We evaluated competing models explaining the occurrence of five stream fishes in an urbanizing watershed to determine the relative importance of (a) impervious surface and other indicators of current land use, (b) historic land use (e.g., agriculture, impoundments), and (c) hydrogeomorphic characteristics (e.g., stream size, elevation, geology). For four of five species, the best-supported models were those that included both current effective impervious cover and historic land use predictor variables, although models with only effective impervious cover were equally well supported for two of those species. For the best-supported models for three species, occurrence probability was predicted to approach zero at levels of development equivalent to about 2%–4% effective impervious cover in the surrounding region. Data were drawn from 357 fish collections made in the Etowah River basin, Georgia, USA, between 1998 and 2003 and analyzed using hierarchical logistic regression accounting for imperfect species detection. This is the first study we know of to examine the response of individual fish species to both increasing impervious cover and historic land use. Such individual species assessments will be increasingly necessary to guide policies for managing urban effects and preventing extirpations of sensitive species.