To assess the value of estrogen receptor protein (ERP) as a predictor of tumor recurrence, 556 patients treated by mastectomy between 1973 and 1978 for primary operable breast cancer had ERP determination of their tumors. All patients had histologically negative nodes. Two hundred fifty-six patients were ERP-positive, 233 were ERP-negative, and 67 were ERP-borderline. ERP-borderline patients showed recurrence and survival figures similar to ERP-negative patients and were grouped with them in the analysis. With a median follow-up of 75 months, overall survival for the entire group at 72 months was 93% with a disease-free survival of 85%. No difference in either overall survival or DFS was noted between the ERP-positive (94% and 834, respectively) and ERP-negative (91 % and 86%, respectively) groups. The incidence of recurrence in premenopausal women was 12% (14/115) in the ERP-negative patients versus 17% (9/48) among the ERP-positive patients. This contradicts the current impression that ERP-negative, premenopausal patients have a poorer prognosis than ERPpositive patients. Postmenopausal patients had a 16% recurrence in both ERP-negative (27/171) and ERP-positive (31/202) categories. On the basis of 6-year median follow-up, it was concluded that ERP status is not an indicator of recurrence. In particular, patients with negative nodes should not be considered for adjuvant chemotherapy on the basis of negative estrogen receptor status of their primary tumors.