Workable criteria are proposed that allow recognition of artifacts in the constant infusion manometric test for the detection of cerebrospinal fluid absorptive defects, thus improving the reliability of the test. A retrospective analysis was made of 47 infusion tests; when artifactual infusion curves were eliminated, the analysis showed excellent correlation between abnormal infusion data and ventricular reflux of greater than 48 hours on cisternography. Less satisfactory correlation was found with present pneumoencephalographic criteria. Patients with artifactual tests had recent lumbar punctures more often than those in the satisfactory group, indicating that a higher success rate may be achieved by performing the infusion as the first lumbar puncture or delaying it at least a week after prior puncture.