This study prospectively evaluated the usefulness of thoracoscopy for staging non-small cell lung cancer in 105 consecutive patients. A comparison was made of TNM stage grouping classification according to clinical disease, thoracoscopic data, and pathological findings. In 40 (38%) patients, thoracoscopy was unreliable for assessing extent of disease due to pleural symphysis. In 13 T1 clinical lesions, thoracoscopy was unreliable in 5, clinical and thoracoscopic staging concurred in 4, but 4 cases changed to T2. In 62 T2 clinical lesions, thoracoscopy was not feasible due to technical difficulties in 21 (34%); however, in the remaining 41 cases, 6 lesions changed to T3 and 1 to T4. In the group of 23 T3 or T doubtful clinical disease, thoracoscopy was conclusive, whereas in 12 T4 clinical lesions, thoracoscopy contributed for tailoring treatment strategies. With regard to N stage, 72 N0 clinical cases, thoracoscopy revealed false negatives in 25%. N1 clinical lesions were not evaluated due to the small number of patients. In 30 N2 clinical lesions, thoracoscopy was incomplete in 11. In another 11 cases, mediastinal node involvement at nodal groups not accessible by mediastinoscopy was confirmed by thoracoscopy. Clinical and thoracoscopic findings were not concurrent in eight cases, therefore in clinical N2 lesions, the diagnostic accuracy of thoracoscopy was 63%. Only one case of unsuspected pleural metastasis was detected. Thoracoscopy-related complications occurred in nine cases. In summary, video-assisted thoracoscopy was useful for staging T3, T4, and T doubtful clinical disease as well as N2 lesions especially for the surgical exploration of lymph nodes at the lower paratracheal level (region 4), aortopulmonary window (region 5), paraaortic (region 6), posterior subcarinal space (region 7), paraesophageal (region 8), and inferior pulmonary ligament (region 9).