Abstract The problem of applying laboratory silicate-flow data to the mantle, where conditions can be vastly different, is approached through a critical review of high-temperature flow mechanisms in ceramics and their relation to empirical flow laws. The intimate association of solid-state diffusion and high-temperature creep in pure metals is found to apply to ceramics as well. It is shown that in ceramics of moderate grain size, compared on the basis of self-diffusivity and elastic modulus, normalized creep rates compare remarkably well. This comparison is paralleled by the near universal occurrence of similar creep-induced structures, and it is thought that the derived empirical flow laws can be associated with dislocation creep. Creep data in fine-grained ceramics, on the other hand, are found to compare poorly with theories involving the stress-directed diffusion of point defects and have not been successfully correlated by self-diffusion rates. We conclude that these fine-grained materials creep primarily by a quasi-viscous grain-boundary sliding mechanism which is unlikely to predominate in the earth's deep interior. Creep predictions for the mantle reveal that under most conditions the empirical dislocation creep behavior predominates over the mechanisms involving the stress-directed diffusion of point defects. The probable role of polymorphic transformations in the transition zone is also discussed.