We understand little about the factors shaping the diversity of arbuscular mycorrhizal fungi (AMF) on a single host plant and how that diversity influences plant performance. To attain that understanding, we must consider whether co-occurring AMF species occupy different niche space and whether the importance to the host of any one AMF species varies with soil conditions. Because there is some evidence that two AMF species can be non-additive in their effects on plant growth, we propose that the appropriate measure of AMF function is how much plant growth is affected when that particular AMF species is deleted from the community. By repeating this design under different soil conditions we can determine if different AMF species provide different services to the plant. These greenhouse experiments used two grass species: Andropogon gerardii and Sorhastrum nutans and a subset of their natural AMF community in soils differing in nitrogen, phosphorus, and nickel, which is naturally high in the plants' native serpentine soils. We find that AMF species with a neutral or negative effect under one soil condition can promote plant growth in another, suggesting that community diversity is useful in coping with heterogeneous soils. By comparing results to the traditional method of evaluating the function of individual AMF species, we find that on average the traditional method overestimates the benefit provided by a single species. These results suggest an explanation for why any one plant species supports several species of these fungi.