Several researchers have hypothesized that, through various mechanisms, loss of species and functional group richness from a plant community will affect the magnitude and interannual variability of productivity. To test this hypothesis, I conducted a microcosm study of California grassland communities that differed in species richness. I grew cohorts of microcosms that simulated undisturbed grassland (in one year) and gopher-disturbed grassland (in two consecutive years). As the number of species per functional group decreased from 4 to 1, biomass production remained constant in all three cohorts. As species richness decreased from 16 to 1 (or 8 to 1, in either case including a drop in functional group richness), productivity declined in one of the cohorts. In this cohort, productivity of one polyculture marginally exceeded that of the most productive monoculture. Resource complementarity and a type of selection effect may have each contributed to the observed diversity-productivity relationships. Results suggest the existence of a selection effect that involves species that are highly productive in mixtures, rather than in monoculture. Over two seasons, species and functional group richness did not affect the interannual variability of biomass production. Comparisons of interannual changes in the productivity of monocultures and polycultures suggested that, in some polycultures, increased water availability might have relieved interspecific competition more than intraspecific competition. Based on results from this experiment and other manipulative experiments, I develop a framework to explain the relationship between species richness and productivity in terrestrial plant communities. The framework highlights the importance of environmental variation in shaping the diversity/productivity relationship.