Fifty-seven children ages 1 month to 12 years hospitalized because of community-acquired pneumonia were compared with age-matched controls who had acute asthma without pneumonia to test the value of rapid bacterial antigen detection and clinical and radiographic criteria for diagnosis of bacterial pneumonia. Bacterial pneumonia, defined on the basis of positive cultures of blood or pleural fluid, was diagnosed in 4 children (7%), 1 of whom also had viral pneumonia. Viral pneumonia, defined as a positive nasopharyngeal sample or positive serology, was diagnosed in 20 children (35%). Serum and concentrated urine were tested by latex agglutination (Wellcogen) for Haemophilus influenzae type b and pneumococcal antigens and by countercurrent immunoelectrophoresis for pneumococcal antigens. Pneumococcal antigen could not be detected in serum or urine from 3 children with culture-proved pneumococcal pneumonia, indicating poor sensitivity of the tests. In contrast apparent H. influenzae type b antigenuria was detected by latex agglutination in 4 of 40 children with pneumonia but also in 5 of 57 controls, and a sensitive enzyme-linked immunosorbent assay for polyribosyl ribitol (PRP) phosphate antigen showed that all 9 cases were false positives. The specificity of H. influenzae type b antigen detection was thus poor. Children with viral and bacterial pneumonia could not be distinguished by radiographic or clinical criteria (symptoms, fever) or by total or differential white blood cell counts, serum C-reactive protein or nasal or serum interferon levels. It is not possible to distinguish reliably childhood viral from bacterial pneumonia clinically or by rapid diagnostic tests.