The hypothesis of S-wave scattering by three-dimensional heterogeneity in the lowermost mantle is shown to provide a plausible alternative to the spherically stratified models recently proposed by Lay and Helmberger to explain certain observed complexities in long period S-wave data. A result of central importance is that, contrary to the conclusions of Lay and Helmberger and, most recently, Young and Lay, sharp Increases of S velocity with depth are not required to explain their data. ¿ American Geophysical Union 1987