The calculation of reflection and transmission coefficients of plane waves at a plane interface between two homogeneous anelastic media may become ambiguous because it is not always obvious how to determine the sign of the vertical component of the slowness vector of the scattered waves. For elastic media, the sign is determined by applying socalled radiation condition when the slowness vector is complex-valued, but it has long been known that this approach does not work satisfactorily for anelastic media. Other approaches have been suggested, e.g., by requiring that the reflection and transmission coefficients should vary continuously with increasing incident angles, or by relating the sign to the direction of the energy flux. In the present paper, it is shown that these approaches may give different results, and that the results can be inconsistent with the elastic case even for weak attenuation. Instead, it is demonstrated that the ambiguity in the reflection coefficient can be resolved by expressing the seismic response of a point source over an interface as a superposition of plane waves and their reflection coefficients, and solving the resulting integral by the saddle point approximation. Although the saddle point itself (point of stationary phase) does not provide new insight, the ambiguity is removed by considering the steepest descent path through the point. Ray synthetic seismograms computed by this method compare well with synthetics computed by the reflectivity method, which does not suffer from the above-mentioned ambiguity since the integration path is taken along the real axis. This paper concentrates on the isotropic case, but it is discussed how the result may be extended to layered transversely isotropic media. The suggested approach, derived for a point source and plane layers, does not directly apply to 2-D or 3-D laterally inhomogeneous media, or to media of general anisotropy. A generalization of the result found is that the sign of the vertical slowness components should be chosen according to the energy flux direction for subcritical incidence and according to the radiation condition for supercritical incidence, even if this creates a discontinuity in the coefficients at the critical incidence angle. Such a discontinuity is sometimes necessary to get results which are consistent with the elastic case. It is discussed how the generalized result can be obtained by applying certain continuity criteria for the sub-and supercritical angle intervals, but the validity of this approach for general models remains to be proved.