1. Wild large herbivores are declining worldwide. Despite extensive use of exclosure experiments to investigate herbivore impacts, there is little consensus on the effects of wild large herbivores on ecosystem function. 2. Of the ecosystem functions likely impacted, we reviewed the five most-studied in exclosure experiments: ecosystem resilience/resistance to disturbance, nutrient cycling, carbon cycling, plant regeneration, and primary productivity. 3. Experimental data on large wild herbivores’ effects on ecosystem functions were predominately derived from temperate grasslands (50% grasslands, 75% temperate zones), and experiments that may not be of adequate size (median size 400m2 despite excluding all experiments below 25m2) or duration (median duration 6 years) to capture ecosystem-scale responses to these low-density and wide-ranging taxa. 4. Wild herbivore removal frequently impacted ecosystem functions; for example, net carbon uptake increased by three times in some instances. However, the magnitude and direction of effects, even within a single function, were highly variable. 5. A focus on carbon cycling highlighted challenges in interpreting effects on a single function. While the effect of large herbivore exclusion on carbon cycling was slightly positive when its components (e.g. pools vs. fluxes of carbon) were aggregated, effects on individual components were variable and sometimes opposed. 6. Given modern declines in large wild herbivores, it is critical to understand their effects on ecosystem function. However, this synthesis highlights strong variability in direction, magnitude, and modifiers of these effects. Some variation is likely due to disparity in what components are used to describe a given function. For example, for the carbon cycle we identified eight distinctly meaningful components, which are not easily combined yet are potentially misrepresentative of the larger cycle when considered alone. However, much of the observed difference in responses likely reflects real ecological variability across complex systems. 7. To move towards a general predictive framework we must identify where variation in effect is due to methodological differences and where due to ecosystem context. Two critical steps forward are 1) additional quantitative synthetic analyses of large herbivores’ effects on individual functions, and 2) improved, increased systematic exclosure research focusing on effects of large herbivores’ exclusion on functions.