Using both experimental and theoretical methods, we examine the contribution of different parts of the head to near-IR (NIR) signal. Time-resolved spectroscopy is employed to measure the mean optical path length (PL), and the absorption (mu(a)) and reduced scattering (mu(s)') coefficients in multiple positions of the human head. Monte Carlo simulations are performed on four-layered head models based on an individual magnetic resonance imaging (MRI) scan to determine mu(a) and mu(s)' in each layer of the head by solving inverse problems, and to estimate the partial path length in the brain (p-PL) and the spatial sensitivity to regions in the brain at the source-detector separation of 30 mm. The PL is closely related to the thickness of the scalp, but not to that of other layers of the head. The p-PL is negatively related to the PL and its contribution ratio to the PL is 5 to 22% when the differential path length factor is 6. Most of the signal attributed to the brain comes from the upper 1 to 2 mm of the cortical surface. These results indicate that the NIR signal is very sensitive to hemodynamic changes associated with functional brain activation in the case that changes in the extracerebral tissue are ignorable.