The diagnosis of small peripheral lung cancer is difficult to achieve by non-invasive methods. We hypothesized that in these patients induced sputum might ncrease the diagnostic yield over spontaneous sputum, representing a good diagnostic alternative in selected patients. We prospectively evaluated 60 patients with peripheral lung lesions and normal bronchoscopic evaluation. Six samples of sputum (three spontaneous and three induced with nebulization of hypertonic saline) before bronchoscopy and six samples of sputum after bronchoscopy (three spontaneous and three induced) were obtained in each subject. Forty-two out of the 60 patients included were finally diagnosed with lung cancer. Eighteen patients were diagnosed with different benign conditions of the lung. Overall, malignant cells in sputum were observed in 21 patients and in all but one, the final diagnosis of lung cancer was achieved. Only one patient with a pseudoinflammatory tumour of the lung had a false-positive result in one spontaneous sputum sample. The diagnosis of lung cancer was obtained in 18 patients with the induced sputum (43%) and in 14 patients with spontaneous sputum (31%) (P=NS). Samples of induced sputum were more adequate for cytological analysis than samples of spontaneous sputum (P < 0.001). Of 13 patients with peripheral lung neoplasms of 2 cm or less in diameter, five were diagnosed using induced sputum (38%) and only one using spontaneous sputum (8%) (P<0.05). In conclusion, induced sputum is a valuable technique for the diagnosis of peripheral lung cancer. Induced sputum gives better quality specimens and better diagnostic yield in small lesions than the spontaneous sputum and may be indicated in selected patients with disseminated disease, inoperability or severe co-morbities.