NIRS has been widely utilized for monitoring oxygen concentration of cerebral blood flow (CBF). However, meanings of signals measured by NIRS still have many unclear points. One of the factors is that the physiological mechanism of coupling between neuronal activity, metabolism and CBF is not clarified enough. In this study, we evaluate NIRS data based upon numerical simulation of oxygen transport to cerebral tissue. With a 2-dimensional mathematical model of oxygen transport from an arteriole to its surrounding tissue, we simulate the activity-dependent oxygenation changes. On the basis of calculated oxygen tension distribution, we derive quantities of two kinds of hemoglobin in the arteriole by using the oxygen dissociation curve, and theoretically decompose each hemoglobin change into its factors. This decomposition has revealed that NIRS data can reflect two types of physiological phenomena: a qualitative change caused by oxygen dissociation and a quantitative change caused by an increase of CBF. These results indicate that cellular oxygen consumption can be reflected more in the time courses of deoxygenated hemoglobin than those of oxygenated hemoglobin. It will be desirable to focus not only on oxygenated hemoglobin but also on deoxygenated hemoglobin when conducting evaluation of a brain function.