The ongoing increase in atmospheric CO2 concentration ([CO2]) can potentially alter litter decomposition rates by changing: (i) the litter quality of individual species, (ii) allocation patterns of individual species, (iii) the species composition of ecosystems (which could alter ecosystem-level litter quality and allocation), (iv) patterns of soil moisture, and (v) the composition and size of microbial communities. To determine the relative importance of these mechanisms in a California annual grassland, we created four mixtures of litter that differed in species composition (the annual legume Lotus wrangelianus Fischer & C. Meyer comprised either 10% or 40% of the initial mass) and atmospheric [CO2] during growth (ambient or double-ambient). These mixtures decomposed for 33 weeks at three positions (above, on, and below the soil surface) in four types of grassland microcosms (fertilized and unfertilized microcosms exposed to elevated or ambient [CO2]) and at a common field site. Initially, legume-rich litter mixtures had higher nitrogen concentrations ([N]) than legume-poor mixtures. In most positions and environments, the different litter mixtures decomposed at approximately the same rate. Fertilization and CO2 enrichment of microcosms had no effect on mass loss of litter within them. However, mass loss was strongly related to litter position in both microcosms and the field. Nitrogen dynamics of litter were significantly related to the initial [N] of litter on the soil surface, but not in other positions. We conclude that changes in allocation patterns and species composition are likely to be the dominant mechanisms through which ecosystem-level decomposition rates respond to increasing atmospheric [CO2].