In the philosophy of language, I’m mainly interested in how we use language to attribute psychological states (like beliefs and desires) to ourselves and others.

Take the following pair of sentences:

S1   Aristotle believed that water was heavier than air.

S2   Aristotle believed that H2O was heavier than air.

These differ in truth value, despite being composed of terms that ordinarily stand for the same features of reality. This fact has been a major source of headache for those attempting to construct a tractable formal semantics for natural language. Our use of verbs like “to believe” appears to violate the principle of compositionality, which says that the semantic value of a complex expression is a function of the semantic values of its component expressions together with their manner of composition. In a compositional language, the truth conditions of a declarative sentence can be calculated from the customary referents of its constituent terms plus their mode of composition in the sentence. Given that we cannot calculate the truth conditions of a propositional attitude report this way, the question arises whether there is some other systematic basis on which we can calculate its truth conditions.

The standard way to deal with the challenge that S1 and S2 pose for compositionality is to hold that although “water was heavier than air” and “H2O was heavier than air” normally have the same semantic value (e.g., express the same proposition), they have different semantic values as they occur in S1 and S2. I deal with the problem differently, accounting for the different truth values of S1 and S2 by reference to a difference in the content of the word “believed” as it occurs in these sentences. So, while standard accounts see the structures of the sentences as:

Bel(Aristotle, φ1).

Bel(Aristotle, φ2).

 I see their structures as:

Bel1(Aristotle, φ).

Bel2(Aristotle, φ).

Philosophically, the most important result of this analysis is that it casts doubt on the “new fact thesis,” according to which it is impossible to acquire new truth-apt knowledge (“knowledge-that”) without also gaining knowledge of some fact previously unknown to the individual acquiring the knowledge. This thesis  underlies a number of influential arguments in the philosophy of mind and metaethics. (For more on this, see the entry on the knowledge argument.)

In graduate school I also wrote a couple of papers on the semantics of proper names. At the time, I was heavily influenced by Kripke’s attack on Russellian descriptivism, and the purpose of the papers was to defend Kripke’s non-descriptivist theory of names from some of the major objections to it. (See this and this.) These papers now strike me as poorly-motivated, since I no longer find Kripke’s objections to Russell convincing. When Russell held that in order to refer to a thing with which you aren’t directly acquainted, you must believe, of some condition which that thing uniquely satisfies, that it is uniquely satisfied, he didn’t mean consciously or occurrently believe. He meant to include dispositional beliefs as well. But once we take dispositional beliefs into account, Kripke’s objections to descriptivism fall flat. (If no amount of Socratic inquiry can lead me to judge, of any condition that Cicero uniquely satisfies, that that condition is uniquely satisfied, then I just don’t have thoughts about Cicero.)

A lot of my work in the philosophy of language focuses on indexicality, especially subtle forms of indexicality. In this paper about family resemblance, I argue that a family-resemblance predicate is best understood as an indexical that is sensitive to discursive features of contexts of utterance, in a way that is constrained but not fully determined by linguistic precedent.