Proteins and polyphenols were combined in model systems, and the resulting hazes were measured by light scattering. The amount of haze formed depends both on the concentrations of protein and polyphenol and on their ratio. A conceptual model in which a protein molecule has a fixed number of polyphenol binding sites explains the observed behavior and has implications for turbidimetric methods for estimating haze-active protein and haze-active polyphenol in beverages. The ranking of haze-forming activity of the test polypeptides was different with tannic acid than with catechin; this indicates differences in binding site availability, bridging ability, or specificity for the two polyphenols. More haze was observed when model systems were heated, suggesting that polyphenol binding sites are exposed when protein hydrogen bonds are broken. Freshly formed haze dissolved when dimethylformamide or dioxane was added; this may be useful for recovering compounds from isolated hazes for analysis. Keywords: Haze-active protein; haze-active polyphenol; beverages; model systems; gelatin; tannic acid