The diversity and structure of Magnaporthe oryzae populations was described in many countries during the last 20 years. The expected clonal structure of the populations has been illustrated in various studies. The relationship between the genetic structure, deduced from neutral molecular markers, and the pathotypic structure, showed a range of situations varying from a one lineage-one race to almost a one genotype-one race correspondence. These studies have helped in choosing appropriate isolates for genetic characterization and to propose strategies to breed for durable resistance to blast disease. This background information is the basis to understand rice blast population evolution. But, to date, our understanding of how new virulent races appear and spread is limited. For example, the relative importance of short and long distance migration in the spreading of new virulent races is unknown. However, this information is needed to determine at which scale the deployment of a resistance strategy is likely to be effective and durable. We recently developed and used a set of 18 microsatellite markers for population studies (Adreit et al. 2007). Studies on populations from the Central part of Madagascar and from France show that populations can be differentiated at a local scale (10-20 km or less). These results suggest limited migration. However, we also showed host specialization. Therefore, the selection pressure exerted by resistant cultivars on blast populations is likely to also favor differentiation at a local scale. We also studied the distribution of the genotypes of the cloned avirulence gene ACE1 at the worldwide scale. We determined the ACE1 genotype of more than 800 isolates. Avirulent isolates were the most frequent, were sampled all over the world, and shared the same ACE1 allele. Two major virulent genotypes were identified. Their frequencies vary with geographic origin. These genotypes appeared by a complex duplication/deletion event of ACE1. These two genotypes are widely distributed over different continents. Altogether, these results suggest a unique selection event followed by long distance migration(s). Population structure at the worldwide scale also confirmed the hypothesis of intercontinental migrations. Our apparently contradictory results from studies at two different geographic scales are explained by two distinct modes of dissemination and by selection. Structuration at a local scale is consistent with short distance spore dispersal (1-5 m) observed during natural epidemics and local adaptation. Long distance migrations, including intercontinental, are possible through the transport of infected materials (probably seeds).