Objective: To compare the lumbar infusion test and the cerebrospinal fluid (CSF) tap test for predicting the outcome of shunt surgery in patients with suspected normal pressure hydrocephalus. Methods: 68 patients with suspected normal pressure hydrocephalus were studied. The absence of preceding history indicated idiopathic disease in 75% of these. All patients were assessed twice with walking and psychometric tests before lumbar infusion test and tap test assessments. The lumbar infusion test was done using a constant infusion rate (0.80 ml/min) and regarded as positive if the steady state CSF plateau pressure reached levels of > 22 mm Hg (resistance to outflow > 14 mm Hg/ml/min). The tap test was regarded as positive if two or more of four different test items improved after CSF removal. As the variability in baseline test results was large, the better of two evaluations was used in comparisons with the results after CSF removal, as well as to evaluate the outcome after shunt surgery. Only patients with a positive lumbar infusion test or a positive tap test had surgery. Results: The results of the CSF tap test and the lumbar infusion test agreed in only 45% of the patients. Of the total cohort, 47 (69%) had positive test results and were operated on; 45 (96%) of these reported subjective improvement, and postoperative assessments verified the improvements in 38 (81%). Improvements were highly significant in walking, memory, and reaction time tests (p < 0.001). Most of the patients improved by surgery (84%) were selected by a positive lumbar infusion test, and only 42% by a positive tap test. Positive predictive values were 80% for lumbar infusion test and 94% for tap test. The false negative predictions in the operated group were much higher (58%) with the tap test than with the lumbar infusion test (16%). Conclusions: Both the lumbar infusion test and the tap test can predict a positive outcome of shunt operations in unselected patients with suspected normal pressure hydrocephalus. The two tests are complementary and should be used together for optimal patient selection.