The Sapir-Whorf Hypothesis

I've just been reading this old post from the Language Log on the Sapir-Whorf Hypothesis, and it reminded me of a question I wanted to think through a while back and never got around to. The question is this: does Burge-style anti-individualism underwrite a principled philosophical defense of the Sapir-Whorf Hypothesis? Here is why you might think that it does.

According to Burge, mental states have externalist individuation conditions, not just with respect to the local environment, but also with respect to the linguistic community in which one is situated. We individuate beliefs (and other propositional attitudes) via their content, and we individuate content in accordance with constraints on the publicity of meaning. Suppose, for example, that there are two concepts associated rhuemetoid ailments. Arthritis is the concept of a rheumetoid ailment restricted to the joints; tharthritis, by contrast, is the concept of a rhuemetoid ailment not so restricted. The English word "arthritis" expresses the concept of arthritis. Now consider an individual, Oscar, who tells his doctor that he believes that his arthritis is spreading to his thigh. What Oscar says, namely, "I believe that my arthritis is spreading to my thigh," is false since by fiat arthritis can't occur in the thigh.

But what about Oscar's belief itself? There are two possibilities. (1) Oscar has the false belief that his arthritis is spreading to his thigh; (2) Oscar has the (possibly) true belief that his tharthritis is spreading to his thigh. Burge argues convincingly that the correct answer is (1). Oscar's concept is the concept of arthritis, and he (as it were) mistakes that concept for the concept of tharthritis.

Now consider a situation in which everything in Oscar's linguistic history is held fixed but, entirely outside the realm of his experience, the medical community has fixed "arthritis" to mean tharthritis. Nothing internal to Oscar's psychology has changed, only external facts about the language community. In this case, the proposition Oscar expresses when he utters "I believe that my arthritis is spreading to my thigh" is the (possibly) true proposition that Oscar believes his tharthritis is spreading to his thigh. Moreover, his belief is no longer a belief about arthritis, but about tharthritis! But clearly the only thing that has changed between the two cases is which linguistic community Oscar is embedded in. Thus, our mental states (i.e., what thoughts we have) are individuated in part by our linguistic community.

As the case has been described, we have held everything fixed except the external linguistic environment. This allows Burge to locate the shift in belief in the shift in the linguistic environment. But once the conclusion has been established, it would seem to hold more or less across the board: under normal circumstances the contents of our thought is dependent upon the linguistic community in which we happen to be located.

Now here is the million dollar question: do the semantics of at least some linguistic communities differ significantly enough from one another that members of the two communities would typically exhibit Burge-like divergences in their thoughts? If the answer here is "yes", then we seem to have the makings of a relatively robust defense of the Sapir-Whorf Hypothesis (or something like it). Put generally (and putting to one side further anti-realist considerations), the Sapir-Whorf Hypothesis is the thesis that culture, via its influence on language, substantially affects our thoughts about the world.

The answer to this question is, at least in part, an empirical matter. But I do think that there are some general considerations in philosophical semantics that might well lend support to the S-P Hypothesis. In particular,there are good philosophical reasons for thinking that many (perhaps most) words have various types of categorial information associated with them and that this categorial information might vary culture-wise. One such bit of information is whether or not the meaning of the word is functional or (micro)structural. For example, most philosophers of language are now convinced that the concept expressed by the word "water" is microstructural, in the sense that nothing could be water (i.e., fall under that concept) unless it is composed of H2O molecules. But it is difficult to believe that this is the only reasonble way of handling the semantics of a word of the "water" category. [By this, I just mean a word which is at least roughly translatable as by the English word "water".] In particular, there seems to be no principled barrier to having a "water" word which is categorially functional/descriptivist rather than microstructural. Call the resulting concept macro-water (Korman). Now, if the anti-individualist thesis is correct, then whether your thoughts are water thoughts or macro-water thoughts will depend (in part) on facts about your language community. And it isn't unreasonable to think that different cultures might well differ quite generally in their propensity to take one or the other (or, if you are fond of two-dimensionalism, both) of these tracks. The former we might broadly characterize as atomistic cultures; the latter, as functionalistic ones. [Brian Weatherson had a link to a paper that argued something like this, but alas I couldn't find the reference.]

Does this really give us anything like a substantial S-W Hypothesis? Well, consider the concept expressed by a functionally oriented community who conceives of time as cyclical or nonlinear or whatever!

Comments

[Kerim Friedman]

It is important to note that the version of the Sapir-Whorf Hypothesis you propose here is much closer to the one actually proposed by Whorf than the kind of arguments Mark Abley seems to be making.

Whorf drew from his experience working for an insurance company reviewing insurance claims in developing his theory of how linguistic categories influence behavior. For instance, he argued that someone smoking a cigarette next empty oil canisters was misled by the category "empty" into believing that they did not pose a danger (even though empty canisters are actually more dangerous than full ones, which are sealed).

Unfortunately, a lot of people, including Steven Pinker, have found it useful to portray Whorf as making much stronger claims than he was actually making.

[Danny Korman]

Marc –

I think you’re exactly right about the possibility of categorially descriptive “water”-words. And I suspect that many linguistic communities *do* have words roughly translatable into English as ‘water’ that express something like the concept of being a wet, transparent, drinkable ... stuff.

That said, if there really are “reference magnets” – that is, properties (or concepts) that are especially eligible to be the referents (or meanings) of the expressions of a language – then the Burgean defense of linguistic relativity might be in trouble. So suppose that our externalist water-concept really is more eligible than the corresponding internalist concept. In that case, while it’s *possible* to introduce a word that expresses the internalist concept, this requires very deliberate stipulatory work in order to prevent the more eligible concept from being meant. On the other hand, when a “water”-word is introduced less cautiously – for instance, by simply pointing at water and saying ‘water’ (‘agua’, etc.) – the most eligible concept in the vicinity wins out.

So, although there are in all probability plenty of linguistic communities that wouldn’t have the Kripkean intuition when you ask them whether their “water”-word would refer to the stuff on XYZ-world, this probably wouldn’t move a proponent of reference magnets (“reference suckers”?) To these guys, intuitions are a pretty good guide to what concept we have, but intuitions must ultimately be weighed against the naturalness or eligibility of candidate meanings of our words in order to determine what's really meant. (That’s why Brian Weatherson thinks that knowledge is justified true belief – he argues for that in his Counterexamples paper.)

So, while it’s *possible* for there to be twin communities that mean twater rather than water by their “water”-word (and likewise for tharthritis), the existence of reference magnets would make it pretty unlikely that there actually are any such communities.

DZ

[marc]

DZ -

There is more room in here than your proposal suggests. Specifically, Lewis articulates his theory of eligibility against a background descriptivist semantics (and associated descriptivist referential intentions). Because of this, it is possible to put to one side referential intentions as part of the eligibility equation--they are assumed to be descriptivist.

However, when we consider the concept of eligibility without such background commitments, it seems to me that it is best construed as a match between referential intentions and the world. Thus, if x has descriptivist intentions, then the most eligible concept will be a "macro-concept"; but if one has nondescriptivist intentions, then the most eligible concept may be a microstructural or de re concept. And so on. [Maybe Lewis and/or Putnam would maintain that the only relevant referential intentions are our intentions "to refer in such a way as to come out right." I am presupposing that we have more fine-grained referential intentions than that--though the story I want to tell here is complex.]

Two questions: (1) Is only one sort of referential story possible for all languages? (2) If not, do natural languages in fact differ with respect to their referential semantics (and associated referential intentions), perhaps in part because of cultural differences? If the answers to (1) and (2) are "no" and "yes" respectively, then the point in the post would seem to hold. [If the answer to both is "no", then the S-W Hyp stands as an unrealized possibility.]

When you say "when a “water”-word is introduced less cautiously – for instance, by simply pointing at water and saying ‘water’ (‘agua’, etc.) – the most eligible concept in the vicinity wins out," I am not sure what your conception of eligibility amounts to. Are some concepts, as it were, intrinsically more eligible to be referents /simpliciter/ (regardless of one's referential intentions)?

[Danny Korman]

I guess there are various grades of eligibility theories. On one extreme, we have highly fine-grained referential intentions which completely constrain the referents (meanings) of our expressions, and according to which the eligibility of a property (concept) is not intrinsic, but rather varies with our referential intentions. Eligibility only comes into play at the point that our intentions underdetermine the referent.

On the other extreme, certain properties are intrinsically more eligible than others, and manage to become the referents of our terms *regardless* of any of our referential intentions. This of course is too extreme, since on this picture *every* expression ends up referring to the single most eligibile property (i.e., The Good). Less extreme versions maintain that certain properties are intrinsically more eligible (e.g., because the members of their extensions are objectively more similar to one another), but admit that our referential intentions play *some* role in determining what our expression denote. But the intrinsic ineligibility of what we’re intending to refer to might sometimes result in our ending up referring to something that explicitly violates certain of our referential intentions, but which it more eligible. (Cf, again, Weatherson on knowledge and JTB.)

I guess that whether or not eligibility theories pose a problem for the Burgean defense of S-W wholly depends on (a) how fine-grained our referential intentions really are, and (b) the extent to which eligibility is able to contravene referential intentions. I have my doubts about “intention-contravening eligibility”; but it does seem like that’s how some of these reference suckers are thinking about eligibility. And if there is such a thing, then there’s a chance that – despite the mere possibility of languages with different referential semantics – no natural languages differ with respect to their referential semantics (of, say, their ‘water’-terms).

Incidentally, I had in mind intrinsic eligibility when I gave the ‘water’ (‘agua’) baptism case; but it may be that a certain kind of extrinsic eligibility might work just as well. Perhaps, because of general facts about the way our brains are wired, the relevant externalist concept is more eligible to be possessed by us than is the associated internalist concept. Nevertheless, I might, on some dispositional level, intend to express the internalist concept, but (because of the contravening-strength of more eligibile concepts in the vicinity) I might still end up expressing the externalist concept.

[Clayton]

Marc and Danny,

First, Marc, congratulations on the new place. Second, Danny when are you going to open shop?

I wanted to know more about this contrast between water and macrowater and the contrast between functional vs. mirco kinds. Are we supposed to think that this categorial distinction is exhaustive? Are we supposed to think that matching the category with the term is related to conceptual and linguistic competency?

If we think of the distinction as exhaustive, then it seems that kinds at the most basic level should be conceived of as functional--but this doesn't seem to be something that fits our actual practice. In fact, those who are competent enough to be ascribed attitudes with the term don't need to know whether 'atom', for example, is a functional kind or not, what its function is, etc., as evidenced by the fact that prior to Dalton, 'atom' was treated as that which played a certain role but subsequently we were able to say before Dalton scientists wrongly thought that atoms were the smallest bit of indestructible matter from which larger bits are made.

[viagra online]

It is an interesting point of view, at the end it is uncertain.