What is the Role of a Truth Theory in a Meaning Theory?

Kirk Ludwig

The suggestion that a truth theory, in the style of Tarski, can play a central role in a compositional meaning theory is a familiar one, due to Donald Davidson. However, it is often not clear from Davidson's work, or from that of his followers, exactly how we are to conceive of the connection. It has sometimes been thought that the truth theory is supposed to replace a meaning theory, to provide the most we could provide in the way of a compositional meaning theory. On this view, the truth theory does not serve as a compositional meaning theory, but as a more philosophically and scientifically respectable replacement of it. I do not believe that this was Davidson's intent, though there are certainly things he says that would lead one to believe this, and I think many people have been misled. In this paper, I want to explain what I think the connection is, and make much more explicit the role of a truth theory in a compositional meaning theory than Davidson has. I will do this by stating explicitly the form of a theory that entails all instances of (M),

(M) s means in L that p

for a language L in a way that makes central use of a truth theory, and which I believe meets at least in spirit the other requirements on an adequate compositional meaning theory. Even if I am mistaken in thinking that this represents Davidson's line of thought (one will not find what I say explicitly in anything he has written), it illustrates one way (an important way, I think) of seeing how to exploit the recursive machinery of   a truth theory to meet the goal of a compositional meaning theory. And it has some unexpected benefits. In particular it enables us to see how the theory can achieve its aims even though the language contains defects that spell trouble for a truth theory simpliciter. I have in mind specifically the semantic paradoxes and semantic vagueness, both of which spell trouble for truth theories for natural languages. I first develop the account for a context insensitive language, where the issues will be clearest. Then I extend the account to context sensitive languages. Finally, I show how this helps us out of what have been taken to be some very serious difficulties for the truth-theoretic approach to meaning.