We reviewed tumors from two groups of patients with breast cancer, distinguished by differences in outcome. One group (85 cases) survived more than 8.5 years without tumor recurrence; the other 85 cases had recurrent disease within 2 years. Histologic and immunocytochemical studies on all cases were performed without patient identifiers and prior to review of clinical prognostic factors. As expected, lymph node and estrogen receptor status differed substantially between the groups, but menopausal status and family history for breast cancer did not. We noted that 27% of node-negative patients died within 5 years, and nine patients with four or more tumor-containing nodes were symptom-free for over 8.5 years. Histologic grade (degree of tubule formation) and nuclear grade (including mitotic rate) differed significantly between the groups, as did vascular invasion, including both lymphatics and blood vessels. Prognostic value attached to tumor border only when fat was invaded without fibroblastic or inflammatory response (P = .012). Subgrouping cases of infiltrating ductal carcinoma (not otherwise specified) was prognostically informative in the B subgroup, 69% of whom were in the rapidly recurrent tumor group. Immunocytochemical staining for c-erbB-2 was positive in 19.3% of cases, but was equally distributed between the two outcome groups. We conclude that traditional histologic parameters are highly informative, and that c-erbB-2 studies do not increase the value of histologic diagnosis.