No single theory so far proposed gives a wholly satisfactory account of the origin and maintenance of bird-song dialects. This failure is the consequence of a weak comparative literature that precludes careful comparisons among species or studies, and of the complexity of the issues involved. Complexity arises because dialects seem to bear upon a wide range of features in the life history of bird species. We give an account of the principal issues in bird-song dialects: evolution of vocal learning, experimental (...) findings on song ontogeny, dialect descriptions, female and male reactions to differences in dialect, and population genetics and dispersal.We present a synthetic theory of the origin and maintenance of song dialects, one that accommodates most of the different systems reported in the literature. The few data available suggest that large, regional dialect populations are genetically differentiated; this pattern is correlated with reduced dispersal between dialects, assortative mating by females, and male-male exclusion. At the same time, “subdialects” may be formed within regional dialects. Subdialect clusters are usually small and may represent vocal mimicry among a few adjacent territorial males. The relative importance of genetic and social adaptation may contribute to the emergence of subdialects; their distinctiveness may be correlated with the degree of polygyny, for example. Thus, subdialect formation is linked to one theory of the evolution of repertoire size, but data are too fragmentary to examine this idea critically. (shrink)