Symptomatic pulmonary disease occurred in 20 per cent of 93 patients with anaplastic gliomas being treated with carmustine (BCNU). An analysis of the variables has revealed a relation between the occurrence of pulmonary toxicity on the one hand, and the total cumulative dose of BCNU, the number of cycles over which the BCNU was administered, the history of lung disease, the patient's age, and the platelet-count nadir after the first course of BCNU on the other. An equation has been generated that allows prediction of pulmonary toxicity during the course of therapy with BCNU with 80 per cent accuracy. Pretreatment analysis of individual cases should allow safe use of BCNU and prevention of most of the serious pulmonary complications caused by this drug.