Grazing and fire are disturbances integral to the evolution and maintenance of savanna grasslands. Humans are altering or completely eliminating these disturbance regimes at a global scale, with important consequences for savanna ecosystem structure and function. It is unknown whether the alteration of these disturbance regimes will have similar effects on grass communities of savanna grasslands in different geographic regions that vary in their biogeographic and evolutionary histories, as well as in the diversity of extant grazers. Here, we examined the effects of large herbivore loss on different aspects of grass community structure – taxonomic, phylogenetic and functional – across a range of fire frequencies in C4-dominated mesic savanna grassland sites of North America (Konza Prairie Biological Station, Kansas, USA) and South Africa (Kruger National Park). The goal of the study was to determine whether the loss of large herbivores exerted a consistent effect on the grass communities of two physiognomically similar grasslands with different biogeographic and grazing histories. The removal of large herbivores resulted in divergent responses in the grass communities at Konza and Kruger that was consistent across fire treatments. At Konza, there was a rapid and significant response to grazing exclusion while the response was muted and transient at Kruger. Functional syndromes associated with grazing resistance were generally conserved across sites, and it was the functional strategies of the dominant species at each site that drove the divergent responses. Further, our study supports the hypothesis that grazing and aridity may be selective forces that act in parallel as those species that were grazing resistant also occupied drier niches. Synthesis. Our study demonstrates that savanna grassland communities with different biogeographic and grazing histories respond differently to the removal of large herbivores and that climate, fire and grazing are interactive forces in maintaining savanna grassland diversity and function. We show that the functional attributes of the dominant grasses, which are in part driven by the biogeographic and grazing history experienced, are the most relevant in predicting the response of savanna ecosystems to the loss of large herbivores.