The human faculties for communicating acoustically with words and music rely on cognitive mechanisms that are strikingly similar, and which a long series of alternating theoretical and empirical arguments have suggested were adapted either for language, music, or a common antecedent of both. One of the most-enduring proposals for selective pressures that may have shaped our cognition specifically for music is a pre-linguistic role it may have held for social cohesion, traces of which may be still manifest today in music’s ability to mediate social attitudes such as cooperativeness and trust. However, one still does not know on what cognitive capacity these effects rely, whether this capacity is functionally distinct from that utilized to communicate similar attitudes linguistically, and whether it was indeed selected specifically for music. In this talk, we will show that spontaneous dyadic musical interactions can not only mediate, but directly communicate complex positive and negative social attitudes such as conciliation, disdain or insolence, and that such communication relies for a significant part on extra-linguistic cues linked to temporal and harmonic coordination. However, striking differences between musicians and non-musicians show that the capacity to process such cues is not strongly innately constrained, suggesting it wasn’t a specific adaptation for music. While casting doubts on social cohesion scenarios for the origins of music, these results nevertheless establish that more diverse and complex social behavior is possible with music than previously believed, opening new avenues for diagnosis and remediation of social processing disorders such as the ones found on the autism spectrum.