The structure of a plant community in terms of functional traits can strongly affect community productivity. Two components may contribute to this, community‐wide trait means (mass‐ratio hypothesis) or community‐wide trait variations (diversity hypothesis) across species and individuals. We compared the explanatory power of the two hypotheses for explaining biomass variation among individuals and communities as measures of productivity. We set up experimental communities of tree seedlings that ranged in species richness from 1 to 8. We measured above‐ and belowground biomass, leaf area, specific leaf area and height of all individuals separately after 16 months of growth and calculated species‐level values by averaging across all individuals of each species grown in monoculture. Using species‐ and individual‐level trait data, we calculated community‐ (CWM) and neighbourhood‐weighted mean (NWM) values to assess the mass‐ratio hypothesis and community functional dissimilarity (FD) and average trait dissimilarity of target individuals to neighbours (ATD), i.e. trait variations, to assess the diversity hypothesis. CWMs explained much more variation in community biomass than did species richness or FD, supporting the mass‐ratio hypothesis. The explanatory power of CWMs increased when calculated with individual‐level data, demonstrating positive effects of trait adjustments of individuals in response to species richness, i.e. plasticity in the broad sense. FD calculated with individual‐level data also increased community biomass, lending some support to the diversity hypothesis. The effects of trait means and variations on community biomass were related to effects of trait means and variations on the biomass of individuals. Overall, our results suggest that trait means rather than trait variations may drive community biomass productivity at the earliest stage of forest development.