Functional near-infrared spectroscopy (fNIRS) signals originate in hemoglobin changes in both the superficial layer of the head and the brain. Under the assumption that the changes in the blood flow in the scalp are spatially homogeneous in the region of interest, a variety of methods for reducing the superficial signals has been proposed. To clarify the spatial distributions of the superficial signals, the superficial signals from the forehead during a verbal-fluency task were investigated by using ten source–detector pairs separated by 5 mm, whereas fNIRS signals were also detected from two source–detector pairs separated by 30 mm. The fNIRS signals strongly correlated with the superficial signals at some channels on the forehead. Hierarchical cluster analysis was performed on the temporal cross-correlation coefficients for two channels of both the NIRS signals, and the analysis results demonstrate spatially heterogeneous distributions and network structures of the superficial signals from within the forehead. The results also show that the assumption stated above is invalid for homogeneous superficial signals from any region of interest of 15-mm diameter or larger on the forehead. They also suggest that the spatially heterogeneous distributions may be attributable to vascular networks, including supraorbital, supratrochlear, and superficial temporal vessels.