While ethnomethodological conversation analysis has argued for the fundamentally situated deployment of communicative resources, the field of linguistics has broadly maintained the existence of a decontextualized structure that is assumed to reside in individual minds. In this talk, I will ask the question what kind of advances in understanding patterned ways of speaking (syntax) are possible when taking co-present embodied interaction as their starting point. By looking closely at interactants' coordination of multimodal resources at single instances of interaction, and building collections, also across diverse languages, we can arrive at conclusions about the true power of syntax as a dialogic tool of mutual engagement. Regardless of obvious methodological hinders – such as lack of video-recordings in the earlier history of languages – it is even possible to hypothesize about novel developmental grammaticalization paths, based on the analysis of present-day multimodal interaction. Empirical work on local episodes of interaction can reveal the full diversity of syntactic practices that may or may not emerge as grammar-as-we-know-it.