An important worry about what Simon Blackburn has called "quasi-realism" is that it collapses into realism full-stop. Edward Harcourt has recently pressed the worry about collapse into realism in an original way. Harcourt presents the challenge in the form of a dilemma. Either ethical discourse appears to ordinary speakers to express representational states or not. If the former then expressivism means that this appearance is not saved after all, in which case quasi-realism fails in its own terms. If the latter, then we lose our grip on the idea that ordinary descriptive utterances are somehow paradigmatic assertions of fact and with it our ability to explain why such utterances really do express representational states. Finally, Harcourt argues that the expressivist's only hope for meeting these concerns relies on (a) a distinction between states of mind with a "world-directed" direction of fit (representational states) and those with a "state-directed" direction of fit (desire-like states), (b) the thesis that no state of mind has both directions of fit, and (c) the thesis that a state-directed direction of fit entails no truth-conditions. Previous critics have attacked (a) and (b), but Harcourt takes a novel turn and attacks (c). The challenge is a bracing one. If Harcourt is right, then the quasi-realist project inevitably destroys the very distinction in terms of which his position is cast â?? roughly, the distinction between beliefs and desires. I argue that Harcourt's challenge can be met. To anticipate, I argue that Harcourt's critique presumes that the quasi-realist's fundamental distinction is best understood semantically, whereas I shall argue that it is better understood functionally.