The applicability of a model of specific and non-specific binding sites for the binding of organic ligands to serum albumin was examined statistically. It is proposed in this model that there are a limited number of a single class of high affinity sites and an unlimited number of low affinity sites. Ligands bind to the former sites independently, as in the Scatchard model for a single class of binding sites, and to the latter sites in a partition-like manner, the amounts bound increasing linearly with the concentration of free ligands. In this study, data on the binding of warfarin to human serum albumin reported by Wilting et al. [J. Wilting et al., J. biol. Chem. 255, 3032 (1980)] and on the bindings of indomethacin and salicylic acid to human serum albumin reported by Hultmark et al. [D. Hultmark et al., Acta pharm. suecica 12, 259 (1975)] were used. These binding data were analyzed according to the model of specific and non-specific sites as well as the Scatchard model of two classes of binding sites. Statistical analyses showed that the model of specific and non-specific binding sites fitted the data for binding of all compounds very well, indicating the applicability of this model. An index of the upper limit of ligand concentrations necessary for accurate analysis of the binding data was also discussed.