Results obtained by competitive protein binding assay (PBA) and a solid-phase radioimmunoassay (RIA) for cortisol were compared in 157 samples from 100 psychiatric patients given a dexamethasone suppression test (DST). Cortisol levels in plasma samples obtained at 8:00 a.m. or 4:00 p.m. the day following 1.0 mg dexamethasone orally at bedtime ranged from 0 to 30 micrograms/dl and correlated closely (r = 0.96). However, RIA gave values that were consistently and significantly lower (average = 8.9%) than those obtained by PBA. When samples were further assayed by a specific RIA for corticosterone, there was a strong correlation between cortisol and corticosterone RIA values (r = 0.79), and corticosterone (7.8% of cortisol levels) accounted for most of the difference between PBA and RIA for cortisol. The relationship between results of the two cortisol assay methods can be expressed (in micrograms/dl) by the equation: RIA = 0.92(PBA) - 0.10, based on findings obtained in a separate analysis of 127 samples with cortisol values in the 0-10 micrograms/dl range, critical to the valid interpretation of the DST in melancholia. A reported criterion of a "positive" DST in psychiatry, of plasma cortisol of greater than or equal to 5.0 micrograms/dl has been suggested by use of a PBA. Use of the present RIA required that this value be adjusted downward, at least to 4.5 micrograms/dl; application of this criterion increased the clinical sensitivity of the DST by 10%. We urge local, independent verification of criteria to define the DST as "positive" in each laboratory and with each method of assay.