Results/Conclusions: We find that variable trait structures, species numbers among functional groups, and assembly processes each affect the BEF relationship. In particular, lower costs associated with increasing one trait over another, or variable costs as suggested from global meta-analyses, lead to stronger BEF relationships owing to their greater probability for including species highly efficient at biomass accumulation. Loss of some functional groups (e.g. perennial bunchgrasses) leads to consistent declines in function while loss of others (e.g. perennial forbs) often had variable effects based on the functional groups that remained and their constituent richnesses. This general pattern of ecosystem responses to plant functional diversity loss seems corroborated by observations from our ongoing biodiversity removal field experiment, although some important specific differences exist. Our findings suggest that simple models like ours, parameterized with easily measured species trait values, can capture the essential features of complex BEF relationships, and may help generate scenarios to predict the effects of future biodiversity decline.