Whether serum C-reactive protein (CRP) can be used to distinguish bacterial from viral pneumonia was studied in 193 paediatric patients who were identified in a prospective, population-based study. The proportion of patients < 5 y of age was 51%, 53% of these and 12% of the older patients were treated in hospital. Pneumococcal aetiology of infection was studied in paired sera by antibody and immune-complex assays, and chlamydial, mycoplasmal and viral aetiologies by routine antibody assays. CRP concentration was measured by immunoturbidometry. Pneumococcal infection (mixed infections with other agents included) was present in 57 cases, mycoplasmal and/or chlamydial infection (pneumococcal infections excluded) in 43, and viral infection (pneumococcal, mycoplasmal and chlamydial infections excluded) in 29 cases. The mean CRP concentrations (95% confidence interval) in these groups were 26.8 mg/l (20.1-33.5 mg/l), 31.8 mg/l (20.5-33.1 mg/l) and 26.1 mg/l (19.1-33.1 mg/l), respectively, and 24.9 mg/l (18.8-31.0 mg/l) in patients with no aetiological findings. When CRP values were compared between the 2 diagnostic groups of pneumococcal infections (antibody and immune-complex positive) no difference was found. In infants < 12 months of age the mean CRP concentration was 14.6 mg/l, and in 11 (65%) of them it was unmeasurable (< 10 mg/l). No significant differences were seen between hospitalized patients and outpatients. In conclusion, CRP concentration had no significant association with the microbial aetiology of pneumonia.