Development of theory has outpaced experimental tests for most maintenance of diversity mechanisms. Here we demonstrate how data from biodiversity–ecosystem functioning experiments can be used to determine the mechanisms that maintain plant species diversity. We hypothesized that grassland plant diversity is maintained by two classes of mechanisms: (1) equalizing mechanisms, which reduce asymmetric competition by reducing differences in monoculture biomass production among species in mixture, and (2) species interaction mechanisms, which increase overyielding by increasing niche partitioning and facilitation among species in mixture. Specifically, equalizing mechanisms reduce the coefficient of variation in monoculture biomass production among species in mixture. Species interaction mechanisms increase species overyielding in mixture, especially for low‐biomass species. We tested these predictions with a seven‐year data set from an experiment that varied grassland plant species evenness and richness. We used path analysis to model effects of these mechanisms on annual and multiyear changes in diversity. We found that diversity was frequently maintained by species interaction mechanisms and was infrequently maintained by equalizing mechanisms. Species interaction mechanisms maintained diversity by allowing the species that produced the least biomass in monoculture to benefit the most from species interactions in mixture. Equalizing mechanisms infrequently maintained diversity because asymmetric competition infrequently resulted in competitive exclusion. We propose that this mechanistic framework be used to better understand the specific processes that influence diversity.