Saliva analysis is an accepted non-invasive alternative to plasma in pediatric endocrinology. Although commercial saliva collectors are available, the reliability of these devices for the analysis of salivary hormones has not been proved. We investigated the recovery and linearity of salivary steroids (cortisol, cortisone, 17-hyroxyprogesterone, testosterone, androstenedione) being relevant in endocrine research and therapy control. Pooled saliva was spiked with ascending concentrations of the steroids and applied onto a variety of absorbents, such as the cotton and the polyester (PE) Salivette (Sarstedt), the foam-tip applicator (Whatman) and strips of blood-spot collection paper (Whatman). Analysis was performed by LC-MS/MS. Best results were achieved using the PE Salivette, yielding recoveries (%) of 99.8 (cortisol), 98.7 (cortisone), 91.8 (17OHP), 96.3 (testosterone), 98.9 (androstendione) with a volume recovery of 98+/-1%. Using the blood-spot paper, recoveries (%) were 92.0 (cortisol), 89.1 (cortisone), 72.0 (17OHP), 70.3 (testosterone) and 77.1 (androstendione). The recovery of glucocorticoids was significantly higher compared to androgens (p<0.001). The recovery of liquid volume was 95+/-2%. The cotton Salivette yielded weak recoveries of 88.7 (cortisol), 86.2 (cortisone), 60.9 (17OHP), 62.0 (testosterone) and 72.4 (androstendione). The recovery of the glucocorticoids differed significantly from the androgens (p<0.001). Liquid recovery was most variable with 89+/-8%. The weakest recoveries were found in the foam-tips being 76.2 for cortisol, only 41.8 for cortisone, 31.1 for 17OHP, 38.5 for testosterone and 36.1 for androstendione. The volume recovery here was 97+/-1%. We assume only the PE version of the Salivette suitable for salivary steroid analysis. The weak recovery from the cotton version is a severe problem due to lacking comparability with values obtained with the polyester wads and the weak homogeneity as observed over a physiological concentration range.