The understanding of joint action typically involves several levels. At the institutional (macro) level, it often involves rules, laws, institutions, functions that typically provide background constraints and common meanings. At the personal (meso) level, it involves a number of psychological attitudes (goals, beliefs, commitments, intentions…) that must be suitably intertwined. At the psychological or cognitive (micro) level, it involves mechanisms that facilitate or trigger such attitudes and allow for coordination. Coordination and cooperation in the context of musical performances has long been a core focus for the literature on joint action – starting with Bratman’s (2014) example of singing a duet together. Indeed, musical performance seems particularly relevant when discussing the possible links between levels of understanding. In the case of an execution of a pre-composed piece of music, a key factor which explains the success of joint action at the meso level is a shared intention to follow a same score (or different individual parts of said score). This factor is determinant to explain how the musicians end up building together a complex, temporally extended musical form. Of course, this is not the only determinant. Some small-scale mechanisms of motor-synchronization, joint attention, and perception-action matchings also contribute to the success of the joint action (Wöllner & Keller 2017), as well as general principles defining the regulative norms of interpretation of musical works, embedded in the general musical culture and institutions (Cook 20013). That being said, the meso level of shared representations and intentions (which is often obtained through planned coordination) seems decisive in explaining the possibility and success of such a complex and temporally extended joint musical action. However, in the practice of musical improvisation, and in particular free improvisation, a good deal of the shared representations and intentions found at this level seem to be missing. In collective free improvisation (Bailey 1992), musicians aim at performances whose very structure and conditions for success are unknown from the get-go. In other words, the content of the performance is supposed to emerge from the very interaction of the musicians. When free improvisation is successful, it provides an interesting case where a complex joint action occurs in the absence of any detailed shared representations and intentions. Even though the existence of emergent – as opposed to planned – coordination is well documented in the literature on joint action (Knoblich et al. 2011; Butterfill forthcoming), the case of collective free improvisation is noteworthy for its level of complexity and duration (compared to the classic examples of synchronized rocking chair movements or applauses), since such complexity is often thought to be achievable only through planned coordination. In that respect, it is remarkable that the complexity of collective musical improvisations is more often than not comparable to the complexity of performances consisting in the mere execution of a pre-composed work. This phenomenon naturally raises the following question: how is the absence of these shared representations and intentions compensated in order to give rise to a successful joint action? According to a “bottom-up” hypothesis, the shape of the musical performance emerges from the mechanisms operative at the micro-level. In that perspective, there are no crucial differences between the case of collective free improvisation and the standard cases of emergent coordination. Even though the musicians do not start the performance with a shared representation of the music they will play together – and not even with a shared representation of the music they will play individually – successive short-term shared intentions emerge step by step in virtue of the combination of basic processes of emergent coordination (e.g. entrainment, affordances, action-perception matchings, common predictive models). In other words, the skillful combination of these processes among the members of the group could explain the temporally extended construction of a complex musical form, without any pre-defined shared representation of the resulting piece of music (see for example Canonne 2013 on how expert improvisers use salient musical events to “move” swiftly from one sequence of the improvisation to the next). But one might wonder if the basic processes of emergent coordination provide sufficient strong constraints to guarantee the convergence of the short-term musical intentions of the musicians to yield the observed complexity. According to a competing “sandwich” hypothesis, the successful coordination observed in collective free improvisation is to be explained both by the skillful combination of micro-level processes and the existence of a musical common ground between the improvisers which belongs to the third institutional level. In other words, in order to be able to improvise together successfully, musicians would need to share compatible regulative norms regarding the practice of improvisation. Evidence for this alternative hypothesis is provided by ethnographic work on free improvisation ensembles (Canonne, forthcoming). This hypothesis, if further confirmed, would vindicate the relevance of the third institutional layer of collective intentionality in order to understand even small-scale coordination phenomena such as musical improvisation. Collective free improvisation is paradigmatic of the spontaneous and creative joint actions that we often encounter in our everyday life (Ingold & Hallam 2007). The detailed analysis of this phenomenon should greatly help us in precisely understanding how such complex and unscripted joint actions can occur – particularly in showing how the interaction of bottom-up and top-down processes, and the combination of micro-level psychological mechanisms and macro-level regulative norms can compensate for the absence of planned coordination and pre-established shared representations or intentions.