The authors evaluate 1455 consecutive breast cancers detected 1985-1987, undergoing mammography 6 months before diagnosis at the most. One hundred and seventy-eight cases reported as negative/benign were assumed as false negatives. Overall sensitivity was 88%, and was influenced by tumor stage (TIS = 80%, T1 = 83%, T2 = 91%, T3-4 = 95%) and patients' age (less than 40 = 76%, 40-49 = 78%, 50-59 = 91%, greater than 59 = 94%). Sensitivity varied also with readers' experience full- or part-time involved in reading mammograms (total cases = 92% vs 88%, T1 cases = 86% vs 74%). No correlations were evidenced between mammographic sensitivity and histologic type of cancer. Palpation and cytology allowed the correct identification of 98% of mammographic false negatives. One hundred and seventy-eight false negatives and 142 more cases showing evidence of a mammographic error more than 6 months before diagnosis were reviewed to assess the most common types of radiologic fault. A benign-like lesion was the most frequent finding (49.4%), whereas suspicious signs were infrequent at review (16.6%). Failure to encompass the lesion in the mammographic field or poor technique accounted for 3.8 or 1.3% of cases, respectively. 29% of false-negative subjects (mostly young or with a dense breast) showed no lesion even at review. This study confirms the good sensitivity of mammography but suggests a less confident diagnostic approach in younger women, stresses the need for other tests whenever a minimal doubt is present, and encourages the radiologist's full-time involvement in mammography and breast diagnosis to achieve greater experience and improve his diagnostic sensitivity.