Fibreoptic bronchoscopy was performed in 43 of 52 consecutive patients with opportunistic pneumonia in the acquired immune deficiency syndrome (AIDS). The 15 patients in whom a likely pathogen was not found by bronchoscopy (including the nine not having the procedure) were treated for Pneumocystis carinii pneumonia (PCP) alone and all responded. In 11 of these a diagnosis of AIDS was confirmed because of an alternative opportunistic infection within 6 months. PCP was confirmed in 38 of the 52 patients and cytomegalovirus (CMV) was isolated from 15 patients. The lower the partial pressure of arterial oxygen (PaO2) on admission the more likely was a pathogen to be found by bronchoscopy. The admission PaO2 while the patient was breathing room air was the single most reliable prognostic indicator. The mean PaO2 for survivors was 9.6 kPa and 6.7 kPa for non-survivors (P less than 0.01 Student's t-test), with 50% mortality for patients with a PaO2 of less than 8 kPa on admission. Temperature and pulse rate were sensitive indicators of response to treatment, obviating the need for frequent arterial gas measurements and chest radiography. Our findings suggest that although fibreoptic bronchoscopy contributed little to the treatment and final outcome of the infection, it identified the causative pathogen in most patients.