The need to protect crop genetic resources has sparked a growing interest in the genetic diversity maintained in traditional farming systems worldwide. Although traditional seed management has been proposed as an important determinant of genetic diversity and structure in crops, no models exist that can adequately describe the genetic effects of seed management. We present a metapopulation model that accounts for several features unique to managed crop populations. Using traditional maize agriculture as an example, we develop a coalescence-based model of a crop metapopulation undergoing pollen and seed flow as well as seed replacement. In contrast to metapopulation work on natural systems, we model seed migration as episodic and originating from a single source per population rather than as a constant immigration from the entire metapopulation. We find that the correlated origin of migrants leads to surprising results, including a loss of invariance of within-deme diversity and a parabolic relationship between F(ST) and migration quantity. In contrast, the effects of migration frequency on diversity and structure are more similar to classical predictions, suggesting that seed migration in managed crop populations cannot be described by a single parameter. In addition to migration, we investigate the effects of deme size and extinction rates on genetic structure, and show that high levels of pollen migration may mask the effects of seed management on structure. Our results highlight the importance of analytically evaluating the effects of deviations from classical metapopulation models, especially in systems for which data are available to estimate specific model parameters.