The authors evaluate 261 consecutive mammographic false positives observed from 1985 to 1987. Histological evidence of benign lesion followed in all cases. The comparison with the actual number of cancers and of the whole of mammographic examinations performed in the study period allowed specificity and positive predictive value of mammography to be assessed as 99.5% and 83%, respectively. Specificity and predictivity are lower in younger women, but this is more likely to depend on a different age-related incidence of cancer and benign lesions than on an intrinsic limitation of the method. The reader's diagnostic aggressivity, more than his experience, seems to affect both specificity and predictivity. At review, false positives were mostly due to asymmetric densities (49) or to circumscribed opacities with clear-cut (44) or blurred (62) outlines, whereas irregular star-like opacities or distortions (19) were infrequent. Microcalcifications were, in most cases, apparently benign (39) or dubious (76); strong suspicion was rare (4). Overall, one-fourth to one-third (27.9%) of the cases were reported as strongly suspicious at review. Palpation and cytology were also falsely suspicious--that is, co-responsible for unnecessary biopsies in over 50% of cases. Our results suggest that further improvement in the specificity or positive predictive value of mammography seems unlikely. Moreover, the benign/malignant biopsy ratio (0.2:1) presently achieved in suspicious mammographic cases appears quite satisfactory.