As experience accumulates on the use of sentinel node biopsy in breast cancer, it is becoming clear that the method can reliably predict the state of the axilla and thus be used to decide whether to perform complete axillary dissection. Ongoing controlled trials will soon provide definitive evidence on the latter point. The key issue regarding sentinel node biopsy is pathologic evaluation of the biopsied node, which should be done intraoperatively whenever possible. In our initial experience with a conventional intraoperative frozen section method, the false-negative rate was 19% compared to examination of permanent sections of the biopsied node. We therefore devised a new intraoperative method in which pairs of sections are obtained every 50 mm for the first 15 sections and every 100 mm for any remaining node, which essentially samples the entire node; the method takes about 40 minutes. Sentinel node metastases were found in 119 of 295 (40%) of T1N0 breast cancer patients examined by this new method. This high rate of positivity indicates that the new method is reliable. In all cases, metastases were identified on hematoxylin-eosin (HE)-stained sections, although in 4% of positive cases the HE sections were doubtful, and cytokeratin immunostaining on the adjacent section was useful for confirming malignancy. Of 295 patients, 8 (2.7%) had a negative sentinel node but another axillary node metastasis. In conclusion, we found that extensive intraoperative frozen section examination of sentinel nodes correctly predicts a metastasis-free sentinel node in 95.4% of cases (negative predictive value), it is therefore suitable for identifying patients in whom axillary dissection might be avoided. Immunocytochemical staining for cytokeratins or other epithelial markers may be helpful for reducing the risk of missing micrometastatic foci.