This paper is an advertisement for a problem. Preference -- what is better than what -- is (somehow) tied to what we hafta do and so (somehow) to obligation-describing language. But sometimes our predicaments are less specific -- more chunky -- than any preference relation could be. Even in such predicaments our obligation-describing language isn't murky: there are clear facts about what we hafta do. And so we need to understand chunky preference in a way that plays nice with our obligation-describing language. But here's the problem: (i) the easy answer (roughly: steal from the mushy credence crowd) won't work, (ii) probing why suggests a natural constraint that leads us straight into a theorem (following from a version of Arrow's theorem) that says that only the most un-chunky preferences can matter, but (iii) any picture like that is at odds with our obligation-describing language.