ABSTRACT A Sephadex assay technique, in use in our laboratory for the study of steroid-protein interactions, allows expression of the molar concentration of a testosterone-binding component of human serum of high binding affinity in terms of the molar concentration of testosterone-binding sites: examples with model calculations are provided. The presence in the assay system of competing protein of low binding affinity (i. e. albumin) and of competing steroid (i. e. endogenous, radioinert testosterone) is taken into account when estimating the testosterone-binding component. With the aid of some approximations, the distribution of bound testosterone between the testosterone-binding component and albumin is readily calculated. From this estimate and the total testosterone concentration, the serum concentration of unbound testosterone (Su) is calculated. The concentration of the testosterone-binding component is thus estimated to be of the order of 8 × 10−7M and 1 × 10−7M in late pregnancy serum and adult male serum, respectively, assuming that there is only one binding site for testosterone in the protein molecule. The testosteronebinding component appears to play a role at least equal to that of albumin in binding endogenous testosterone in adult male serum; the testosterone-binding component is clearly dominant in this role in late pregnancy serum. The values for Su at 25° are calculated to be of the order of 3 × 10−10M and 1 × 10−11M in adult male serum and late pregnancy serum, respectively, corresponding to 1.7% and 0.27% of the total testosterone concentration, respectively. It is suggested that the physiologically effective level of circulating androgen may correlate better with the serum concentration of unbound testosterone (Su) than with that of total testosterone.