We examined the estrogen receptor (ER) content of 124 primary breast cancers by hormone binding and immunohistochemical (ER-ICA) assays. Both assays were in agreement in 110 tumors (89%; P less than .0001); 68 tumors were positive and 42 were negative. In 14 cases (11%), the assays yielded discordant results. Three tumors showed hormone binding in the absence of immunohistochemically detectable ER; the false positive hormone binding resulted from the presence of normal epithelium adjacent to ER-ICA negative malignant cells. Eleven tumors failed to show hormone binding but were ER-ICA positive. Four of these were from premenopausal patients whose circulating endogenous estrogen may occupy the receptor, giving rise to false negative hormone binding assays. In four cases, the discrepancy of negative hormone binding assay and positive ER-ICA assay was attributed to scant tumor cells in the tissue sample. The remaining three discrepancies could not be resolved with certainty, but possibly resulted from alteration of the hormone binding site with preservation of the immunoreactive epitope on the ER molecule. These results indicate that the ER-ICA assay is more accurate than the hormone binding assay in identifying the presence of ER in cancer cells. The heterogeneous immunostaining of ER in tumor sections, which may reflect mosaicism of tumor cells, rate of cell proliferation, or phase of cell cycle, remains unexplained.