This retrospective study presents the diagnostic, prognostic and therapeutic problems raised by axillary recurrences (AR). 1589 cases of breast cancer measuring less than 3 cm, treated at the Institut Curie between 1981 and 1987, were studied by a combination of surgery and radiotherapy. Treatment of the breast always included wide local excision associated with irradiation. The axilla was treated either by dissection (865 cases) or by irradiation (724 cases) and 159 patients received chemotherapy. 26 patients (2%) developed AR, confirmed by fine needle aspiration cytology in 92% of cases. None of these 26 patients had initially received chemotherapy. The treatment of the AR was variable, adapted to the initial treatment. 22 patients retained their breast during treatment of the AR and none subsequently developed a local recurrence. 4 mastectomies were performed and histological examination revealed a subclinical local recurrence in 2 cases. The TNM classification, menopausal status, size of the tumour and hormonal receptor status were not risk factors for AR. Young age (P = 0.01) and high histological grade (P = 0.03) were significant risk factors for AR. The AR rate was similar whether axillary dissection or axillary irradiation had been performed. The overall 5-year survival after initial treatment was 85% for AR and 95% for the reference population. The overall 4-year survival after recurrence was 69% and the incidence of metastasis was markedly increased (P = 0.002). 2 of the 26 patients developed lymphoedema of the arm after treatment of AR. We confirm that AR worsens the prognosis, but not significantly more than local recurrence. Young age and the modified histological grading of Scarff Bloom and Richardson were risk factors for AR. Although excision of the AR is necessary to ensure local control, mastectomy is unnecessary when clinical examination and mammography are normal.