The intraoperative frozen section technique has proven to be valid and accurate in the diagnosis of palpable breast lesions, while its use in non-palpable lesions of the breast has been recently discussed. In order to estimate the accuracy of frozen section technique in non-palpable breast lesions we reviewed our series of 672 cases of non-palpable lesions examined intraoperatively between January 1977 and December 1991. Frozen section diagnoses were compared with diagnoses obtained at review of the permanent paraffin sections. A review of the mammographic pattern of each lesion was also carried out. Frozen section diagnosis was replaced by permanent paraffin sections in 22 cases (3.3%). Benign or malignant (in situ and invasive carcinomas) frozen section diagnoses were accurate in 623 of 650 cases (95.8%). Frozen section diagnosis was modified on the basis of permanent sections in 27 cases (4.2%), with 3 false positives and 24 false negatives. Sensitivity and specificity of frozen section technique were 91.7% and 99.2%, respectively. Comparing frozen section and definitive diagnosis according to the mammographic pattern, sensitivity was lower in patients with microcalcifications as the only alteration (88.8%) compared with patients with opacities (94.9%). On the basis of our results the frozen section technique in non-palpable breast lesions is to be considered a valid and reliable diagnostic procedure, particularly in lesions detectable at mammography as an opacity and identifiable on gross examination of the surgical specimen.