The optimal management strategy for ventilator-dependent patients who develop symptoms suggestive of lung infection remains controversial. Proponents of the empirical approach prefer to treat most patients with fever and pulmonary infiltrates with one or more new antibiotics, even if it may be difficult (1) to determine whether pneumonia has developed in such patients, (2) in case of infection, to precisely identify the responsible microorganisms and thereby select the optimal antimicrobial treatment, and (3) to avoid resorting to broad-spectrum drug coverage in patients without true infection. Our personal bias is that using bronchoscopic techniques to obtain protected specimen brush and bronchoalveolar lavage specimens from the affected area in the lung permits to devise a therapeutic strategy superior to the one based only on clinical evaluation. These bronchoscopic techniques, when they are performed before new antibiotics are administered, enable physicians to identify most patients who need immediate treatment and select optimal therapy, in a manner that is safe and well tolerated by patients. Furthermore, they frequently permit the clinician to withhold antimicrobial treatment in patients without infection, minimizing the risk of the emergence of resistant microorganisms in the intensive care unit. In patients with clinical evidence of severe sepsis, the initiation of antibiotic therapy should not, however, be delayed while awaiting bronchoscopy, and patients should be given immediate treatment with antibiotics. In that case, "simplified" non-bronchoscopic diagnostic procedures might allow obtaining reliable distal pulmonary secretions for quantitative cultures on a 24-hour basis just before the initiation of a new antimicrobial therapy.