The need for fiberoptic bronchoscopy in the patient with hemoptysis and a normal or nonlocalizing chest roentgenogram remains a subject of debate. Currently, diagnostic fiberoptic bronchoscopy is recommended as the investigative procedure of choice. To develop predictors that identify the patient in whom fiberoptic bronchoscopy is most likely to be diagnostic, we reviewed our community's experience with this population over a five-year period. We identified 196 patients with hemoptysis and a normal or nonlocalizing chest roentgenogram who underwent fiberoptic bronchoscopy. Three quarters were active or previous smokers. We examined the relationship of advancing age, sex, smoking, nonspecific roentgenographic findings and the amount, duration, and previous bouts of hemoptysis to the incidence of a diagnostic fiberoptic bronchoscopy. Twelve patients (6 percent) had bronchogenic carcinoma and 33 (17 percent) another specific cause for the hemoptysis identified by fiberoptic bronchoscopy. By univariate and discriminant analyses, we found that the three factors of age of 50 years or more, male sex, and smoking of 40 pack-years or more best predicted a diagnosis of malignancy. Bleeding in excess of 30 ml daily was associated with an increase in overall diagnostic yield. The presence of two of the three factors associated with malignancy or bleeding in excess of 30 ml daily (or both) identified 100 percent of the patients with bronchogenic carcinoma and 82 percent of all of the diagnostic fiberoptic bronchoscopic procedures. use of these criteria in selecting the patient for fiberoptic bronchoscopy could have reduced our use of the bronchoscope by 28 percent, with the remaining patients safely observed.