Background: Accurate false negative rates for endobronchial ultrasound (EBUS) and esophageal ultrasound (EUS) for mediastinal (N2) lymph nodes are unknown. Methods: A retrospective review of patients with non-small cell lung cancer (NSCLC) underwent integrated positron emission tomography and computed tomography (PET/CT) and CT scan. All suspicious N2 lymph nodes were biopsied before thoracotomy. The EBUS was performed for suspicious nodes in stations 2R, 4R, 4L, 5, and 7; the EUS was performed for 4L, 4R, 5, 7, 8, and 9. Mediastinoscopy was performed in selected patients if they were negative by EUS/EBUS; if N2 negative, all patients underwent thoracotomy with complete thoracic lymphadenectomy. Results: There were 425 patients over a 2-year period, and 234 had suspected N2 disease. Of these patients, 72 had an EBUS; 16 were positive for N2 disease and 12 were false negative (7 patients at station 4R/4L, 4 patients at station 7; patient sensitivity 57%, negative predictive value 79%, accuracy 83%). Seventy-nine patients had EUS; 20 patients were positive for N2 disease and 12 were false negative (4 patients at station 4R/4L, 4 patients at station 7; patient sensitivity 63%, negative predictive value 80%, accuracy 85%). One hundred and forty-six patients had mediastinoscopy, which revealed N2 or N3 disease in 42 patients, and 7 were false negative (patient sensitivity 88%, negative predictive value 93%, accuracy 95%). Conclusions: Both EBUS and EUS are useful initial tests to biopsy suspicious N2 mediastinal lymph nodes; however, as EBUS and EUS have high false negative rates, especially at stations 4R and 7, mediastinoscopy is still required for patients with suspicious nodal disease in these stations. © 2010 The Society of Thoracic Surgeons.