The accuracy of intraoperative frozen-section diagnosis of small in situ and invasive breast cancers is uncertain. We reviewed the results of 1,490 consecutive wire-localized breast biopsies from 1,439 patients with nonpalpable mammographically detected abnormalities examined by frozen section at Mayo Clinic over a 3-year period. The mammographic abnormalities included benign calcifications (61 cases), indeterminate calcifications (422 cases), worrisome calcifications (54 cases), well-circumscribed nodules (115 cases), irregular nodules (473 cases), architectural distortions (52 cases), asymmetry (39 cases), well-circumscribed nodules with calcification (12 cases), irregular nodules with calcification (75 cases), architectural distortions with calcification (35 cases), and asymmetry with calcifications (12 cases). We detected 457 carcinomas, including 135 in situ and 322 invasive cancers. The invasive carcinomas had a mean diameter of 1.07 cm (range, 0.1-3.0 cm), including 191 with diameters of 1 cm or less (59.3% of cases). In 77 cases (5.2% of total), the diagnosis was deferred to permanent sections. Frozen-section accuracy was 97.7%. False-negative diagnoses were rendered in 0.5% of cases, and there were no false-positive diagnoses. There was no correlation of infiltrating carcinoma diameter and error rate. These results indicate that intraoperative frozen section of mammographically directed breast biopsies provides accurate and reliable diagnoses.