The acoustic adaptation hypothesis was reviewed. This predicts that the sounds used by birds singing in a dense habitat should be of lower frequency than those in a more open habitat, and that sounds should be spaced out more in time in denser habitats to avoid degradation by reverberations. These predictions were tested by recording chaffinch (Fringilla coelebs) songs in open scrub, coniferous plantations, and a natural Scots pine forest, but the results obtained gave little support to the predictions made. Transmission of white noise and the songs of blue (F. teydea) arid Canary Islands chaffinches (F. c. tintillon) through laurel and Canarian pine forests on Tenerife revealed a sound window of less excess attenuation than expected in both habitats at frequencies of 2-3kHz. The song of the blue chaffinch appears to be better adapted for transmission through both habitats. Computer simulations of the formation of dialects by random copying of the songs possessed by neighbours predicted a strong effect of both the number of neighbours available to learn from, and the repertoire size, upon the songs' longevity and the number of birds sharing a particular song type. If the song type which was commonest amongst those sung by neighbours was learnt, larger groups of birds were found to share song types. At low copying error rates these groups approached the size of those described for the white-crowned sparrow (Zonotrichia leucophrys). A computer program which uses the dynamic programming algorithm to compare objectively shapes digitised from sonagrams was developed and tested with syllables from chaffinch songs.