C-reactive protein (CRP) is an unreliable diagnostic tool in the early diagnosis of neonatal septicaemia. However, serial measurements have been shown to be useful in monitoring the effectiveness of treatment. The aim of the present study was to investigate whether a specific CRP response pattern to different groups of pathogens could be identified during treatment of neonatal septicaemia. Serial CRP measurements from day 1 to 4 in monomicrobial blood culture-proven episodes of septicaemia were reviewed. In 4416 admissions, 180 out of 206 positive blood cultures were monomicrobial; 121 monomicrobial septic episodes were eligible for final analysis of the CRP response during treatment. A low median (M) value (day 1 to 4) was identified in coagulase-negative staphylococci (CONS) (M=23 mg/l), contrasting with high median values in Staphylococcus aureus (M=58 mg/l), group B streptococci (M=51 mg/l), Escherichia coli (M=51 mg/l) and Candida species (M=76 mg/l) (p<0.001). Median CRP values in the two groups were different for each of the treatment days 1 to 4 (p<0.001). An increase (p<0.001) in CRP during the 24 h before initiation of treatment was a sign of late-onset CONS septicaemia. In episodes where antimicrobial treatment failed, CRP levels were moderately elevated the day prior to treatment start and increased continuously thereafter, whereas successful treatment was generally accompanied by a decline in CRP in less than 4 days. The CRP response to CONS is significantly less pronounced than to other commonly encountered pathogens in neonatal septicaemia. A rise in CRP beyond the third day of empirical treatment should give rise to a suspicion of fungal infection or ineffective antibacterial treatment.