Highly variable effects of legumes have been observed in biodiversity experiments, but little is known about plant diversity effects on N₂ fixation of legume species. We used the ¹⁵N natural abundance method in a non-fertilized regularly mown 6-year biodiversity experiment (Jena Experiment) to quantify N₂ fixation of 12 legume species. The proportion of legume N derived from the atmosphere (%N_dfa) differed significantly among legume species. %N_dfa values were lower in 2004 after setting-up the experiment (73 ± 20) than in the later years (2006: 80 ± 16; 2008: 78 ± 12). Increasing species richness had positive effects on %N_dfa in 2004 and 2006, but not in 2008. High biomass production of legumes in 2004 and 2006 declined to lower levels in 2008. In 2006, legume positioning within the canopy best explained variation in %N_dfa values indicating a lower reliance of tall legumes on N₂ fixation. In 2008, larger %N_dfa values of legumes were related to lower leaf P concentrations suggesting that the availability of phosphorus limited growth of legumes. In summary, diversity effects on N₂ fixation depend on legume species identity, their ability to compete for soil nutrients and light and may vary temporally in response to changing resource availability.