A prospective study of 250 consecutive neonatal admissions to a regional perinatal referral centre and of 10 additional consecutive cases with culture-proven neonatal septicaemia was undertaken. Quantitative C-reactive protein (CRP) determination, white cell count and differential were performed on blood samples obtained from all babies on admission, as well as 10-14 h and 22-26 h later. Using clinical signs, chest X-rays, blood cultures, tracheal aspirates obtained within 4 h of delivery and an abnormal immature/total neutrophil ratio (I/T), infected babies were defined as belonging to one of the following groups: culture-proven septicaemia (n = 19); clinical septicaemia (n = 35); congenital pneumonia (n = 28). The sensitivity, specificity, positive and negative predictive value of CRP were calculated for each sampling time and patient group. No baby had a rise in CRP (greater than 6 mg/l) before an abnormal I/T ratio was first detected. A delayed rise in CRP concentration in the majority of infected babies occurred approximately 12-24 h after the abnormal I/T ratio was first detected. The overall specificity of a CRP level of greater than or equal to 10 mg/l remained approximately constant (97%-94%) while sensitivity increased from 22%-61% with increasing time after admission. The same pattern emerged if each patient group was considered separately. The positive predictive value for a CRP level of greater than or equal to 10 mg/l 22-26 h after admission was 83% and the negative predictive value 82%. CRP had no value in the early diagnosis of neonatal infection. Its main role lies rather in the exclusion or confirmation of infection 24 h after the first clinical suspicion.