In 1847, Karl Bergmann proposed that temperature gradients are the key to understanding geographic variation in the body sizes of warm-blooded animals. Yet both the geographic patterns of body-size variation and their underlying mechanisms remain controversial. Here, we conduct the first assemblage-level global examination of ‘Bergmann’s rule’ within an entire animal class. We generate global maps of avian body size and demonstrate a general pattern of larger body sizes at high latitudes, conforming to Bergmann’s rule. We also show, however, that median body size within assemblages is systematically large on islands and small in species-rich areas. Similarly, while spatial models show that temperature is the single strongest environmental correlate of body size, there are secondary correlations with resource availability and a strong pattern of decreasing body size with increasing species richness. Finally, our results suggest that geographic patterns of body size are caused both by adaptation within lineages, as invoked by Bergmann, and by taxonomic turnover among lineages. Taken together, these results indicate that while Bergmann’s prediction based on physiological scaling is remarkably accurate, it is far from the full picture. Global patterns of body size in avian assemblages are driven by interactions between the physiological demands of the environment, resource availability, species richness and taxonomic turnover among lineages.