Recent investigations have indicated that different solvation methods give qualitatively different results for the nonpolar solvation contribution to ligand-binding affinities. Therefore, we have calculated the nonpolar solvation contribution to the free energy of benzene binding to the T4 lysozyme Leu99Ala mutant using thermodynamic integration (TI) and three approximate methods. The total binding free energy was calculated with TI and then decomposed into contributions from the solvent and the solute. The nonpolar contribution from the solute was compared to approximate methods within the framework of the molecular-mechanics and generalized Born with surface area method (MM/GBSA). First, the nonpolar solvation energy was calculated with a linear relation to the solvent-accessible surface area (SASA). Second, a recent approach that divides the nonpolar solvation energy into cavity and dispersion parts was used, and third, the nonpolar solvation energy was calculated with the polarized continuum model (PCM). Surprisingly, the simple SASA estimate reproduces the TI results best. However, the reason for this is that all continuum methods assume that the benzene cavity is filled with water for the free protein, contrary to both experimental and simulation results. We present a method to avoid this assumption and then, PCM provides results that are closest to the results obtained with TI.