Though I've thought a lot about inscrutability and indeterminacy (well, I wrote my PhD thesis on it) I've always run a bit scared from the literature on Kripkenstein. Partly this is because the literature is so huge and sometimes intimidatingly complex. Partly it's because I was a bit dissatisfied/puzzled with some of the foundational assumptions that seemed to be around, and was setting it aside until I had time to think things through.
Anyway, I'm now thinking about making a start on thinking about the issue. So this post is something in the way of a plea for information: I'm going to set out how I understand the puzzle involved, and invite people to disabuse me of my ignorance, recommend good readings or where these ideas have already been worked out.
To begin with, let's draw a rough divide between
three types of facts:
- Paradigmatically naturalistic facts (patterns of assent and dissent, causal relationships, dispositions, etc).
- Meaning-facts. (Of the form: “+” means addition, “67+56=123” is true, "Dobbin" refers to Dobbin.)
- Linguistic norms. (Of the form: One should utter “67+56=123” in such-and-such circs).
Kripkenstein’s strategy is to ask us to show how facts of (A) can constitute facts of kind (B) and (C). (An oddity here: the debate seems to have centred on a “dispositionalist” account of the move from A to B. But that’s hardly a popular option in the literature on naturalistic treatments of content, where variants of radical interpretation (Lewis, Davidson), of causal (Fodor, Field) and teleological (Millikan) theories are far more prominent. Boghossian in his state of the art article in Mind seems to say that these can all be seen as variants of the dispositionalist idea. But I don't quite understand how. Anyway...)
One of the major strategies in Kripkenstein is to raise doubts about whether this or that constitutive story can really found facts of kind (C). Notice that if one assumes that (B) and (C) are a joint package, then this will simultaneously throw into doubt naturalistic stories about (B).
In what sense might they be a joint package? Well, maybe some sort of constraint like the following is proposed: unless putative meaning-facts make immediately intelligible the corresponding linguistic norms, then they don’t deserve the name “meaning facts” at all.
To see an application, suppose that some of Kripke’s “technical” objections to the dispositionalist position were patched (e.g. suppose one could non-circularly identify a disposition of mine to return the intuitively correct verdicts to each and every arithmetical sum). Still, then, there’s the “normative” objection: why are those the verdicts the ones one should return in those circumstances? And (right or wrongly) the Kripkenstein challenge is that this normative explanation is missing. So (according to the Kripkean) these ain’t the meaning-facts at all.
There's one purely terminological issue I'd like to settle at this point. I think we shouldn’t just build it into the definition of meaning-facts that they correspond to linguistic norms in this way. After all, there’s lot of other theoretical roles for meaning other than supporting linguistic norms (e.g. a predicative/explanatory role wrt understanding, for example). I propose to proceed as follows. Firstly, let’s speak of “semantic” or “meaning” facts in general (picked out if you like via other aspects of the theoretical role of meaning). Secondly, we'll look for arguments for or against the substantive claim that part of the job of a theory of meaning is to subserve, or make immediately intelligible, or whatever, facts like (C).
Onto details. The Kripkenstein paradox looks like it proceeds on the following assumptions. First, three principles are taken as target (we can think of them as part of a "folk theory" of meaning)
- the meaning-facts to be exactly as we take them to be: i.e. arithmetical truths are determinate “to infinity”; and
- the corresponding linguistic norms are determinate “to infinity” as well; and
- (1) and (2) are connected in the obvious way: if S is true, then in appropriate circumstances, we should utter S.
The “straight solutions” seem to tacitly assume that our story should take the following form. First, give some constitutive story about what fixes facts of kind (B). Then (supposing there’s no obvious counterexamples, i.e. that the technical challenge is met). Then the Kripkensteinian looks to see whether this “really gives you meaning”, in the sense that we’ve also got a story underpinning (C). Given our early discussion, the Kripkensteinian challenge needs to be rephrased somewhat. Put the challenge as follows. First, the straight solution gives a theory of semantic facts, which is evaluated for success on grounds that set aside putative connections to facts of kind (C). Next, we ask the question: can we give an adequate account of facts of kind (C), on the basis of what we have so far? The Kripkensteinian suggests not.
The “sceptical solution” starts in the other direction. It takes as groundwork facts of kind (A) and (C) (perhaps explaining facts of kind (C) on the basis of those of kind (A)?) and then uses this in constructing an account of (something like) (B). One Kripkensteinian thought here is to base some kind of vindication of (B)-talk on the (C)-style claim that one ought to utter sentences involving semantic vocabulary such as " '+' means addition".
The basic idea one should be having at this point is more general however. Rather than start by assuming that facts like (B) are prior in the order of explanation to facts like (C), why not consider other explanatory orderings? Two spring to mind: linguistic normativity and meaning-facts are explained independently; or linguistic normativity is prior in the order of explanation to meaning-facts.
One natural thought in the latter direction is to run a “radical interpretation” line. The first element of a radical interpretation proposal is identify a “target set” of T-sentences, which the meaning-fixing T-theory for a language is (cp) constrained to generate. Davidson suggests we pick the T-sentences by looking at what sentences people de facto hold true in certain circumstances. But, granted (C)-facts, when identifying the target set of T-sentences one might instead appeal to what person’s ought to utter in such and such circs.
There’s no obvious reason why such normative facts need be construed as themselves “semantic” in nature, nor any obvious reason why the naturalistically minded shouldn’t look for reductions of this kind of normativity (e.g. it might be a normativity on a par with that involved with weak hypothetical imperatives, e.g. in the claim that I should eat this food, in order to stay alive, which I take to be pretty unscary.). So there's no need to give up on reductionist project in doing things this way. Nor is it only radical interpretation that could build in this sort of appeal to (C)-type facts in the account of meaning.
One nice thing about building normativity into the subvening base for semantic facts in this way is that we make it obvious that we’ll get something like (a perhaps restricted and hedged) form of (iii). Running accounts of (B) and (C) separately would make the convergence of meaning-facts and linguistic norms seem like a coincidence, if it in fact holds in any form at all.)
Is there anything particularly sceptical about the setup, so construed? Not in the sense in which Kripke’s own suggestion is. Two things about the Kripke proposal (as I suggested we read it): it’s clear that we’ve got some kind of projectionist/quasi-realist treatment of the semantic going on (it’s only the acceptability of semantic claims that’s being vindicated, not "semantic facts" as most naturalistic theories of meaning would conceive them). Further, the sort of norms to which we can reasonably appeal will be grounded in practices of praise and blame in a linguistic community to which we belong, and given the sheer absence of people doing very-long sums, there just won't be a practice of praise and blaming people for uttering "x+y=z" for sufficiently large choices of x, y and z. The linguistic norms we can ground in this way might be much more restricted than one might at first think: maybe only finitely many sentences S are such that something of the following form holds: we should assert S in circs c. Though there might be norms governing apparently infinitary claims, there is no reason to suppose in this setup that there are infinitely many type-(C) facts. That'll mean that (2) and (3) are dropped.
In sum, Kripke's proposal is sceptical in two senses: it is projectionist, rather than realist, about meaning-facts. And it drops what one might take to be a central plank of folk-theory of meaning, (2) and (3) above.
On the other hand, the modified radical interpretation or causal theory proposal I’ve been sketching can perfectly well be a realist about meaning-facts, having them “stretch out to infinity” as much as you like (I’d be looking to combine the radical interpretation setting sketched earlier with something like Lewis’s eligibility constraints on correct interpretation, to secure semantic determinacy). So it's not "sceptical" in the first sense in which Kripke's theory is: it doesn't involve any dodgy projectivism about meaning-facts. But it is a "sceptical solution" in the other sense, since it gives up the claims that linguistic norms "stretch out" to infinity, and that truth-conditions of sentences are invariably paired with some such norm.
[Thanks (I think) are owed to Gerald Lang for the title to this post. A quick google search reveals that others have had the same idea...]
12 comments:
Would you get linguistic facts out of radical interpretation? I thought Davidson acknowledged that some indeterminacy of reference would remain in his system even if not as much as in Quine's.
Hi Ken,
I guess it depends on what you mean by "linguistic facts". If you mean "complete determinacy in what each word picks out/applies to" then you're unlikely to get any theory that gives you linguistic facts: some small bits of indeterminacy seem almost inevitable whatever theory of meaning you adopt (e.g. about which of the cat-like fusions in the vicinity of Tibbles is picked out by "Tibbles").
So I think the relevant question is not: whether you get linguistic/semantic facts; but rather: what sort of determinacy of linguistic/semantic facts you get. And there's going to be a sliding scale from utter determinacy (Williamson?) to almost complete indeterminacy (Kripkenstein's sceptic).
Davidson and Quine both seem to believe that there may be no determinate facts of the matter at all about what subsentential items refer to, though probably a great deal more determinacy than Kripkenstein would have over the truth conditions of whole sentences. But I don't think that widespread indeterminacy of meaning is an inevitable consequence of radical interpretation: Lewis's version, for example, attempts to avoid it.
So I think the relevant question is not: whether you get linguistic/semantic facts; but rather: what sort of determinacy of linguistic/semantic facts you get. And there's going to be a sliding scale from utter determinacy (Williamson?) to almost complete indeterminacy (Kripkenstein's sceptic).
That seems right to me.
I confess that I'm even more ignorant about this stuff than you claim to be, but one thought occurred to me: It seems that there is a discrepancy between facts about what T-sentences people de facto hold and facts about what T-sentences people ought to hold, namely that the former are observable (given that, as you say in (A) patterns of assent and dissent are observable) and the latter are not. Is this a problem?
I suppose it depends on what we read into the demand that we must *demonstrate*, which fact it is that constitutes our meaning plus rather than quus (see for instance Kripke, pp. 9-11). But at any rate, I'm not sure how much is left of the radical interpretation scenario once you allow for unobservable facts.
I guess if you take a naturalist-reductionist line on the normative facts, as you suggest is available, then you might point out that the facts we are reducing to are observable. But this would only answer the above worry, if there is one, to the extent that we know beforehand *which* are the right facts to reduce to (presumably there are different candidates), and it seems we cannot know this without knowing what facts we are reducing.
Well, it's not so clear that what T-sentences people de facto hold is "observable" in any direct sense. What Davidson really works with is facts about what is held-true in what scenario. But you need to do considerable work from distribution of patterns of holding-true to a pattern of T-sentences.
Saying that we can arrive at facts about what T-sentences people ought to hold is to write a pretty big cheque. Not clear it can be cashed, I admit. I was thinking of something like this: suppose that as well something in the vicinity of patterns of assent and dissent, we also allow the radical interpreter to take as basic patterns of praise and blame directed at particular events. (Potential objection: this is overly intentional).
Then the idea is that this radical interpreter should concentrate only on the regularly praised assenting and dissentings when constructing the target T-sentences which the semantic theory is then constrained to generate.
This *really does* need more working through. But I don't feel too pessimistic.
I'm not sure, though, I quite get the worry you express in the last paragraph (about which facts we're reducing too). Could you expand a bit on your idea here?
I was assuming that the worry of observability was a real one. Maybe it isn't, given your remarks about praise and blame. However, if there *is* a worry about observability, then it might seem that one way of answering it is to say, look, I can observe these naturalistic facts and since the normative facts reduce to them, that's good enough.
But the problem is that this line of response presupposes that you know what the normative facts are - and as long as you're only allowed observable facts as evidence (and still assuming that the normative facts aren't observable) these won't be available.
In addition, I was trying to point out that there will presumably also be many different candidate naturalistic fact-types that you might reduce the normative ones to. But choosing among them likewise requires knowing knowing what the normative facts are.
Hope that helps.
Ah. I think I begin to see.
Here's the way I was thinking of the enterprise. (1) Pick a range of "problem" facts. Say, representation ones, or normative ones. Suppose these are in good order. (2) Propose a reduction of these two facts of a less problematic sort: as it may be, naturalistic ones.
This way of thinking of things presupposes that we have as our starting point a grip on what the facts of kind (1) are. And the ensuing theory is a metaphysical one.
I know that one way of reading the Kripkenstein stuff is as a epistemological puzzle: how do we ever know what we mean? How could we know it in the distinctive way we seem to? I don't mean to diss the epistemology: ultimately we need an integrated story that not only explains how there can be meaning-facts, but also how we know them. But the way I was taking the puzzle is really as a metaphysical one.
Ok. If this is the project, then I think it is crucial to distinguish it from the epistemological one. Exegetically, my preferred reading of Kripke is one which puts the epistemological puzzle in the foreground.
But I agree that there are two different puzzles - also in Kripke.
Fair enough. I think the metaphysical one is interesting too, though, and the normativity stuff in particular is worth thinking about.
Oh, yeah, I agree - I certainly didn't mean to imply that I don't find the metaphysical problems interesting. Anyway, I'll shut up about this now.
uh "patterns of assent and dissent" are not "paradigmatically naturalistic facts" (unless you are Quinefused)
Could you please put the DATE on the comments added to your blog? The TIME alone is useless! (There is an option: settings/comments/comments timestamp format).
Post a Comment