A prospective, blinded study of pulmonary findings in hospitalized patients with HIV infection compared auscultatory percussion (AusP) with conventional percussion (ConP) and conventional auscultation (ConA) using chest radiographs as the gold standard. Sixty-three patients had chest radiographs and were examined by one to three examiners. Seventy of the 126 lungs had radiographic abnormalities (55.6 percent). Auscultatory percussion proved to be the most sensitive of all techniques for each examiner (range, 51.0 to 69.6 percent) for detecting radiographic abnormalities and also had higher likelihood ratios for two of the three examiners; AusP also had the highest likelihood ratio pooled across examiners. Of the 166 abnormal results of lung examinations, the combination of AusP and ConA detected 31 more abnormalities than ConP and ConA combined, with 14 of these being diagnosed with Pneumocystis carinii pneumonia. No abnormalities were detected by ConP that were not detected by AusP. These findings suggest that AusP, a rapid clinical maneuver, is more sensitive and specific than ConA and ConP in determining pulmonary abnormalities in HIV-infected inpatients.