Numerous experiments have demonstrated that diverse plant communities use nitrogen (N) more completely and efficiently, with implications for how species conservation efforts might influence N cycling and retention in terrestrial ecosystems. However, most such experiments have randomly manipulated species richness and minimized environmental heterogeneity, two design aspects that may reduce applicability to real ecosystems. Here we present results from an experiment directly comparing how realistic and randomized plant species losses affect plant N use across a gradient of soil depth in a native-dominated serpentine grassland in California. We found that the strength of the species richness effect on plant N use did not increase with soil depth in either the realistic or randomized species loss scenarios, indicating that the increased vertical heterogeneity conferred by deeper soils did not lead to greater complementarity among species in this ecosystem. Realistic species losses significantly reduced plant N uptake and altered N-use efficiency, while randomized species losses had no effect on plant N use. Increasing soil depth positively affected plant N uptake in both loss order scenarios but had a weaker effect on plant N use than did realistic species losses. Our results illustrate that realistic species losses can have functional consequences that differ distinctly from randomized losses, and that species diversity effects can be independent of and outweigh those of environmental heterogeneity on ecosystem functioning. Our findings also support the value of conservation efforts aimed at maintaining biodiversity to help buffer ecosystems against increasing anthropogenic N loading. Key words: biodiversity; ecosystem functioning; environmental heterogeneity; nonrandom diversity loss; nutrient cycling; resource use; species loss.