Kaplan observed in Demonstratives that in indirect discourse the Content but not the Character of the reported utterance can be preserved. Thus whether John says: 'My pants are on fire' or: 'His pants are on fire' (where his refers to John himself), both speech act can be truly reported by: 'John says that his pants are on fire'. Kaplan gave a semantics on which 'John says that his pants are on fire' is true just in case for some character C asserted by John, the content of C in John's context of speech is that John's pants are on fire. This yielded the result that the report should be true in case Kaplan said 'My pants are on fire', or 'His pants are on fire', as is desired. We modify Kaplan's account. In particular: (a) Whereas Kaplan's analysis applied solely to 'say', we extend his analysis to all attitude verbs; (b) We provide an account of what it means to 'assert' or 'believe' a Character. In a nutshell, an individual i believes a Character C just in case each of i's epistemic alternatives belongs to the diagonal of C. (c) In addition, we depart from Kaplan's analysis of Characters in that we replace his world-time index with a second context parameter, which yields in certain ways a more elegant system and more powerful system. We then show that our version of Kaplan's solution is, on closer inspection, inadequate. Despite its intuitive plausibility, it turns out to predict trivial truth conditions. We will prove that whenever a subject x believes a true proposition p at some index, then x has to believe any other proposition q that is true at that index. In order to solve the problem, we restrict the existential quantification over Characters to suitable Characters. It is not clear what the restriction should be. The rather natural option that the Character must be expressible in the agent's language can be trivialised under certain assumptions on expressibility. The puzzle raised is an important one because it seems to us that the overall approach to the semantics and syntax of attitudes is the correct one...