We conducted a prospective study to determine the effectiveness of protected bronchoalveolar lavage (PBAL) in diagnosing pneumonia in ventilated patients and the usefulness of bronchoscopic data in treating ventilated patients. Entrance criteria were (1) fever and a new or progressive infiltrate on chest roentgenogram with either leukocytosis or a macroscopically purulent tracheal aspirate, and (2) no antibiotic therapy for at least 48 h before bronchoscopy. Twenty-five ventilated patients met entrance criteria for the study and completed the protocol. PBAL was effective in retrieving distal airway secretions with a minimal degree of contamination as indicated by a specificity and a negative predictive value of 100 percent. Bacterial isolates grew in all patients with pneumonia at a concentration greater than or equal to 100,000 cfu/ml, with a median growth of 500,000 cfu/ml. The presence of a two-log difference between the highest quantitative culture count in patients without pneumonia and the lowest quantitative culture count in patients with pneumonia allowed a clearer determination of a patient's status, with regard to pneumonia, compared with the significant overlap in unprotected BAL. Gram and Giemsa stains of the PBAL were positive in all patients with pneumonia and negative in those without pneumonia. All but one patient with pneumonia received narrow-spectrum antibiotic therapy. All patients without infection had no antibiotic administered. Clinical and roentgenographic criteria could not discriminate between patients with and without pneumonia, confirming the findings of previous investigations. The results of microscopic and culture analyses of the PBAL effluent proved useful in directing antibiotic treatment in patients with pneumonia and in avoiding unnecessary antibiotic use in those patients without pneumonia.