The authors evaluated 127 cases (31 infiltrating carcinomas, 22 intraductal carcinomas, 74 benign lesions) of breast microcalcifications with no palpable lesions. The patients had undergone mammography, stereotaxic cytology and direct magnification (87 cases). Blind interpretation of standard films and of direct magnification was performed by two independent readers (A, B), who classified the cases according to radiologic pattern (annular, punctate, granular, linear, branching), degree of suspicion and need for surgical biopsy. The study confirmed the association of different types of microcalcifications with breast cancer (predictivity: annular, punctate, granular, linear, branching; A = 25%, 15%, 40%, 92%, 86%; B = 14%, 26%, 39%, 90%, 100%). However, diagnosis was not very accurate (sensitivity: A = 83%, B = 71.1%; specificity: A = 83%, B = 78.4%); it was also observed that most cancers can be diagnosed only at a relatively high cost in terms of unnecessary biopsies. Interobservers' agreement was high relative to diagnosis (negative/dubious vs. suspicious/positive = -88.9%; K = 58.5) and low for pattern attribution (52%). Direct magnification allowed better detail definition but if sensitivity improved (A = 87% vs. 80.6%; B = 64.5% vs. 51.6%), specificity was negatively affected (A = 73.2% vs. 80.4%; B = 69.6% vs. 80.3%), due to an excess of false positives. Stereotaxic cytology [inadequacy rate = 26%, sensitivity and specificity (ex inadequates) = 80% and 100%, respectively] was more accurate than mammography as far as the decision for surgical biopsy was concerned, and was correct in identifying most (A = 6/7, B = 7/11) of the cancers which had been misdiagnosed as benign at conventional and magnification mammography. Stereotaxic cytology should be routinely performed in cases of microcalcifications with no associated palpable lesions.