Summary • Habitat structure has been considered as a main factor shaping the evolution of bird song acoustics. • Based on expected differential patterns of sound degradation in different habitats, the acoustic adaptation hypothesis (AAH) proposes that songs with lower frequencies, narrower frequency ranges and longer inter-element intervals should occur more frequently in densely vegetated compared with herbaceous habitats. • Empirical tests of the AAH have provided mixed results. Here, we review for the first time the literature on this topic using a meta-analytical, quantitative approach. • Maximum, minimum, peak frequency and frequency range were found to be significantly lower in closed compared with open habitats, but the mean size of the effect of habitat was small. Inter-element intervals were not affected by habitat structure. • The AAH implicitly assumes that birds are selected to maximize song broadcast range and minimize degradation, while it neglects the potential role of energetic costs of singing and selection by eavesdroppers (e.g. predators and parasites). • Our meta-analysis supports the AAH, but habitat structure only weakly predicts the acoustical properties of bird songs. Thus, other potentially relevant factors should be included in realistic models of the evolution of bird song acoustics. Functional Ecology (2007) 21, 134–142 doi: 10.1111/j.1365-2435.2006.01207.x