In oscine songbirds, song phenotypes arise via gene-culture coevolution, in which genetically transmitted learning predispositions and culturally transmitted song forms influence one another’s evolution. To assess the outcome of this process in a population of chestnut-sided warblers (Dendroica pensylvanica), we recorded songs at intervals over a 19-year period. These recordings revealed the pattern of cultural evolution of songs in our study area, from which we inferred likely learning predispositions and mechanisms of cultural transmission. We found that the species’ two song categories form two distinct cultural traditions, each with its own pattern of change over time. Unaccented-ending songs have undergone continual, rapid turnover of song and element types, consistent with a model of neutral cultural evolution. Accented-ending songs, in contrast, persisted virtually unchanged for the entire study period, with extraordinarily constant song form and only one appearance of a new song type. Our results indicate that, in songbirds, multiple independent cultural traditions and probably multiple independent learning predispositions can evolve concurrently, especially when different signal classes have become specialized for different communicative functions.