The aims of this study were to determine the suitability for radiographic interpretation, interobserver variability and validity of mammography after breast-conserving treatment, Initial and post-treatment mammograms of 100 consecutive patients treated between 1982 and 1987, with a minimal follow-up of 5 years, were independently selected for review by two radiologists. Mammograms were classified according to suitability for interpretation and radiological diagnosis based on the presence of characteristics of malignancy. The interobserver variability was expressed in kappa values, the validity in a receiver operating characteristic (ROC) plot. 534 post-treatment and 86 initial mammograms of 92 patients were obtained. Suitability for interpretation was not different from pre-treatment mammograms and was significantly associated with age, being better in the age group over 50 years. No association was observed between suitability for interpretation and treatment-related factors, even if irradiation was combined with concurrent chemotherapy. Reliability of conclusions regarding sensitivity and specificity in this study are limited due to the small number of events. Interobserver agreement concerning classification was moderate (weighted kappa=0.49). ROC analysis showed an optimal decision threshold between the ''uncertain'' and ''suspect'' categories of malignancy, resulting in a sensitivity of 86% and a specificity of 98%. The appearance of new pathological microcalcifications with or without tumour mass seemed to be the most important characteristics of malignancy predicting local relapse. No clear alteration in suitability for interpretation was observed in the mammograms after breast-conserving treatment, even if irradiation was combined with concurrent chemotherapy. Mammography after breast conserving-treatment may be slightly less sensitive but is equally specific compared with mammography in the screening situation.