Rice, as a widely and intensively cultivated crop, should be a target for parasite host shifts and a source for shifts to co-occurring weeds. Magnaporthe oryzae, of the M. grisea species complex, is the most important fungal pathogen of rice, with a high degree of host specificity. On the basis of 10 loci from six of its seven linkage groups, 37 multilocus haplotypes among 497 isolates of M. oryzae from rice and other grasses were identified. Phylogenetic relationships among isolates from rice (Oryza sativa), millet (Setaria spp.), cutgrass (Leersia hexandra), and torpedo grass (Panicum repens) were predominantly tree like, consistent with a lack of recombination, but from other hosts were reticulate, consistent with recombination. The single origin of rice-infecting M. oryzae followed a host shift from a Setaria millet and was closely followed by additional shifts to weeds of rice, cutgrass, and torpedo grass. Two independent estimators of divergence time indicate that these host shifts predate the Green Revolution and could be associated with rice domestication. The rice-infecting lineage is characterized by high copy number of the transposable element MGR586 (Pot3) and, except in two haplotypes, by a loss of AVR-Co39. Both mating types have been retained in ancestral, well-distributed rice-infecting haplotypes 10 (mainly temperate) and 14 (mainly tropical), but only one mating type was recovered from several derived, geographically restricted haplotypes. There is evidence of a common origin of both ACE1 virulence genotypes in haplotype 14. Host-haplotype association is evidenced by low pathogenicity on hosts associated with other haplotypes. W IDELY distributed, high-density populations, such commonly follows (Tibayrenc et al. 1991; Maynard as those of cultivated crops or humans, are large Smith et al. 1993; Brasier and Buck 2001). While we targets for host shifts by parasites. After crossing host would expect that virulence determinants in a pandemic species boundaries, successful shifts culminate with speclone would remain associated and modified only cialization and genetic divergence on the new host through mutation, any genetic exchange with recombi- (Antonovics et al. 2002). If the parasite crosses a species nation would shuffle these determinants into novel combarrier only once, virulence determinants would derive binations across the parasite population. Within large from this common origin. In contrast, if the target host pathogen populations, recombination would increase species is invaded in multiple, independent host shifts, the rate at which multiple virulence factors could comvirulence could have multiple, fundamentally different bine to generate "super parasite" genotypes (Crow and determinants among host populations (MacLeod et al. Kimura 1970). 2001.The rice blast pathosystem is a model system not only After successful establishment on a new host, espefor plant-pathogen interactions (Valent 1990; Talbot cially one that is abundant and relatively genetically 2003), but also for host ...