Several elegant experiments have demonstrated positive relationships between species diversity and biomass production. The underlying causes of this relationship have been hotly debated, with complementarity in resource use and facilitation suggested on the one hand, and sampling effect on the other, according to which species-richer communities work better because they are more likely to contain productive species. Based on careful experimental design and statistical analysis, sampling effect could be refuted for some experiments, including the one used in the new study by Caldeira et al. [1xMechanisms of positive biodiversity–production relationships: insights provided by δ13C analysis in experimental Mediterranean grassland plots. Caldeira, M.C. et al. Ecol. Lett. 2001; 4: 439–443CrossRef | Scopus (61)See all References][1]. They studied the role of water use for the positive diversity–production relationship observed in their experimental Mediterranean (i.e. frequently water-stressed) grasslands.Caldeira et al. looked at the ratios of stable carbon isotopes in leaves of five important species of this system, representing grasses, legumes and forbs. The ratios indicate differences between ambient and intercellular CO2 levels, which, in turn, are affected by stomatal behaviour and water use. For all species, water use was elevated at high diversities, significantly so for at least one species per group. Moreover, during an intermittent rainy period, soil moisture in the upper soil layer, where most of the roots were situated, was lowest in bare soil, intermediate in monocultures and higher in diverse plots. This indicates that stomatal behaviour was triggered by higher water availability in diverse communities. The higher biomass in more diverse plots, which, early in the season, could have been caused initially by complementary resource use, could have reduced subsequent evaporation or percolation, or increased dew capture, thus facilitating growth via reduced water loss. Plot biomass explained significant variation in the carbon isotope ratios; however, we are not told whether this explains all of the species-richness effect. Thus, the diversity effect on water use could be direct or indirect via increased biomass.This study clearly demonstrates mechanisms beyond the sampling effect that could affect the diversity–production relationship. Moreover, it suggests feedback between early- and late-season complementarity and facilitation, and implies that the biodiversity–production relationship also involves changes in soil moisture and water use, which in turn might affect organisms other than these primary producers. Regardless of whether the observed diversity effect on water use was ‘just’ a biomass effect initially started by earlier complementarity, the study suggests that the biodiversity–production question has left the pattern stage for the mechanism stage.