We present a method for predicting permeability from sonic and density data. The method removes the porosity effect on the ratio upsilon (sub p) /upsilon (sub s) of dry rock, and it addresses the specific surface as an indirect measure of permeability. We look at ultrasonic data, porosity, and the permeability of 114 carbonate core plugs. In doing so, we establish an empirical relationship between the specific surface of the solid phase (as calculated by Kozeny's equation) and upsilon (sub v) /upsilon (sub s) (linearly transformed to remove the porosity effect). One must view the specific surface derived by using Kozeny's equation as an effective specific surface because Kozeny's equation only holds for homogeneous rock with interconnected pores. The ratio upsilon (sub p) /upsilon (sub s) of dry rocks, on the other hand, seems to be controlled by the true specific surface, pointing to an inherent limitation in the method. The 114 carbonate plugs originate in three geological settings and comprise 83 calcitic and 31 dolomitic samples. Their depositional texture varies from mud-dominated to grain-dominated and recrystallized types. Our research applies the relationship to 137 carbonate samples from two different depositional settings. We find a reasonable match between predicted and measured permeability. The match is better for samples with carbonate mud-filled depositional textures than for carbonate mud-poor depositional textures. Diagenetic factors such as vuggy porosity decrease the predictability of permeability.