Wolfgang Schwarz

Thanks for that. It's interesting that "S" is true and S aren't even intersubstitutable in the scope of temporal operators ("Madagaskar is an island" used to be false, but is now true even though Madagaskar always was an island.)

I wasn't quite sure how the discussion supported the conclusion that 'it is true that S' is not intersubstitutable with 'S' though. I think there's a legitimate reading of 'it's true that' in which:

(2) If 'pig' meant bird, then it would be the case that pigs can fly.

(2') If 'pig' meant bird, then it would be true that pigs can fly.

are equivalent and both false (even though (1) is true.) This sense of 'it's true that' is also immune from the paradoxes, since it can always be consistently interpreted with the trivial identity truth table.

Dear Wo,

I think Andrew Bacon is right: one has to clearly distinguish between the first-level operator "It is true that (sentence used)" and the metalinguistic predicate "(sentence mentioned) is true".
Also, I do not quite see that your general lesson follows from your observation. As for the metalinguistic predicate, I want to hold that
(*) Snow is white iff "Snow is white" is true
is NOT analytic, not in the required sense, that is. It is apriori for a speaker of English, to be sure. Perhaps it is epistemically analytic, because its truth is knowable apriori just on the basis of linguistic competence. But it is not "semantically" or "metaphysically" analytic.
As for the introduction and elimination rules, it seems to me that we have to take into account introductions INTO intensional contents.
Thus, for "now" we have:
(**) p, therefore: It was/will be the case that now p.
Doesn't that count as an introduction rule that distinguishes between "p" and "now p"?
True, the operator "now" doesn't have largest scope in the sentence introduced. But is this essential of meaning-constituting introduction rules?
Likewise:
(***) p, therefore: It is necessary that actually p.
(for the relevant use of "actually", the one that in arbitrary contexts takes us home to the actual world).

@Andrew: right, the sentence operator 'it is true that' looks transparent. I meant to be talking about the exchangeability of 'S is true' and S. I mentioned 'it is true that' later because (unlike for the sentence predicate) the corresponding T-schema is unproblematic here, so it is natural to think that it somehow gives the whole meaning of 'it is true that'. But this can't be right, because 'now' and 'actually' also satisfy the schema, but have a rather different meaning. Sorry for the confusion.

Good point about 'it used to be that'.

@Ralf: I'd say that (*) ('Snow is white iff "Snow is white" is true') is metaphysically analytic, but maybe we mean slightly different things by 'metaphysically analytic'. For any utterance of (*), the semantics of English entails that the utterance is true. So (*) is true solely in virtue of its meaning.

I don't have any clear opinions about introduction and elimination rules. You're right that one might extend the notion to rules like (**) and (***). It's a good question whether any finite set of rules like these could settle the meaning of 'actually' and 'now'. (**) and (***) obviously don't do that, although they do distinguish the two from 'it is true that'.

Hey wo,

I think your inference from substitution intuitions about (1) and(2) to the claim that S and 'S is true' don't have the same meaning is a bit quick. Of course, English speakers share the intuition that (1) and (2) differ in truth-value. But this in itself (even in conjunction with compositionality) does not entail your claim about 'S is true' and its subject term: Speaker intuitions about (1) and (2) may be determined, not by the compositional meanings of (1) and (2) but rather by information they pragmatically communicate in using (1) and (2) in contexts;and it may be that these pieces of information differ in truth-value although (1) and (2) are synonymous.
To put this point differently: It may be that counterfactuals are highly context-dependent, and so (1) and (2) have different meanings in different contexts. Speaker intuitions about (1) and (2) are compatible with (1)and (2) being synonymous in every context and with being determined by the fallacy of evaluating (1) and (2) with respect to different contexts. Again, speakers' substitution intuitions alone (in conjunction with compositionality) do not entail your claim.
There is also a lesson here (coming from Kripke): Substitution phenomena are not truisms we can use freely in reasoning about semantic issues (a lot of Fregeans still won't agree with me in this regard). Rather, more often than not, their explanation is what is at issue, be it semantic or pragmatic. However, this task in itself is so difficult and controversial that we should not appeal to substitution phenomena alone (together with our favourite explanations of them)in deciding intricate semantic questions.