Metapopulation theory provides a framework for understanding population persistence in fragmented landscapes and as such has been widely used in conservation biology to inform management of fragmented populations. However, classical metapopulation theory [Levins, R. (1970) Lect. Notes Math. 2, 75-107] ignores metapopulation structure and local population dynamics, both of which may affect extinction dynamics. Here, we investigate metapopulation dynamics in populations that are subject to different migration rates by using experimental metapopulations of the annual plant Cardamine pensylvanica. As predicted by classical metapopulation theory, connected populations persisted longer than did isolated populations, but the relationship between migration and persistence time was nonlinear. Extinction risk sharply increased as the distance between local populations increased above a threshold value that was consistent with stochastic sim- Science 290, 1360 -1364], a metric that predicts synchronous cycles, indicated that continuous populations should cycle in phase, resulting in an increased extinction risk. Determining empirically the optimal migration level to improve survival chances will be challenging for any natural population. Migration rates that would not increase migration above the threshold value would be ineffectual, but migration rates that would homogenize local densities could increase the risk of coherent oscillations and enhance extinction risk.experimental populations ͉ metapopulation capacity ͉ population coherence ͉ migration ͉ population persistence H abitat fragmentation is implicated in the decline of many endangered wild species (1-4). To offset the problems associated with habitat fragmentation, conservation biologists have called for the construction of corridors to facilitate migration between fragmented habitats (5), but the construction of corridors and their attendant increase in migration rates is not without risks (6). Predicting how migration affects population persistence is important for species conservation. Metapopulation theory provides a framework for understanding population persistence in fragmented landscapes and as such has been used in conservation biology to inform management of fragmented populations. Originally conceived by Levins, metapopulation dynamics are determined by the balance between local colonization and extinction events (7). When colonization events exceed extinctions, metapopulations are predicted to persist, because migration allows an extinct local population to become recolonized (7-9). Absent from this theory is any consideration of how differences in patch suitability (area, distance to other local patches) and local dynamics (such as reproductive rate and intraspecific competition) influence metapopulation persistence. However, recent theoretical work has included more realistic aspects of metapopulations. Hanski and Ovaskainen (10) examine how differences in patch area and in metapopulation structure affect predicted metapopulation persistence ( refs....