Axillary dissection in early breast cancer remains controversial because of its substantial side-effects and because its value with respect to recurrence or survival has not been unequivocally proven. Between 1982 and 1987, 658 patients were included in a prospective randomised comparison of lumpectomy alone with lumpectomy plus axillary dissection. All patients had a unilateral breast tumour not exceeding 3 cm in diameter and lymph-node involvement or metastases. Radiation therapy was given to both groups. The two groups of patients were similar with respect to mean age, TNM stage, and presence of hormonal receptors. Median follow-up was 54 months. 5-year survival of the patients was 94.2% (95% Cl: 92.1-96.4). There was a significant advantage in survival in the axillary dissection group (p = 0.014). Recurrence of tumour in the breast was similar in the two groups but visceral metastases, supraclavicular metastases, and lymph-node recurrences were less frequent in the axillary dissection group. Survival was related to the age of the patients (p = 0.005), the presence of positive nodes (p = 0.006), the histological grading (p less than 0.0001), and the presence of hormonal receptors (progesterone p = 0.0008, oestrogen p less than 0.0001). Treatment-adjusted relative risk was 2.4 (95% Cl: 1.3-4.2). The findings show that axillary dissection is justified for treatment of small breast cancers, although whether the better survival is due to axillary clearance itself or to adjuvant treatment for lymph-node involvement is unclear.