A series of nine simulations has been made with the Lund–Potsdam–Jena Dynamic Global Vegetation Model (LPJ-DGVM) in order to explore the impacts of climate variability and Holocene changes in variability (as simulated by the Fast Ocean-Atmosphere Model, FOAM) on vegetation in three forest-dominated regions of China and in the semi-arid Sahelian region of northern Africa. The simulations illustrate that changes both in the magnitude of climate variability and in the persistence of above/below average conditions have the potential to modify the vegetation response to changes in mean climate. Simulated changes in moisture availability affect vegetation through drought stress or through changing the fuel availability in semi-arid regions where lack of fuel often limits the incidence of fire. Increasing moisture availability causes trees to replace grasses in China by reducing drought stress; increasing moisture availability in the Sahel increases the available fuel and hence reduces fire return times, favouring grasses. The modelling results imply that climate variability is important to vegetation dynamics; that not only the magnitude, but also the temporal structure of variability is important; and that correctly simulating vegetation changes in response to climate variability requires a realistic “baseline” simulation of plant community composition. They further indicate that the impacts of climate change on ecosystems can sometimes derive as much from changes in variability as from changes in mean climate.