The accurate measurement of testosterone remains a challenge. The determination of the blood testosterone concentrations in serum by conventional immunoassays is inaccurate in men and even more so in females and children. A new luminescence enzyme immunoassay (LIA) has been developed and validated. The high analytical (8.7 pmol/L) and functional (17.3 pmol/L) sensitivity allows the quantification of the very low concentration in saliva, as well as in serum, after 1/40 dilution. This study measured salivary testosterone levels and compared the results with the free levels calculated from total testosterone and sex hormone-binding globulin in eugonadal and hypogonadal men. Salivary testosterone concentrations in healthy men in morning hours were 369 pmol/L (mean), range 263–544 pmol/L, which was statistically significantly higher than that in men with androgen deficiency, 215 pmol/L (mean), range 51–249 pmol/L.Repetitive determination of free testosterone concentrations in saliva (once a week for 5 weeks) showed high stability of results over time, with coefficient of variation 9% (range 5–23%).In this study we showed that free salivary testosterone levels in morning samples correlated well with calculated free testosterone in blood, both in healthy men (R = 0.754, P = 0.001), and in patients with androgen deficiency (R = 0.889, P = 0.0001), though in cases with very low testosterone, salivary concentrations were systematically higher than calculated free testosterone levels in blood.