Although it is possible to accurately predict angle-resolved scattering from surfaces covered with multilayer dielectric films and some of the scattering from metal-coated surfaces, there are still areas of disagreement between theory and experiment. In particular, the ratio of p- to s-polarized scattered light at a given scattering angle should be a function only of the optical thicknesses and optical constants of the surface layers (for dielectric films) or optical constants (for opaque metal films) , and the angles of incidence and scattering; it should be independent of surface topography. It was found experimentally that the p/s polarization ratio for 60-degree incidence, 60-degree retroscattering at λ6328 Å varied widely from the theoretical values for silver- and aluminum-coated glass samples and for polished and diamond-turned copper samples. In addition, theory predicts no cross-polarized scattered light (p-incident/s-scattered or s-incident/p-scattered), but in some cases large amounts were measured experimentally. It was found that the polarization ratio could be increased to approach the theoretical value by heating the films in vacuum, with consequent recrystallization and roughening of the films. It was hypothesized that some of the measured effects could be caused by fluctuations in the optical constants in the plane of the film; a first-order vector perturbation theory was derived to cover this situation. Calculations made using the theoretical equations look encouraging, but the theory does not predict any cross-polarized scattered light. More theoretical work needs to be done and, in addition, some structural characterization of the films.