Background/Question/Methods While biodiversity loss due to species extinction is permanent, smaller-scale biodiversity loss due to local extirpation may be reversible. Understanding the biotic and abiotic factors that lead to native biodiversity recovery following local extirpation may inform future conservation and restoration efforts. In this study, we examine the effects of prior small-scale diversity loss and soil depth on native plant community recovery following five years of plot-scale experimental biodiversity manipulation in a California serpentine grassland. During the manipulation phase of this experiment, plots were maintained with 0, 2, 5, 8, 12, or 16 native plant species, with species loss following either a realistic, drought-based species loss scenario or a randomized species loss scenario. The recovery phase of the study lasted another five years, during which time neither the plots nor the surrounding matrix were manipulated. At the end of the recovery phase, we resurveyed the plots to quantify native percent cover, native species richness, native productivity, invasive cover, and invasive productivity. Results/Conclusions After five years of recovery, plots that were maintained as bare (zero species) plots during the experimental phase had nearly 50% higher native plant cover than any of the other treatments (P=0.002); shallow soil-depth plots had higher native species richness than medium or deep plots (P=0.01). In addition, the species loss scenario during the experimental phase had a significant effect on native plant community recovery. Plots within the realistic species-loss scenario overall had higher native species richness following recovery than did the plots in the randomized species-loss scenario. In addition, there was a significant interaction between species-loss scenario and plot species richness on native community recovery: low diversity plots in the randomized species loss scenario led to lower native cover and richness following recovery than did high diversity plots, while low diversity plots in the realistic species loss scenario led to higher native cover and richness following recovery than did high diversity plots. Overall, the legacy of experimental biodiversity manipulation persisted for at least five years, with prior realistic diversity loss facilitating native community recovery.