The Rayleigh solution for reflection from a corrugated surface, and some possible modifications, are examined. For small slopes these methods produce similar analytical expressions and almost identical numerical results. When the first nonradiating term is included in the computations, the magnitude of the scattered amplitudes is essentially the same, but an estimate of the phase change as a function of frequency is provided. For larger slopes, differences appear in the analytical form as well as in the numerical results. The Rayleigh expression has advantages above cutoff, while a new approach seems better before and just after cutoff.