If publication rates are an accurate guide to popularity, then the two currently most popular approaches to belief ascription are Millianism and Contextualism. The former rules out ordinary Frege cases such as Lois believing that Superman flies while failing to believe that Kent flies. The latter approach, as I will define it for the purposes of this essay, includes all theories according to which standard uses of belief reports of the form 'A believes that a is F' and 'A believes that b is F', where 'a' and 'b' are coreferential proper names or general terms, may differ in truth-value even though the proposition that a is F is identical to the proposition that b is F. For those who enthusiastically concur with current conventional wisdom, the primary argument of this essay, centring on a puzzle about belief reports, is modest; for those who don't buy the fashionable views, the argument is challenging. Facts about the puzzle case and commonsensical principles about belief entail that one of the two approaches must be correct, although the puzzle is perhaps best thought of as a test case for all theories of belief ascription. If one thinks that any version of either Millianism or Contextualism must flout important intuitive principles regarding belief, then one is left with the conclusion that any adequate theory of belief ascription will have to be significantly revisionary with regard to semantic intuitions. If so, then we can forget about letting those intuitions have the weight that they are commonly accorded in theory construction for belief ascription (and, as a consequence, belief). I present the argument by altering, elaborating and examining a puzzle that I introduced in a rough-and-ready way in earlier papers (1998, 1999). If the reader has had enough of puzzles about belief or belief ascription, I