We studied CT and mediastinoscopy as methods for preoperatively diagnosing mediastinal lymph node metastases (N2) in 133 resected lung cancers, and determined the optimal criterion for a CT diagnosis based on node size. All 133 patients were examined with CT, with a resulting sensitivity of 57% and specificity of 81%. The optimal CT criterion for metastasis was a short node axis of greater than or equal to 10 mm. Where nodes with short axes of greater than or equal to 20 mm, for squamous cell carcinoma, and greater than or equal to 15 mm, for adenocarcinoma, were selected, 100% specificity was obtained. It can thus be considered that nodes of this size on CT show definite metastatic disease. Thirty-three patients who satisfied the selection criteria out of a total 80 patients underwent mediastinoscopy as a clinical trial, with a sensitivity of 70% and a specificity of 100%. Of these, the nine CT false positives (eight squamous cell carcinomas) and three out of the six CT false negative (all adenocarcinomas) were properly diagnosed. We compared 80 cases diagnosed as N2 by CT alone and by CT plus mediastinoscopy, and obtained the following results: accuracies of 67.5 and 82.5%, sensitivities of 54 and 67% and specificities of 73 and 89%, respectively, showing the addition of mediastinoscopy significantly to improve the diagnosis of N2 disease (P = 0.03). We now routinely include a mediastinoscopy except in cases where greatly enlarged nodes are visible on X-ray or in patients who are not candidates for neoadjuvant chemotherapy.