Tumor samples of 240 patients with primary breast cancer were biochemically and immunohistochemically investigated for estrogen receptors (ER) and, in 130 of the samples, for progesterone-receptors (PgR) in order to examine reasons for discordant findings. The biochemical (DCCA) and immunohistochemical assays (ICA) yielded positivity in 71% for ER, and in 44% for PgR. Concordant ER-DCCA and ER-ICA results were obtained in 84%; two thirds of the discordant ER-findings manifested as DCCA-neg/ICA-pos. Concordance in the case of PgR amounted to 72%, and of the discordances 60% were DCCA-neg/ICA-pos. Significant association with postmenopausal status existed only for ER positivity in ICA (p = 0.01), whereas ER-DCCA, PgR-DCCA and PgR-ICA were all more or less independent of the menopausal status. The frequency of discordances was independent of menopausal status. Discordance for ER-assays increased significantly near the respective cut-off point; this was not unequivocally true for PgR-assays. The correlation of tumor types of sparse cellularity, as well as prominent stroma content ('scirrhous carcinoma') with increased frequency of the constellation DCCA-neg/ICA-pos was of borderline significance for PgR (p = 0.06), but not for ER. The percentage of discordant ER-findings, figuring as DCCA-neg/ICA-pos, was statistically significantly increased in locally advanced breast cancer (p = 0.03). Fibrocystic disease in peritumoral breast tissue had no impact on receptor-assay discordance. In any case, the models derived from theoretical thought, laboratory data and singular observations can only in part explain the discordance in steroid receptor values measured with different methods.