Abstract

Frege's puzzle about identity, Kripke's puzzle about belief, and Mates's puzzle about synonymy can be solved with a semi-quotational account of belief ascriptions. On this account, belief ascriptions have one sense in which they mean that the subject believes something about a word. This is preferable to a purely quotational view because it allows for the possibility of ascribing beliefs to someone who does not speak the same language. In support of the semi-quotational account, I propose a set of tests for detecting when a sentence or proposition is about a certain word. The tests show that the puzzling sentences are indeed understood quotationally. ;The semi-quotational account provides a semantic solution to puzzles about substitutivity. Others have defended a pragmatic solution, according to which substitutions of co-referring names or synonyms in belief contexts cannot change the truth value of the sentence. In Chapter 2 I show that the arguments which have been offered for pragmatic theories are unconvincing, and that my semantic theory explains the data at least as well. ;In Chapter 3 I argue that the semi-quotational account is compatible with a broadly Fregean view of language. Moreover, the puzzling sentences pass some standard tests for ambiguity, as would be expected. The tests also show that many puzzling belief ascriptions involving imperfectly understood words depend on a quotational sense. ;Chapter 4 looks at substitution failures of synonyms with different structures, and specifically at the paradox of analysis. I argue that most common versions of this paradox are not generated by analyses per se, but by any pair of synonyms. These versions have a straightforward quotational solution. I then try to answer the related question of how learning an analysis can be useful. ;Finally, it is natural to think that, despite appearances, proper names must be substitutable because their function is simply to refer. Against this I propose a more plausible account of the function of names. Specifically, I argue that by using names we maximize the amount of information conveyed. I explain what this information is and how it is maximized