Numerical calculations, based on a plausible capillary-wave model and assuming the existence of a local surface of well-defined slope in the neighbourhood of an evaporating molecule, do not yield acceptable distributions of velocity or angle for molecules leaving the surface. Calculated correlation functions, for the slopes of the surface at two points a short distance apart, fall to zero over distances which are comparable with or less than the dimensions of solvent molecules, so the assumption of a surface of well-defined slope is incorrect. At high energies, where surface motions are effectively frozen, this assumption remains valid.