Missed breast cancer continues to account for the highest percentage of medical malpractice cases in the United States. A retrospective, computer-aided study was performed to investigate the mechanisms of missed breast carcinomas, missed either by mammography or by clinical exam. In a consecutive series of 509 breast cancers found in patients registered at a University Comprehensive Breast Cancer Clinic, no tumour that was 5 mm or less in maximal diameter was clinically palpable. This subgroup consisted of seven in situ and 32 invasive carcinomas. The incidence of palpable tumours in 5 mm increments increased so that when a tumour was between 11 and 15 mm in size, 48% of the lesions were palpable, and with tumours greater than 20 mm in size, 84% were palpable. There was a good correlation between size of the tumour as judged by mammography and the eventual size determined by histologic examination. Smaller breast cancers were detected by mammography than by physical examination. In a separate analysis of 553 consecutive cases of breast cancer examined by mammography, there were 50 (9%) cases in which the cancer was not read from the mammogram. In retrospect, 10 of these mammograms were abnormal for a misinterpretation rate of 1.8%. Cancers associated with false negative mammograms occurred more often in younger women and in dense breast parenchyma than cancers detected by mammography. Cancers missed by mammography were smaller than palpable cancers detected by mammography, more often had negative nodes and presented with a lower stage of disease. Breast augmentation implants were associated more frequently with missed breast cancers, with 5/8 clinically detected breast carcinomas being undetected by mammography. An asymmetric mass was more often associated with cancers missed by mammography, accounting for the sole sign of malignancy in 3% of all cancers, but was the source of 14% of false negative exams. Architectural distortion, ill-defined or well circumscribed masses or calcifications as mammographic signs of malignancy were not associated with an increased frequency of missed cancers. Three 'interval' breast cancers occurred in this series and are included in the false negative mammograms. It is concluded that the threshold of clinically detected breast cancers is 6 mm and experienced clinicians do not detect the majority of breast cancers until the lesions are greater than 16 mm. Mammography has a defined misinterpretation and false negative rate. Likewise, asymmetric mammographic densities that are greater than 16 mm and are not palpable may be followed, since most breast cancers are palpable in this range.