To investigate the role of serum C-reactive protein (CRP) in the diagnosis of typhoid fever, we studied 227 febrile Malaysian children hospitalized during a 12-month period. The children were: culture-positive for Salmonella typhi (Group 1; n = 108); culture-negative but with typical clinical features of typhoid fever (Group 2; n = 60); or had non-typhoidal illness (Group 3; n = 59). Group 1 children had the highest serum CRP concentrations (geometric mean [SD range]; 43 [12-150] mg/l vs. 26 [8-85] mg/l in Group 2 and 21 [4-110] mg/l in Group 3;p <0.001). In regression analysis, age, patient group and fever duration were independently associated with serum CRP (p <0.05) but gender was not. In Group I patients, there was a significant positive association between serum CRP and Widal O and H agglutinin titres. In receiver-operator characteristic (ROC) analysis of serum CRP for Groups 1 and 2 combined, compared with Group 3, the area under the curve (AUC) was 0.65. These data show that the serum CRP is highest in culture-positive children with enteric fever and reflects the immune response to the infection in this group. Nevertheless, serum CRP had relatively low sensitivity and specificity for confirmed or clinically diagnosed typhoid fever (68 and 58 per cent, respectively at 'cut-off' concentration 30.0 mg/l), and an AUC value only moderately above that associated with no predictive power (0.5). Although of limited use as a primary diagnostic test, a raised serum CRP may still have a place as one of a range of features that facilitate assessment of a febrile child in a typhoid-endemic area.