To address the question of the nature and function of conventions, this essay explores a possible systematic link between Heidegger's hermeneutical phenomenology and Foucault's historical ontology. It argues that these analyses establish, contra to the cognitivist interpretation, dominant then and now, that conventions are properly understood not as "common knowledge" (Lewis) nor even as "jointly accepted beliefs' (Gilbert)—that they are not, at least at their most fundamental layer, cognitive states at all—but are instead (individual and social) bodily dispositions that are forged by historically shifting entwinements of practices of power and forms of knowledge.