The effect of the probe arrangement on the reproducibility of topographic images of the concentration changes in oxygenated hemoglobin and deoxygenated hemoglobin is evaluated by a virtual head phantom. A virtual head phantom consists of five types of tissue the 3D structure of which is based on a magnetic resonance imaging scan of an adult head. Localized and broadened brain activation is assumed in a virtual head phantom. The topographic images are obtained from the reflectance detected by the standard probe arrangement and the double-density probe arrangement. The uneven thickness of the superficial layer, which cannot be evaluated by the previous slab model, affects the distribution of measured activation in the topographic image, and this reduces the position reproducibility of near-infrared (NIR) topography with the standard probe arrangement. The overlapping measurements by the double-density probe arrangement can improve the reproducibility of the image obtained by NIR topography.