Paradox in Thought and Natural Language

Around 600BC, Epimenides, a Cretan apparently discontented with thehonesty of his compatriots, lamented that all Cretans are liars.Together with a few innocent assumptions, well-entrenched principlesof logic entail that Epimenides' lamentation cannot be true, and yetcannot be untrue---a flat contradiction. What's gone wrong? In thisdissertation, I argue that the source of the problem has beenmisdiagnosed as one about language (especially formal languages). Theproblem runs deeper, and stems from the structure of thought itself. The dissertation proceeds in two main stages. The first stage(Chapter 2) makes the case that that the intuitions that underlie theparadoxes come from natural languages, not from formal/mathematicalones. The Liar and related paradoxes are generally presented asconstraints on the latter. Their lesson, the story goes, is that noformal theory strong enough to represent the primitive recursivefunctions can include a satisfactory truth predicate. I argue thatit's our natural-language competence with the truth predicate thatunderlies our understanding of what 'satisfactory' means here, whichshifts the focus of the project to natural language semantics. In thisdomain, it's tempting to think (and many have thought) that theproblem with Epimenides' utterance is that it fails to express aproposition, and this failure explains why we have trouble assigningit a truth-value. Or, perhaps it does express a proposition, but notthe one that it seems to express. Or, perhaps it can express aproposition, but which proposition it expresses depends on context. Iargue that all such responses fail, in part because they cannot makesense of related attitude attributions. I can believe or disbelieveEpimenides, which wouldn't be possible if his utterance didn't expressthe proposition it seems to express.In the second stage, I argue that such paradoxes arise, not from thelanguage/thought interface, but rather from thought itself. The firststep in this argument concerns knowledge attributions (Chapter 3),where I develop and defend a novel solution to the Knower paradox.Then I move from attitude attributions to attitudes themselves(Chapter 4). Just as sentential truth and knowledge predicates givesrise to paradoxical sentences, seemingly innocent combinations ofbeliefs and desires give rise to paradoxical propositions---even whenthose beliefs and desires are not expressed in language. Thepossibility of such pathological combinations isn't accounted for byany extant theory of mental content, and, I argue, provides supportfor a non-classical theory. Finally (Chapter 5) I consider anobjection to these putative combinations of desires. I introduce whatI call /advisory/ desire reports, which seem to exhibit the radicallyexternalist behavior that the previous chapter rejects. I conclude byoffering reasons to think that the availability of these readings doesnot undermine the case for non-classical accounts of attitudes