Sixty-one children were treated in hospital from 1981 to 1982 because of both radiologically and microbiologically verified viral or bacterial pneumonia. The chest radiographs were interpreted by two radiologists, not familiar with the clinical data, on two occasions three years apart, and only those patients with a definite alveolar (n = 27) or interstitial (n = 34) pneumonia at both evaluations were included in the present analysis. In addition, all patients had viral (n = 20), mixed viral-bacterial (n = 21) or bacterial (n = 20) infections diagnosed by viral or bacterial antibody or antigen assays. Viral infection alone was seen in 7 (26%), mixed viral-bacterial infection in 8 (30%) and bacterial infection alone in 12 (44%) of the 27 patients with alveolar pneumonia. The respective figures were 13 (38%), 13 (38%) and 8 (24%) for the 34 patients with interstitial pneumonia. C-reactive protein concentration was greater than 40 mg/l (a screening limit for viral and bacterial infections) in 15 (56%) of the patients with alveolar and in 11 (32%) of the patients with interstitial pneumonia. Thus 74% of the patients with alveolar and 62% with interstitial pneumonia had bacterial infection, either alone or as a mixed viral-bacterial infection. Our results suggest that the presence of an alveolar infiltrate in a chest radiograph is a specific but insensitive indicator of bacterial pneumonia. We conclude that patients with alveolar pneumonia should be treated with antibiotics. In patients with interstitial pneumonia, however, both viral and bacterial aetiology are possible. In those, the decision concerning antibiotic treatment should be based on clinical and laboratory findings.