Determination of the optical pathlength of light in tissue is important to quantitate NIRS data. However, the inhomogeneity of the illuminated tissues increases the difficulty of determining the relevant optical pathlength in the tissue. For instance, in the head, the contribution of the tissues overlying the brain to the total optical pathlength cannot be ignored in the monitoring of cerebral oxygenation with NIRS. In this study, time-of-flight measurements of an inhomogeneous phantom are carried out in the laboratory to examine the contribution of the overlying tissue to the optical pathlength. The phantom consists of two homogeneous components, the boundaries of which are two concentric cylinders. The TPSF is measured with a picosecond laser and a streak camera, and the change of TPSF with the distance between source and detection fibers is examined. The experimental TPSF and mean time of flight are compared with the results of a Monto Carlo simulation and a finite element model based on the diffusion equation. A comparison of the accuracy of prediction of the pathlength by each model is presented as a function of the spacing between source and detection fibers. The intensity photon measurement density functions in each of the cylinders were estimated from the Monte Carlo simulations. The results provide estimates for the amount of the NIRS signal arising from overlying tissues in the head.