The incidence of lymphoedema was studied in 200 patients following a variety of treatments for operable breast cancer. Lymphoedema was assessed in two ways: subjective (patient plus observer impression) and objective (physical measurement). Arm volume measurement 15 cm above the lateral epicondyle was the most accurate method of assessing differences in size of the operated and normal arm. Arm circumference measurements were inaccurate. Subjective lymphoedema was present in 14 per cent whereas objective lymphoedema (a difference in limb volume greater than 200 ml) was present in 25.5 per cent. Independent risk factors contributing towards the development of subjective late lymphoedema were the extent of axillary surgery (P less than 0.05), axillary radiotherapy (P less than 0.001) and pathological nodal status (P less than 0.10). The risk of developing late lymphoedema was unrelated to age, menopausal status, handedness, early lymphoedema, surgical and radiotherapeutic complications, total dose of radiation, time interval since presentation, drug therapy, surgery to the breast, radiotherapy to the breast and tumour T stage. The incidence of subjective late lymphoedema was similar after axillary radiotherapy alone (8.3 per cent), axillary sampling plus radiotherapy (9.1 per cent) and axillary clearance alone (7.4 per cent). The incidence after axillary clearance plus radiotherapy was significantly greater (38.3 per cent, P less than 0.001). Axillary radiotherapy should be avoided in patients who have had a total axillary clearance.