I doubt that a deflationist about truth—someone who understoods truth at base in terms of Tarski's Schema (T)—who does not make use of propositions can make sense of the following claim:
- There was a probability at least 0.01 that somewhere there would evolve beings most of whose empirical beliefs are true.
- There was a probability at least 0.01 that somewhere in the universe there would evolve beings most of whose empirical beliefs b satisfy the condition that there is a sentence S of English such that b is a belief that S, and in fact S,
But one way to see that (1) and (2) differ is this. Let (1a) be a straightforward translation of (1) to Chinese—what a competent translator will produce given (1). Then, (1a) will be synonymous with (1). Now consider the claim:
- There was a probability at least 0.01 that somewhere in the universe there would evolve beings most of whose empirical beliefs b satisfy the condition that there is a sentence S of English such that b is a belief that S, and in fact S.
So, we conclude by transitivity of cognitive equivalence, that according to our deflationist, (2) and (3) are cognitively equivalent. But they are not. The reason is simple: Chinese and English developed in sufficiently different cultural milieux that there will surely be some concepts in the one language that have no equivalent in the other. (Think for instance of the impossibility of translating the English "nice" to various other languages, or the impossibility of translating the Hebrew "khesed" to English.) As a result, the substitutional quantification over all sentences of Chinese will pick up some sentences that have no English equivalents, and vice versa, and so different claims will be made by (2) and (3). (Indeed, if the probability of the evolution of linguistic beings most of whose beliefs are true turns out to be very, very close to 0.01, then (2) and (3) might differ in truth value, because the probabilities in these two claims will be slightly different.) But once we accepted that (1) and (2) do not differ cognitively, we had to accept that (2) and (3) don't differ cognitively. Hence, (1) and (2) do differ cognitively.
A second way to see that (1) and (2) differ is this: Plainly there could have evolved linguistic beings most of whose beliefs are true but few of whose beliefs can be translated into English. (Think of plasma-based beings most of whose beliefs are about aspects of plasma-based existence that English lacks the ability to express.)
But maybe one will say: "English and Chinese nowadays include the whole language of science, and are highly extensible, etc., so in fact anything an alien could believe is something we could say in either English or Chinese." I don't know about that. But non-propositional deflationism about truth, if true, should be true of all languages. Now, imagine a race of aliens who spend most of their life playing games and philosophizing. They live in a nutritionally rich environment such that they do not need to have as many well-developed empirical and scientific concepts as we do. Their language is highly deprived on the empirical side. For instance, they do not know anything about light or any other form of electromagnetic radiation. But they're superb players of a game very much like chess but played by smell, and they love talking about the nature of language. They have a concept of truth that functions just as ours does. Now, if non-propositional deflationism is true, it is true for these aliens, too. Thus, their equivalent to (1) had better be equivalent to something like (2) or (3), but with the name of their language in the place of "English" and "Chinese". But it is clear that the switch from "English" and "Chinese" to that alien language does in fact change things. In particular, because of the empirical impoverishment of their language, it is harder for their equivalent to (2) or (3) to be correct, because there are fewer empirical beliefs of yet other aliens that can be translated into their language than can be translated into ours.