The relative importance of granivorous rodents, birds and ants in four North American arid ecosystems was examined using two methods: an energetics approach (Population Daily Energy Budgets, or PDEBs); and an experimental seed-dish technique. Rodent, bird and ant population densities, combined with species-specific Daily Energy Budgets (DEBs), were used to compute the PDEBs (Kcal/ha/day). The PDEBs were assumed to be an index of the potential seed consumption ‘pressure’ from each taxocene on the seed pool. Seed consumption at sites in the Chihuahuan Desert/Grassland transition zone, the Sonoran and Mojave Deserts, and the Intermountain Basin Shrub-steppe, was estimated by offering each granivore taxocene exclusive access to seed-dishes containing either a native-seed mix or commercial millet. The ratios of rodent, bird and ant PDEBs were compared with the amounts of seeds actually taken by each taxocene from the seed-dishes. The following conclusions were made. (1) Rodents were more ‘important’ (energetically) than birds on the Sonoran, Mojave and Intermountain Basin study sites, but birds were of equal or greater importance as rodents on the Chihuahuan study sites. (2) Birds removed far fewer seeds from seed-dishes than predicted from their PDEBs. (3) Rodents, birds and ants removed more millet and used more seed-dishes containing millet than those containing the native-seed mix. (4) Seed-dish data should be interpreted with caution, due to differential responses by rodents, birds and ants. (5) An energetics approach to desert granivory can provide reliable information about granivore impacts on seed reserves, but requires large data sets and biologically realistic estimators to produce accurate, high-precision results.