There is a "production-first" way of thinking about the project of semantics - in particular, about what it is for a sentence to have truth-conditions - that makes expressivism seem at least very puzzling, and more likely impossible or incoherent. I'll look at an anti-expressivist argument from Frank Jackson and Phillip Pettit that presupposes this production-first way of thinking about semantics, and at David Lewis's articulation of a production-first picture of the project of semantics in "Languages and Language". I then argue that the choice of a production-first framework is unforced, and that adopting a different picture (either context-first or acceptance-first) allows us to recognize the possibility of a coherent expressivism.