An increasing number of studies are utilizing saliva sampling as a method of assessing adrenal steroid secretion. Saliva samples have certain advantages over plasma, being non-invasive and easily collected. However, some methods of collection may compromise the accuracy of the assay, particularly those which employ aids to stimulate saliva production. We sought to compare the accuracy of cortisol and dehydroepiandrosterone (DHEA) measurement by examining the association between plasma levels, saliva and saliva collected using a citric acid-treated salivette device. Twenty six healthy male volunteers were recruited for the study. To increase the range of steroid levels in the samples collected, half the subjects were pre-treated with hydrocortisone (20mg, twice a day for 7 days) and half with placebo. Saliva samples were then collected from each subject using both a 'passive drool' method and a citric acid-treated salivette. A plasma sample was also collected. Cortisol and DHEA levels were measured by radioimmunoassay. For cortisol levels, both methods of saliva collection correlated highly with plasma levels and with each other (r 0.85; R(2) 0.72 for all). For DHEA levels, only saliva samples collected using the unstimulated collection method correlated with plasma levels. DHEA collected using the salivette device did not correlate significantly with either plasma or the unstimulated saliva (r 0.2;R(2) 0.04). It is crucial that future studies are aware of these issues and are cognizant of the effects of the method of collection when examining steroid levels in saliva.