A prospective study was undertaken of all children referred to the Hospital for Sick Children with a provisional diagnosis of shunt blockage over a 5-month period. Fifty-two admissions were recorded, relating to 45 children, 5 of whom had multiple admissions. Only 19 of the 52 admissions led to a final diagnosis of shunt malfunction. No source of referral, whether by the child's general practitioner or from another hospital, was found to be more accurate than direct referral by the parents to the neurosurgical ward. Headache, vomiting and irritability were not significant indicators as to whether the child's shunt was actually blocked, and nor was the duration of the symptoms. Drowsiness was a significant, but not definite, indicator of shunt blockage, while pyrexia made it more likely that the patient had an alternative diagnosis. In 35 of the admissions a computed tomographic scan was performed: a normal scan, unchanged from previous scans, did not reliably exclude the diagnosis of shunt blockage. Percutaneous manometry via the reservoir of the shunt system was performed during 26 admissions: this investigation produced no false positives nor false negatives, but was equivocal in 5 cases, all of which were found at surgery to have a definite shunt blockage. The accuracy of the diagnosis of shunt blockage made prior to referral to a neurosurgical unit is discussed, together with the implications for resource use.