There seems to be something special about sentences like ‘all bachelors are unmarried’ and ‘red is a colour’. Philosophers have claimed that this is because they are analytic, where this is to say that they are true in virtue of meaning, and that anyone who understands one can know that it is true. Some have also claimed that the notion of analyticity can be used to solve problems in epistemology. However, in the last century the work of Quine and Putnam led many to doubt such claims, and to suspect that there is no analyticity, only an illusion of analyticity to be explained.

I think that a number of Quine and Putnam’s insights were important and correct. Moreover, they can help us in formulating a better account of analyticity. The core of my dissertation is that account. I use recent developments in the philosophy of language--including Russellian semantics, distinctions between different kinds of meaning, two-dimensional modal logic and the distinctions between analyticity, necessity and a priority--to defend a theory of analyticity which covers both the traditional necessary analytic truths such as ‘all vixens are female’ and Kaplan’s examples of the contingent analytic such as ‘I am here now’, dthat[] = and ‘Snow is white iff actually, now, snow is white.’

I use the account to provide solutions to problems in philosophical logic concerning the correct accounts of vagueness, inconsistency and modality. I also show how it can be used to provide a solution to the Liar paradox.

Chapter 1 begins with the Liar paradox. Routley and Priest have proposed that intractable paradoxes such as this show that there are true contradictions. Though there are good arguments for this view--known as dialetheism--few philosophers in the analytic tradition take it seriously. One reason for this is that they don’t believe the view makes sense. Suppose there were true contradictions: what would that mean the world was like?

I present a version of dialetheism which offers an anodyne reply to that question: the truth value of any sentence is determined in part by the way the world is, but also in part by what it means; semantic dialetheism is the view that some contradictory sentences are true, but not as a result of some unimaginable strangeness in the world, only as a result of strangeness in our language. I use the notion of an over-defined predicate (a predicate whose extension and anti-extension overlap) to show how language can generate paradoxes.

The resulting position resembles Tarski’s view that semantically closed languages are inconsistent and I show how semantic dialetheism can handle Herzberger and Soames’ objections to Tarski. The view also has similarities to the view of vagueness according to which there is no genuine vagueness in the world, only vagueness as a result of our language. Some philosophers take a similar linguistic approach to modal notions.

But you might find something about such views puzzling. Take the case of vagueness. There is a language, consisting of a set of sentences, and there is the world, which is in a certain state. This world state is such as to leave some of the truth-values of the sentences indeterminate. But what is the significance of saying that it is language that is responsible for this indeterminacy?

The worry is similar to one which Quine presents in “Carnap and Logical Truth”. He points out that even in the case of the most promising examples of analytic sentences, such as ‘(x),(x = x)’ and ‘all bachelors are unmarried’, there is a state of the world “corresponding” to the truths: everything in the world is self-identical and all the bachelors out there in the world really are unmarried. So then why, when there is such a state of the world available, should we say in the case of these sentences that the language is responsible for the truth-value? Why single them out as true in virtue of meaning?

Quine thinks his observation shows that the even the most promising examples of analytic sentences are really just like synthetic sentences. But it doesn’t. Consider the binary multiplication function on the natural numbers, and in particular the following instance:

Just as the truth value of many sentences is determined in part by the way the world is, and in part by the meaning of the sentence, so the value of the multiplication function is determined in part by the value of its first argument (had it been 2 the answer would have been 10), and in part by the value of its second argument. Now consider a different instance:

There is a good sense in which this case is special. In a non-trivial sense the value of the function is determined by the value of the first argument. And this remains true in the face of the observation that there is a second argument--namely 5. Similarly, the observation that the bachelors of the world are all unmarried does not undermine the claim that ‘all bachelors are unmarried’ has its truth-value determined by its meaning.

In the rest of the dissertation I examine the analytic/synthetic distinction with a view to resolving the puzzle about the distinction between semantic and ontic approaches to contradictions, vagueness and modality.

In chapter 2 I review the history of positive accounts of analyticity, concentrating on the treatment in Locke, Hume, Kant, Frege, the writings of the positivists, and modern day defenders such as Katz and Boghossian. In chapter 3 I discuss the case against the distinction. I cover the major points from Quine’s “Truth by Convention”, “Carnap and Logical Truth”, “Two Dogmas of Empiricism”, and Word and Object, Putnam’s work on natural kind terms and his argument that the sentence ‘cats are animals’ is not analytic, and the work of Harman, especially in “The Death of Meaning”, “Against Conceptual Analysis” and “Analyticity Regained?”

In chapter 4 I argue that the debate over analyticity has been hampered by an intuitively appealing but false view of language, which I call ‘The Language Myth’. The Language Myth consists in the conflation of four different kinds of meaning: referent/extension, content (that which an expression contributes to the proposition expressed by a sentence containing it), character (that which a speaker must know in order to count as competent with the expression) and reference determiner (that which determines the referent/extension of the expression). I present the myth and explain how it supports a bad version of the analytic/synthetic distinction. I then show why it is false, drawing on the work of Kripke and Kaplan to prise apart the different kinds of meaning.

Chapters 5 and 6 contain my positive account of analyticity. I claim that the kind of meaning which is important for analyticity is reference determination. I give a three-dimensional modal characterisation of what it is for a sentence to be true in virtue of meaning, but then show that this characterisation only approximates that given by a more important “underlying picture”, which posits non-modal elements to explain the modal properties of analytic sentences and which allows us to distinguish analytic sentences from non-analytic sentences with the same modal profile. (In a similar way, the theory of direct reference posits non-modal elements as the referents of some singular terms to explain why they have the modal property of being a rigid designator.)

It is because analytic sentences have seemed to have interesting epistemological properties that they are of interest to philosophers outside of the philosophy of language, and in chapter 6 I turn to epistemological issues. According to tradition, analytic truths are knowable a priori. However, the notion of a priority is itself difficult and controversial. I propose that we put the issue of a priority aside and consider whether analytic sentences can provide some distinctive kind of analytic justification for belief (regardless of whether that justification is strictly a priori or just a special kind of a posteriori justification). I discuss the problems posed by externalism about content and argue that a restricted class of sentences which are true in virtue of meaning provide us with a distinctive kind of justification.

In chapter 7 I address the arguments against analyticity one by one, and defend my positive proposal against them. I take it to be an advantage of my rejection of the Language Myth in chapter 4 that I can dispute Quine and Putnam’s rejection of analyticity and still do justice to some of their insights. From my point of view the traditional account of analyticity was constructed on a bad theory of meaning. Quine and Putnam’s arguments exploit the problems with this view of meaning to attack the analytic/synthetic distinction. But a better theory of meaning--inspired in part by those arguments--allows a better account of analyticity as well.

In the final chapter I use the distinctions and tools developed in chapters 4 and 5 to give a solution to the problem about the significance of claiming that vagueness, necessity or inconsistency can arise due to language. A central idea is to use the notion of reference determination from chapter 4 to give an account of what it is for a predicate to be absolutely perfect (as opposed to perfect relative to some domain of discourse). The ontic dialetheist thinks that there can be true contradictions couched in an absolutely perfect language. The semantic dialetheist denies this. With this problem solved I propose semantic dialetheism as the correct solution to the Liar paradox.
Primary Advisor: Scott Soames
Second Advisor: Gilbert Harman