Studies of log body size/log energy use in local animal assemblages reveal a range of distributional patterns from those with a negative upper bound slope (NUBS), showing energy use to be dominated by smaller bodied species, to those with a positive upper bound slope (PUBS), indicating that larger bodied species dominate energy use. We investigated the log body size/log energy use relationships for a local termite assemblage at the Mbalmayo Forest Reserve in Cameroon, and compared the patterns for wood and soil feeding groups. We measured the biomass densities of 75 species of termites across three forest sites. Termite species metabolic rates were derived either directly (25 species), or by regression estimates (50 species), from Warburg manometry measurements of termite xO2 at 28°C. The combined data from the three sites revealed a marked difference in log body size/log energy use distributions between wood and soil feeding groups. While wood feeders showed a polygonal distribution, with a negative upper bound slope (NUBS), soil feeders showed a (statistically very significant) positive distribution. The data for soil feeders are also exceptional, in comparison with the vast majority of other animal assemblage data, in showing significantly greater 'total' energy use for larger, compared with smaller, body size categories. Furthermore, the modal body size category for number of individual termites does not coincide with that for number of termite species. We interpret these results in the context of the likely constraints imposed by the food substrates of termites and the economies of scale. Specifically, soil feeders utilise a low energy, universally abundant substrate compared with wood feeders which feed on a patchily distributed, higher energy resource. Digestive efficiency is likely to be at a premium in soil feeders, driving selection towards larger body size (and hence larger gut volume and transit time), compared with wood feeders.