Following functional extirpation in Missouri, American black bear (Ursus americanus) populations in this state have been increasing in recent years through recolonization from re-established populations in northern Arkansas. To increase our understanding of resource selection by recolonizing black bears in the Ozark Highlands of the United States, we attached Global Positioning System (GPS) transmitters to 54 black bears during May–August 2010–2013, and used location data based on biological seasons. We constructed models with anthropogenic (distance to nearest development, distance to nearest road), biological (sex, age class, season), and environmental (distance to nearest water, land cover) categories. We used infinitely weighted logistic regression to approximate the inhomogeneous Poisson point process model for presence-only (i.e., GPS locations) data to fit models. We used Bayesian Information Criterion and found that the best-performing model in the set of 81 models included all independent variables except sex and all combinations of 2-way interactions except those between biological covariates. Forested areas generally were more strongly selected than non-forested areas and bears generally selected areas distant from roads and other human development. However, selection for areas proximate to roads in the composite cover type (e.g., shrub–scrub, woody wetlands) occurred, where roads may have been used as travel corridors in unsuitable cover during the breeding season (ad) or dispersal (subad), or alternatively as a potential barrier, depending on road type and traffic volume. Use of apparent lower quality non-forested areas by bears suggests that the current level of human development in southern Missouri is unlikely to halt their recolonization.