Since the introduction of mediastinoscopy, there has been a great deal of discussion regarding indications for this technique and the significance of positive findings. We undertook this study to determine the role of clinical staging and the value of routine mediastinoscopy in the treatment selection of patients with primary lung cancer. From 1975 to 1983, 1,259 consecutive patients with proven and operable lung cancer underwent preresection mediastinoscopy. Nodes were sampled at three levels, and findings were recorded by location, invasiveness, and histology. There were no operative deaths, but 3 patients had a major complication. Mediastinoscopy was positive in 339 (27%) patients and negative in 920 (73%). In the group with positive findings, 303 patients had no operation because a curative resection was not possible (extranodal metastases, 180; location, 76; histology, 47). No patient survived 5 years, and only 4% survived 2 years. Of the 36 patients considered to have operable disease, 28 underwent resection with a projected 5-year survival of 18%. In the group with negative findings, 89% had a curative resection with a hospital mortality of 3.2% and 5-year survival of 53%. When results of mediastinoscopy were correlated with findings at thoracotomy, the sensitivity of the test was 93% on nodes in the superior mediastinum and the specificity, 100%. This study shows that mediastinoscopy is safe and is an accurate indicator of the presence or absence of tumor in superior mediastinal nodes. If positive nodes are found, a curative resection is generally not possible, thoracotomy is avoided, and the overall survival is low.