BACKGROUND AND OBJECTIVES: Interactions between plants and beneficial soil organisms (e.g. rhizobial bacteria, mycorrhizal fungi) are models for investigating the ecological impacts of such associations in plant communities, and the evolution and maintenance of variation in mutualisms (e.g. host specificity and the level of benefits provided). With relatively few exceptions, variation in symbiotic effectiveness across wild host species is largely unexplored. METHODS: We evaluated these associations using representatives of several legume genera which commonly co-occur in natural ecosystems in south-eastern Australia and an extensive set of rhizobial strains isolated from these hosts. These strains had been previously assigned to specific phylotypes on the basis of molecular analyses. In the first of two inoculation experiments, the growth responses of each host species was evaluated with rhizobial strains isolated from that species. The second experiment assessed performance across genera and the extent of host specificity using a subset of these strains. RESULTS: While host growth responses to their own (sympatric) isolates varied considerably, rhizobial phylotype was a significant predictor of symbiotic performance, indicating that bacterial species designations on the basis of molecular markers have ecological importance. Hosts responded in qualitatively different ways to sympatric and allopatric strains of rhizobia, ranging from species with a clear preference for their own strains, to those that were broad generalists, through to species that grew significantly better with allopatric strains. CONCLUSION: Theory has focused on trade-offs between the provision of benefits and symbiont competitive ability that might explain the persistence of less beneficial strains. However, differences in performance among co-occurring host species could also drive such patterns. Our results thus highlight the likely importance of plant community structure in maintaining variation in symbiotic effectiveness.