In order to perform the dexamethasone suppression test (DST) with saliva as an alternative to serum, we assayed directly the cortisol concentrations in 25 microliters saliva samples, using a commercial radioimmunoassay kit for serum cortisol with minor modifications. Cortisol in saliva showed a diurnal rhythm parallel to that of cortisol in serum samples collected simultaneously. Saliva cortisol levels increased significantly after ACTH injection, but with a 60 min delay in reaching their peak compared to peak serum cortisol levels. The increase in saliva cortisol was five-fold, while that in serum was two-fold. Saliva cortisol levels continued to increase in some subjects while serum total cortisol levels already had begun to decline. In those subjects, the correlation of saliva with serum cortisol was greater when a quadratic curve was fitted than when calculated for a linear correlation. Considerable variation was observed for within-subject correlations, ranging from + 0.48 to + 0.999. The DST with saliva sample collection was performed on 43 inpatients with affective disorders. Sensitivity, specificity and diagnostic confidence of the DST for major depressive episode with melancholia were 33%, 91%, and 78%, respectively, at the criterion value of 0.3 microgram/100 ml for saliva cortisol, which are similar to those most often reported for the DST with serum cortisol determination. These results indicate that saliva cortisol levels do not always parallel serum cortisol levels and thus are not an unequivocal substitute. The findings for the DST in psychiatric patients, however, support the practical clinical usefulness of saliva cortisol measurements.