Abstract: When we design and interpret speech acts, we need to rely on background information. Which background information do we rely on? Which should we rely on? An influential tradition says that the answer to both of these questions is “common ground,” which is a body of information that is, in some sense, shared by the speaker and their audience. I will defend the alternative view that communicators do and should bring to bear all of their beliefs, and not just those that are common ground, for the purposes of audience design and pragmatic reasoning. I will respond to two traditional ways of defending common ground, one rooted in the philosophy of language and another rooted in psycholinguistics. Of course, a communicator’s beliefs include their beliefs about interlocutors beliefs, including beliefs about interlocutors’ beliefs about their own beliefs, and so on. But, I will argue, each step in this metarepresentational hierarchy requires more scarce cognitive resources than the last, and there is evidence that we make intelligent tradeoffs when deciding how to invest these resources. My conclusion is that common ground is best understood as part of a model of human communication that idealizes away from the limitations that necessitate such tradeoffs.