BACKGROUND Survival following pulmonary resection for primary lung cancer is considered to be principally dependent on the clinical stage of the disease. A study was undertaken to verify this and to identify other contributing factors. METHODS The case records of all patients who underwent surgery for lung cancer over a two year period between January 1987 and December 1988 were reviewed retrospectively. RESULTS One hundred and forty-seven lobectomies and 60 pneumonectomies were performed with 2.8% and 5.3% operative mortality, respectively. Squamous carcinoma was the commonest pathology (60%) followed by adenocarcinoma (30%). The overall five year survival was 45.5% (95% CI 44.1% to 57.9%). There were 123 patients with stage I disease, 40 with stage II, and 37 in stage IIIa with five year survival of 59.4% (95% CI 50.8% to 68%), 30% (95% CI 15.9% to 44.1%), and 16.2% (95% CI 3.5% to 31%), respectively. There were no differences in survival with respect to sex, extent of resection, or cell type. In patients with stage II disease the five year survival of those with T1 lesions (50%, 95% CI 37.3% to 62.9%) was better than those with T2 (28.1%, 95% CI 16.9% to 39.3%). Of eight patients over the age of 70 with stage IIIa disease none survived more than 24 months. CONCLUSIONS Stage at operation is the most accurate predictor of long term survival in early lung cancer and surgery remains an effective treatment, particularly in stage I and II disease. Further study is needed to assess the prognostic value of subdividing stage II disease into T1 and T2 lesions. Major resection for locally advanced disease in older patients may be relatively ineffective.