Plant populations may experience local extinction and at the same time new populations may appear in nearby suitable locations. Species may also colonize the same site on multiple occasions. Here, we examined the impact of local extinction and recolonization on the genetic structure of wild populations of lima beans (Phaseolus lunatus) in the Central valley of Costa Rica. We compared genetic diversity from the samples taken from the populations before and after extinction at 13 locations using microsatellite markers. Locations were classified according to the occurrence of extinction episodes during the previous five years into three groups: 1) populations that experienced extinction for more than one year, and were later recolonized (recolonized), 2) populations that did not experience local extinction (control), and 3) populations that did not experience local extinction during the study, but were cut to experimentally simulate extinction (experimental). Our data did not show a clear tendency in variation in allele frequencies, expected heterozygosity, and effective number of alleles within and between groups of populations. However, we found that the level of genetic differentiation between samples collected at different times at the same location was different in the three groups of populations. Recolonized locations showed the highest level of genetic differentiation (mean Fst= 0.2769), followed by control locations (mean Fst= 0.0576) and experimental locations (mean Fst= 0.0189). Similar findings were observed for Nei s genetic distance between samples (di,j= 0.1786, 0.0400, and 0.0037, respectively). Our results indicate that genetic change in lima beans depends on the duration and frequency of local extinction episodes. These findings also showed that control populations are not in equilibrium. Implications of these results for the establishment of conservation strategies of genetic resources of lima beans are discussed. Rev. Biol. Trop. 56 (3): 1023-1041. Epub 2008 September 30.