In order to investigate the validity of the single compartment model in measuring CBF with the use of 15O-labeled water (H2 15O), dynamic positron emission tomography (PET) was performed following bolus injection of H2 15O. Careful attention was paid to accuracy in the measurement system (especially for the input function). In the region of the putamen, which includes the smallest mixture of gray and white matters in addition to the smallest contamination of cerebrospinal fluid (CSF) spaces, the partition coefficient obtained was 0.88 +/- 0.06 (ml/g). The discrepancy from the prediction estimated from the brain/blood water content ratio was only 7%. This finding suggests that there is no more complicated model than the usual single compartment one to describe the physiological behaviour of 15O water. On the other hand, in the other cortical regions, the discrepancy was larger (e.g., about 12% for the insular cortex and 26% for the frontal cortex) than in the region of the putamen, and a significant fit-interval dependence was observed in the calculated parameters. These observations suggest a significant effect of tissue heterogeneity and/or contamination with nonperfusable spaces in actual clinical PET data.