We welcome the target article's main hypothesis: Music may originally have served as a credible signal, with diverse musical forms and features subsequently emerging via cultural evolution. What may be missing from the authors' analysis is an explicit model of the processes by which musical systems emerged and evolved. Such a model would build on the idea of signaling and would be applicable to simulate and study experimentally how patterns of musical behavior appear and change over time.

A widely-accepted proposal in cultural evolution theory is that population-level patterns partly emerge from individual cognitive biases amplified by cultural transmission (Kirby, Dowman, & Griffiths, Reference Kirby, Dowman and Griffiths2007). This population-level thinking is now supported by a growing number of studies using theoretical or experimental models of cultural transmission, studied using diffusion chains or more complex societies of interacting agents (Mesoudi, Reference Mesoudi2011). These methods aim to capture microevolutionary mechanisms at work at small-scale and over short time periods, reproducing data that matches actual historical patterns. Cultural transmission experiments (CTEs) offer valuable means to empirically test theoretical predictions with human participants, while maintaining much of the rigor and factor control of theoretical models (Mesoudi & Whiten, Reference Mesoudi and Whiten2008). In section 5.2 of their article, Mehr et al. do not mention this line of research, and they do not advocate using CTEs to test their claims on music evolution. To evaluate some of the hypotheses proposed in their article, Mehr et al. would need a model of signaling behavior that lends itself to experimental and computational analyses of how universal and diverse musical structures (Mehr et al., Reference Mehr, Singh, Knox, Ketter, Pickens-Jones, Atwood and Glowacki2019; Savage, Brown, Sakai, & Currie, Reference Savage, Brown, Sakai and Currie2015) emerge from individual behaviors and their neurocognitive underpinnings. A model that meets this requirement, and that embodies the idea of music as a credible signal, is the signaling game (Lewis, Reference Lewis1969; Skyrms, Reference Skyrms2010).

Signaling games in dyads or structured populations of senders and receivers have been widely applied to model coordination behaviors (Galantucci, Reference Galantucci2009). Recently, our team has adapted two-player signaling games in dyads and linear diffusion chains (Moreno & Baggio, Reference Moreno and Baggio2015; Nowak & Baggio, Reference Nowak and Baggio2016) to test hypotheses on the cultural transmission and evolution of music (Lumaca & Baggio, Reference Lumaca and Baggio2016, Reference Lumaca and Baggio2017; Lumaca, Haumann, Vuust, Brattico, & Baggio, Reference Lumaca, Haumann, Vuust, Brattico and Baggio2018, Reference Lumaca, Kleber, Brattico, Vuust and Baggio2019). In two studies (Lumaca & Baggio, Reference Lumaca and Baggio2017; Lumaca et al., Reference Lumaca, Haumann, Vuust, Brattico and Baggio2018), senders and receivers were arranged in linear diffusion chains of several generations each: Each game between two players modeled an interaction between two adjacent generations of learners. The signals were tone sequences, paired with affective meanings: The pairing emerged as a result of player coordination in a game. At the end of each game, the receiver (generation n) became the sender in the next game (n + 1). Senders were asked to transmit the musical code (signals and mappings to meanings) learned in the previous game. In cooperative signaling games, there is an incentive to signal honestly and credibly in order to coordinate more rapidly with the other player: In all our studies, receivers eventually learned the mapping that senders used in a game. However, more relevant from the point of view of the target article is evidence of convergent evolution. Small transmission errors, likely driven by individual biases, accumulated in the musical code. Each transmission chain developed its own “musical culture” based on patterns of melodic and rhythmic structure. Thus, we demonstrated experimentally that individual biases, brought out by intergenerational signaling, can lead to convergence toward attested musical patterns.

Modifications of the signaling games paradigm could be useful to test other hypotheses from the target article. One is the cumulative increase in the complexity and diversity of signals, particularly in groups where signalers have conflicting interests (sect. 5.2). Signaling games are flexible enough to accommodate several network structures – from simple dyads to games with many senders and receivers – and payoff structures – from shared to conflicting interests between signalers. To address the former hypothesis, one could organize senders and receivers into “microsocieties” (Baum, Richerson, Efferson, & Paciotti, Reference Baum, Richerson, Efferson and Paciotti2004) of several interacting individuals, where player payoffs would either differ (experimental groups) or not (control groups). The generational progression would be recreated by replacing the longest-standing members of the groups with naive players. Finally, the complexity and diversity of signals could be quantified (Miton & Charbonneau, Reference Miton and Charbonneau2018) and compared between groups and across generations.

Importantly, some of our experimental results diverge from a core proposal by Mehr et al.: the music-specificity of cultural attractors (sect. 3.1). In two studies (Lumaca & Baggio, Reference Lumaca and Baggio2016; Lumaca et al., Reference Lumaca, Haumann, Vuust, Brattico and Baggio2018), we used signaling games in combination with electroencephalogram (EEG) to test the idea that music adapts to auditory perception mechanisms (Trainor, Reference Trainor2015). We recorded participants' brain responses in an auditory oddball task, which evoked an ERP signature of auditory scene analysis (ASA): the mismatch negativity (MMN) (Näätänen, Gaillard, & Mäntysalo, Reference Näätänen, Gaillard and Mäntysalo1978). Another day, participants played in one signaling game as receivers and in a subsequent game as senders of a musical code. We showed that individual MMN latencies, which reflect ASA efficiency, predict the degree of melodic (Lumaca & Baggio, Reference Lumaca and Baggio2016) and rhythmic structures (Lumaca et al., Reference Lumaca, Haumann, Vuust, Brattico and Baggio2018) introduced in the code. These findings trace the origins of core musical structures to neural mechanisms of ASA, which are arguably phylogenetically older than human musicality and are fairly widely conserved across species.

We argue that Mehr et al. could take advantage of the signaling games model to refine, constrain, and empirically test their hypothesis on the origins of music as a credible signal. Our experiments are a highly simplified model of signaling behavior and music transmission, yet they tap into the essential mechanisms which we suspect are at work in the emergence and evolution of music as a cultural symbolic system. Ultimately, the study of music's origins demands a joint effort across different disciplines and methods, including behavior and neuroscience. But, a unification of methods and results is unlikely to happen in the absence of a model and paradigm that can guide research. Signaling games can take on such a unifying role, especially if we accept the idea that human symbolic systems, including music, are systems of culturally transmitted credible signals.