Accounts of common knowledge typically focus on how to represent what agents would know when they achieve common knowledge, instead of asking how agents achieve common knowledge. In this paper, by modifying a definition proposed by David Lewis, I generate a set of conditions the satisfaction of which would entail that a group of agents has common knowledge, properly so-called. I then ask what it would take to satisfy these conditions. I argue that these conditions can be satisfied and common knowledge achieved in the course of a conversation if, but only if, we adopt a particularly strong interpretation of Austin’s claim that the successful performance of a speech act requires the securing of uptake. I then argue that the generalization of this result to conversations involving more than two people requires that the audience to whom a speech act is addressed be able to recognize themselves as the audience of the speech act, in an essentially plural act of recognition. I close by arguing that one of the so-called paradoxes of common knowledge (the coordinated attack problem) in fact gives us some reason to favor Austin’s account of speech acts and therefore some reason to believe that common knowledge can, on occasion, arise.