Near-infrared (NIR) spectroscopy has been used in studies of the cerebral hemodynamic response to visual processing. In this paper, we present theoretical results from finite element and Monte Carlo modeling in order to help understand the contribution to the NIR signal from different parts of the head. The results from the models show that at the typical optode spacings used in these studies, an infrared spectroscopy measurement ofintensityis sensitive to the outer 1–2 mm of the cortical gray matter and the partial optical path length in the gray matter is approximately 10 mm, compared with a total optical path length of 400 mm. When the NIR measurement is ofchange in mean photon arrival time(orphase shift), the signal comes from the upper 2–4 mm of the cortical surface and there is an increased lateral spread of the contributing tissue. We predict that for a 4-cm separation of input and detection optodes at 800 nm, a 1 μM change in hemoglobin concentration in the cortex corresponds to an attenuation change of approximately 0.001 OD (optical density) or 1 ps mean time change. Movement of the brain caused by this increase in volume will cause an absorption change of approximately half this magnitude, but does not affect the photon arrival time at 4-cm spacing. A discrepancy between the predicted and the experimentally measured intensities may support the supposition that the NIR signal is actually very sensitive to changes occurring in the pial cerebral vessels lying on the brain surface.