treatment with antidepressants: 1) Spontaneous (n = 72), 2) 2h after oxazepam suppression (45 mg, n = 28; 60 mg, n = 37) and 3) 16 h after dexamethasone suppression (2 mg, n = 71). In addition, spontaneous cortisol was measured after 3 weeks' treatment (n = 55) and 5 weeks' treatment (n = 36). Both spontaneous and suppressed cortisol levels seemed to have a predictive value in the endogenously depressed patients: complete responders had significantly lower pretreatment cortisol levels compared to poor responders. However, other covarying factors such as distress and age may as well account for the differences in treatment effect. During treatment a significant decrease of spontaneous cortisol was found from about 400 nM in poor responders and 325 nM in complete responders to about 300 nM in all groups. There was a positive correlation between pre- and post-treatment cortisol levels and between pretreatment levels and per cent fall in spontaneous cortisol levels.