Species abundances (evenness or identity of the dominant species in mixtures) usually are not rigorously controlled when testing relationships between plant production and species richness and may be highly dynamic in disturbed or early successional communities. Changes in species abundances may affect the yield of mixtures relative to yields expected from species monocultures [the net biodiversity effect (NBE)] by changing how species that differ in function are distributed in the plant community. To test the prediction that variation in species abundances affects the NBE via changes in the expression of functional differences among species (the complementarity effect), we grew perennial grasses and forbs in field plots in central Texas, USA, as equal-density monocultures and two-species mixtures in which relative abundances of species were varied. Function should differ more consistently between species of different growth forms than of the same growth form. We predicted, therefore, that the complementarity effect and influence of species abundances on the NBE would be more pronounced in grass/forb mixtures than in mixtures with species of the same growth form (grass/grass and forb/forb mixtures). The NBE varied with species evenness in two of the six species pairs studied and with identity of the dominant species in a third species combination. The NBE was sensitive to species proportions in both grass/grass and grass/forb assemblages. In all combinations in which the NBE differed with either evenness or identity of the dominant species, the variation resulted largely from change in the complementarity effect. Our results suggest that the NBE of mixtures is sensitive to effects of species ratios on complementarity.