Over the past 2 decades, we have begun to accumulate a basic understanding of the roosting and foraging ecology of temperate insectivorous bats in forests. As our understanding improves, it is not surprising there should be attempts at synthesizing our knowledge to prioritize future research directions (e.g., Hayes 2003-,-->Miller et al. 2003). -->Miller et al. (2003) reviewed results of 56 papers (1980–2001) and concluded that current data were unreliable because of small sample sizes, the short-term nature of studies, pseudoreplication, inferences beyond the scale of data collected, study design, and limitations of bat detectors and statistical analyses. Our concern is that this type of narrative synthesis that highlights limitations ignores any quantitative patterns that may exist. In this study we assess whether general patterns in North American bat use of roost trees and stand characteristics are robust enough to distill from the published literature. We used a series of meta-analyses on the same set of studies cited by Miller et al. (2003) to assess whether limitations of the current data warrant exclusion of bats from management recommendations. We used a second series of meta-analyses incorporating more recent data to determine the best current synthesis of knowledge on bat use of forests for roosting. In a third and fourth series of meta-analyses, we separated studies done on bats roosting in cavities versus roosting in foliage. In general, we found that, relative to other trees in the forest, the roost trees of bats were tall with large DBH in stands with open canopy and high snag density. In contrast, roost trees of bats did not differ from random trees with respect to live-tree density. The main differences we detected between foliage and cavity-roosting bats were in percent canopy cover and distance to water. The roost trees of cavity-roosting species had more open canopies and were closer to water than random trees. Our results clearly show that significant patterns can be detected from the literature when data sets are combined using a meta-analytic approach.