We report a radioimmunoassay sensitive enough to determine 17 alpha-hydroxyprogesterone concentrations in 200 microliter of parotid fluid or mixed whole saliva. Because the correlation of concentrations in matched samples of parotid fluid and saliva was excellent (r = 0.98), we exclusively used saliva, which is easier to collect, in later studies. The assay is specific; saliva samples assayed with and without thin-layer chromatographic purification showed no significant difference. The assay is also precise, and has a lower limit of sensitivity of 4 pg per assay tube. In 14 patients having congenital adrenal hyperplasia from a C21-hydroxylase enzyme deficiency, all of whom were receiving cortisol replacement therapy, the range in 17 alpha-hydroxyprogesterone concentrations observed in saliva (67-26,300 pmol/L) was about 20-fold that seen in 32 healthy children (90-1520 pmol/L). The close correlation (r = 0.91) between 17 alpha-hydroxyprogesterone concentrations in matched samples of saliva and plasma from these patients indicates that determination of steroids in saliva could well replace determination in plasma. This concept is supported by 17 alpha-hydroxyprogesterone concentrations monitored throughout 24 h from one patient and following stimulation with synthetic corticotropin in another patient.