Some difficulties in testing explanations for the occurrence of bird song dialects