A retrospective study of 207 febrile episodes by infants less than 3 months old was performed. Most infants had rectal temperatures ranging from 38 to 39.9 degrees C and only 6 (3%) infants had fever above 40 degrees C per rectum. One hundred and nine (53%) of 207 visits resulted in subsequent hospitalizations including most by infants under 1 month of age. Urinary tract infection was the most common bacterial infection and upper respiratory tract infection the most common viral infection observed. A total of 112 infants were assigned into high-risk or low-risk group for serious bacterial infection by the application of the criteria proposed by Dagan et al. in 1985. The overall incidence of serious bacterial infection was 20.5% (23 patients). Seven (9.2%) of the 76 infants in the low-risk group had serious bacterial infection, compared with 16 (44.4%) of the 36 in the high-risk group (p < 0.05). The negative predictive value of meeting the criteria was 90.8%, a value unacceptably lower than those of previous studies. For maximizing the negative predictive power of the criteria, the addition of another low-risk factor, a normal stool analysis, is suggested. Besides, urine cultures should be taken concomitantly with urinalyses in febrile infants, when considering nonpyuria in infants with urinary tract infections. We demonstrate the pitfalls and difficulties in the application of the criteria for excluding serious bacterial infection in febrile young infants and suggest prospective studies involving more patients should be undertaken before its wider application in clinical practice.