In order to address the problem of three-body interactions in gas–surface scattering, we considered the collision of a He atom with the (0001) surface of graphite coated by a monolayer of Xe. To eliminate the uncertainties connected with errors in the two-body He–Xe interaction, we determined the latter by crossed-beam differential collision cross-section measurements performed at two energies (67.2 and 22.35 meV). These scattering data together with room-temperature bulk diffusion data are then fitted with a Hartree–Fock–dispersion–type function to yield an interaction potential that explains most of the properties of this system within the experimental errors and represents an improvement on previously published He–Xe potentials. Helium diffraction measurements are then carried out from the Xe overlayer and the dependence of the specular intensity from the angle of incidence is carefully determined. Further, a He–surface potential is constructed by adding together the following terms: (1) the He–Xe pairwise sum, (2) the long-range He–(0001)C interaction, (3) the three-body contribution generated by the Axilrod–Teller–Muto term, (4) the so-called surface-mediated three-body interaction He–Xe–(0001)C first considered by A. D. McLachlan [Mol. Phys. 7, 381 (1964)], and finally (5) a small correction which is meant to take into account the nonstationary nature of the surface. Using this potential, well-converged close-coupling scattering calculations are carried out, and their results compared with the data. In general, good agreement is obtained. The agreement can, however, be improved by (a) an increase of about 30% in the contribution of three-body forces, (b) the lowering of the He–graphite long-range attraction coefficient by about 15%, or (c) a reduction of the two-body interaction well depth of 1.6% (the experimental error) together with any combination of the factors under (a) and (b) reduced by an adequate amount. Elimination of the contribution of the graphite surface by studying Xe multilayers is hindered by the uncertainties in the ‘‘thermal correction’’ [point (5) above] which, due to the multilayer increased ‘‘softness,’’ becomes an appreciable source of uncertainty.