Humpback whales (Megaptera novaeangliae) are seasonal breeders, annually migrating from high-latitude summer feeding grounds to low-latitude winter breeding grounds. The social matrix on the winter grounds is a loose network of interacting individuals and groups and notably includes lone males that produce long bouts of complex song that collectively yield an asynchronous chorus. Occasionally, a male will sing while accompanying other whales. Despite a wealth of knowledge about the social matrix, the full characterization of the mating system remains unresolved, without any firm consensus, as does the function of song within that system. Here, I consider and critically analyse three proposed functions of song that have received the most attention in the literature: female attraction to individual singers, determining or facilitating male–male interactions, and attracting females to a male aggregation within the context of a lekking system. Female attraction suggests that humpback song is an advertisement and invitation to females, but field observations and song playback studies reveal that female visits to individual singers are virtually absent. Other observations suggest instead that females might convey their presence to singers (or to other males) through the percussive sounds of flipper or tail slapping or possibly through vocalizations. There is some evidence for male–male interactions, both dominance and affiliative: visits to singers are almost always other lone males not singing at that time. The joiner may be seeking a coalition with the singer to engage cooperatively in attempts to obtain females, or may be seeking to disrupt the song or to affirm his dominance. Some observations support one or the other intent. However, other observations, in part based on the brevity of most pairings, suggest that the joiner is prospecting, seeking to determine whether the singer is accompanying a female, and if not soon departs. In the lekking hypothesis, the aggregation of vocalizing males on a winter ground and the visits there by non-maternal females apparently for mating meet the fundamental definition of a lekking system and its role though communal display in attracting females to the aggregation, although not to an individual singer. Communal singing is viewed as a form of by-product mutualism in which individuals benefit one another as incidental consequences of their own selfish actions. Possibly, communal singing may also act to stimulate female receptivity. Thus, there are both limitations and merit in all three proposals. Full consideration of song as serving multiple functions is therefore necessary to understand its role in the mating system and the forces acting on the evolution of song. I suggest that song may be the prime vector recruiting colonists to new winter grounds pioneered by vagrant males as population pressures increase or as former winter grounds become unavailable or undesirable, with such instances documented relatively recently. Speculatively, song may have evolved historically as an aggregating call during the dynamic ocean conditions and resulting habitat uncertainties in the late Miocene–early Pliocene epochs when Megaptera began to proliferate. Early song may have been comprised of simpler precursor sounds that through natural selection and ritualization evolved into complex song.