Birdsong is a complex cultural and biological system, and the selective forces driving evolutionary changes in aspects of song learning vary considerably among species. The extent to which repertoire size, the number of syllables or song types sung by a bird, is subject to sexual selection is unknown, and studies to date have provided inconsistent evidence. Here, we propose that selection pressure on the size and complexity of birdsong repertoires may facilitate the construction of a niche in which learning, sexual selection, and song-based homophily may co-evolve. We show, using a review of the birdsong literature and mathematical modeling, that learning mode (open-ended or closed-ended learning) is correlated with the size of birdsong repertoires. Underpinning this correlation may be a form of cultural niche construction in which a costly biological trait (for example, open-ended learning) can spread in a population (or be lost) as a result of direct selection on an associated cultural trait (for example, song repertoire size).