Background/Question/Methods Dispersal comprises a complex life history syndrome that strongly influences the demographic dynamics of especially those species that live in fragmented landscapes. On the other hand, the demographic dynamics and the spatial structure of the landscape may impose natural selection on dispersal. The question is how strong these effects might be? We have constructed an individual-based evolutionary sexual model of dispersal for species occurring as metapopulations in habitat patch networks. The model assumes correlated random walk dispersal with edge-mediated behavior (habitat selection) and spatially correlated stochastic local dynamics. The model is parameterized with extensive data for the Glanville fritillary butterfly. Based on empirical results for a SNP in the phosphoglucose isomerase (Pgi) gene, we assume that dispersal rate in the landscape matrix, fecundity, and survival are affected by a locus with two alleles, A and C, individuals with the C allele being more mobile. Results/Conclusions The model was successfully tested with two independent empirical data sets on spatial variation in Pgi allele frequency. First, at the level of local populations, the frequency of the C allele is highest in newly-established isolated populations and lowest in old isolated populations. Second, at the level of sub-networks with dissimilar numbers and connectivities of habitat patches, the frequency of C increases with decreasing network size and hence with decreasing average metapopulation size. The frequency of C is highest in landscapes where local extinction risk is high and where there are abundant opportunities to establish new populations. Our results indicate coupling of the ecological and evolutionary dynamics involving dispersal. The strength of the coupling depends on the spatial scale and is asymmetric, demographic dynamics having a greater immediate impact on genetic dynamics than vice versa.