Most recent research on the roosting habits of temperate, forest-living bats has focused on species that use enclosed cavities, but less has been done to address roosting by foliage-living species, which are assumed to have more flexible roost requirements. Numerous studies have suggested that bats select roosts on the basis of microclimate, yet few have tested this hypothesis empirically and none have addressed the use of foliage roosts in this context. We used radiotelemetry to locate roost sites of reproductive female hoary bats (Lasiurus cinereus) then compared a variety of physical features of these trees with randomly selected nonroost trees. We also recorded ambient temperature and wind speed at roost and nonroost sites to test the hypothesis that physical features associated with foliage roosts provide energetic benefits. Hoary bats selected roost sites on the southeast side of mature white spruce trees (Picea glauca; X̄ orientation 158.6 ± 6.3° SSE). Roost trees were more likely than random trees to be the same height as the surrounding forest canopy; had less canopy cover facing out from the tree in the direction of the roost branch; and had lower forest density on their southeast side. Wind speed was significantly lower at roosts sites compared with opposite sides of the same trees, presumably due to increased protection from prevailing west winds. Incorporating an estimate of convective cooling due to wind, we predicted daily thermal energy expenditure for normothermic bats and found that selected roost sites provided statistically significant energy savings (up to 1.60 ± 0.99 kJ/day) relative to the predicted expenditures if bats had roosted on the opposite sides of trees. Our results provide direct evidence that hoary bats select forest roosts on the basis of microclimate and suggest that roost requirements of foliage-roosting species may be more specific than has been previously assumed.