We evaluated the genetic consequences and efficiency of conservation practices in Oryza rufipogon using microsatellite DNA markers. Spatial autocorrelation analysis from 12 microsatellite loci revealed that microsatellite alleles were exclusively distributed in patches within the population, indicating that large populations were unlikely to be homogeneous. An in situ conserved stand of O. rufipogon, which has been protected by a concrete wall from a large population, captured only 67.9% of the total genetic variation of the previous large population. The concrete wall was built to protect the wild rice, but it acted more as a physical barrier to gene exchanges between the two sides. An assignment test revealed only 11.1% putative seed exchanges across the wall. A reintroduced population was found to be genetically very diverse. About 76.3% of the total genetic variation detected in other populations was captured in this reintroduced population, and 24.8% of the total genetic variation in this population was not found in other populations. These results display two important findings for conservation of O. rufipogon. First, conserving one part of a large population of O. rufipogon will not preserve an adequate sample of the genetic variability, since populations are not homogeneous, and genotype distribution varies among localities. Second, a reintroduced population is not genetically depauperate, but it is too early to assess its long-term survival.