We conducted a field experiment that manipulated landscapes by mowing so that the amount of unfavorable habitat (low cover) for prairie voles (Microtus ochrogaster) increased while the number and size of favorable patches (high cover) remained constant. Distance between favorable patches increased as the amount of unfavorable habitat increased, so we could test two current hypotheses concerning the effect of habitat fragmentation on local populations: 1) increased distance between favorable habitat patches reduces successful per capita dispersal (emigration and immigration) because dispersers suffer greater exposure to predators (the predation hypothesis); and 2) per capita dispersal is inversely density dependent in voles because increased aggression at higher density inhibits movements (the social fence hypothesis). As predicted by the predation hypothesis, increased distance between favorable habitat patches led to decreased successful dispersal among patches and increased per capita mortality, particularly among subadult and adult males (the categories of voles most likely to emigrate). As predicted by the social fence hypothesis, dispersal was inversely density dependent, and dispersing voles displayed a greater frequency of wounding (an indication of increased aggressive interactions) than did residents. The amount of wounding in general did not increase with density, however, and, as distance between patches increased to 60 m, successful dispersal became rare and erratic. Nevertheless, our overall results supported current hypotheses regarding the effects of increased habitat fragmentation on patterns of dispersal and mortality. Examining the impact of these effects on local population dynamics within different landscapes will require longer periods of observation.