To better understand the relationship between diversity and productivity, experimental designs need to include ecologically realistic rarity and extinction scenarios, rather than assuming random extinction. We established experimental plots in Iowa to test whether reduced evenness and variance in plant height affects complementarity and aboveground net primary productivity (ANPP). Plots were planted with equal-sized seedlings so that species richness was varied from 1 to 4, and evenness from 0.44 to 0.97, for two different rarity-extinction scenarios: 1) increased rarity of two short species as dominance of the tall and ubiquitous grass Andropogon gerardii increased (dissimilar-height), and 2) increased rarity of two tall species as dominance of Andropogon gerardii increased (tall). Mean ANPP was significantly greater in mixtures than in monocultures. The complementarity effect was large in both rarity-extinction scenarios, but did not differ between scenarios or among evenness treatments. This suggests that height dissimilarity was not responsible for the complementarity effect in these mixtures. ANPP increased with evenness in tall, but not in dissimilar-height communities, due to the selection effect. During the first three years of the study, local extinctions were more frequent in dissimilar-height than in tall plots, and in low evenness than in high evenness plots. Only the rarest species in each scenario went locally extinct. We suggest that changes in ANPP due to changes in evenness will depend on the species becoming rare or extinct, and that maintaining high species evenness will facilitate the maintenance of species richness in tallgrass prairies.