Saturday, December 03, 2005

Language-Dependence

Here's a common claim schema:

If ... is dependent on our language, then were our language to be different in relevant respects then so would ... be.

E.g.:

(1) If vagueness depends on our language, then were all semantic vagueness to be eliminated then there would be no vagueness.

A worry about this is that it seems a semantic theorist of vagueness (someone who thinks vagueness is entirely due to semantic features of our language, and has nothing to do with how the world is independently of our linguistic representations of it) might sensibly reject (1), and instead assert:

(2) Although vagueness depends on our language, if all semantic vagueness were eliminated, this apple would still be borderline red.

For the consequent of the counterfactual is couched in the very language which is supposed to be suitable for creating vagueness. Asserting (2) therefore does nothing to undermine the thought that it is the semantic features of our actual word 'red' language which are responsible for the borderlineness, in other worlds, of the apple's redness.

What do people think?

5 comments:

Richard Zach said...

This strikes me as parallel to the dependence of necessary facts (in part) on language. E.g., compare "If 'bachelor' had meant something other than it does, some bachelors might married" and "Even if 'bachelor' meant something other than it does, all bachelors would still be unmarried." I think it's pretty widely accepted that the right way to think about this is the way that endorses the latter.

Daniel Elstein said...

I agree that (2) is correct. The way I've seen the underlying principle stated is that when evaluating counterfactuals you take all terms *used* in the counterfactual to have their actual meanings. I suppose there's a question about whether any counterfactual can express the semantic theory of vagueness. Perhaps:

(3) If all semantic vagueness were eliminated, no predicates would have borderline cases.

The reason I think that (2) is false and (3) is true is that (2) only seems true if "red" is interpreted non-actually, whereas the truth of (3) depends on what predicates there are in the closest worlds where semantic vagueness is eliminated. But perhaps (3) doesn't really say that vagueness originates in language not in the world. So you want something more like:

(4) If all semantic vagueness were eliminated, there would be no need for predicates with borderline cases.

Elizabeth said...

Hi Carrie,

As I've mentioned to you, I think the best way to go is a more general:

(5)a case of vagueness would be ontic iff were all the representational content precisified there would still be no (determinate) polar truth value.

Doing it this way, you're giving a general schema, so you're not using (or mentioning) any specific (and worryingly vague) terminology. What you do use are notions like polar truth values and determinacy. Some people, admittedly, think determinacy is subject to semantic vagueness, but I doubt it's sort of vagueness that would be problematic in this case, since it's generally inherited from the vague predicates it's applied to, and with the counterfactual on offer you're not applying it to any specific predicates.

Carrie Jenkins said...

Richard,

That's exactly the sort of parallel I had in mind, and I agree that in this area lots of people have noticed that the obvious counterfactual is no good as a characterization of language-dependence.

Daniel and Elizabeth,

Thanks for those suggestions. They seem (to me, anyway) quite close to something I am already working on, namely:
(6) If all semantic vagueness were eliminated, it would still be vague whether or not the (determinate) truth-condition for ‘There are red patches’ obtained.
Prima facie it looks as though, if you accepted (6), that would be good enough reason to reject (3) or accept the RHS of the biconditional in (5).
But I also suspect that a semantic theorist can accept (6), and blame the vagueness in question on semantic features of our word ‘obtains’. How can ‘obtains’ be vague in the requisite way? Well, consider the kind of Sorites series of (determinate) states of affairs which the ontic theorist will use to characterise the vagueness of ‘There are red patches’ in situations where the obtaining of this sentence’s (determinate) truth-condition – call it t – is an ontically vague matter. Then say that ‘obtains’ in our actual language is vague in that the truth-condition of our actual-language sentence ‘Truth-condition t obtains’ is not fixed as between the various determinate states of affairs in the ontic theorist’s Sorites series. This vagueness in truth-condition is not due to vagueness as to which truth-condition t is, since that has been specified precisely, but it can be due to vagueness semantic features of our word ‘obtains’.
On Daniel's (4), there are lots of reasons why we might need a bit of language which have nothing to do with what's going on out there in the world, so we'd have to specify the kind of need that's relevant. Can this be done non-circularly?

Daniel Elstein said...

Hi Carrie,
I may be misunderstanding, but I think that your suggestion for reconciling the semantic theorist with (6) has some odd consequences. If 'obtains' is vague, then the claim 'The truth condition for "Tom has 157 hairs on his scalp" obtains' is vague (because it is analogous to the counterfactual scenario you give with a precisified 'red'). But presumably there is some kind of disquotation schema which allows that that claim is vague iff the claim 'Tom has 157 hairs on his scalp' is also vague, though none of the words used in it is. So the semantic theorist is left without any words to attribute the relevant semantic features to. But perhaps your idea was that this kind of semantic theorist will deny the disquotation schema that I'm relying on; if so, that looks like a problematic oddity in the view.

One reason I wanted to move from (3) to (4) is that it looks as though psychological theorists might well accept (3). So I agree that there could be a non-worldy need for vague predicates: that's something that psychological theories hold. Unless you were taking psychological theories to be a subclass of semantic theories, in which case I'd agree that (4) isn't precise enough. Maybe 'non-psychological need'?

On the other hand, I'm worried that even those who believe in worldly vagueness could accept (3), because (3) seems to be a claim about language, not about vagueness. And maybe the same goes for Elizabeth's (5). Perhaps I have a funny idea of what worldly vagueness would be - in my fevered imagining it's the view that even if you could precisify your language, that would just mean that there were some facts about the world that you'd be missing out: the vague facts. If you get rid of the vague predicate 'red' that just means that you fail to have a way of talking about the vague property of redness. But you would have restricted your language to such an extent that all your claims had determinate truth values, so (5) would be false. Which is another count in favour of (4) as a criterion. But I have the sinking feeling that I'm thinking about a view which no sane person has ever held.