AbstractPlant productivity often increases with species richness, but the mechanisms explaining this diversity–productivity relationship are not fully understood. We tested if plant–soil feedbacks (PSF) can help to explain how biomass production changes with species richness. Using a greenhouse experiment, we measured all 240 possible PSFs for 16 plant species. At the same time, 49 plant communities with diversities ranging from one to 16 species were grown in replicated pots. A suite of plant community growth models, parameterized with (PSF) or without PSF (Null) effects, was used to predict plant growth observed in the communities. Selection effects and complementarity effects in modeled and observed data were separated. Plants created soils that increased or decreased subsequent plant growth by 25% ± 10%, but because PSFs were negative for C3 and C4 grasses, neutral for forbs, and positive for legumes, the net effect of all PSFs was a 2% ± 17% decrease in plant growth. Experimental plant communities with 16 species produced 37% more biomass than monocultures due to complementarity. Null models incorrectly predicted that 16‐species communities would overyield due to selection effects. Adding PSF effects to Null models decreased selection effects, increased complementarity effects, and improved correlations between observed and predicted community biomass. PSF models predicted 26% of overyielding caused by complementarity observed in experimental communities. Relative to Null models, PSF models improved the predictions of the magnitude and mechanism of the diversity–productivity relationship. Results provide clear support for PSFs as one of several mechanisms that determine diversity–productivity relationships and help close the gap in understanding how biodiversity enhances ecosystem services such as biomass production.