Recent Posts

The Bertrand Russell Show

Feminist Philosophers

fragments of consciousness

Gender, Race and Philosophy: The Blog


Leiter Reports: A Philosophy Blog

Long Words Bother Me

semantics etc. highlights

Thoughts Arguments and Rants



Thursday, July 13, 2006

A Problem for Kaplan-style Semantics

John Hawthorne presented the following argument as a problem for a certain kind of Kaplan-style semantics that I find attractive (U is any utterance/sentence-in-context):

U is true
U means that p
So, p

Hawthorne's hunch is that it is valid. The problem is this. In Kaplan-style semantics, proposition truth is relative, i.e. propositions are true or false only relative to circumstances of evaluation. But utterances (and sentences-in-contexts) are true or false simpliciter. Assume that temporalism is right (i.e., sentences-in-contexts/utterances without explicit time determination express temporally neutral propositions). Let U be my utterance 3 years ago of "I am hungry", and suppose I was hungry then. Since utterances are true or false simpliciter, the first premise is true. The second premise is true; U means the temporal proposition that Brit is hungry. Yet the conclusion is false. I am not hungry.

Assume that eternalism is right (i.e. sentences-in-contexts/utterances without explicit time determination express propositions with times in them). Let U be my utterance at t in a non-actual world w of "I am hungry", and suppose I am hungry at t there. Since utterances are true or false simpliciter, the first premise is true. The second premise is true; U means the eternal proposition that Brit is hungry at t. Yet the conclusion may be false. For I may not be hungry at t in the actual world @.

What are we to do? If we reject the assumption that utterance truth is absolute, we get relativistic semantics (MacFarlane, Egan, etc.). Are we really forced to become relativists?

I think the solution to the problem is to treat object-language and meta-language occurrences of 'is true' differently. The argument is invalid if formulated in the object-language, but valid if formulated in the meta-language. I think there are good reasons to take object-language occurrences of 'S as uttered in c is true' to be true iff the proposition expressed by S-in-c is true at the circumstance of evaluation determined by the context in which 'S as uttered in c is true' is uttered (which is not c). I offer some reasons in my paper "Moral Contextualism and Moral Relativism".


Robbie said...


Does anything turn on the indexing to time? (which I guess is independently controversial).

I could have been seven foot tall. Let's suppose I would then have uttered "I'm seven foot tall". Call this merely possible utterance U. The proposition U expresses is true at that world, false at this world.

So we (in this world) have the triad: (1) U is true; (2) U expresses the proposition p (3) not p.

Seems a suspiciously strong argument to me, if it gets us into trouble even with the view that what utterances express determines truth conditions!

Anyway, it seems a bit weird to me to think that object-language `truth' claims work such that, were I seven foot tall and uttering ``I'm seven foot tall'', I'd be speaking falsely. But the proposition expressed by the sentence is false relative to the circs of evaluation determined by our actual context, so it seems to me that's what your view has to say. Any thoughts?

(p.s. in your proposed resolution, isn't it that the argument is unsound rather than invalid (it has a false premise: that U is true)? Or maybe I'm missing the trick).

Brit Brogaard said...

Hi Robbie. Thanks for your comments! Here is what I had in mind.

In Kaplan-style semantics, propositions are true at a world (or at a world, time pair). But utterances are true or false simpliciter (actually, "sentences-in-contexts", but never mind). Of course, 'truth simpliciter' (for utterances) can be defined in terms of relative proposition truth. If semantic eternalism is true, utterances express propositions with times in them (if the utterance is present-tensed, the time will be the time of speech). So, if eternalism is true, then propositions will be true or false only relative to worlds.

Now you say "So we (in this world have the triad: (1) U is true; (2) U expresses the proposition p (3) not p" (where U is a merely possible utterance that is true).

I agree that that follows from standard semantics. The problem is that the argument

U is true
U means that p
So, p

seems valid; we need a way of explaining away the apperance of invalidity. I want to say that if the argument is formulated in the object-language, then there is no way for the premises to be true and the conclusion false. For if 'U is true' is true, then it expresses a proposition that is true at the circumstance determined by the person who uttered 'U is true'. So, p is true as well.

If the argument is formulated in the meta-language, then there is a way for the premises to be true and the conclusion false (as you point out).

In a world where you are 7 feet tall and utter "I am 7 feet tall", you are not speaking falsely; you are speaking truly (there no object-language occurrences of 'is true' here).

Hope that makes sense.

matt weiner said...

Hi Brit,

I don't know if this makes a big difference to your argument, but I don't find it natural to say "U is true" if U is your utterance three years ago of "I am hungry." I'd say "U was true." In some circumstances I might even say "U was true, but isn't any more." This seems to militate in favor of temporal propositions, or maybe just temporalism about utterance-truth. (And in the counterfactual I'd be more likely to say "U would be true.")

Brit Brogaard said...

Thanks, Matt! I share your intuitions. And I am all for temporalism. But I recently said something similar to MacFarlane (who, by the way, is sympathetic to temporalism). He wasn't impressed. He replied that we also say things like 'A: houses are cheap. B: no, that's true in Missouri but not in Berkeley' or 'A: Missouri Quarterly is a pretty good place to publish. B: No, that's true for philosophers in Missouri but not for philosophers on the East Coast' (his examples were better. I just don't remember them at this precise moment). I think, however, that there are a bunch of other reasons to favor temporalism over eternalism. So I think you right.

Brit Brogaard said...

I found MacFarlane's example:

220 is a good bowling score.

That may be true for the pros, but it isn't true for ordinary folks.

Robbie said...

Hi Brit---thanks for the reply (been away for a couple of weeks, hence the delay in getting back to this).

I'm sympathetic to the project of finding a way to explain the seductiveness of the argument! I just found the story about why it's seductive less plausible in the worlds-version than in the times-version (I think I expressed that badly, though). In particular, I thought that your story would be committed to there being *a* sense (the object-language one) in which my merely possible utterance is false (if you like: there should be a reading on which the following sentence, uttered here and now, is false: "my merely possible utterance is false".) And (unlike Matt and you, perhaps) I thought there wasn't such a reading...

Anyway, one issue might be that I don't properly understand the meta/object language distinction that is required for your proposal, so perhaps I should ask about that!

Suppose at w,t I use the word "true" to talk about English utterances. Suppose, in particular, I utter "U is true" (where U is an utterance, in English, taking place at w' and t'). Is my use of "w" meta-linguistic or object-linguistic (relative to U)? It's in the same natural language (English) as U, as much as any use of semantic vocabularly can be... that might suggest it's in the object language. It's a predicate of sentences... that might suggest that it's in the metalanguage. But if either of those were decisive considerations, they'd go for *any* use of "is true", so we wouldn't be able to distinguish two separate usages. So I guess you're not using either of these as criteria for dividing tokenings of semantic vocabularly into object-linguistic and meta-linguistic.

Other than this, my only thought on how to tell when expressions are object-languagey and when they are meta-languagey, is project-relative. E.g. when I'm writing formal (Kaplanian) semantics for a language L, I can tell you what words are taken to be in L, and which I'm using to talk about L. And I can change this by changing my choice of L. But I don't really see how this would help?

Here's a final query: if English turns out to be semantically closed, then (in at least one sense!) the metalanguage would be part of the object language. Can we still distinguish two "is true"'s as you want?

Brit Brogaard said...

Hi Robbie,
I agree that there is a real difficulty determining whether we are speaking in the meta-language or the object-language. But it seems clear to me that most ordinary occurrences of 'is true' in English are occurrences in the object-language. Or maybe we cannot make sense of this distinction, in which case it may just be that 'is true' plays two different roles in the language.

I think we can still distinguish two senses of 'is true'. It certainly seems to me that there is a tensed and a tenseless form of 'is true'. I took the tenseless form to be meta-linguistic.