Previous studies of learned bird song have suggested the existence of species-universal patterns in song organization: clear clusters in produced songs that do not vary within a species. Here the authors combine a computational method of comparing songs with statistical methods of assessing cluster structure to investigate this issue in a more quantitative manner. The authors first analyze song phonology and then examine song syntax at a population level in 3 species with very different song structure: chaffinches (Fringilla coelebs), zebra finches (Taenopygia guttata), and swamp sparrows (Melospiza georgiana). The authors used a dynamic time-warping algorithm to compare song elements, which closely matched the judgments of human observers. Clustering tendency and validation statistics showed that broad phonological categories existed in all 3 species, but these categories explained no more than half of the overall phonological variation. The authors developed a novel statistic to assess syntactical structure, which indicated that element transitions were not randomly distributed. In the clearest case, in chaffinches, this could be explained by syllables being linked to certain positions within the song. These results demonstrate measures of song organization that can be applied across species, enhancing the potential of comparative studies.