To assess the utility of bronchoalveolar lavage as a technique for diagnosing lung cancer, 850 lavages from 421 patients were reviewed. Biopsy-proven lung carcinoma was present in 35 cases. Of these, 24 (68.6%) had cells diagnostic of malignancy on cytologic preparations of the bronchoalveolar lavage fluid. Agreement between cancer subtypes determined by lavage and by tissue biopsy was 79.1%; variation usually occurred between large cell undifferentiated carcinoma and adenocarcinoma, not with small cell anaplastic carcinoma. The subtype of tumors was most accurately determined by examination of Papanicolaou-stained slides. Reactive bronchial epithelium often mimicked carcinoma, but could be correctly identified by its characteristic cytomorphology. No false-positive diagnoses of lung cancer occurred in 386 patients. The sensitivity of bronchoalveolar lavage for the diagnosis of lung carcinoma is similar to that of transbronchial biopsy and Wang needle biopsy. Because bronchoalveolar lavage may detect opportunistic infections, interstitial lung diseases and malignant cells with a low morbidity, it is a useful tool to assess patients with pulmonary infiltrates.