For the purpose of testing the validity of the new TNM classification (fourth edition) for lung cancer, data from 1086 patients with surgical treatment were analysed prospectively. Several items were examined: (1) the agreement between clinical (TNM) and pathologically confirmed classification (pTNM); (2) the value of the various diagnostic tests in estimating the pathologically confirmed classification; (3) the influence of the TNM definitions on separating distinct prognostic groups. With regard to the primary tumour (T), clinical and pathological classification were identical in 64% of the cases. With regard to lymph node involvement (N), the agreement was 48%, for distant metastasis 90% and for the staging 55%. As for the primary tumour (T), the accuracy of radiography (59%) was nearly identical with that of computed tomography (58%). Both these diagnostic techniques were less precise in determining the extent of lymph node involvement (computed tomography 50% correct assessments, radiography 43%). The statistically significant differences in the prognoses for the various pT, pN and pM categories as well as for the pathological stages and the categories of the new R classification could be confirmed. The new 1987 TNM definitions for lung cancer make possible international conformity; the classification is also practically useful and the prognostic relevance improved. The new classifications thus provide a more reliable basis for establishing guidelines for individual oncological therapy strategies and for the exchange of information between different centres on the progress made in diagnosis and therapy of lung cancer.