Field, Hartry. “Tarski’s Theory of Truth,” Journal of Philosophy 1972 vol. 69 (13) pp. 347-375
The short explanation is, “Blame Hartry Field”. I have been haunted by Field’s work on deflationism for some time now, and will continue to be in the years to come. At least in a manner of speaking, I am coming around to deflationism.
When you first encounter Tarski’s argument for a semantic conception of truth for an object language, it is tempting to dismiss it as being entirely useless for the purposes of answering the most interesting questions in epistemology. We want to know what makes a claim that is true for a language also true for reality, what truth means for our extralinguistic day to day practices, and how we can make sense of claims about the normativity of truth-claims in future contexts which involve the use of radically different languages. On first glance, it is not obvious that these expectations have been satisfied when you try to define truth in terms of its role in a semantic theory.
If you’re like me, anyway, those were the kinds of feelings you had when you first read Tarski. That’s not necessarily a slight against Tarski, mind you; he does double-duty in his initial article, to be both polemical and technical in roughly equal amounts. But sometimes in philosophy the success of an idea comes from its ability to be explained in unique ways by different kinds of thinkers, which is why it’s interesting to go over Field’s gloss on Tarski.
As Field explains, the advantage of Tarski’s theory is that it is economical. The theory is simple, in the sense that it provides an account of truth that contains no unexplicated semantic terms, like “applies to” or “denotes” (p.354, 360). Part of the genius of the system is that most of the intractable projects that have stopped philosophers from getting at a proper theory of truth (for example, the issue of explicating processes of denotation) are just offloaded onto the meta-language’s treatment of an object language. The point is that these concepts — of denoting, of application — are assumed by a theory of truth for an object language, but are not a part of that theory. If this helps, it might help to think of Tarski’s innovation by thinking of it in terms of the social science concept of a ‘division of labor’: Tarski, in effect, says, “Here is what truth explains. Here is what denotation explains. They are related, but are answering different questions.” By setting up a division of labor between the concept of truth and the concept of denotation, we know how to get on with truth without having to worry about it whether or not it will suit the latest tastes in philosophy.
Of course, these exported concepts (of application and denotation) do have to be given some kind of account eventually. (That is to say, our success in vindicating the truth of a statement is a function of how primitive names in a token of an expression of a language are referentially satisfied. If we have no theory of names, we will be unable to vindicate any truths.) But even if we never come up with a decent theory of denotation, that will not be because we have misunderstood truth, or misunderstood what it takes to say true things.
Perhaps that sounds like a Pyrric victory. After all, we learn from Field that Tarskian semantics is only slightly more powerful than a substantive characterization of truth — i.e., one that contains singular terms that can be explained as a function of primitive denotation. (p.370, 350) The advantage of Tarski’s system is that it is slightly clearer in its formulation of how definitions of denotation, application, and characterization of function symbols than its charming competitors. But this is not philosophically interesting, and Tarski’s attempts at providing a theory of denotation for particular languages (DE, DG) are not especially helpful (p.365-369). We haven’t solved the hard questions, we’ve just reframed hard questions about truth by turning them into questions about denotation and so on.
Field suggests that we might want to adopt a causal theory of denotation. As it happens, I agree with Field that a theory of denotation ought to be articulated partly in terms of causation. What this means, I do not know; the devil is in the details. Many causal theories of reference — say, correspondence theories — seem to set a very high bar on what counts as ‘true’ in an object language. Even so, notice that this is a worry is best expressed in terms of limitations we are placing upon the meta-language. The deflationist’s division of labor goes unchallenged.
Anyway — whatever happens, what we don’t want is for our meta-language to be too powerful. sp., we don’t want the meta-language to be built up on the basis of a conception of satisfaction that is powerful enough to explain tokenings of imperative sentences in an object language. I mean, a meta-language that gives us an articulation of the satisfaction conditions of tokenings of imperative sentences is not a meta-language that is explaining truth. e.g., the utterance of “Pass me the cup” has no truth-conditions — it has execution conditions. Any meta-language that assigns truth-values to the tokening of an imperative sentence will have completely messed up the semantics of truth for a language. John Searle has been banging on about this for more than half his life, and it’s one of the things he’s right about.