Near-infrared spectroscopy (NIRS), which can be used to detect changes in the concentration of oxygenated hemoglobin (oxy-Hb) and deoxygenated hemoglobin (deoxy-Hb) in tissue by using illumination at two different wavelengths, is often applied to noninvasive measurements of human brain functions. It is common to use two wavelengths that are on opposite sides of the point where the optical absorptions of oxy- and deoxy-Hb are equal (about 800 nm) but an optimal wavelength pair has not yet been determined. In this study, we conducted simultaneous recordings at five wavelengths (678, 692, 750, 782, and 830 nm) to determine the best wavelength for pairing with 830 nm. A theory suggests that pairing a shorter wavelength with 830 nm can provide more sensitivity because of the larger difference in absorption coefficients of hemoglobin between two wavelengths. The changes measured in four cortical areas (frontal, occipital, parietal, and temporal) showed that the noise level when the 678-, 692-, and 750-nm wavelengths were paired with 830 nm was usually lower than when the 782-nm wavelength was paired with 830 nm, which is consistent with theoretical prediction. Moreover, the signal-to-noise ratios (S/Ns) and wavelength dependencies of the power detected in all areas and subjects together suggest that the 692-nm pairing had the highest S/N. This suggests that the optimal wavelengths depend on not only the difference in the absorption coefficients of hemoglobin but also on the optical properties in the measurement area, which affect the strength of the attenuation data. The 692-nm wavelength is thus a more optimal choice than wavelengths around 780 nm for pairing with 830 nm to measure Hb changes induced by cortical activation. The improved S/N enables more sensitive statistical analysis, which is essential to functional mapping with NIRS.