According to intentionalism, the meaning of my words is determined by my intentions in using them. This theory, while both appealingly simple and surprisingly accurate, quickly runs into a problem: according to intentionalism, it looks like my words mean whatever I want them to mean. But surely there are some limits! We should, for instance, hardly want to follow Lewis Carroll's Humpty Dumpty in saying that 'glory' can mean 'a nice knock-down argument' if only one really wants it to. And yet, when we strive to offer a semantics for terms like demonstratives (e.g. 'this' and 'that') and pronouns (e.g. 'he', 'she', and 'it'), appealing to speakers' intentions has struck many philosophers not just as an appealing option, but as the only viable option around.
The primary goal of this thesis is to introduce a satisfying, and genuinely non-Humpty Dumptian, version of intentionalism. The view for which I argue---what I call the 'constraint theory'---avoids two serious problems facing earlier attempts to overcome Humpty Dumptyism. First, many earlier intentionalist theories make implausible predictions in cases where the speaker is sufficiently confused about the context. And, second, most earlier theories fail to generalize beyond demonstratives and pronouns. That is, these theories have been so tailored to account for the particular characteristics of these terms that they make inaccurate predictions when extended to account for other sorts of referential terms, like names. In contrast, the constraint theory not only makes accurate predictions in cases of speaker confusion, it looks well-situated to generalize beyond demonstratives and pronouns---so as to provide a unified account of the meanings of referential terms in context.
Chapter 1 introduces the problem of reference fixing in more detail, and discusses a bit of its history. While the Humpty Dumpty Problem was first introduced as a objection to the claim that there are distinctly referential uses of definite descriptions, I explain why the problem turns out to be even worse for intentionalist accounts of demonstratives. These terms will thus serve as the focus of the three main chapters.
Chapter 2 considers the prospects for defending a non-intentionalist theory of how the reference of demonstratives is fixed in context. By far the most popular theory of this sort is 'salientism', or the view that it whatever is maximally salient in context is what a use of a demonstrative refers to. Maximal salience, in turn, can be spelled out in several different ways. I argue that each of these ways exhibits serious flaws, and that there are reasons to think that no version of salience will be able to account for certain sorts of cases---involving speakers who are tying to deceive their listeners in a particular sort of way.
Chapter 3 introduces a classic hard case for intentionalists, the 'Carnap-Agnew' case. The force of the case requires our imagining a speaker who is confused about the world in a particular sort of way---and it is this confusion, in particular, that makes the case a hard one for intentionalists. I consider three intentionalist theories that purport to account for this sort of case: the perceptual grounding theory, the coordination account, and the neo-Gricean account. I argue that, while each can account for at least some versions of the Carnap-Agnew case, each suffers from some serious drawbacks. The perceptual grounding theory is extremely limited in its scope, the coordination account relies on several implausible assumptions about idealized listeners, and the neo-Gricean theory seems forced to collapse linguistic and nonlinguistic knowledge in order to make accurate predictions. I then introduce my own proposal, the constraint theory, and show how it can account not just for the Carnap-Agnew case and its many variants, but also for analogous cases involving names.
Chapter 4 focuses on the neo-Gricean theory of reference. One possible concern with the criticism of that theory offered in chapter 3 is that it might appear to rest on what amounts to a mere technical flaw. That flaw might, in turn, seem to be fixable. In this chapter, I therefore endeavor to introduce a deeper problem for the neo-Gricean---one that should prove instructive for the theory of meaning more generally. Contrary to what the neo-Gricean predicts, speakers can both intend to and succeed in using names and demonstratives to deceive their listeners not just about the world, but about the very content of what they say. I call these 'sneaky' uses of referential terms, and I spend most of the chapter working through the ramifications of these cases for the Gricean theory of meaning and its successors. I argue that these cases present a serious challenge to that class of theories, and one that is not easily avoided. The constraint theory, in contrast, neatly accounts for sneaky cases. What's more, it does so while still preserving many of the virtues of the Gricean theory.
Chapter 5 concludes the thesis by briefly tying up two loose ends: showing how the theory can be extended to account for so-called 'true' indexicals like 'I', 'here', and 'now', and explaining how the theory can endorse the claim that sneaky uses of referential terms partially ground the standing meaning of these very terms while simultaneously endorsing the claim that sneaky uses of referential terms asymmetrically depend on cooperative uses.