Our linguistic and inferential practices are said to implicate a kind of abstract object playing various roles traditionally attributed to propositions, and our predictive and explanatory success with this ‘‘proposition-talk’’ is held to underwrite a realistic interpretation of it. However, these very same practices pull us in different directions regarding the nature of propositions, frustrating the development of an adequate unified theory of them. I explain how one could retain proposition-talk, and the advantages of interpreting it as being purportedly about (...) propositions, even if problems about the identity conditions for propositions motivated a Quinean rejection of them. The non-error-theoretic solution is to understand proposition-talk in terms of semantic pretense. On this approach, talking as if there were propositions lets us put readily available logical and linguistic devices to new expressive purposes, providing a way to make indirectly certain complicated, genuinely true assertions we cannot make directly. Proposition-talk thus extends the expressive capacity of a language in a logico-syntactically conservative way. (shrink)