Between January 1980 and October 1983, 417 patients were randomized for mastectomy followed by axillary node sampling or full axillary clearance. The aim of the study was to determine whether a standard 'four-node' axillary sample, followed by careful dissection of removed tissue, could accurately indicate the extent of local treatment required. Axillary radiotherapy was given only to patients with histological involvement of sampled nodes and not to any having axillary clearance. The incidence of involved nodes was similar for both groups, as were distant relapse and survival rates. Currently 62.6 percent are alive after clearance and 65.0 percent after sampling. A non-significant increase in the rate of locoregional relapse was observed for those treated by axillary node clearance, this being due mainly to increased relapse on the unirradiated chest wall (clearance 21 percent versus sampling 12 percent in patients with node-positive disease). There was only a minor difference in axillary relapse, favouring axillary clearance (3.0 versus 5.4 percent). In patients with operable breast cancer, mastectomy with axillary node sampling gives equal control to mastectomy with axillary node clearance but, as morbidity is greater, surgical clearance of the axilla is the preferred option.