In a retrospective survey of 337 patients who had a resection for bronchial carcinoma data were obtained on 14 preoperative and two operative variables. Factors significantly adversely influencing survival were weight loss, chest pain, raised erythrocyte sedimentation rate, tumour larger than 5 cm, cell type other than squamous, pneumonectomy and tumour spread to hilar lymph nodes. Multiple regression analysis showed that four variables, weight loss, pain, tumour size and cell type had significant independent effects on survival. From the regression equations predicted median (50%) survival, varying from 9 to 95 months was calculated and tested on a series of 49 patients undergoing resection since completion of the survey. The correlation between overall observed and predicted median survival was highly significant (P less than 0.001). The prediction equation was more accurate for patients with a predicted survival of less than 30 months if cell type was known. Classification using the prediction equations may improve patient selection for surgery.