A large data set of relative food-consumption estimates (Q/B) of marine and freshwater fish populations (n = 108 populations, 38 species) is documented and used to derive a predictive model for Q/B, using asymptotic weight, habitat temperature, a morphological variable and food type as independent variables. Salinity is shown to have no effect on Q/B in fish well adapted to fresh or salt water (other things being equal), while mortality (Z), has a strong, positive effect on Q/B and on gross food-conversion efficiency (defined by GE = Z/(Q/B)), by affecting the ratio of small:large fish. The empirical models thus derived should be useful for parameterization of trophic models of ecosystems and similar applications.