It has become clear that plants can create soils that affect subsequent plant growth. However, because plant-soil feedbacks (PSFs) are typically measured in monoculture experiments, it remains unclear to what extent PSFs affect plant growth in communities. Here we used data from a factorial PSF experiment to predict the biomass of 12 species grown in 162 plant community combinations. Five different plant growth models were parameterized with either monoculture biomass data (Null) or with PSF data (PSF) and model predictions were compared to plant growth observed in communities. For each of the five models, PSF model predictions were closer to observed species biomass in communities than Null model predictions. PSFs, which were associated with a 28% difference in plant biomass across soil types, explained 10% more variance than Null models. Results provided strong support for a small role for PSFs in predicting plant growth in communities and suggest several reasons that PSFs, as traditionally measured in monoculture experiments, may overestimate PSF effects in communities. First, monoculture data used in Null models inherently includes “self ” PSF effects. Second, PSFs must be large relative to differences in intrinsic growth rates among species to change competitive outcomes. Third, PSFs must vary among species to change species relative abundances.