Models of source–sink population dynamics have to make assumptions about whether, and eventually how, demographic parameters in source habitats are dependent on the demography in sink habitats. However, the empirical basis for making such assumptions has been weak. Here we report a study on experimental root vole populations, where estimates of demographic parameters were contrasted between source patches in source–sink (treatment) and source–source systems (control). In the presence of a sink patch (simulated by a pulsed removal of immigrants), source-patch populations failed to increase over the breeding season, mainly due to a high spatially density-dependent dispersal rate from source to sink patches. The per capita recruitment rate was almost two times higher in source–sink than in the source–source systems, but this did not compensate for the loss rate due to dispersal from source to sink patches. Sex ratio in the source–sink systems became less female biased, probably as a result of an enhanced frequency of dispersal movements in females. Good knowledge of the degree of density-and habitat-dependent dispersal is critical for predicting the dynamics of source–sink populations.