2.1 Ancestral bonding mechanisms

Why was social bonding adaptive for our ancestors, and in what ways does music improve or increase social bonding? Group living comes with costs (e.g., increased local competition for food and mates) and benefits (e.g., safety in numbers and cooperative hunting/defense). Animals that live in groups, particularly primates, have evolved mechanisms that help balance these costs and benefits by forging strong affiliative bonds: good quality, persistent, differentiated inter-individual commitments that require investment of time and energy (Dunbar, Reference Dunbar1991). Strong social bonds enhance individuals' prospects of receiving support through coalitions, which, in certain primate species, influence dominance rank and reproductive performance (Silk, Reference Silk2007). These coalitions form the backbone of successful cooperative hunting, child rearing, and joint defense against predators or competitors (Dunbar & Shultz, Reference Dunbar and Shultz2010). Ecological factors typically constrain the size of a group, but larger groups of well-coordinated, strongly bonded humans enabled exploitation of new forms of resources (e.g., larger prey), and more reliable protection from predators (Dunbar, Reference Dunbar2012b).

ABMs in other primates include grooming, play, and – in some species – non-procreational sex. These ABMs are essentially dyadic (or for play, very small groups mostly limited to young animals), and require substantial time commitments even in small groups if all individuals in the group are to invest in all others. Although vocal duets are present in tropical birds and some primates (Farabaugh, Reference Farabaugh, Kroodsma and Miller1982; Haimoff, Reference Haimoff1986; Mann, Dingess, Barker, Graves, & Slater, Reference Mann, Dingess, Barker, Graves and Slater2009; Thorpe, Reference Thorpe1972), group vocal choruses that are both differentiated and coordinated appear nearly unique to humans (but see Mann, Dingess, & Slater [Reference Mann, Dingess and Slater2006] for the fascinating example of the group-chorusing plain-tailed wren).

As Dunbar (Reference Dunbar1993) has argued, the steady increases in group size, complexity, and fluidity that occurred during hominin evolution put increasing strain on ABM-based social bonds. Beyond group sizes of 20 or so, dyadic bonding based on ABMs such as grooming became unsustainably time-consuming, so supra-dyadic bonding mechanisms were needed. Dunbar (Reference Dunbar and Bannan2012a) suggests that another ABM in great apes and humans was laughter (Davila Ross, Owren, & Zimmerman, Reference Davila Ross, Owren and Zimmerman2009), which facilitates social bonds among reasonably large groups. However, there are limits to a bonding mechanism based on laughter: Unlike music, which people can intentionally choose to engage in at any time, large group laughter can be difficult to elicit and to sustain for long periods. Music may have provided our ancestors with a novel system that, like laughter, allowed for simultaneous bonding with a larger group of individuals, but across a broader set of times and contexts, and for longer periods of time than otherwise possible (Dunbar, Reference Dunbar and Bannan2012a; Launay, Tarr, & Dunbar, Reference Launay, Tarr and Dunbar2016). This new system augmented the smaller-scale ABMs that became less robust in larger groups. Specific design features of human musicality – particularly our capacity and proclivity to produce repetitive, synchronized, harmonized music for extended periods – provided a flexible toolkit for bonding, allowing our ancestors to achieve social bonding on a large scale.

2.2 Design features of musicality

2.3 Gene–culture coevolution

These specific design features and their interactions – dancing to an isochronous beat with a metrical hierarchy, singing learned melodies based on discrete scales in harmony, using predictable, repetitive musical structures, and using musical performances as cues for social identity – are widespread throughout the world's musical systems (Savage et al., Reference Savage, Brown, Sakai and Currie2015; see sect. 3.1). These features have clear functions for group performance, but little or no function in solo performance (hence their rarity in birdsong, whale song, and certain solo human music genres such as lament; Frigyesi, Reference Frigyesi1993; Tolbert, Reference Tolbert1990). These design features are therefore predicted a priori by the MSB hypothesis, but not by solo signaling hypotheses such as sexual selection for mate attraction (Miller, Reference Miller, Wallin, Merker and Brown2000) or maternal singing to infants (Mehr & Krasnow, Reference Mehr and Krasnow2017; Mehr et al., target article). Although these features promote coordination in dyadic music (e.g., duets) and memorability/communicative power in solo music (e.g., lullabies; Cirelli & Trehub, Reference Cirelli and Trehub2020; Corbeil, Trehub, & Peretz, Reference Corbeil, Trehub and Peretz2016), their added value in supporting extended, coordinated group performances is most evident for larger groups.

MSB posits an extended timeline in which different core mechanisms of musicality arose through a coevolutionary “virtuous spiral.” Although many of the specific design features above could in principle function independent of the others, and would prove adaptive independently at any proto-musical stage, over evolutionary time we hypothesize that isochronous beats coevolutionarily enabled meter and dance, and that pitched singing enabled scale-based melody and harmony. Each new feature added value in supporting extended, coordinated, harmonious group performance. Each feature may have been initially based on behavioral innovations involving synchronization of the ancestrally individualistic displays seen in other great apes (e.g., chimpanzee pant-hoot displays and fruit tree “carnival” displays, cf. Merker, Reference Merker1999; Merker et al., Reference Merker, Morley, Zuidema and Honing2018). However, each innovation opened a new cognitive/musical niche selecting for independent specialization of relevant neural circuitry (see sect. 4).

Early instantiations of music provided selective preconditions for later cognitive and neurobiological changes underlying human musicality, analogous to the well-documented examples of gene–culture coevolution involving fire and dairy farming. Cultural innovations created a variety of proto-musical behaviors, with musical knowledge becoming a potential cue to social history and cultural (sub-)group membership. For example, this could have created selective feedback favoring individuals who used music as cues to group membership. Together, biological and cultural coevolution created a framework for the coordinated, harmonious, emotional group performances that are evident today throughout the world's musical cultures. The major inter-relationships among these components of human musicality are summarized in Figure 2 (but see sect. 6.3 for caveats regarding causality in our proposed coevolutionary mechanisms).

Figure 2. Proposed coevolutionary relationships among multiple musical features and mechanisms, indicating their contributions to ultimate functions by facilitating social bonding in multiple ways, their proximate neurobiological underpinnings in prediction and reward systems, and feedback loops among these different levels.

2.4 Benefits of social bonding

We hypothesize that musicality increased the number of “simple” relationships (e.g., “friends”), and increased the quality (depth and complexity) of existing relationships. The opportunity for many individuals to participate productively in social interaction through proto-musical behaviors facilitates an efficient bonding mechanism for groups of varying sizes, thereby conferring associated benefits (as outlined in sect. 2.1). However, we must consider the nature of the subsidiary relationships and social structures in which they operate. Many vertebrate species live in large groups (e.g., fish schools, bird flocks, and ungulate herds), but do not exhibit strong social bonds with more than a small number of individuals, and/or the relationships are undifferentiated. Indeed, the “number of differentiated relationships” (Bergman & Beehner, Reference Bergman and Beehner2015) can vary independently from raw group size. For example, a monogamous pair with bi-parental care involves two differentiated relationships (sexual mate, and caregiving partner) or even three (adding joint territory defense), a situation typical in many birds. The social bonding design features we have identified can operate at multiple levels simultaneously, in the same way that a couple dancing at a party can intensify their own relationship, and their relationship with the broader social group.

2.5 Participatory versus presentational music

For most of hominin evolution, the only way to experience music was to make it oneself, or to observe others making music in real time. But as music-making technology culturally evolved, opportunities for solo listening increased (e.g., recording technology and personal music-playing devices) and individual virtuosity became increasingly emphasized. Cross-cultural analyses suggest that forms of music-making coevolved in parallel with social structures: larger-scale, more hierarchical societies tend to emphasize “presentational” music made by small numbers of performers for large numbers of passive (or virtual) audiences. Conversely, smaller-scale, more egalitarian societies tend to emphasize “participatory” music in which large groups sing, dance, and play instruments together with little or no distinction between performers and audience (Lomax, Reference Lomax1968; Turino, Reference Turino2008). Once group size increases substantially, it may not be feasible for all individuals to participate actively in a coordinated manner, but music can facilitate bonding via passive (including digital) participation. This enables music (e.g., national anthems) to help construct social identities even among massive “imagined communities” (Anderson, Reference Anderson1991) whose members may never physically interact with one another.

The participatory mode of musical performance is hypothesized to be the ancestral one that operated over long time scales. It is imperative to avoid conflating pervasive technology-driven aspects of contemporary musical practice (e.g., static audiences, solo listening, and control by global corporations) with the conditions under which humans experienced music during most of our evolutionary history. As a result, testing predictions of the MSB hypothesis should favor contexts such as drumming circles, campfire singalongs, and folk dances over solo-listening via headphones, or collective, static listening at a Mozart performance. Even in societies dominated by presentational music, participatory contexts retain their social and emotional potency, as highlighted by the collective singing of Italians from their balconies during the coronavirus lockdown (Grahn, Bauer, & Zamm, Reference Grahn, Bauer and Zamm2020; Horowitz, Reference Horowitz2020; Kornhaber, Reference Kornhaber2020).

2.6 Summary

Summarizing, the MSB hypothesis argues that music is a derived bonding mechanism, akin to but augmenting previous ABMs such as grooming and laughter. This augmentation occurs via the provision of a shared framework for individual participants to establish and maintain strong bonds with more than one individual (or a small group of individuals) at a time, thus bridging the “bonding gap” problem posed during human evolution by increasing group size and complexity (Dunbar, Reference Dunbar1993, Reference Dunbar2012b). Proto-musical features may initially have arisen as behavioral innovations that later initiated a process of gene–culture coevolution. Crucially, the design features of musicality discussed above make music better suited than ABMs or language for coordinating behavior and facilitating social bonding in larger and more complex groups.

3.1. Cross-cultural evidence

One line of evidence for the MSB hypothesis comes from the study of cross-cultural musical universals (Brown & Jordania, Reference Brown and Jordania2013; Lomax, Reference Lomax1968; Mehr et al., Reference Mehr, Singh, Knox, Ketter, Pickens-Jones, Atwood and Glowacki2019; Nettl, Reference Nettl2015; Savage, Reference Savage2018, Reference Savage and Sturman2019b; Savage & Brown, Reference Savage and Brown2013; Stevens & Byron, Reference Stevens, Byron, Hallam, Cross and Thaut2016; Trehub et al., Reference Trehub, Becker, Morley and Honing2018). Music, like language, is a human universal found in all known cultures (Brown, Reference Brown1991; Mehr et al., Reference Mehr, Singh, Knox, Ketter, Pickens-Jones, Atwood and Glowacki2019). Few if any specific musical features are found in all known musics, just as few specific linguistic features are found in all known languages (Evans & Levinson, Reference Evans and Levinson2009). However, researchers have identified dozens of “statistical universals” that predominate throughout diverse samples of the world's music, relating both to functional context and to musical structure (Mehr et al., Reference Mehr, Singh, Knox, Ketter, Pickens-Jones, Atwood and Glowacki2019; Savage et al., Reference Savage, Brown, Sakai and Currie2015; Table 1). These cross-cultural similarities suggest selection by biological and/or cultural evolution.

Table 1. Cross-culturally widespread musical structures and functions

Functional contexts were found by Mehr et al. (Reference Mehr, Singh, Knox, Ketter, Pickens-Jones, Atwood and Glowacki2019) to be associated with singing in ethnographic descriptions of the 60 societies from the Human Relations Area Files Probability Sample (Lagacé, Reference Lagacé1979). Musical structures were found by Savage et al. (Reference Savage, Brown, Sakai and Currie2015) to predominate (items 1–18) or to co-occur (item 19) consistently in each of nine world regions across a sample of 304 audio recordings from the Garland Encyclopedia of World Music (Nettl, Stone, Porter, & Rice, Reference Nettl, Stone, Porter and Rice1998–2002). Nested relationships are indicated with indented italics.

^a Indicates associations that were only significant using one of the two methods reported by Mehr et al. (Reference Mehr, Singh, Knox, Ketter, Pickens-Jones, Atwood and Glowacki2019) (Mehr et al. [Reference Mehr, Singh, Knox, Ketter, Pickens-Jones, Atwood and Glowacki2019] used two methods to examine universal associations with singing: “topic annotations from the Outline of Cultural Materials [‘OCM identifiers’] and automatic detection of related keywords.” The second method was needed “because some hypotheses correspond only loosely to the OCM identifiers (e.g., ‘love songs’ is only a partial fit to ARRANGING A MARRIAGE [the OCM identifier used] and not an exact fit to any other identifier).” Similarly, “group bonding” is only a partial fit to the OCM identifier “SOCIAL RELATIONSHIPS AND GROUPS,” which covers a broader range of social behaviors than simply “group bonding.” After adjusting for ethnographer bias and multiple comparisons, Mehr et al. found “support from both methods for 14 of the 20 hypothesized associations between music and a behavioral context, and support from one method for the remaining six.” See Mehr et al. [Reference Mehr, Singh, Knox, Ketter, Pickens-Jones, Atwood and Glowacki2019] for further details).

Crucial to our hypothesis, music performs similar social bonding functions across cultures. All of the 20 widespread functional contexts supported by at least one analysis in Mehr et al. (Reference Mehr, Singh, Knox, Ketter, Pickens-Jones, Atwood and Glowacki2019) summarized in Table 1 relate to social bonding, particularly through the ubiquitous use of music in communal ceremonies and rituals (e.g., healing, procession, mourning, storytelling, greeting visitors, praise/religion, and weddings). Even the secular use of music as art or entertainment is itself often a form of communal ritual. For example, aspects of Western art music concert attendance function to cement social bonds between participants and exclude non-participants in similar ways to other elite rituals throughout history (Nooshin, Reference Nooshin2011; Small, Reference Small1998). Other non-ritual contexts have social bonding functions in bringing together parents and infants (lullabies and play songs), mates (love songs), or coordinating activities among multiple individuals (work songs and dance music). Finally, regulation of moods/emotions is one of the key components of our definition of social bonding (“…synchronizing and harmonizing the moods, emotions, actions, or perspectives of two or more individuals”). Even mood regulation via solo music can support social functions or evoke social contexts. For example, people may ease the pain of separation from loved ones by listening to or playing music that evokes shared memories (Kornhaber, Reference Kornhaber2020), or use music to prepare their mood for an effective social interaction, allowing them to regulate their behavior and behave in the socially-expected manner (Erber, Wegner, & Therriault, Reference Erber, Wegner and Therriault1996; Greenwood & Long, Reference Greenwood and Long2009).

Similarly, most of the widespread structural aspects of music support coordinated music-making. Throughout the world, humans tend to sing, play percussion instruments, and dance to simple, repetitive music in groups, and this is facilitated by the widespread use of simple-integer pitch and rhythm ratios, scales based on a limited number of discrete pitches (≤7), and isochronous beats grouped in multiples of two or three (Bowling & Purves, Reference Bowling, Hoeschele, Gill and Fitch2015; Jacoby & McDermott, Reference Jacoby and McDermott2017; Jacoby et al., Reference Jacoby, Polak, Grahn, Cameron, Lee, Godoy and McDermottPreprint; Kuroyanagi et al., Reference Kuroyanagi, Sato, Ho, Chiba, Six, Pfordresher and Savage2019; Ravignani, Delgado, & Kirby, Reference Ravignani, Delgado and Kirby2017; Savage et al., Reference Savage, Brown, Sakai and Currie2015). The widespread use of simple, discrete meters and scales also enables multiple people to memorize and coordinate their performances. These widespread musical properties have few direct parallels in language. Group coordination provides a common purpose that unifies the cross-cultural structural regularities of human music (Savage et al., Reference Savage, Brown, Sakai and Currie2015).

3.2 Fossil and archeological evidence

Although music itself leaves no fossil record, inferences can be drawn from evidence about the evolution of musicality, the role this played in early human society, and its relationship to other evolutionary developments such as brain size, language, group size, and sociality (Mithen, Reference Mithen2005; Morley, Reference Morley2013). The fossil record for human evolution indicates that capacities for sophisticated and diverse vocalizations and body language, including dancing, were present before there is credible evidence for compositional language (as reviewed in Mithen, Reference Mithen2005). Archeological evidence from the Paleolithic indicates increasing group size and long-distance contacts (Gamble, Reference Gamble, Dunbar, Gamble and Gowlett2010; Read & Van der Feeuw, Reference Read, Van der Feeuw, Coward, Hosfield, Pope and Wenban-Smith2015), suggesting that ABMs had become insufficient by at least 2 million years ago. The earliest surviving musical instruments – bone flutes – have been dated to over 35,000 years ago and are speculated to have functioned to support larger social networks (Conard, Malina, & Münzel, Reference Conard, Malina and Münzel2009). Prehistoric rock art often appears to be positioned with regard to the acoustic properties of either the cave or cliff face on which it is located (e.g., Fazenda et al., Reference Fazenda, Scarre, Till, Pasalodos, Guerra, Tejedor and Foulds2017; Rainio, Lahelma, Aikas, Lassfolk, & Okkonen, Reference Rainio, Lahelma, Aikas, Lassfolk and Okkonen2018), suggesting that music played a role in the social-bonding rituals associated with that art. Similarly, prehistoric and early historic architecture used for social-bonding ceremonies often appears to have been designed with regard to its acoustic properties and to facilitate music making (e.g., Göbekli Tepe: Notroff, Dietrich, & Schmidt, Reference Notroff, Dietrich, Schmidt, Renfrew, Boyd and Morely2015; Stonehenge and other Neolithic monuments in Britain: Banfield, Reference Banfield2009; Watson & Keating, Reference Watson and Keating1999; and Ancient Mayan temples: Sanchez, Reference Sanchez2007).

3.3. Developmental evidence

Extensive evidence demonstrating spontaneous and early development of social functions of music also supports the MSB hypothesis. Adults around the world produce infant-directed songs, such as lullabies, with similar, cross-culturally recognizable acoustic features (Mehr, Singh, York, Glowacki, & Krasnow, Reference Mehr, Singh, York, Glowacki and Krasnow2018; Trehub, Unyk, & Trainor, Reference Trehub, Unyk and Trainor1993). Song is highly effective at emotional modulation in infants – reliably more effective than speech, with infants exhibiting longer visual fixations and greater reductions in stress and body movement to maternal singing than to speaking (Cirelli & Trehub, Reference Cirelli and Trehub2020; Corbeil et al., Reference Corbeil, Trehub and Peretz2016; Ghazban, Reference Ghazban2013; Nakata & Trehub, Reference Nakata and Trehub2004; Trehub, Reference Trehub, Hallam, Cross and Thaut2016). Infants also respond differently to songs sung in different styles (e.g., lullaby vs. playsong; Cirelli, Jurewicz, & Trehub, Reference Cirelli, Jurewicz and Trehub2019; Rock, Trainor, & Addison, Reference Rock, Trainor and Addison1999). Singing to infants thus appears to serve a communicative function, allowing parents to communicate specific emotional messages to infants before they can understand the semantic content of language (Rock et al., Reference Rock, Trainor and Addison1999; Trainor, Clark, Huntley, & Adams, Reference Trainor, Clark, Huntley and Adams1997; Trehub et al., Reference Trehub, Unyk, Kamenetsky, Hill, Trainor, Henderson and Saraza1997). Singing and musical interactions also directly improve parent–infant social bonds: Interventions promoting singing and musical interaction between parents and infants strengthen parents' attachment to their infants, more so than nonmusical play (Vlismas, Malloch, & Burnham, Reference Vlismas, Malloch and Burnham2013). Music thus facilitates both parent–infant communication and parent–infant bonding from early in life, before extensive experience or opportunities for learning.

Beyond infancy, musical activities continue to promote bonding: Across a range of tasks, group musical involvement increases children's prosocial behavior. Thus, young children act more prosocially (in terms of sharing and fairness) after a musical game than a similar non-musical game (Kirschner & Tomasello, Reference Kirschner and Tomasello2010); after group singing than group art or competitive games (Good & Russo, Reference Good and Russo2016); and after joint synchronized, rhythmic movement than non-synchronized movement (Rabinowitch & Meltzoff, Reference Rabinowitch and Meltzoff2017).

Children (like adults) choose to affiliate with members of their own social group (Bigler, Jones, & Lobliner, Reference Bigler, Jones and Lobliner1997). From early infancy, music serves as a marker of social group membership, allowing for the identification of preferred social partners (Cirelli, Trehub, & Trainor, Reference Cirelli, Trehub and Trainor2018). Shared knowledge of specific songs serves as a particularly informative signal of common group membership: because of the wide range of forms a song can take, knowledge of a particular song implies common social or cultural background (Soley & Spelke, Reference Soley and Spelke2016). Infants accordingly treat shared musical knowledge as socially meaningful from early in life: 5-month-old infants prefer to look at people who sing melodies previously sung by a parent, over people who sing melodies previously sung by an unfamiliar adult (Mehr, Song, & Spelke, Reference Mehr, Song and Spelke2016). These early preferences appear to form the foundation for selective social affiliations based on music: At preschool age, children use knowledge of a familiar song as a social cue to select friends (Soley & Spelke, Reference Soley and Spelke2016), and by 14 months exhibit more prosocial behavior (helping) toward an unfamiliar woman who sings a familiar song (previously sung by a parent) than an unfamiliar song (Cirelli & Trehub, Reference Cirelli and Trehub2018). Together, these results suggest that musical knowledge shapes the formation of children's social bonds, and that the link between shared musical knowledge and social connection is rooted in early infancy.

3.4. Social psychological evidence

Behavioral experiments from social psychology support the MSB hypothesis, suggesting that musical behavior is not only associated with, but may causally support, social bonding. In particular, music provides a foundation for synchronized behavior in large groups (as argued above), and a number of experiments and meta-analyses show that rhythmic synchronization with other individuals promotes increased prosocial behavior (i.e., actions that increase others' well-being; Mogan, Fischer, & Bulbulia, Reference Mogan, Fischer and Bulbulia2017; Rennung & Göritz, Reference Rennung and Göritz2016). Synchrony has been empirically linked to cooperation in economic games (Lang, Bahna, Shaver, Reddish, & Xygalatas, Reference Lang, Bahna, Shaver, Reddish and Xygalatas2017; Launay, Dean, & Bailes, Reference Launay, Dean and Bailes2013; Reddish, Bulbulia, & Fischer, Reference Reddish, Bulbulia and Fischer2014; Wiltermuth & Heath, Reference Wiltermuth and Heath2009), entitativity (feelings of being on the same team; Lakens & Stel, Reference Lakens and Stel2011; Reddish, Fischer, & Bulbulia, Reference Reddish, Fischer and Bulbulia2013), rapport and interpersonal liking (Hove & Risen, Reference Hove and Risen2009; Miles, Nind, & Macrae, Reference Miles, Nind and Macrae2009; Valdesolo & Desteno, Reference Valdesolo and Desteno2011), and helping behavior (Cirelli, Einarson, & Trainor, Reference Cirelli, Einarson and Trainor2014; Kokal, Engel, Kirschner, & Keysers, Reference Kokal, Engel, Kirschner and Keysers2011; Valdesolo & Desteno, Reference Valdesolo and Desteno2011). Similarly, dancing in synchrony increases participants' feelings of connectedness to the group with which they are dancing, as well as their liking and assessment of similarity with co-dancers (Tarr, Launay, Cohen, & Dunbar, Reference Tarr, Launay, Cohen and Dunbar2015; Tarr, Launay, & Dunbar, Reference Tarr, Launay and Dunbar2016). These prosocial effects of synchrony are robust in different contexts (Mogan et al., Reference Mogan, Fischer and Bulbulia2017). Although demand characteristics have been suggested as possible confounds underlying these effects (Atwood, Mehr, & Schachner, Reference Atwood, Mehr and Schachner2020; Rennung & Göritz, Reference Rennung and Göritz2016), significant prosocial effects of synchrony remain after potential confounds of suggestion, competence, and shared intention are eliminated (e.g., in a virtual reality setting; Tarr, Slater, & Cohen, Reference Tarr, Slater and Cohen2018). However, meta-analyses implied inconclusive results regarding the precise roles of “music” and of synchrony to an isochronous beat, as opposed to more generally synchronized or coordinated non-musical behaviors such as gaze synchrony, affect synchrony, and motor synchrony (Mogan et al., Reference Mogan, Fischer and Bulbulia2017; Rennung & Göritz, Reference Rennung and Göritz2016). In sect. 5, we propose clearer predictions and tests of specific mechanisms by which music promotes social bonding.

More broadly, behavioral studies indicate varied social bonding effects associated with music-based activities, even those that do not explicitly involve constant synchrony. Young children randomly assigned to activities incorporating music exhibit elevated levels of empathy compared to non-musical controls in longitudinal studies (Rabinowitch, Cross, & Burnard, Reference Rabinowitch, Cross and Burnard2013), and adults singing in regular group sessions develop feelings of social closeness toward co-participants more quickly than people engaged in other (non-musical) group activities (Pearce, Launay, & Dunbar, Reference Pearce, Launay and Dunbar2015). Feelings of inclusion, connectivity, and positive affect emerge in small and large singing groups, with participants in large choirs (>80 participants) reporting greater changes in these measures compared to smaller choirs (Weinstein, Launay, Pearce, Dunbar, & Stewart, Reference Weinstein, Launay, Pearce, Dunbar and Stewart2016). These findings highlight the relevance of music-based activities for large-scale social bonding.

4.1 Perception–action coupling

Perception–action coupling refers to anatomical and/or functional connectivity between brain regions involved in sensory perception (e.g., of pitch or rhythm) and those that are involved in movement (e.g., vocalization and dance). Specifically, auditory–motor coupling is a key neural mechanism that underlies social bonding through music because it enables individuals to synchronize and/or harmonize their own music and actions with others, which is crucial for coordinated group music making. Even during the perception of solo music, the tight coupling between perceptual and motor regions leads to spontaneous and obligatory activity in premotor and supplementary motor areas, classic motor areas that are also part of the action observation network that drives physical and observational learning (Cross, Kraemer, Hamilton, Kelley, & Grafton, Reference Cross, Kraemer, Hamilton, Kelley and Grafton2008).

Rhythm and beat consistently activate the premotor area, supplementary motor area, and basal ganglia, regions commonly thought to belong to the motor system (Grahn & Brett, Reference Grahn and Brett2007). Furthermore, the auditory system is strongly coupled with areas in the motor system during rhythm perception (Grahn & Rowe, Reference Grahn and Rowe2009), and rhythmic oscillatory activity in both the auditory and motor systems tracks the rhythm of music (Fujioka, Ross, & Trainor, Reference Fujioka, Ross and Trainor2015). Some observations show that neural phase-locking activity is even higher in music than in speech (Vanden Bosch der Nederlanden, Joanisse, & Grahn, Reference Vanden Bosch der Nederlanden, Joanisse and Grahn2020). This process of “neuronal entrainment” (neural activity changing its frequency, amplitude, and/or phase in response to external stimulation) is a proposed mechanism through which rhythm in sensory stimuli affects the brain by coordinating activity between separate neuronal populations, such as between the auditory and motor systems (Jones, Reference Jones2018; Morillon & Baillet, Reference Morillon and Baillet2017). This neuronal entrainment enables selective attention to specific points in time (Lakatos et al., Reference Lakatos, Karmos, Mehta, Ulbert and Schroeder2008; Large & Jones, Reference Large and Jones1999). In particular, auditory–motor coupling is strongest when perceiving “high groove” music that elicits the pleasurable drive toward action such as in dance (Janata, Tomic, & Haberman, Reference Janata, Tomic and Haberman2012). Groovy music elicits the urge to dance by increasing the auditory cortex's sensitivity and its coupling with the motor cortex (Stupacher, Hove, Novembre, Schutz-Bosbach, & Keller, Reference Stupacher, Hove, Novembre, Schutz-Bosbach and Keller2013), which is particularly evident with medium levels of rhythmic complexity and expectation violation (Koelsch et al., Reference Koelsch, Vuust and Friston2019; Witek et al., Reference Witek, Clarke, Wallentin, Kringelbach and Vuust2014). In this respect, dance – or any movement to music – is inextricably linked to musical experiences. Note, however, that similar to many of the mechanisms proposed here, coding of value in sensory cortices (i.e., a stronger sensory response to more important or rewarding stimuli) is not unique to the auditory domain but is also evident in other sensory domains such as vision (Koelsch et al., Reference Koelsch, Vuust and Friston2019).

An important pathway underlying perception–action coupling is the arcuate fasciculus, a bundle of axonal connections between frontal lobe (including motor areas) and superior temporal lobe (including auditory areas). Abundant neuroimaging evidence supports the role of the arcuate fasciculus in music making, specifically in auditory perception–action coupling (Halwani, Loui, Rüber, & Schlaug, Reference Halwani, Loui, Rüber and Schlaug2011; Loui, Alsop, & Schlaug, Reference Loui, Alsop and Schlaug2009, Reference Loui, Li and Schlaug2011; Moore, Schaefer, Bastin, Roberts, & Overy, Reference Moore, Schaefer, Bastin, Roberts and Overy2017; Sammler, Grosbras, Anwander, Bestelmeyer, & Belin, Reference Sammler, Grosbras, Anwander, Bestelmeyer and Belin2015). This same pathway also plays a role in social functions: more emotionally empathic people have higher microstructural integrity within the arcuate fasciculus (Parkinson & Wheatley, Reference Parkinson and Wheatley2014). In contrast, people on the autism spectrum, who have known impairments in social bonding, have less connectivity in the arcuate fasciculus (Fletcher et al., Reference Fletcher, Whitaker, Tao, DuBray, Froehlich, Ravichandran and Lainhart2010; Wan, Demaine, Zipse, Norton, & Schlaug, Reference Wan, Demaine, Zipse, Norton and Schlaug2010). By enabling perception–action coupling, the arcuate fasciculus thus provides one possible shared neural mechanism between music and social bonding.

4.2 Prediction and the dopaminergic reward system

Musical perception–action coupling sets up repeated cycles of prediction, expectation violation, and resolution (Huron, Reference Huron2006). In these hierarchical perception–action trajectories, the predictive context surrounding pitch and rhythm are established, violated, and then resolved (Clark, Reference Clark2013; Fitch, von Graevenitz, & Nicolas, Reference Fitch, von Graevenitz, Nicolas, Skov and Vartanian2009). Successful predictions become rewarding to the brain by activating neurons of the dopaminergic system and its related areas (caudate, nucleus accumbens, amygdala, and ventromedial prefrontal cortex) that code for fundamental evolutionary rewards such as food and sex, and also learned rewards such as money (Friston, Reference Friston2010; Knutson, Westdorp, Kaiser, & Hommer, Reference Knutson, Westdorp, Kaiser and Hommer2000; Schultz, Dayan, & Montague, Reference Schultz, Dayan and Montague1997). The same dopaminergic reward system is also active during the anticipation and perception of pleasurable music (Blood & Zatorre, Reference Blood and Zatorre2001; Blood, Zatorre, Bermudez, & Evans, Reference Blood, Zatorre, Bermudez and Evans1999; Cheung et al., Reference Cheung, Harrison, Meyer, Pearce, Haynes and Koelsch2019; Salimpoor, Benovoy, Larcher, Dagher, & Zatorre, Reference Salimpoor, Benovoy, Larcher, Dagher and Zatorre2011, Reference Salimpoor, Zald, Zatorre, Dagher and McIntosh2015; Zatorre, Reference Zatorre2018; Zatorre & Salimpoor, Reference Zatorre and Salimpoor2013), supported by the functional coupling between auditory areas in the superior temporal lobe and reward-sensitive areas such as the nucleus accumbens (Salimpoor et al., Reference Salimpoor, van den Bosch, Kovacevic, McIntosh, Dagher and Zatorre2013). Manipulating expectations for pitch-related musical features, such as consonance and dissonance, can modulate activity in the nucleus accumbens and amygdala. Thus, music can provide its own reward prediction error and motivate learning (Cheung et al., Reference Cheung, Harrison, Meyer, Pearce, Haynes and Koelsch2019; Gold et al., Reference Gold, Mas-Herrero, Zeighami, Benovoy, Dagher and Zatorre2019). Additionally, people who frequently experience chills when listening to music show high white matter connectivity between auditory, social, and reward-processing areas (Sachs, Ellis, Schlaug, & Loui, Reference Sachs, Ellis, Schlaug and Loui2016). Chills from music are also related specifically to increased binding to dopamine receptor D2 (Salimpoor et al., Reference Salimpoor, Benovoy, Larcher, Dagher and Zatorre2011). In contrast, people with musical anhedonia, who find music unrewarding, have decreased functional connectivity and altered structural connectivity between auditory and reward-related areas (Loui et al., Reference Loui, Patterson, Sachs, Leung, Zeng and Przysinda2017; Martínez-Molina, Mas-Herrero, Rodríguez-Fornells, Zatorre, & Marco-Pallarés, Reference Martínez-Molina, Mas-Herrero, Rodríguez-Fornells, Zatorre and Marco-Pallarés2016; Mas-Herrero, Zatorre, Rodriguez-Fornells, & Marco-Pallarés, Reference Mas-Herrero, Zatorre, Rodriguez-Fornells and Marco-Pallarés2014).

Because humans are social animals, the predictions we make and the rewards we receive are often tied to social stimuli. Thus, the brain has to learn from social cues by associating social stimuli with reward predictions (Atzil, Gao, Fradkin, & Barrett, Reference Atzil, Gao, Fradkin and Barrett2018). Indeed, the same areas in the dopaminergic reward system – the caudate, nucleus accumbens, and ventromedial prefrontal cortex – are causally linked to cooperative behavior as well as prediction and reward. The reward system is activated when we share information with others about ourselves (Tamir & Mitchell, Reference Tamir and Mitchell2012), when we view loved ones (Bartels & Zeki, Reference Bartels and Zeki2004), and when mothers bond with their infants (Atzil et al., Reference Atzil, Touroutoglou, Rudy, Salcedo, Feldman, Hooker and Barrett2017). Prosocial behaviors commonly engage the reward system (Zaki & Mitchell, Reference Zaki and Mitchell2013); these include cooperating (Decety, Jackson, Sommerville, Chaminade, & Meltzoff, Reference Decety, Jackson, Sommerville, Chaminade and Meltzoff2004), perspective taking (Mitchell, Banaji, & Macrae, Reference Mitchell, Banaji and Macrae2005), and empathizing with others (Beadle, Paradiso, & Tranel, Reference Beadle, Paradiso and Tranel2018). Together, these results suggest that the dopaminergic reward system is involved causally in the link between music and social bonding through the mechanism of prediction.

4.3 Oxytocin and the endogenous opioid system (EOS)

We propose that opioids released in the EOS, and oxytocin, are also part of the mechanistic underpinnings linking prediction, reward, and social bonding (Chanda & Levitin, Reference Chanda and Levitin2013; Launay et al., Reference Launay, Tarr and Dunbar2016; Tarr, Launay, & Dunbar, Reference Tarr, Launay and Dunbar2014). The nucleus accumbens and ventral tegmental area are key regions that overlap between the dopaminergic reward system and the EOS (Dölen, Darvishzadeh, Huang, & Malenka, Reference Dölen, Darvishzadeh, Huang and Malenka2013; Le Merrer, Becker, Befort, & Kieffer, Reference Le Merrer, Becker, Befort and Kieffer2009), and dopamine is thought to be a salience processing mechanism regulated by oxytocin (Love, Reference Love2014; Shamay-Tsoory & Abu-Akel, Reference Shamay-Tsoory and Abu-Akel2016).

The EOS likely plays a mechanistic role in music-related prosociality. This system has been implicated in the maintenance of social bonds in primate social networks (Keverne, Martensz, & Tuite, Reference Keverne, Martensz and Tuite1989; Maestripieri, Reference Maestripieri, Platt and Ghazanfar2010; Ragen, Maninger, Mendoza, Jarcho, & Bales, Reference Ragen, Maninger, Mendoza, Jarcho and Bales2013; Schino & Troisi, Reference Schino and Troisi1992). Intervention studies in humans indicate that, compared to a placebo, naltrexone (an opioid blocker) can reduce feelings of social connections with others (e.g., Inagaki, Reference Inagaki2018; Inagaki, Ray, Irwin, Way, & Eisenberger, Reference Inagaki, Ray, Irwin, Way and Eisenberger2016), and lower affiliative behavior and desire for interpersonal closeness (Tchalova & Macdonald, Reference Tchalova and Macdonald2020). Listening to music influences mu-opiate receptor expression in the EOS (Stefano, Zhu, Cadet, Salamon, & Mantione, Reference Stefano, Zhu, Cadet, Salamon and Mantione2004) and can reduce the need for pain medicationFootnote ³ (e.g., Bernatzky, Presch, Anderson, & Panksepp, Reference Bernatzky, Presch, Anderson and Panksepp2011; Lepage, Drolet, Girard, Grenier, & DeGagné, Reference Lepage, Drolet, Girard, Grenier and DeGagné2001). Elevated pain thresholds are experienced after singing (Pearce et al., Reference Pearce, Launay and Dunbar2015; Weinstein et al., Reference Weinstein, Launay, Pearce, Dunbar and Stewart2016) and synchronized dancing (Tarr et al., Reference Tarr, Launay, Cohen and Dunbar2015, Reference Tarr, Launay and Dunbar2016), but not after administration of naltrexone (Tarr, Launay, Benson, & Dunbar, Reference Tarr, Launay, Benson and Dunbar2017), suggesting that pain threshold is an appropriate proxy-measure of endorphin uptake in these experiments. There is some evidence of endorphin-mediated synchrony effects on cooperation (e.g., when dancing; Lang et al., Reference Lang, Bahna, Shaver, Reddish and Xygalatas2017), further demonstrating links between music, the EOS, and social bonding.

Although more empirical research is needed, there is evidence that oxytocin levels are elevated after taking part in a singing class (Grape, Sandgren, Hansson, Ericson, & Theorell, Reference Grape, Sandgren, Hansson, Ericson and Theorell2003), or following a group jam session of improvised singing (Keeler et al., Reference Keeler, Roth, Neuser, Spitsbergen, Waters and Vianney2015). Elevated oxytocin levels have been correlated with increased generosity (Fujii, Schug, Nishina, Takahashi, & Okada, Reference Fujii, Schug, Nishina, Takahashi and Okada2016; Zak, Stanton, & Ahmadi, Reference Zak, Stanton and Ahmadi2007), empathy (Domes, Heinrichs, Michel, Berger, & Herpertz, Reference Domes, Heinrichs, Michel, Berger and Herpertz2007; Hurlemann et al., Reference Hurlemann, Patin, Onur, Cohen, Baumgartner, Metzler and Kendrick2010), and possibly trust (Kosfeld, Heinrichs, Zak, Fischbacher, & Fehr, Reference Kosfeld, Heinrichs, Zak, Fischbacher and Fehr2005; Zak, Kurzban, & Matzner, Reference Zak, Kurzban and Matzner2005; but see Nave, Camerer, & McCullough [Reference Nave, Camerer and McCullough2015] and Declerck, Boone, Pauwels, Vogt, & Fehr [Reference Declerck, Boone, Pauwels, Vogt and Fehr2020]). Furthermore, intranasal administration of oxytocin promotes in-group cooperation (e.g., De Dreu and Kret, Reference De Dreu and Kret2016) and increases synchrony in dancing (Josef, Goldstein, Mayseless, Ayalon, & Shamay-tsoory, Reference Josef, Goldstein, Mayseless, Ayalon and Shamay-tsoory2019) and finger-tapping behavior (Gebauer et al., Reference Gebauer, Witek, Hansen, Thomas, Konvalinka and Vuust2016), suggesting a reciprocal feedback loop between music-based activity and social cohesion. Although evidence linking oxytocin specifically with music remains limited, and the strength of oxytocin's relationship with cooperation more generally is debated (particularly studies based on administering intranasal oxytocin; e.g., Walum, Waldman, & Young, Reference Walum, Waldman and Young2016), current evidence suggests that music engages the oxytocin and EOS systems in ways that facilitate social bonding, as predicted by the MSB hypothesis. Combined with the reward system, these pathways offer a positive-feedback loop following music engagement, enabling groups of individuals to synchronize their moods, emotions, actions and/or perspectives, and providing motivation to continue engaging with others in social and musical contexts.

4.4 Learning and vocal imitation

The capacity to learn and reproduce complex motor movements, including vocalizations (songs), is central to the cultural transmission of music. Although humans are the only primates capable of learning complex, novel vocalizations, this ability has evolved independently at least seven times in evolutionary history (Fitch & Jarvis, Reference Fitch, Jarvis and Arbib2013; Nowicki & Searcy, Reference Nowicki and Searcy2014; Syal & Finlay, Reference Syal and Finlay2011), allowing us to make inferences about how and why it evolved. Some vocal learning clades (seals, baleen whales, and some songbirds) show a strong male bias in vocal learning abilities consistent with sexual selection. However, such a bias is absent in most other vocal learners (parrots, elephants, toothed whales, many tropical bird species, and humans), suggesting that sexual selection cannot be the only factor driving the evolution of vocal learning (Fitch, Reference Fitch2006). Instead, learned animal songs (solo or duet) appear to serve multiple evolutionary functions within the umbrella of social bonding, including mate attraction, cementing and affirming social bonds within pairs or groups, and territorial functions including advertising the bonded group's ability to repel outsiders (Geissmann, Reference Geissmann1999; Haimoff, Reference Haimoff1986; Wickler, Reference Wickler1980).

In vocal learning species, vocal imitation and song production are likely based on similar neurobiological mechanisms (Mercado, Mantell, & Pfordresher, Reference Mercado, Mantell and Pfordresher2014). Learning to reproduce pitches and rhythms accurately engages reward mechanisms, as shown by evidence that dopamine neurons encode performance error in songbirds (Gadagkar et al., Reference Gadagkar, Puzerey, Chen, Baird-Daniel, Farhang and Goldberg2016). Furthermore, the presence of a conspecific (of the opposite sex in this case) leads the male zebra finch to decrease variability of sung syllables; this syllabic structure is attributed to perception–action circuits analogous to the human superior temporal and motor structures (Fitch & Jarvis, Reference Fitch, Jarvis and Arbib2013; Sakata & Brainard, Reference Sakata and Brainard2008). Once individuals learn to produce musical features, they not only reproduce learned patterns of features, but also deviate from predicted combinations of features, for example by inventing new melodies (Wiggins, Tyack, Scharff, & Rohrmeier, Reference Wiggins, Tyack, Scharff, Rohrmeier and Honing2018).

5.1. Cross-domain predictions (e.g., music, language, ritual)

The MSB hypothesis predicts that music (including dance) is better-suited to social bonding of large, complex groups than ABMs (grooming and laughter), language, or other non-acoustic bonding mechanisms such as shared decorations or non-musical ritual behaviors (e.g., praying together without music). Music should be more effective and/or efficient relative to other methods as group size and complexity increase, such that while making music in pairs might only produce a small increase in dyadic bonding relative to conversation, making music in larger, more complex groups of people (dozens or hundreds organized into differentiated sub-groups) should be more effective for collective bonding than language, laughter, grooming, and so on.

In a social species such as humans, many activities can develop and enhance social bonding, but we predict that bonding via non-musical methods such as language, ritual, or sports should be enhanced by the addition of musical components (e.g., religious services with group singing will result in stronger bonding than those that only involve group prayer). Different musical components are predicted to have synergistic effects such that – all things being equal – including more of these components (e.g., synchronized, harmonized singing and dancing in groups) will tend to increase bonding more than activities that only use one or a few (e.g., conversations or recitation in pairs).Footnote ⁴ We also predict that participatory musical performances will have significantly stronger effects than either non-participatory (e.g., performance for a static audience) or solo musical experiences (e.g., listening alone to recordings). Group size and complexity should have independent effects (e.g., singing in large choirs should produce greater bonding than singing in small choirs).

These predictions can be tested in controlled experiments and/or field studies along the lines of those discussed in sect. 3. Designing studies that control for specific similarities and differences between closely related domains such as music, language, and dance is challenging but not impossible. For example, to control for the fact that languages have their own (non-isochronous) rhythms, Savage et al. (Reference Savage, Yamauchi, Hamaguchi, Tarr, Kitayama and Fujii2020) had groups of participants simultaneously recite the lyrics to “Twinkle, Twinkle, Little Star” to an isochronous beat or in non-isochronous free rhythm. Savage et al. (Reference Savage, Yamauchi, Hamaguchi, Tarr, Kitayama and Fujii2020) also propose additional manipulations that would allow this paradigm to test other specific predictions of the MSB hypothesis regarding the social bonding effects of melody, harmony, and dance (cf. Fig. 3 in Savage et al., Reference Savage, Yamauchi, Hamaguchi, Tarr, Kitayama and Fujii2020).

5.2. Cross-cultural predictions

The MSB hypothesis predicts that music's social bonding functions should be distributed widely in space and time. Hence, the kinds of predictions described in sect. 5.1 regarding music's superior social bonding power in large groups should apply consistently across cultures and throughout history. Furthermore, it predicts that musical contexts and structures that promote social bonding (e.g., coordinated, participatory group performances) will be more common across cultures than music produced by and for individuals. At the same time, the relative importance of participatory versus presentational music-making is predicted to vary cross-culturally as a function of social structure (because of limitations on simultaneous coordinated performance discussed in sect. 2.5). Smaller-scale, more egalitarian cultures should thus perform and value participatory music more than larger-scale, hierarchical cultures where presentational music should be more common and valued. Participatory versus presentational distinctions are analogous to those found in “imagistic” (high-intensity, small-scale) versus “doctrinal” (low-intensity, large-scale) religious rituals, respectively (Whitehouse, Reference Whitehouse2004), and are predicted to covary cross-culturally with these modes of religiosity. Even in cultures where music is often consumed passively by individuals (e.g., in Western culture, over headphones on personal listening devices), MSB predicts that music will be more effective than non-musical alternatives for social bonding purposes (cf. Rentfrow & Gosling, Reference Rentfrow and Gosling2006). These predictions about cross-cultural use of music for social bonding could be tested in cross-cultural behavioral experiments (cf. Henrich et al., Reference Henrich, Boyd, Bowles, Camerer, Fehr, Gintis, McElreath, Alvard, Barr, Ensminger, Henrich, Hill, Gil-White, Gurven, Marlowe, Patton and Tracer2005; Jacoby et al., Reference Jacoby, Polak, Grahn, Cameron, Lee, Godoy and McDermottPreprint; Polak et al., Reference Polak, Jacoby, Fischinger, Goldberg, Holzapfel and London2018) or analysis of cross-cultural databases of recordings, artifacts, ethnographies, or questionnaires (cf. Lomax, Reference Lomax1968; Mehr et al., Reference Mehr, Singh, Knox, Ketter, Pickens-Jones, Atwood and Glowacki2019; Savage, Reference Savage, Shanahan, Burgoyne and Quinn2019c; Savage et al., Reference Savage, Brown, Sakai and Currie2015; Whitehouse et al., Reference Whitehouse, François, Savage, Hoyer, Feeney, Cioni and Turchinpreprint; Wood et al., Reference Wood, Kirby, Ember, Silbert, Daikoku, McBride and SavagePreprint).

5.3. Cross-species predictions

The MSB hypothesis proposes that human musicality has been shaped by biological and cultural selection, and that the features of music are particularly well-suited for social bonding functions because they support extended, coordinated group performances on a large scale. The MSB account does not claim that music's social bonding function is a unique biological adaptation specific to human musicality. Instead, it argues that music-like behaviors should enhance existing bonding mechanisms in other species as well. Thus, it predicts that, rather than an all-or-nothing divide between human and non-human “music,” species will vary continuously in the degree to which they share specific features of human musicality. The social bonding functions associated with different components of musicality should operate similarly across species, depending on the specific subcomponent, its suitability for group coordination, and the importance of social bonding to their species.

Thus, melodic, learned songs among songbirds, whales, or other vocal learners are predicted to enhance social bonding in these species in a manner analogous to song in humans. These effects may be limited in many non-human species by their lack of ability and/or interest in performing in coordinated groups (e.g., some primates appear motivated to conduct group displays but are unable to synchronize to a beat, whereas some birds appear able to move to a beat but are unmotivated to do so in groups in the wild; Hoeschele, Merchant, Kikuchi, Hattori, & ten Cate, Reference Hoeschele, Merchant, Kikuchi, Hattori, ten Cate and Honing2018). However, such effects should be pronounced in species that perform duets (e.g., many birds, and duetting primates such as gibbons or titi monkeys; Haimoff, Reference Haimoff1986; Hall, Reference Hall2004). Conversely, social primates that do not typically perform in coordinated groups may nonetheless experience social bonding effects of “group” music when exposed to versions of their own vocalizations that have been artificially manipulated to be in synchrony/harmony. Such production/perception dissociations and other nuances of musicality could be tested in controlled cross-species experiments (cf. Hoeschele et al., Reference Hoeschele, Merchant, Kikuchi, Hattori, ten Cate and Honing2018; Merchant, Grahn, Trainor, Rohrmeier, & Fitch, Reference Merchant, Grahn, Trainor, Rohrmeier and Fitch2015).

The MSB hypothesis posits that music and musicality provided a major means by which humans could coordinate behavior on a larger scale than dyads or small groups, allowing for the formation of larger socio-cultural groups. If true, and if different species share components of musicality to differing degrees, then across species, production or proficiency in “musical” behaviors should predict both the number and complexity of social bonds. For example, gelada baboons live in unusually large and complex groups for primates, and they also exhibit rhythmic and melodic vocal features that are unique among primates (Bergman, Reference Bergman2013; Gustison, Aliza, & Bergman, Reference Gustison, Aliza and Bergman2012; Richman, Reference Richman1978, Reference Richman1987). Similar to geladas, many parrot species live in large fission–fusion social groups, and members of the parrot clade show vocal imitation, call convergence, duetting, and the capacity for rhythmic synchronization (Balsby & Scarl, Reference Balsby and Scarl2008; Bradbury, Reference Bradbury2001; Scarl & Bradbury, Reference Scarl and Bradbury2009; Schachner et al., Reference Schachner, Brady, Pepperberg and Hauser2009). In both of these clades, pairs or mating “harems” form stronger bonds than those they share with the larger groups in which they are embedded (cf. Balsby & Scarl, Reference Balsby and Scarl2008; Wanker, Sugama, & Prinage, Reference Wanker, Sugama and Prinage2005). Other species that live in complex fission–fusion groups and could provide evidence of specific design features are elephants and some odontocetes (e.g., orcas and bottlenose dolphins). Such species live in large, complex fission–fusion groups, and are documented vocal learners, but their possession of other design features of music (e.g., synchronization) have not been tested rigorously.

For many species, evidence for design features of musicality would count as evidence against our hypotheses. Examples include solitary species (e.g., many reptiles), species for whom groups consist only of mothers and dependent young (e.g., many carnivores), or group living species that do not have differentiated social bonds with other group members (e.g., schooling fish, larger herds, and swarming insects).

The MSB hypothesis further predicts that if a species does not follow this pattern (e.g., by having a larger social group size than predicted by their features of musicality), then that species will have evolved other non-musical but effective means of coordinating behavior that likely do not appear in human behavior (e.g., reproductive suppression in naked mole rats or pheromonal queen control in eusocial insects; Alaux, Maisonnasse, & Le Conte, Reference Alaux, Maisonnasse and Le Conte2010; Dengler-Crish & Catania, Reference Dengler-Crish and Catania2007). Thus, although the social bonding design features seen in human musicality are not the only way to achieve large, well-bonded groups, they are effective enough that we predict them to evolve convergently (cf. Fitch, Reference Fitch2006).

5.4. Neurobiological predictions

The MSB hypothesis predicts that each of the mechanistic factors proposed above (Fig. 3) contributes to the effects of music on social bonding. Alterations of these mechanistic pathways should therefore produce specific, quantifiable results on bonding. For example, music's perceived social bonding functions should correlate with oxytocin/EOS production, and disrupting the oxytocin/EOS pathway via blocking oxytocin or opioid receptors should disrupt its social bonding effects. Furthermore, because of the dopaminergic reward system is at the center of prediction for musical features, populations with deficient dopaminergic activity may have impaired predictions, which could affect their ability to synchronize or harmonize with others. On the contrary, drugs that restore dopaminergic functions are hypothesized to restore these abilities, and because of the reciprocal nature of these interactions, activities that enhance predictions (such as dancing and harmonizing) may in turn restore dopaminergic functions. These predictions are being tested in the case of Parkinson's disease, which is a special population with deficient dopaminergic activity (Cameron et al., Reference Cameron, Pickett, Earhart and Grahn2016; Grahn & Rowe, Reference Grahn and Rowe2009).

Another prediction is that special populations with high sociability may respond well to musical features especially when coupled with social stimuli, as in the case of children with Williams syndrome (Järvinen-Pasley et al., Reference Järvinen-Pasley, Vines, Hill, Yam, Grichanik, Mills and Bellugi2010; Lense, Gordon, Key, & Dykens, Reference Lense, Gordon, Key and Dykens2014). At a neural level, music's social bonding function should correlate with the degree of neural connectivity between the perception–action and prediction–reward networks, and disruptions to this network (e.g., lesions or genetic syndromes) should accordingly disrupt music's social bonding effect. For example, people with musical anhedonia, who have disrupted connectivity between auditory prediction and reward networks (Belfi & Loui, Reference Belfi and Loui2020), are predicted to have weaker social bonds, and genetic differences (e.g., in DRD2) may predict variation in bonding experienced through musical activities. Although some of these predictions may be difficult to test ethically in humans through controlled experiments, many can be tested using neuroimaging combined with neuropsychological testing in special populations, as well as correlational, longitudinal, or intervention (including brain-stimulation) studies, genome-wide association studies, and/or animal models that share specific neurobiological endophenotypes (Finlay, Darlington, & Nicastro, Reference Finlay, Darlington and Nicastro2001; Fitch & Jarvis, Reference Fitch, Jarvis and Arbib2013; Gingras, Honing, Peretz, Trainor, & Fisher, Reference Gingras, Honing, Peretz, Trainor, Fisher and Honing2018; Hoeschele et al., Reference Hoeschele, Merchant, Kikuchi, Hattori, ten Cate and Honing2018; Niarchou et al., Reference Niarchou, Sathirapongsasuti, Jacoby, Bell, McArthur, Straub and Gordon2019).Footnote ⁵

6.1 Music, language, and domain-specificity

The key criticism that we anticipate regards the degree to which the evolution of musicality and social bonding are uniquely and causally linked. Few would deny that music can facilitate social bonding via neurobiological mechanisms that are evolutionarily adaptive. However, whether music is a domain-specific evolutionary adaptation for social bonding, as opposed to a byproduct of the evolution of other adaptations, is open to debate. Language, in particular, has been proposed as an evolutionary adaptation that led to musicality as a byproduct (Pinker, Reference Pinker1997).Footnote ⁶ Importantly, many researchers have noted that, although there are clear differences in the structure and processing of music and language, there is extensive overlap ranging from structural content (e.g., “musilinguistic continua” between speech and song including intermediate forms like poetry and chant) to neurobiological substrates (e.g., similar neural substrates for processing of pitch, rhythm, and syntax; Brown, Reference Brown, Wallin, Merker and Brown2000b, Reference Brown2017; Fitch, Reference Fitch2006; Patel, Reference Patel2008; Peretz & Coltheart, Reference Peretz and Coltheart2003; Peretz, Vuvan, Lagrois, & Armony, Reference Peretz, Vuvan, Lagrois, Armony and Honing2018; Savage, Merritt, Rzeszutek, & Brown, Reference Savage, Merritt, Rzeszutek and Brown2012). Indeed, many have proposed that the evolution of musicality may have paved the way for the evolution of language (Brown, Reference Brown, Wallin, Merker and Brown2000b, Reference Brown2017; Darwin, Reference Darwin1871; Fitch, Reference Fitch2010; Mithen, Reference Mithen2005; Shilton, Breski, Dor, & Jablonka, Reference Shilton, Breski, Dor and Jablonka2020).

We accept that our present level of understanding is insufficient to demonstrate conclusively that music coevolved uniquely with social bonding independent from language or other social behaviors. Accordingly, in sect. 5, we proposed future investigations of such relationships. However, the fact that music and language are both found universally in all known societies (Brown, Reference Brown1991; Mehr et al., Reference Mehr, Singh, Knox, Ketter, Pickens-Jones, Atwood and Glowacki2019) suggests that both music and language independently fulfill more fundamental adaptive functions than technologies or cultural artifacts that are not cross-culturally universal.

We make no claim that the mechanisms discussed here are entirely specific to music, or that “musicality” is modular in either the cognitive or neuroscientific senses of this term. For example, prediction and predictive coding are ubiquitous features of vertebrate brains (Clark, Reference Clark2013; Schultz & Dickinson, Reference Schultz and Dickinson2000), by no means specific to musicality. However, music affords a uniquely effective scaffolding framework, including rhythm and harmony, within which neural prediction (and occasional expectation violations) can unfold (Fitch et al., Reference Fitch, von Graevenitz, Nicolas, Skov and Vartanian2009; Hanslick, Reference Hanslick1858; Huron, Reference Huron2006; Koelsch, Vuust, & Friston, Reference Koelsch, Vuust and Friston2019). Similarly, synchrony is widespread in human sociality (including phenomena such as gaze synchrony, affect synchrony, the chameleon effect, and others), but the isochrony of musical rhythm provides an unusually effective affordance for synchronization. Furthermore, phenomena such as “groove” seem to be mainly evoked by musical stimuli, and therefore are relatively domain-specific. Thus, musicality encompasses multiple mechanisms that vary in their domain-specificity, but combines them into a uniquely effective package.

6.2 Group selection

Most previous social bonding theories of music evolution have relied on an evolutionary mechanism incorporating some form of group selection, in which genetic variants are selected for because of their effects on the reproductive success of entire groups (e.g., Brown, Reference Brown, Tonneau and Thompson2000a; Wiltermuth & Heath, Reference Wiltermuth and Heath2009). Group selection has been largely dismissed for decades (Williams, Reference Williams1966), and while it is re-emerging in the form of multi-level selection (Traulsen & Nowak, Reference Traulsen and Nowak2006; Wilson & Wilson, Reference Wilson and Wilson2007) and cultural group selection (Richerson et al., Reference Richerson, Baldini, Bell, Demps, Frost, Hillis and Zefferman2016), it remains controversial (Pinker, Reference Pinker2012; see also commentary accompanying Richerson et al., Reference Richerson, Baldini, Bell, Demps, Frost, Hillis and Zefferman2016).

The MSB hypothesis does NOT require group selection (any more than grooming, play, or laughter do): fitness advantages accrue to individuals who are able to bond more effectively with larger numbers of individuals. Although there are often advantages to well-bonded groups for various activities (e.g., group hunting or foraging, jointly repelling enemies), even for such activities the key fitness advantages accrue to individuals.

6.3 Gene–culture coevolution and causality

Some evolutionary psychologists have been critical of social bonding theories of music evolution because they consider them circular arguments that fail to explain the ultimate causal mechanism by which music could have evolved as a biological adaptation:

Our preceding account provides a priori arguments detailing why and how specific design features of human musicality have social bonding effects, the mechanisms underlying these effects, and how and why these may have evolved. In particular, we provided specific reasons that behaviors with the design features of music would have social bonding effects: because such behaviors allow people to predict, synchronize, share goals, distinguish individual contributions, experience shared positive emotions, and make social decisions more than other human behaviors (ABMs or language). This explains why music should produce “[social bonding] effects and not others”: behaviors that allow us to align in time and frequency, coordinate behaviors in large groups while distinguishing individual contributions, share emotions and goals, and choose appropriate social partners have tangible and predictable social bonding effects. Music is a particularly effective cognitive “technology” (Patel, Reference Patel2008, Reference Patel and Honing2018) that fulfills these design criteria, making musicality an effective toolkit for social bonding functions, shaped by both biological and cultural evolution.

Our hypothesis differs from most traditional social bonding theories because we do not argue that musicality necessarily originated as a biological adaptation. Instead, components of musicality may have arisen initially as cultural inventions and/or byproducts of other adaptations, later exapted and modified through gene–culture coevolution for their social bonding functions in a musical context (e.g., beat synchronization initially as a byproduct of the evolution of vocal learning, as argued by Patel, Iversen, Bregman, & Schulz [Reference Patel, Iversen, Bregman and Schulz2009] and Schachner et al. [Reference Schachner, Brady, Pepperberg and Hauser2009]; although cf. Merker et al. [Reference Merker, Morley, Zuidema and Honing2018] for an alternative interpretation). The initial social cohesion functions may not have begun as genetic adaptations. In this sense, we largely agree with Mehr et al., who write:

But in a social environment in which social bonding already enhanced individual reproductive fitness, the subsequent cultural evolution of musical behaviors would lead to biological selection on musicality (e.g., to promote motivation to engage in/attend to musical behaviors), because of the adaptive consequences of musicality for social bonding. In this way, just as social bonding is crucial in most primates, generating selection on the mechanisms that achieve it, social bonding functions of “proto-musical” mechanisms may have played important roles in hominin evolution long before today's full-blown musicality evolved.

We emphasize that past adaptive function, although important, should not be the sole criterion by which to judge theories of the evolution of musicality. As previously argued at length (e.g., Fitch, Reference Fitch2006, Reference Fitch, Toivonen, Csúri and van der Zee2015b; Honing et al., Reference Honing, Cate, Peretz and Trehub2015), Tinbergen's (Reference Tinbergen1963) multi-factorial perspective, which seeks understanding of traits at the four interlinked explanatory levels of mechanism, ontogeny, phylogeny, and adaptive function, is a fruitful method for understanding the evolution of musicality. We may never know with certainty the precise ancestral adaptive conditions or specific genetic mutations involved in the evolution of musicality. Even so, the comparative method provides a key tool for empirically testing evolutionary hypotheses (Fitch, Reference Fitch, Toivonen, Csúri and van der Zee2015b). Section 5 lists a variety of testable empirical predictions of the MSB hypothesis.

6.4. Parochial altruism and out-group exclusion

Enhanced social bonding between some individuals inevitably means a relative decrease between others. In-group social bonding has a dark side of increasing hostility toward out-groups (Gelfand, Caluori, Jackson, & Taylor, Reference Gelfand, Caluori, Jackson and Taylor2020; Whitehouse, Reference Whitehouse2018), as exemplified in the use of music in warfare by the Nazis and other groups throughout history (Brown & Volgsten, Reference Brown and Volgsten2006). The traditional Māori haka “Ka Mate” is famously used by New Zealand's national rugby team to simultaneously bind team-mates together and intimidate the opposing team through coordinated dancing and vocalization (Jackson & Hokowhitu, Reference Jackson and Hokowhitu2002). The ability of music to exclude out-group members might appear to be an argument against its function in bonding in-group members, but out-group exclusion is entirely consistent with the social bonding hypothesis. Because the creation or strengthening of a social bond between some (participating) individuals by definition excludes others, the observation that particular forms of music can cause emotional dissonance or fear in others is compatible with a social bonding function.

Earlier expositions of the social bonding hypothesis (Brown, Reference Brown, Tonneau and Thompson2000a; Freeman, Reference Freeman, Wallin, Merker and Brown2000) noted that “bonding is always exclusionary” and “individuals who do not ‘belong’ become enemies … The process is similar to sexual jealousy, which manifests the exclusionary nature of the pair bond” (Freeman, Reference Freeman, Wallin, Merker and Brown2000, pp. 421–422). This observation is mirrored in the recent literature on oxytocin which, far from being an indiscriminate “love drug,” simultaneously exerts affiliative effects among in-group members and exclusionary effects toward out-group individuals (cf. Beery, Reference Beery2015; Shamay-Tsoory & Abu-Akel, Reference Shamay-Tsoory and Abu-Akel2016). The use of music to exclude others is no argument against its social bonding origins.

6.5. Solo music, sexual selection, and individual signaling

Although coordinated group performances predominate throughout the world, various widespread musical genres are not necessarily performed in coordinated groups. In particular, lullabies and love songs are found throughout the world and are often performed by a lone singer (Mehr & Singh et al., Reference Mehr, Singh, York, Glowacki and Krasnow2018; Trehub et al., Reference Trehub, Unyk and Trainor1993). This is perfectly consistent with the MSB hypothesis, as lullabies and love songs are often dyadic: sung by a soloist to bond with another person (by soothing an infant or wooing a potential mate).

More generally, some may wonder why, if social bonding is so important to the evolution of musicality, do people enjoy playing or listening to music alone? We emphasize that even solo music listening can support social bonding goals (Trehub et al., Reference Trehub, Becker, Morley and Honing2018). A young adult meeting a new person in an online chat discusses music preferences more often than other topics, and based on music preferences alone, people draw social inferences about others (Rentfrow & Gosling, Reference Rentfrow and Gosling2006). Thus, music preferences developed during solo listening can be used as social cues, displayed and evaluated when establishing new social bonds.

Solo listening may serve other, non-social functions (e.g., mood regulation, staying awake while driving; DeNora, Reference DeNora2000; North, Hargreaves, & Hargreaves, Reference North, Hargreaves and Hargreaves2004; Sloboda, O'Neill, & Ivaldi, Reference Sloboda, O'Neill and Ivaldi2001). We do not argue that social bonding is the only possible function of music. By analogy, language's primary function may be to communicate information between people, but it is also useful in private thought, or to allow one to preserve thoughts for the future (particularly after the invention of writing). Similarly, the same auditory–motor–reward connections that make music so socially powerful also allow people to enjoy playing or listening to music alone. Often, solo music was experienced previously in a social context, which is re-evoked by solo listening/playing.

Related to the idea of virtuosic solo music-making is the distinction between social bonding and theories such as sexual selection or honest signaling that emphasize music as a signal of individual fitness. The MSB hypothesis does not reject such theories. Instead, it emphasizes that individual signaling theories are insufficient to explain all of the broader social functions of music, whereas social bonding provides more explanatory power (although we concede that the MSB hypothesis cannot explain all possible functions of music; Oesch, Reference Oesch2019). For example, in contemporary Western night clubs and traditional non-Western societies, all-night music and dance rituals function both to bond participants and as opportunities to find potential mates (Merriam, Reference Merriam1964; Thornton, Reference Thornton1995). In such contexts, dancing, singing, and/or playing instruments can function to bond with same and opposite-sex partners and to advertise evolutionary fitness to potential mates. Bonding and signaling hypotheses are not mutually exclusive, but rather complementary.

The complementarity of the MSB and alternative hypotheses makes it challenging to falsify the MSB hypothesis. However, we have provided a number of specific predictions, each of which is potentially falsifiable and would count as evidence against the MSB hypothesis, particularly if alternative hypotheses better predict the data. For example, our hypothesis and Hagen and Bryant's (Reference Hagen and Bryant2003) coalitional signaling hypothesis make predictions regarding synchrony: we argue that synchrony should enhance social bonding, whereas Hagen and Bryant argue that synchrony should enhance perceived coalitional quality. To differentiate between these and other competing hypotheses, our predictions regarding the effects of synchrony (or other aspects of musicality) on social bonding could be compared directly against perceived coalition quality or other competing predictions (e.g., attractiveness; Miller, Reference Miller, Wallin, Merker and Brown2000, parental investment; Mehr & Krasnow, Reference Mehr and Krasnow2017; Mehr et al., target article) in future research. If synchrony increases perceived bonding relative to perceived coalition quality, attractiveness, or parental investment, it would constitute evidence favoring the MSB hypothesis over competing alternatives. Another example of predictions that differentiate among alternative hypotheses is the MSB prediction that social bonding functions will be common cross-culturally but the relative frequencies of specific genres and sub-functions (e.g., lullabies vs. love songs vs. group dancing) will vary across societies. In contrast, theories that focus on infant-directed song or sexual selection predict instead that these categories should be more common and consistent cross-culturally than the other categories of social bonding. Furthermore, phylogenetic or other cross-species analyses (e.g., Hoeschele et al., Reference Hoeschele, Merchant, Kikuchi, Hattori, ten Cate and Honing2018; Schruth, Templeton, & Holman, Reference Schruth, Templeton and HolmanIn press; Shultz, Opie, & Atkinson, Reference Shultz, Opie and Atkinson2011) could allow us to quantify the relative effects of group size, sexual competition, parental investment strategies, or other factors on the evolution of vocal learning, beat perception, or other aspects of musicality. We encourage tests of MSB predictions against those of competing hypotheses.