Afforestation projects using species mixtures are expected to better support ecosystem services than monoculture plantations. While grassland studies have shown natural selection favoring high-performance genotypes in species-rich communities, this has not been explored in forests. We used seed-family identity (known maternity) to represent genetic identity and investigated how this affected the biomass accumulation (i.e. growth) of individual trees (n = 13 435) along a species richness gradient (1-16 species) and over stand age (9 yr) in a forest biodiversity experiment. We found that among the eight species tested, different seed families responded differently to species richness, some of them growing relatively better in low-diversity plots and others in high-diversity plots. Furthermore, within-species growth variation increased with species richness and stand age, while between-species variation decreased with stand age. These results indicate that seed families within species and their reaction norms along the species richness gradient vary considerably and thus can explain a substantial proportion of the overall variation in tree growth. Our findings suggest that the growth and associated ecosystem services of species-rich mixtures in afforestation projects can be optimized by artificially selecting seed families with high mixture performance in biodiversity experiments.