After introduction of the Salivette device for saliva sampling, an increase of approximately 50 ng/l was noted normal values of salivary 17-hydroxyprogesterone in our laboratory compared with previous results obtained by collection of saliva by direct spitting. To further investigate this observation systematically, 17-hydroxyprogesterone was measured in saliva of 22 healthy adults after sampling without and with the Salivette and in 15 samples of distilled water rinsed over the cotton wool swab of the Salivette. The measurements were performed with a commercially available [125I]17-hydroxyprogesterone radioimmunoassay. All steps in assay performance were exactly the same. In both series, salivary 17-hydroxyprogesterone concentrations significantly increased by 40 ng/l (p < 0.05) with use of the Salivette. The variance of the differences covered a wide range. Our results confirmed that the use of the Salivette increased salivary 17-hydroxyprogesterone values by approximately 40 ng/l. This bias is unacceptable for the establishment of reference values, as these are in the range of 20-70 ng/l in the morning. Thus, at least with our assay, the use of the Salivette is discouraged. We speculate that the cotton wool swabs contain a material. possibly derived from the biological source of the cotton wool, which cross-reacts with the antibody in the assay or affects binding affinity.