Roost requirements of most North American forest bats are well-documented, but questions remain regarding the ultimate mechanisms underlying roost selection. Hypotheses regarding roost selection include provision of a stable microclimate, space for large colonies, protection from predators, and proximity to foraging habitat, among others. Although several hypotheses have been proposed, specific mechanisms likely vary by species and geographic region. Rafinesque's big-eared bat (Corynorhinus rafinesquii) commonly roosts in trees with large basal hollows in the Coastal Plain of the southeastern United States. Our objective was to weigh evidence for hypotheses regarding selection of diurnal summer roosts by Rafinesque's big-eared bat at 8 study sites across the Coastal Plain of Georgia, USA. We used transect searches and radiotelemetry to locate roosts and measured 22 characteristics of trees, tree cavities, and surrounding vegetation at all occupied roosts and for randomly selected unoccupied trees. We evaluated 10 hypotheses using single-season occupancy models and used Akaike's information criterion to select the most parsimonious models. We located 170 tree roosts containing approximately 870 bats for our analysis. The best supported model predicted bat presence from cavity size, interior wall texture, and number of entrances. Because large cavities allow bats to fly and smooth walls impede attacks by terrestrial predators, our results are consistent with the hypothesis that bats select roosts that allow them to evade predators. However, data on predation rates are needed for a conclusive determination. Because trees suitable as roosts for Rafinesque's big-eared bat are rare in the landscape, protection of suitable forested wetland habitat is essential to provide current and long-term roost tree availability. © 2012 The Wildlife Society.