Little is known about the relationship between animal movements and the emergent structure of populations, especially for species occupying large continuous distributions. Some such mammals disperse disproportionately into habitat similar to their natal habitat, a behavioural bias that might be expected to lead to habitat-conforming genetic structure. We hypothesized that coyotes (Canis latrans) would exhibit such natal-biased dispersal, and used 13 microsatellite loci to test, correspondingly, whether genetic structure conformed to major habitat breaks. First, we used a model-based approach to assign coyote genotypes to distinct genetic clusters irrespective of geographical location. Visualization on a geographical information system revealed a strong concordance between the locations of cluster assignments and habitat bioregions, not explainable in terms of physical dispersal barriers or intervening low-quality habitat. Next, we used a multiple Mantel test, which controlled for effects of geographical distance (itself, marginally significant; P = 0.06), to statistically determine that genetic distance was indeed higher between than within bioregions (P < 0.001). Whereas previously published examples of landscape effects on gene flow have typically been explainable in terms of species-wide habitat affinities or dispersal barriers, our finding that genetic subdivisions were associated with unobstructed boundaries between contiguous habitats suggests a role for intraspecific variability in habitat affinities as a factor underlying genetic structure. In addition, our data combined with previously published data suggest a pattern of genetic isolation-by-distance throughout western North America, consistent with independent evidence that the western half of the coyote range predates European settlement.