Experiments testing biodiversity effects on ecosystem functioning have been criticized on the basis that their random-assembly designs do not reflect deterministic species loss in nature. Because previous studies, and their critics, have focused primarily on plants, however, it is underappreciated that the most consistent such determinism involves biased extinction of large consumers, skewing trophic structure and substantially changing conclusions about ecosystem impacts that assume changing plant diversity alone. Both demography and anthropogenic threats render large vertebrate consumers more vulnerable to extinction, on average, than plants. Importantly, species loss appears biased toward strong interactors among animals but weak interactors among plants. Accordingly, available evidence suggests that loss of a few predator species often has impacts comparable in magnitude to those stemming from a large reduction in plant diversity. Thus, the dominant impacts of biodiversity change on ecosystem functioning appear to be trophically mediated, with important implications for conservation.