A simulation model, recent experiments, and the literature provide consistent evidence that megafauna extinctions caused by human hunting could have played as great a role as climate in shifting from a vegetation mosaic with abundant grass-dominated steppe to a mosaic dominated by moss tundra in Beringia at the end of the Pleistocene. General circulation models suggest that the Pleistocene environment of Beringia was colder than at the present with broadly similar wind patterns and precipitation but wetter soils. These and other observations suggest that the steppelike vegetation and dry soils of Beringia in the late Pleistocene were not a direct consequence of an arid macroclimate. Trampling and grazing by mammalian grazers in tundra cause a shift in dominance from mosses to grasses. Grasses reduce soil moisture more effectively than mosses through high rates of evapotranspiration. Results of a simulation model based on plant competition for water and light and plant sensitivity to grazers and nutrient supply predict that either of two vegetation types, grass-dominated steppe or moss-dominated tundra, could exist in Beringia under both current and Pleistocene climates. The model suggests that moss-dominated tundra is favored when grazing is reduced below levels that are in equilibrium with climate and vegetation. Together these results indicate that mammalian grazers have a sufficiently large effect on vegetation and soil moisture that their extinction could have contributed substantially to the shift from predominance of steppe to tundra at the Pleistocene-Holocene boundary. Our hypothesis suggests a mechanism by which the steppe ecosystem could be restored to portions of its former range. We also suggest that mammalian impacts on vegetation are sufficiently large that future vegetation cannot be predicted from climate scenarios without considering the role of mammals.