Abstract: African ungulate populations appear to be limited principally by their food resources. Within ungulate communities, plains zebras coexist with grazing bovids of similar body size, but rarely are the dominant species. Given the highly effective nutritional strategy of the equids and the resistance of zebras to drought, this is unexpected and suggests that zebra populations may commonly be limited by other mechanisms. Long-term research in the Serengeti ecosystem and in the Kruger National Park suggests that zebra could be less sensitive to food shortage, and more sensitive to predation, than grazing bovids: if this is a general principle, then, at a larger scale, resource availability should have a weaker effect on the abundance of zebra than on grazing ruminants of similar body size (wildebeest and buffalo), and zebras should be relatively more abundant in ecosystems where predators are rare or absent. We test these expectations using data on 23 near-natural ecosystems in east and southern Africa. The abundance of wildebeest is more closely related to resources than is that of zebra; buffalo are intermediate. We show that hyena densities are closely correlated with those of lions, and use the abundance of lions as an index of predation by large predators. The numerical response of lions to increases in the abundance of their prey was linear for mesoherbivores, and apparently so for the three species alone. Finally, the abundance of zebra relative to grazing bovids is lower in ecosystems with high biomasses of lions. These results indicate that zebras may commonly be more sensitive to top-down processes than grazing bovids: the mechanism(s) have not been demonstrated, but predation could play a role. If it is true, then when numbers of the large mammalian predators decline, zebra populations should increase faster than buffalo and wildebeest.