We investigated the cause of poor outcome in patients with normal pressure hydrocephalus (NPH) who did not respond as expected after shunt surgery. Two methods were used to evaluate shunts: radionuclide shunt patency study, or continuous ICP monitoring. 33/52 shunted patients (64%) from 1989 to 1995 had poor outcome, and 28/52 (54%) were investigated. Of those investigated, 9/28 (32%) were never better, and 19/28 (68%) were initially better then worse. Of 9 patients who were never better, ineffective shunt function was seen in 7; 5 had shunt revision (2 declined), and 1 improved. Of 19 patients who were initially better then worse, 15 had ineffective shunts; 15 underwent shunt revision, and 13 improved. Poor clinical outcome occurred in two-thirds of all patients after shunt surgery for NPH, but a potentially treatable cause (i.e. obstruction of the shunt or a shunt system that was patent but did not adequately correct the CSF circulatory disorder) was found in nearly 80% (22/28) of those investigated. The predominant cause of ineffective shunt function was obstruction of the peritoneal catheter. Clinical recovery occurred in 70% (14/20) of patients who had shunt revision surgery. We conclude that ineffective shunt function is a frequent cause of poor outcome after shunt surgery to treat NPH that should be sought and treated. These results have implications for longitudinal studies of the diagnosis and treatment of NPH. The effect of unrecognized shunt ineffectiveness on prior studies is unknown. Future studies should be designed to confirm that shunts are functioning before the diagnosis of NPH is considered incorrect.