The mean cerebral blood flow (CBF) has generally been demonstrated to be lower in normal pressure hydrocephalus (NPH) than in normal controls. We investigated the distribution of the regional peri- and paraventricular white matter CBF (WM CBF) in NPH at baseline and during a controlled rise in intracranial pressure (ICP). Twelve patients with idiopathic NPH (mean age 69 years) underwent a CSF infusion study. CBF was measured by H2(15)O PET at baseline and then during the steady-state plateau of raised ICP. The PET images were co-registered and resliced to 3D structural T1-weighted MRIs. Ten healthy normal volunteers served as control subjects for baseline CBF determination only. Profiles of the regional distribution of the baseline WM CBF and of the percentage change in WM CBF as a function of distance from the ventricles were plotted. The global mean baseline CBF in patients (28.4 +/- 5.2 ml/100 ml/min) was lower than in the control subjects (33 +/- 5.4 ml/100 ml/min) (P < 0.005). In patients, the profile of the regional WM CBF at baseline showed an increase with distance from the ventricles (P < 0.0001), with a maximal reduction adjacent to the ventricles and progressive normalization with distance, whereas in controls no relationship was apparent (P = 0.0748). In 10 patients, the rise in ICP during the infusion produced a fall in cerebral perfusion pressure (CPP) and a significant decrease of the global mean CBF from 27.6 +/- 3.1 to 24.5 +/- 2.9 ml/100 ml/min (P < 0.0001). The profile of the percentage changes in regional WM CBF in patients showed a U-shaped relationship with distance from the ventricles (P = 0.0007), with a maximal decrease skewed on the side of the lateral ventricles at around a mean distance of 9 mm. The WM CBF is reduced in NPH, with an abnormal gradient from the lateral ventricles towards the subcortical WM. An excessive decrease in CBF is brought about by reductions in CPP and appears to be maximal in the paraventricular watershed region. These results are discussed in the light of previous hypotheses concerning the aetiology of periventricular CBF reduction in NPH.