I present results from elastic finite-difference simulations of regional wave propagation conducted in an effort to characterize, in a statistical sense, the nature of lithospheric heterogeneities required to generate scattered wave fields with characteristics consistent with those observed in regional array data. In particular, regional P, S, and Lg wave trains that are comprised not of the occasional coherent deterministic phase emersed in randomly scattered coda, but of a continuous succession of coherent forward-scattered arrivals. My modeling suggests that lithospheric heterogeneities should be parameterized using spatially anisotropic correlation functions. Models containing spatially isotropic heterogeneities inhibit the extent to which energy is forward scattered and trapped in the crustal wave guide and, consequently, produce regional wave fields whose characteristics are inconsistent with array observations. Models containing spatially anisotropic heterogeneities—which preferentially forward scatter energy that is subsequently trapped in the crustal wave guide—produce wave fields whose characteristics are consistent with regional array observations and provide intuitively appealing representations of subsurface structure.