Savage et al. posit that if their main hypothesis is correct, namely that music serves to enhance social bonds, “across species, production or proficiency in ‘musical’ behaviors should predict both the number and complexity of social bonds.”

However, this is not borne out by cross-species evidence of natural vocal production behavior. Across species, social complexity and the need for enhanced social bonding is positively associated with the production of learned contact calls, but not with songs (Nowicki & Searcy, Reference Nowicki and Searcy2014; Sewall, Reference Sewall2015). Production of song appears to be independent from social complexity. Yet, considering their design features, songs are more music-like than learned contact calls. From a signaling theory perspective, these differences in form between songs and contact calls make functional sense.

Both Savage et al. and Mehr et al. highlight vocal learning as an important design feature of music. Hence, we restrict our discussion to learned vocalizations. Fitch (Reference Fitch2006) defines songs as “complex learned vocalizations” and notes that this definition “almost coincidentally” also applies to human song. However, not all complex vocal learners produce song, rather several use their vocal learning abilities to produce vocalizations commonly called contact calls. Marler's (Reference Marler, Peter and Slabbekoorn2004) distinction highlights their design differences: “On a structural level, songs are usually longer and more complex acoustically, involving a variety of different notes and syllables, ordered in statistically reliable sequences; calls are often short, monosyllabic with simple frequency patterning, delivered in what often appears to be a disorderly fashion.” This contrast between songs and calls suggests that songs are music-like in design (i.e., complex, with ordered notes and syllables in reliable sequences), which is why they may actually be called “songs,” whereas calls (i.e., simple, short, and disordered) are not. Savage et al. and Fitch (Reference Fitch2006) highlight repetition as a key design feature of music, distinguishing it from language. However, although animal songs may involve repetition on several hierarchical levels, contact calls do not (Catchpole & Slater, Reference Catchpole and Slater2008).

The distinction between songs and contact calls allows us to explore whether they differ regarding their relationship to social complexity across vocal learners. Savage et al. state that “melodic, learned song among songbirds, whales, or other vocal learners are predicted to enhance social bonding in these species.” However, among cetaceans, baleen whales sing (e.g., humpback whales produce long bouts of complex and hierarchically structured song) to attract mates rather than to enhance social bonds, although spending most of their time solitary or in small groups (Janik, Reference Janik2014; Whitehead & Rendell, Reference Whitehead and Rendell2012). In contrast, toothed whales have not been reported to sing but instead produce short individually distinctive learned contact calls to maintain social cohesion in their complex social systems (e.g., signature whistles in bottlenose dolphins) (King, Sayigh, Wells, Fellner, & Janik, Reference King, Sayigh, Wells, Fellner and Janik2013). Similarly, elephants live in complex fission–fusion societies, produce learned contact calls to maintain individual-specific bonds within changing social groupings, but are not known to sing (Poole, Tyack, Stoeger-Horwath, & Watwood, Reference Poole, Tyack, Stoeger-Horwath and Watwood2005). Many parrot species live in large fission–fusion social groups and produce short and simple, individually distinctive contact calls to mediate social and foraging dynamics. And although several parrot species also produce complex duets (e.g., yellow-naped amazon parrots), their primary function is to defend nest sites rather than to support social cohesion within these groups (Bradbury & Balsby, Reference Bradbury and Balsby2016). Finally, hummingbirds live quite solitary lives, but sing sometimes phonologically and syntactically quite complex learned songs to defend territories and attract mates (Araya-Salas et al., Reference Araya-Salas, Smith-Vidaurre, Mennill, González-Gómez, Cahill and Wright2019; Del Hoyo, Elliot, & Sargatal, Reference Del Hoyo, Elliot and Sargatal1999).

Hence, our qualitative comparative review of complex vocal learners suggests that learned contact calls enhance social cohesion and, correspondingly, are positively associated with social complexity (Nowicki & Searcy, Reference Nowicki and Searcy2014; Sewall, Reference Sewall2015). In contrast, songs function to attract mates and defend territories and do not appear to be associated with social complexity. This, therefore, counts as evidence against Savage et al.'s social bonding hypothesis as songs share important design features with music whereas learned contact calls do not.

From a signaling perspective, the observed form-function associations of these vocalizations make a lot of sense. The shortness, simplicity, and resulting behaviorally low-costliness of contact calls are consistent with the fact that they are cooperative signals, which do not need to persuade, but merely transparently and efficiently communicate relevant information, such as individual identity and group membership. In contrast, the elaborateness and associated costliness of songs (in terms of production and opportunity costs as well as predation risk, cf. Mehr et al.) correspond to the costly signaling prediction that songs should be costly to advertise and credibly indicate covert qualities (e.g., genetic quality). Hence, because music shares costly design features with animal songs rather than with contact calls, our comparative review lends support to Mehr et al.'s argument against Savage et al.'s social bonding hypothesis that, because a low-cost signaling system such as language can efficiently facilitate social coordination within groups, music would be needlessly costly to fulfill that role.

Instead of enhancing social bonding, Mehr et al. argue that music evolved as a costly, credible signal of covert qualities such as coalition strength and parental attention. Although we concur that music has the design features of a credible signal, plausibly for these functions, we are less convinced by both Mehr et al.'s and Savage et al.'s refutation of the sexual selection hypothesis, namely that music also evolved as a credible signal of mate quality. Their main argument against this hypothesis is the lack of musical sex differences. Yet, this counterargument is weakened by the fact that musical ability, production, and perception may be selected in both sexes under mutual mate choice (because of male parental investment), which is well-established in humans (Stewart-Williams & Thomas, Reference Stewart-Williams and Thomas2013). Indeed, also in many bird species both sexes sing and, for instance, the greater vasa parrot demonstrates that female song can function to attract males that take up a provisioning role (Ekstrom, Burke, Randrianaina, & Birkhead, Reference Ekstrom, Burke, Randrianaina and Birkhead2007; Riebel, Odom, Langmore, & Hall, Reference Riebel, Odom, Langmore and Hall2019). Hence, although we do not wish to deny potential weaknesses of the sexual selection hypothesis, the lack of sex differences might not necessarily be one of them.