The influence of a subcutaneous fat layer on measurement of muscle oxygenation using near-IR spectroscopy was quantitatively investigated by two-layered phantom experiments and Monte Carlo simulations, with the aim of developing an algorithm that can correct this influence. The phantom consisted of a fat-like layer, which was a mixture of agar and titanium dioxide powder, and a muscle-like layer, which was suspension of washed bovine blood in Intralipid solution. An LED with 760 and 840 nm elements was used as an optical source, and the backscattered light was detected by photodiodes at source-detector distances of 20, 30 and 40 mm. The relationships between changes in optical density and blood concentrations were obtained at fat-like layer thicknesses of 0.5,10 and 15 mm under fully oxygenated and fully deoxygenated states. It was experimentally found that the change in optical density is significantly decreased and the linearity of measurement characteristics is clearly distorted by the presence of a fat layer. In the simulations, normalized light reflectance and mean optical pathlength in a muscle layer were calculated. The simulation results of the light reflectance agreed well with the experimental results. When the absorption in a muscle layer was relatively high, the mean optical pathlength in the muscle layer, or the measurement sensitivity, was not so dependent on the absorption. Therefore, the modified Beer-Lambert law can still be applied to estimate changes in muscle absorption from changes in optical density, even when a fat layer is involved. The results of simulation also suggested that the influence of a fat layer can be eliminated by correcting the measurement sensitivity using the fat layer thickness.