Depending on details of the composite microstructure, different theories may be needed to obtain good agreement with measured elastic properties. This observation is especially pertinent whenever the composite is porous, as is normally true for rocks. Predictions of three theories are compared to data for porous glass samples. The differential effective medium (DEM) theory and Hashin's composite spheres assemblage (H) do a good job of predicting elastic behavior of a porous foam composed of glass. The self-consistent (SC) effective medium theory does equally well at predicting behavior of a sintered glass-bead sample. The realizable microstructure of each theoretical model is a good analog of the microstructure for one or the other of these two very different porous glasses. Velocities of granular rocks such as sandstones may be estimated accurately using the SC theory, whereas velocities of rocks such as basalts having isolated cracks and pores may be better estimated using either the DEM theory of Hashin's model. ¿ American Geophysical Union 1993