We measured the rate of washout of 15O-labeled water generated from labeled oxygen accumulated in brain after bolus [15O]O2 inhalation, and compared the washout with that of labeled water measured with H215O. Contrary to the original expectation, the radioactive water generated from labeled oxygen failed to leave the brain tissue at the rate predicted by exogenous water. Therefore, the use of a separately measured value for exogenous water clearance led to an error in the calculation of oxygen consumption. A new method presented in this paper eliminated the error by yielding oxygen consumption in a single oxygen study. We used time-weighted integration to estimate three parameters, including the unidirectional clearance from blood to brain (KO2(1)), the fractional clearance of the distribution volume in brain (kO2(2)), and the vascular volume correction (VO). We showed that the clearance of oxygen from blood to brain can be estimated with acceptable precision by this new approach, and that the new method yields a reliable measure of oxygen consumption.