Recent advances in laboratory technology have enabled us to measure C-reactive protein with a higher sensitivity in a short period using a minimal amount of blood. Thus C-reactive protein can be measured easily several times a day. In this study, serial changes in C-reactive protein values were evaluated in 108 term and 240 preterm newborn infants with suspicion of infection, and the changing patterns of C-reactive protein values were compared with clinical outcome. For a diagnosis of infection, the negative predictive values in term and preterm infants were 99.0% and 97.8%, respectively, although the sensitivities were 61.5% and 75.0%, respectively. Antibiotic therapy was started at birth and discontinued when the changing pattern of C-reactive protein and clinical findings did not suggest infection. As a result, mean durations of administration of antibiotics in the term and preterm infants were 3 and 4 days, respectively. Recognition of the changing pattern of C-reactive protein was very useful in excluding infection and minimizing unnecessary antibiotic therapy in managing neonatal infection.