A technique is described that provides information about relative cerebral responses to differing neurobehavioral tasks in normal subjects studied with positron computed tomography and oxygen-15-labeled water. Simulation studies demonstrate that this technique is sensitive to changes in true local CBF within a physiological range and tends to underestimate relative flow changes at high flow values (greater than 30 ml min-1 100 g-1) and to overestimate these changes for flow values of less than 25 ml min-1 100 g-1. Image acquisition times of 60 s following the arrival of oxygen-15-labeled water in the brain were the most accurate for identifying such relative changes between radioisotope administrations and were not limited by statistical noise from total image counts. Studies in normal volunteers indicate that the technique is highly reproducible, demonstrating a coefficient of variation for small (less than 2 cm2) regions of 2.98 between studies in the same state. Visual stimulation studies in normal volunteers demonstrated relative radioisotope concentration changes between control and stimulated states that are in good agreement with similar results obtained using the same stimulation paradigm but with the use of fluorodeoxyglucose to determine cerebral glucose metabolism.