In 72 hospitalized depressed patients afternoon plasma cortisol (15.00-15.20) was measured in three ways: spontaneous (n = 72), 16 h after administration of 2 mg dexamethasone (n = 71) and 2 h after administration of 45 mg oxazepam (n = 28) or 60 mg oxazepam (n = 37). When cortisol values were log-transformed the histograms and probit functions of spontaneous cortisol levels and cortisol levels after oxazepam suppression suggested log-normal distributions. However, for cortisol levels after suppression with dexamethasone the probit function was sigmoidal, suggesting non-normality. In addition the histogram for this parameter had two peaks. The most pronounced suppression with dexamethasone was obtained in patients with low spontaneous cortisol levels. In contrast the somewhat weaker cortisol suppressing effect of 45 or 60 mg oxazepam appeared not to be influenced by the spontaneous cortisol level.