Résumé : Three radioimmunoassays (RIA), with or without preparative HPLC, were applied to the monitoring of plasma dexamethasone (DXM) levels during standard dexamethasone suppression test (DST) in psychiatric patients. Due to the robotic ease of the fully automated HPLC process, precision of the chromatographic assay was equivalent to that of the direct assays, but prepurification improved both sensitivity and specificity. These improvements allowed the elucidation of the following features: (1) half (36) of the patients (68) displayed infranormal DXM levels (less than or equal to 0.40 ng/ml) whatever the cortisol response; (2) 22% (15) patients (68) with DXM levels in the low control range showed a strong inhibition of cortisol suppression. These observations raise some doubts on the validity of the DST test and introduce the following questions. (1) What is the dependence of cortisol suppression upon DXM absorption and catabolism? (2) Does plasma DXM measurement several hours after its physiological action still reflect its effect on the hypothalamo-hypophyseal axis? (3) What is the reliability of DXM direct assays when measuring low DXM levels in the presence of high cortisol?