The environmental adaptation hypothesis (EAH) regarding birdsong dialects or ncighbourhoods states that song similarities between neighbouring individuals arise because of common influences on their songs exerted by the acoustic environment of their habitat. An assumption of the hypothesis is that sounds are distorted differently by different types of habitat. A prediction of the hypothesis is that some songs or parts of songs transmit better than others, depending on the habitat of their origin. We tested the assumption and prediction by comparing the attenuation and differential attenuation of pure tones, decreases in modal frequencies of computer simulated songs of American redstarts (Setophaga ruticilla), and the decay of redstart songs and white noise at deciduous, coniferous and open forest sites. The songs were representative of those used by redstarts living in thc three habitats. Results supported the assumption of acoustic differences between habitats but did not support the prediction that some songtypes transmit with less distortion in specific habitats than in others. The EAH also predicts that individuals which inhabit similar vegetation should share more song features than individuals which inhabitat dissimilar vegetation. To test this prediciton samples of songs were taken from the three habitats in different years. There were significant associations by habitat in both samples, but only one of several variables measured was significant and the discriminating variable was not the same for the two periods. Considering together the tests of the assumption and the two predictions, we conclude that for American redstarts evidence of the influence of the acoustic features of habitat on the formation of song dialects is mixed and not convincing.