Pathology reports for all lobectomy and pneumonectomy specimens at UNC Hospitals between 1991 and 2000 (n ¼ 405) were reviewed for correlation between frozen section and final bronchial margin, gross distance between tumor and margin and tumor type. Frozen section was performed in 268 cases (66%). A total of 243 were true negatives (90.6 %), 16 (6.0%) were true positives, four (1.5%) were false positives and five (1.9%) were false negatives. The site of tumor in true-positive cases was mucosal (11), submucosal (three), lymphatics (one), peribronchial (one). The site of tumor in false-negative cases was submucosal (two), lymphatics (one), peribronchial (two). In 137 cases, no bronchial frozen section was performed; there was one case (0.7%) with positive margin. There was no correlation between final margin positivity and distance between gross tumor and margin. Tumor distance to margin in positive margin cases varied from grossly involved to 3 cm away. There were 72 cases in which wedge resection was performed before lobectomy in which no gross tumor remained in the lobectomy, and in all cases final bronchial margins were negative. In all, 373 of cases (92%) were nonsmall carcinomas. Of these, 10 (2.7%) had positive margins. Tumors other than nonsmall cell carcinoma accounted for a disproportionate number of positive margins. In all, 3/6 of adenoid cystic/ mucoepidermoid carcinoma, 1/7 small cell carcinoma and 1/1 lymphoma cases had positive margins. In conclusion, frozen section evaluation of bronchial margins is helpful in central lung tumors. Mucosal tumor is preferentially identified in frozen section. Gross evaluation of margins is problematic, as intramucosal carcinoma or tumor in lymphatics may not be detected, but 3 cm was a 'safe' distance for gross tumor from margin. In lobectomies following wedge resection in which no gross tumor remained, all had negative margins. Salivary gland-type tumors have a high incidence of positive margins, and frozen section is particularly indicated in these tumors.