Local distributions of avian brood parasites among their host habitats may depend upon conspecific parasite density. We used isodar analysis to test for density-dependent habitat selection in brown-headed cowbirds ( Molothrus ater) among tallgrass prairie adjacent to wooded edges, and prairie interior habitat (>100 m from wooded edges) with and without experimental perches. Eight study sites containing these three habitat treatments were established along a geographical gradient in cowbird abundance within the Flint Hills region of Eastern Kansas and Oklahoma, USA. The focal host species of our study, the dickcissel ( Spiza americana), is the most abundant and preferred cowbird host in the prairie of this region. Cowbird relative abundance and cowbird:host abundance ratios were used as estimates of female cowbird density, whereas cowbird egg density was measured as parasitism frequency (percent of dickcissel nests parasitized), and parasitism intensity (number of cowbird eggs per parasitized nest). Geographical variation in cowbird abundance was independent of host abundance. Within study sites, host abundance was highest in wooded edge plots, intermediate in the experimental perch plots, and lowest in prairie interior. Cowbirds exhibited a pattern of density-dependent selection of prairie edge versus experimental perch and interior habitats. On sites where measures of cowbird density were lowest, all cowbird density estimates (female cowbirds and their eggs) were highest near (< or =100 m) wooded edges, where host and perch availability are highest. However, as overall cowbird density increased geographically, these density estimates increased more rapidly in experimental perch plots and prairie interiors. Variation in cowbird abundance and cowbird:host ratios suggested density-dependent cowbird selection of experimental perch over prairie interior habitat, but parasitism levels on dickcissel nests were similar among these two habitats at all levels of local cowbird parasitism. The density-dependent pattern of cowbird distribution among prairie edge and interior suggested that density effects on perceived cowbird fitness are greatest at wooded edges. A positive relationship between daily nest mortality rates of parasitized nests during the nestling period with parasitism intensity levels per nest suggested a density-dependent effect on cowbird reproductive success. However, this relationship was similar among habitats, such that all habitats should have been perceived as being equally suitable to cowbirds at all densities. Other unmeasured effects on cowbird habitat suitability (e.g., reduced cowbird success in edge-dwelling host nests, cowbird despotism at edges) might have affected cowbird habitat selection. Managers attempting to minimize cowbird parasitism on sensitive cowbird hosts should consider that hosts in otherwise less-preferred cowbird habitats (e.g., habitat interiors) are at greater risk of being parasitized where cowbirds become particularly abundant.