Experimental invasion by legumes reveals non-random assembly rules in grassland communities

1. Although experimental studies usually reveal that resistance to invasion increases with species diversity, observational studies sometimes show the opposite trend. The higher resistance of diverse plots to invasion may be partly due to the increased probability of a plot containing a species with similar resource requirements to the invader. 2. We conducted a study of the invasibility of monocultures belonging to three different functional groups by seven sown species of legume. By only using experimentally established monocultures, rather than manipulating the abundance of particular functional groups, we removed both species diversity and differences in underlying abiotic conditions as potentially confounding variables. 3. We found that legume monocultures were more resistant than monocultures of grasses or non-leguminous forbs to invasion by sown legumes but not to invasion by other unsown species. The functional group effect remained after controlling for differences in total biomass and the average height of the above-ground biomass. 4. The relative success of legume species and types also varied with monoculture characteristics. The proportional biomass of climbing legumes increased strongly with biomass height in non-leguminous forb monocultures, while it declined with biomass height in grass monocultures. Trifolium pratense was the most successful invader in grass monocultures, while Vicia cracca was the most successful in non-leguminous forb monocultures. 5. Our results suggest that non-random assembly rules operate in grassland communities both between and within functional groups. Legume invaders found it much more difficult to invade legume plots, while grass and non-leguminous forb plots favoured non-climbing and climbing legumes, respectively. If plots mimic monospecific patches, the effect of these assembly rules in diverse communities might depend upon the patch structure of diverse communities. This dependency on patch structure may contribute to differences in results of research from experimental vs. natural communities.