To paraphrase a claim is to provide an alternative formulation of it. But what are the adequacy conditions on such reformulations? Kant, for example, claimed that the different formulations of the categorical imperative are different formulations of one and the same law. But was he right about this? How are we to know? The problem is widespread. Moore argued that moral claims cannot be expressed using sentences containing only naturalistic language; platonists argue that reality cannot be adequately characterized without using sentences that refer to abstract objects; 'tensers' in the philosophy of time argue that tensed sentences cannot be adequately paraphrased by tenseless sentences; and dualists such as David Chalmers argue that certain facts about consciousness cannot be expressed using sentences containing only vocabulary amenable to reductive physicalism. Whether these claims are correct makes a considerable difference to the way we should conceive of the world and our place within it. But how are we to determine whether they are correct? Despite the importance of developing an adequate understanding of paraphrase, surprisingly little work has been done on the topic. The present dissertation addresses this deficiency. It exposes a widespread misconception about JohnChristopher Adorno Keller the function of paraphrase: that its purpose is to remove unwanted commitments. I argue that the proper way to understand paraphrase is as revealing the commitments of our discourse. Paraphrase is important precisely because we do not in general know what the entailments of our theories are. The entailment relations our theories stand in, being truths of logic and hence necessary, cannot be removedÌ¢ âÂ' the only commitments paraphrase is able to 'remove' are commitments that were not really there in the first place: merely apparent commitments. An important part of the argument for the revealing conception of paraphrase lies in showing that two presuppositions of the removing conception are false. These are, first, that the 'logical forms' of claims we accept are transparent to us; and, second, that when being truthful we typically believe what we say. The first presupposition, despite the ubiquity with which it is accepted, simply does not stand up to critical scrutinyÌ¢ âÂ' if it were true, it would be trivial to find adequate formulations of our thoughts in logical languages. The problem with the second presupposition is less obvious. My argument draws on recent work in the philosophy of language by Kent Bach and others, which shows that the proposition we communicate is rarely the semantic content of the sentence we use to communicate it. It follows that semantic analysis of the sentences we utter is an unreliable method for determining what we believe. The final chapter of the dissertation shows that an influential objection to Quinean meta-ontology is based on a misunderstanding of the function of paraphrase, and spells out the dire consequences of the position adopted as a result of that misunderstanding.