Between 1984 and 1990, 94 women presenting to the Edinburgh Breast Unit with operable breast cancer of 4 cm or greater in diameter (T2, T3, N0, N1, M0) were given preoperative systemic therapy. Initially, all women received hormone therapy, with CHOP (cyclophosphamide 1 g m(-2), doxorubicin 50 mg m(-2), vincristine 1.4 mg m(-2) to a maximum of 2 mg and prednisolone 40 mg per day orally for 5 days) chemotherapy being administered to those who failed to respond by 3 months. After April 1987, first-line hormone therapy was only offered to women with oestrogen receptor (ER)-moderate/-rich (> 20 fmol mg(-1) protein) tumours, and CHOP was reserved for those women whose tumours failed to respond to hormone therapy and for those with ER-negative/-poor tumours. Response data have been published previously (Anderson et al, 1991). After a median follow-up of 7.5 years, there is no difference in survival between those women given initial hormone therapy and those given chemotherapy, with neither group having yet reached its median survival. The two key factors that predicted for a poor survival were the number of involved axillary nodes after preoperative systemic therapy (P < 0.00001) and a lack of response to preoperative therapy (P < 0.05). These data suggest that many women with ER-moderate/-rich tumours will have a good prognosis after preoperative hormone therapy alone. However, it is possible to identify, by their post-systemic therapy axillary node status, a group of women who still have an appalling prognosis after preoperative chemotherapy or hormone therapy.