One component of entrainment is thus temporal. Temporal entrainment to music can be observed at hierarchical levels of metrical periodicity in the body (Leman and Naveda, 2010; Naveda and Leman, 2010; Toiviainen et al., 2010) and brain (Large and Jones, 1999; Large and Palmer, 2002; London, 2004; Jones, 2009; Nozaradan et al., 2011). People often experience temporal musical entrainment in the automatic, even uncontrollable head-bobbing or toe-tapping that occurs when we listen to musicians play (or the covert activation of motor areas of the brain in the absence of overt movement), while the height of temporal entrainment might be seen in the complex rhythmic timing and exchange between partners or ensemble members in music or dance (Keller et al., 2007; Keller, 2008; Goebl and Palmer, 2009) as portrayed by the band in Figure ​Figure11.

A second component of entrainment derives from the mutual sharing of an affective state between individuals. Authors have even spoken of a kind of affective “synchrony,” although we prefer the term entrainment in order to distinguish the temporal and affective components while encapsulating the essence of both. Therefore, for the purpose of this paper we will refer to both temporal and affective entrainment. Affective entrainment involves the formation of interpersonal bonds (see Feldman, 2007) and is related to the pleasure in moving the body to music and being in time with others (known in the vernacular as “groove”; Pressing, 2002; Madison, 2006; Janata et al., 2012). Janata et al. (2012) have put forth that temporal and affective components of entrainment in music might be inextricably linked, as for example, the quality of sensorimotor synchronization (temporal) can predict the degree of experienced groove (affective), and affiliation (Hove and Risen, 2009). In searching for the biological roots of temporal and affective entrainment, early musical interactions between infants and their caregivers might provide the first experiences in entrainment (e.g., Malloch and Trevarthen, 2009) and thus provide a foundation for the development of skills in musical joint action (Figure ​(Figure11).

In joint action in general, multiple individuals coordinate their movements in space and time, with a goal to communicate, or to effect a change in the environment (see Sebanz et al., 2006). Here we take on the term “coordination” commonly used in this literature because movements between co-actors might be coupled in either a synchronized or complementary fashion, in time, or in space. Studies of the spatiotemporal aspects of interpersonal joint action reveal that coordination in joint action can be emergent or planned (Knoblich et al., 2011). Emergent coordination occurs spontaneously due to automatic processes that are grounded in perception and action couplings in the brain (Knoblich et al., 2011). In a non-musical example, when two conversational partners are engaged, they show a coordination of body sway that is unintentional and does not require planned action (Shockley et al., 2003). Unintentional coordination of body sway can also occur between duetting musicians (Keller and Appel, 2010), an example that will be discussed in more detail later. The emergent process can also be described in terms of concepts from dynamical systems theory (e.g., coupled oscillators; see Schmidt and Richardson, 2008; Oullier and Kelso, 2009; Riley et al., 2011). In contrast, planned coordination requires shared representations of the intended outcome of the joint action and a plan for each individual’s own role in the joint action. This can be seen in a variety of activities in which multiple individuals engage, including spoken conversation, or carrying a heavy object together (e.g., Clark, 1996, 2005; Knoblich and Sebanz, 2008; Bekkering et al., 2009; Sebanz and Knoblich, 2009). In music, both emergent (automatic head-bobbing) and planned (playing and dancing along) coordination reflect entrainment and can exert simultaneous influences on the individuals’ actions.

How do we experience emergent and planned coordination in musical interactions where entrainment involves demands on temporal synchronization and complex rhythmic timing? To understand how these forms of musical joint action arise requires an excursion into several different research fields. The same terminology is occasionally used to different ends in these fields, and part of our task consists of an attempt to reconcile this with the aim of merging these diverse fields on common ground (see Box 1 for a glossary of our terminology and definitions). We begin in the next section by identifying two primary generic modes of musical joint action, chorusing and turn-taking, and their underlying entrainment mechanisms. Then in the section entitled “Ensemble Skills in Musical Joint Action” we consider how these mechanisms are crystallized in specific cognitive and motor skills that support musical ensemble performance. We then discuss in the section entitled “The Beginnings of Musical Joint Action” some of the potential ontogenetic roots for the cognitive and motor systems required in musical joint action. In the final section, we speculate on a possible mimetic relationship between developmental trajectories in social coordination skills and musical ensemble skills.

The distinction between chorusing and turn-taking is based on the temporal relations between actions of interacting individuals. Musical joint actions, however, have behaviorally relevant spatial, as well as temporal, components. Spatiotemporal movements shape dance forms and define the embodiment of metrical hierarchies (Leman, 2007). In a musical melody, the pitch intervals and contours define trajectories through mental representations of pitch space that can correspond cross-modally to physical space (Rusconi et al., 2006; Lidji et al., 2007; Keller and Koch, 2008; Eitan and Timmers, 2010). To account for such spatiotemporal relations, chorusing and turn-taking can be classified with respect to two further general classes of action: imitation and complementary joint action (Figure ​(Figure22).

Imitation entails one individual mirroring the spatiotemporal features of another’s movements or their effects (see Brass and Heyes, 2005), often in a goal-directed manner (Sebanz et al., 2006). Imitation may occur intentionally, such as when a musician or dancer incorporates a pre-existing sequence in a variation on a theme, or unintentionally, as when two duetting players inadvertently copy each other’s body sway or the effects of that sway on expressive parameters of the sound that they are producing (Keller and Appel, 2010). Imitation may occur at varying time lags, ranging from short intervals that give the impression of simultaneity (as in unison chorusing, or moving bodies or instruments to a common musical pulse) to longer intervals at which the model behavior and the copy do not overlap (serial turn-taking). Complementary joint action, which requires more sophisticated partnership skills, occurs when spatiotemporal features of one individual’s behavior are different from, but systematically related to, those of another individual (see Bekkering et al., 2009). This is exemplified in non-unison chorusing and hierarchical turn-taking.

We assume that entrainment supports imitation and complementary joint action through two brain mechanisms that capitalize on links between perception and action. Specifically, imitation is mediated by motor resonance while complementary joint action requires internal simulation processes that may involve mental imagery. Motor resonance occurs when brain regions that control movement – including motor, premotor, and sensorimotor cortices – are activated automatically by the observation of another’s movements (Rizzolatti and Craighero, 2004). This process may lead to non-conscious mimicry, where individuals unwittingly adopt the facial expressions, mannerisms, or postures of interaction partners (Chartrand and Bargh, 1999). Motor resonance may contribute to the perception of musical pulse (van Noorden and Moelants, 1999; Grahn and Brett, 2007; Chen et al., 2008; Large, 2008) and, in both music and dance, may promote readiness for temporally coordinated, planned action.

Motor resonance evoked during the perception of music and dance does not only function in the service of action, but may also modulate aesthetic responses (Calvo-Merino et al., 2008; Kornysheva et al., 2010; Cross and Ticini, 2011). Motor representations of practiced actions result in stronger motor resonance in adults as well as in infants (Calvo-Merino et al., 2006; van Elk et al., 2008), which suggests that motor resonance is tuned by experience and, when combined with perceptual resonance (Schütz-Bosbach and Prinz, 2007a), can contribute to action simulation.

Action simulation, as conceived here, is richer than motor resonance to the extent that it involves the activation of sensory, in addition to movement-related, brain regions (cf. Schütz-Bosbach and Prinz, 2007a). Specifically, action simulation occurs when sensorimotor neural processes that resemble those associated with executing an action are engaged in the absence of overt movement (see Gallese et al., 2004; Wilson and Knoblich, 2005; Decety and Grèzes, 2006). Such covert activity may be triggered by observing – or by merely imagining – an action or its effects, for example, body movements in the case of dance (Cross et al., 2006), or tones in the case of music (Keller, 2008). This highlights what we believe to be a fundamental distinction between motor/perceptual resonance and action simulation: While resonance is driven exogenously and automatically (i.e., preattentively) by the perception of external events, simulation may be generated endogenously in the absence of external stimuli (cf. Grush, 2004) and may be modulated by attention.

Action simulation is a mark of expertise, as it is mediated by experience-based associations between sensory and motor processes (Haueisen and Knösche, 2001; Baumann et al., 2005; Bangert et al., 2006; Zatorre et al., 2007). Auditory–motor simulation is, therefore, especially strong in musicians (Lahav et al., 2007) and visuo-motor simulation is particularly potent in dancers (Calvo-Merino et al., 2005, 2006; Cross et al., 2006). It has been proposed that action simulation facilitates the understanding of others’ intentions and affective states, as well as playing a role in predicting an observed action’s immediate outcome and future course (e.g., Wilson and Knoblich, 2005; Decety and Grèzes, 2006; Schütz-Bosbach and Prinz, 2007b; Sebanz and Knoblich, 2009). In musical joint action, predictions based on simulation can facilitate interpersonal coordination in the context of musical textures characterized by chorusing and turn-taking (e.g., Keller et al., 2007; Keller and Appel, 2010).

Instances of unison chorusing – for example, singing a tune such as Auld Lang Syne, or moving one’s body in matched fashion with other individuals as in a group folk dance – can be considered to be forms of imitation to the extent that co-actors adopt leader or follower roles. Spatiotemporal interpersonal coordination at this level can especially reveal the contribution of motor resonance triggered exogenously by perceptual input, in addition to the simulation, and action-planning required to execute the musical unison. A complex example that illustrates the contribution of multiple cognitive–motor processes is the Afro-Brazilian Congado – a musical ritual of religious significance that entails the simultaneous performance of separate but proximal communal groups of singers, percussionists, and dancers. A rigorous analysis of Congado performances (Lucas et al., 2011) has shown that inter-group entrainment is influenced by proximity and visual contact, similarity of tempo, and intention (i.e., whether proximal groups resist entrainment), factors which reveal contributions of emergent entrainment and imitation, as well as simulation and action-planning in planned coordination. Finally, during turn-taking, the serial or hierarchical ordering of signals produced by different individuals (e.g., when “trading licks” in jazz or exchanging melody and accompaniment roles in chamber music) constitutes complementary joint action. Effective coordination during turn-taking therefore requires the participating individuals to predict the spatiotemporal trajectories of each other’s actions via simulation.

The most fundamental ensemble skill is adaptive timing, or adjusting the timing of one’s movements in order to maintain synchrony in the face of unintended temporal perturbations (e.g., by other players) and intended tempo changes in the music. Adaptive timing is mediated by temporal error correction mechanisms that enable internal timekeepers (i.e., neural oscillations at timescales relevant to musical meter) to remain coupled, or entrained in stable phase relations with external signals (Repp, 2005; Large, 2008; Repp and Keller, 2008). Temporal error correction processes that operate automatically support emergent coordination, while others that require attention may be invoked in contexts requiring planned coordination, such as when an ensemble musician intentionally adapts to the expressively motivated tempo changes of a co-performer (Keller, 2008). Research on adaptive timing has yielded evidence for assimilation in interpersonal action timing during dyadic sensorimotor synchronization. This has been observed in joint finger-tapping tasks that require basic in-phase coordination akin to chorusing (Konvalinka et al., 2010), as well as in tasks that require individuals to produce movements in alternation, which involves turn-taking. In these studies, cross-correlation analyses revealed interpersonal dependencies in the timing of co-actors’ finger taps that were suggestive of mutual assimilation. Temporal assimilation may be a form of imitation that facilitates ensemble cohesion by making multiple individuals sound collectively as one.

A more cognitively advanced skill is prioritized integrative attending, that is, simultaneous attention to one’s own actions (high priority) and those of others (lower priority) while monitoring the integrated ensemble sound (Keller, 1999, 2001). Prioritized integrative attending thus relies both on an individual’s ability to divide attention between different sound sources, and on the group’s joint attention skills (to the extent that multiple individuals must attend to the aggregate structure that results from their coordinated actions). This variety of joint attention is particularly important in music characterized by complex turn-taking. For example, to perform the interlocking rhythms of Central African music, according to Nketia (1962), each player must have a general awareness of the resulting aggregate pattern, as well as the “knack” for coming in at the right moment. It has been proposed that the dynamic allocation of attentional resources that is required for prioritized integrative attending relies on entrainment (Keller, 2008). Specifically, the coupling of internal timekeepers to periodicities associated with the music’s metric structure may provide a temporal schema that allows attention to be modulated in a manner that is optimal for monitoring multiple levels of the musical texture concurrently (Keller, 1999; London, 2004; Jones, 2009). Evidence that such metric schemas facilitate prioritized integrative attending has been found in studies in which musicians are required to memorize one instrumental part, as well as the aggregate structure, of multi-part rhythm patterns (Keller and Burnham, 2005). For example, in one experiment listeners were required simultaneously to memorize a target (high priority) part and the overall aggregate structure (resulting from the combination of parts) of short percussion duets. The key finding was that recognition memory for both aspects of each duet was influenced by how well the target part and the aggregate structure could be accommodated within the same metric framework.

The third ensemble skill is anticipatory imagery, which involves the use of mental imagery in planning the production of one’s own sounds and predicting upcoming sounds of co-performers. Auditory or motor imagery may thus assist in anticipating others’ actions with a view to optimizing ensemble cohesion (Keller, 2008). In support of this hypothesis, it has been shown that individual differences in synchrony between pianists in duos are positively correlated with performance on a task designed to measure the vividness of anticipatory auditory imagery (Keller and Appel, 2010). With regard to the mechanisms underling this relationship, it has been claimed that anticipatory imagery relies on entrainment-based sensorimotor coupling to drive internal models that trigger mental images during action simulation (Keller, 2008). These internal models, which allow sensorimotor transformations between bodily states and events in the immediate environment to be represented in the brain, are harnessed during action simulation to generate predictions about an action’s future course (Wolpert et al., 1998, 2003). The accuracy and stability of sensorimotor synchronization in the context of real and virtual instances of interpersonal coordination have been shown to depend on temporal prediction abilities (Wöllner and Cañal-Bruland, 2010; Pecenka and Keller, 2011), which are, in turn, positively correlated with temporal imagery abilities (Pecenka and Keller, 2009a,b).

The infant is responsive to such expressive musical characteristics in the mother’s (or other caretaker’s) displays, and also participates in “communicative musicality” (Malloch and Trevarthen, 2009; the infant’s behavior has also been referred to as “protomusical,” e.g., Cross, 2003), for example, in predicting the timing of the mother’s expressions and producing responses that are to some degree temporally coordinated (e.g., Crown et al., 2002). Jaffe et al. (2001) define temporal coordination as a form of interpersonal contingency, and they describe interactional (non-periodic) rhythms which allow an infant and parent to each predict the timing pattern of the other’s behavior. This ability to predict timing is necessary in order to eliminate the time lag of a sensorimotor reaction (Merker et al., 2009), and is considered to be essential to bonding within the dyad (Jaffe et al., 2001). In a common game played by adults to engage infants, violation of expectation in timing is used to make the infants laugh, also supporting infants’ ability for predictive timing (Stern, 2007). Infants’ production of musical actions, as in sung tones or rhythmic movements, begins to emerge during the first year (perhaps especially with encouragement) and can be considered in the context of the communication of the pre-verbal infant (Trehub, in press). Yet before full musical actions emerge, the infants’ responsiveness to musicality and the timing of their responses may build upon a repertoire of non-verbal vocal and gestural expressions (Eckerdal and Merker, 2009).

Infants thus engage in a kind of “sympathetic conversation” with their mothers, the timing of which enables them to anticipate and relish in the mothers’ expressive displays, and causes them distress if timing is mismatched (see Trevarthen and Aitken, 2001). Temporally coordinated actions can be simultaneous, dovetailing, or alternating (Feldman, 2007), though these interactions are not typically strictly periodic, and are stochastic and bidirectional in organization (Cohn and Tronick, 1988). When mother–infant interactions take the form of chorusing, they have been referred to as “coaction” (Dissanayake, 2000). In this context, the behaviors that infants display with their caregivers include social gaze, facial expressions, and vocal behaviors (by 3months of age), gesture, and shared attention to objects (after 6months of age; Feldman, 2007). The roughly unison or matched forms of interaction between infants and their caregivers might be attributed to motor resonance (e.g., Meltzoff and Decety, 2003), and reflect emergent coordination behavior. The function of this emergent coordination is to establish affective entrainment, cooperation, and bonding between the infant and her caregiver (Feldman, 2007; Feldman et al., 2007).

Turn-taking in infancy has been studied primarily in the context of observing spoken conversation, and indicates the importance of the development of social cognition and attention. Infants shift gaze or attention as they observe videos of adults engaged in conversational turn-taking (von Hofsten et al., 2005), and gaze shifts become increasingly predictive with age (Bakker et al., 2011). By 4–6months infants are sensitive to cues of social cognition (such as selective attention to face-to-face interactions) in turn-taking (Augusti et al., 2010), which coincides with the development of their ability to use infant-directed speech cues to choose their preferred social partners (Schachner and Hannon, 2011). Before semantics and syntax are shared between conversational partners, cues to turn-taking are manifest in culturally dependent vocal prosody, eye gaze, and body movements (Wilson and Wilson, 2005) but with a universal target of timing in turn-taking (Stivers et al., 2009). A further cognitive skill in interpretation of conversational turn-taking is anticipating transitions in conversation (i.e., when the next speaker will begin to speak), which does not develop until around 3years of age and might be influenced by language development (von Hofsten et al., 2009). This anticipatory timing skill might also correspond to improvements in planned coordination of actions (Figure ​(Figure22).

Spatiotemporal imitation appears in the repertoire of young infants – for example in facial (see Meltzoff and Moore, 1997) and manual gestures (Bekkering et al., 2000), as well as affective mirroring, beginning with the first social smiles at just 6weeks of age (Rochat, 2007) – which are all examples of emergent coordination (Figure ​(Figure2).2). The timing of imitation is limited by cognitive–motor maturity but is taught in part by caregivers to young children, often via imitation games aimed at encouraging joint action (Gergely et al., 2002; Sebanz et al., 2006; Papoušek, 2007). In early stages imitation is automatic and relies primarily on motor resonance (Paulus et al., 2011), even more strongly in practiced behaviors such as crawling than in novel behaviors (i.e., walking; van Elk et al., 2008). While many actions of young infants may rely on resonance, according to von Hofsten (2004), even from birth such actions are not mere reflexes but can be motivated, informed, and goal-directed. Evidence for this claim includes infants’ interest in tracking and imitating the purpose and the outcomes of observed actions (e.g., von Hofsten and Siddiqui, 1993; Gergely et al., 2002; Gergely and Csibra, 2003). The goal-directed nature of infants’ actions, revealing planning, prediction, and motor representations, could enable the progression (with muscular and especially cognitive development) from resonant imitation to more complex forms of joint and complementary joint action.

In the earliest joint actions that are performed by infants with adults, the adult typically helps the infant to achieve her goal, in which case a precise motor representation is not required (Vesper et al., 2010). By around 1year of age, several cognitive changes facilitate joint action. First, joint attention has emerged (see Rochat, 2007), in which individuals knowingly attend to the same object or event. This skill coincides with monitoring of relative shared attention between the infant and his social partner (Rochat, 2007). The 1-year-old understands intentions, which has been argued (Tomasello et al., 2005) to be the basis for understanding beliefs (i.e., theory of mind) – emerging around 15months (Onishi and Baillargeon, 2005) and continuing to develop with experience with language and shifting of perspective (Tomasello et al., 2005). At 1year action-planning is evident, as infants show goal-directed eye movements that reflect motor representations (Falck-Ytter et al., 2006). These motor representations are thought to be supported by the mirror neuron system: a brain network that is recruited similarly during action perception and action execution (Falck-Ytter et al., 2006), and is thought to be involved in the interpretation of music and dance in adults (Stevens et al., 2001; Zatorre et al., 2007; Overy and Molnar-Szakacs, 2009).

Beyond the first year emerge the abilities to interpret goal-directed actions in a rational manner (e.g., Gergely et al., 2002), suppress imitative motor representations, and eventually perform complementary joint actions (Sebanz et al., 2006). Motivation for these changes stems in part from “shared intentionality,” or the sharing of psychological states in order to reach mutual collaboration (Tomasello and Carpenter, 2007). Planned coordination calls upon the more advanced cognitive–motor skills of precise (and shared) sensory and motor representations, action monitoring, and behavioral modification (cf. Keller, 2008). Presumably, concurrent with the maturation of the above-mentioned cognitive and motor skills (hence less reliance on scaffolding by adults) as well as language development (e.g., von Hofsten et al., 2009), the practice of musical activities during childhood may continue to engage attention and foster coordination skills. For example, prioritized integrative attending, when musical ensemble performers monitor the aggregate sounds (which children’s choirs can do to an extent) presumably builds upon the earlier abilities for joint attention and dividing attention between actors, once a shared goal representation is also established. Between the ages of 2.5 and 3years, the skills of coordination (timing) in joint action show substantial improvement even if individual performance on a task improves only marginally (Meyer et al., 2010). Complementary roles in joint action appear to be mastered from the age of 3years (similar to linguistic turn-taking; see von Hofsten et al., 2009), as action-planning and control become refined (Meyer et al., 2010). Improvisation in social contexts may have a role in the building of planned coordination capacities upon emergent coordination capacities. For example, turn-taking has been observed in children’s vocal play, even in improvised and complementary forms (Dissanayake, 2000), and such vocal play is expressed in virtual social contexts, as in imaginary dialogs and play-acting (see Rochat, 2007).

To achieve temporal entrainment in music, it may be necessary to practice the skill of sensorimotor synchronization – that is, the ability to move in time with perceived external events (Repp, 2005). For example, the clapping games and nursery rhymes, songs, and group dances, of school-age children begin to demonstrate the kind of coordination that is required to keep time as a group (e.g., Provasi and Bobin-Begue, 2003), with an external periodic pulse (McAuley et al., 2006), and using multiple levels of metrical hierarchy (Drake et al., 2000). Such music and games, which represent collective social entrainment (Phillips-Silver et al., 2010), may also foster the development of ensemble playing, and play an important role in the improvement of automatic and deliberate adaptive timing skills, attention, and auditory–motor imagery (Keller, 2008).

The ability for precise synchronization seems to mature gradually, probably building on early perceptual abilities for processing the musical beat (Hannon and Trehub, 2005; Phillips-Silver and Trainor, 2005; Winkler et al., 2009). In studies of synchronization of body movement with music or with a musical partner (in children between the ages of 5months and 5years), the children’s motions – or the sounds produced by them – are not tightly synchronized (phase-locked) to the musical beat (Eerola et al., 2006; Kirschner and Tomasello, 2009; Merker et al., 2009; Zentner and Eerola, 2010). This suggests that sensorimotor synchronization in music is not typically developed until sometime later in childhood or near adolescence (Merker et al., 2009; although cases of exceptional childrens’ musical performances can suggest otherwise, e.g., Merker et al., 2009; Sowinski et al., 2009)¹.

We have suggested that musical joint action builds upon early manifestations of affective and temporal entrainment. The affective component of entrainment – which is central to interpersonal synchrony in music – is grounded in emotional resonance and affective mirroring in early infancy (Rochat, 2007). This component is arguably the first and foremost in ontogeny of coordinated action and musical communication. For example, imitative forms of chorusing and turn-taking are mastered more readily than complementary varieties of musical joint action. Imitation may also provide the easiest route to the socio-emotional benefits of joint musical experience. Indeed, synchronous, matched motor activity has been found to foster interpersonal affiliation (McNeill, 1995; Hove and Risen, 2009), cooperation and pro-social behavior (Wiltermuth and Heath, 2009; Kirschner and Tomasello, 2010), and even altruism (Valdesolo and DeSteno, 2011).

The temporal component of entrainment emerges in various forms during childhood: through joint attention activities in infancy, the improvement of adaptive timing, and finally anticipatory processes in school-aged children. Formal music training plays a role in developing such capacities for attention and executive functioning (see Hannon and Trainor, 2007), as well as refined auditory–motor interactions (see Zatorre et al., 2007). Hannon and Trainor (2007) have suggested that joint task goals, action monitoring, and synchronization can contribute to the benefits of musical training on cognitive and motor processing abilities in musical joint action. As increasing capacities for imitation and complementary joint action are developed, the cognitive and motor demands of chorusing and turn-taking can be met in musical joint action.