Frequency distributions of mammal body sizes in large-scale assemblages have often been found to show a positive skew. In an attempt to explain this pattern, a model has been put forward which incorporates energetic constraints on fitness and thereby predicts optimal body sizes corresponding to the mode of the distribution. A key assumption of the model is that energy is unlimited. However, if energy is limited, the input of energy into a herbivorous mammal community should influence the shape of the frequency distribution. Thus, I propose that increases in primary productivity will decrease the variation of body size and increase the mean body size in a distribution. So, in low-productivity environments we should see a predominance of small-sized species, but with a great variation of body sizes due to limitations of resources (energy). I tested this hypothesis using the herbivorous mammal fauna (rodents, bats and marsupials) in seven biomes of Australia. Because herbivorous marsupials generally are fairly large-bodied while rodents and bats are small-sized and because marsupials also have a different mode of reproduction from placental mammals, the hypothesis was also tested on placental mammals and marsupials separately. There was no clear mode for the entire assemblage in any biome, but as primary productivity increased, the variation of body masses decreased and the mean body mass of the distribution increased. Body mass distributions of both placental mammals and marsupials displayed clear modes. Placental mammals also showed an increase in mean body mass. The variation in body mass of marsupials was highest for the intermediately productive biomes. Primary productivity does seem to have some effect on mammalian body mass in this case, but the results here need to be complemented with studies of other assemblages before any general conclusions can be drawn. It is also important to distinguish which taxa are affected in a heterogeneous assemblage like the Australian herbivorous mammal fauna.