An emerging pattern is that population densities of generalist rodents are higher in small compared to large forest patches in fragmented landscapes. We used genetically based measures of migration between patches to test two dispersal-based hypotheses for this negative densityarea relationship: (1) emigration rates from small patches should be relatively lower compared to large patches (''inhibited dispersal hypothesis''), or (2) immigration rates should be higher into small than large patches (''immigration hypothesis''). Neither hypothesis was supported using data on dispersal inferred from eight microsatellite loci for 12 populations of Peromyscus leucopus in six small (1.3-2.7 ha) and six large (8-150 ha) forest patches. Emigration rates were not lower from and immigration rates were not higher into small than large patches. In fact, contrary to both hypotheses, emigration rates were higher from populations of P. leucopus in small compared to large patches. Based on a combination of genetic and field data, we speculate that higher reproduction in smaller patches resulted in higher densities which led to higher emigration rates from those patches. Rates of reproduction (presumably driven by better habitat conditions in smaller patches), rather than dispersal, seems to drive density differences in forest patches. We conclude that smaller forest patches within an agricultural matrix act as a source of individuals, and that migration rates are fairly high among forest patches regardless of size.