Blog Coordinator

Knobe's X-Phi Page

X-Phi Grad Programs

« The next x-phi video (freedom) | Main | Newly Revamped Journal of Theoretical and Philosophical Psychology »

04/28/2009

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

David Chalmers

Both the "use case" and the "metalinguistic case" are metalinguistic, testing subject's intuitions about others' uses of names. For a real use case, you'd need to set up a situation where the experimental subject uses a name such as 'Godel', then ask them questions such as: did Godel prove incompleteness?

Colin Caret

Aren't truth-values evaluated metalinguistically?

Justin Sytsma

Interesting stuff (I am especially intrigued by the new data using the clarified question that Jonathan and I had used).

I am somewhat confused, however, by both the object and the response. Apparently I see the dialectic somewhat differently than Martí and MOB. (Of course, as Edouard played a prominent role in the dialectic, I feel a little strange challenging MOB's reading of MMNS's argument. Nonetheless, I'm going to go ahead and challenge it.) Basically, I do not think that MMNS were arguing against the causal-historical account so much as against (part of) the case that Kripke made for it. Specifically, Kripke gave the Gödel example in support of the causal-historical account, expecting it to elicit the clear intuition that when John uses the name "Gödel" he is talking about Gödel and not Schmidt. If Martí is correct, the intuitions Kripke was expecting to elicit are metalinguistic intuitions. That strikes me as reasonable. Now, if she is also correct that (as MOB put it) "only linguistic intuitions provide evidence for determining how reference is fixed," then she has an objection against *Kripke*; but, I don't see why this is an objection against *MMNS*. MMNS challenged Kripke's assumption that his (metalinguistic) intuition about the Gödel case is representative (that is, they challenged Kripke's assumption of what Jonathan and I termed the Uniformity Conjecture); Martí challenges Kripke's assumption that his (metalinguistic) intuition is relevant. Perhaps Marti's objection trumps MMNS's objection; nonetheless, they don't necessarily seem to be in conflict.

This leaves me confused as to why MOB go along with Martí, granting that the intuitions MMNS needed to test were linguistic intuitions and arguing that the metalinguistic intuitions are in line with the linguistic intuitions. This is puzzling in part because MOB seem to recognize that Martí in fact raises an objection against Kripke (and others): "If linguistic and metalinguistic intuitions really are incongruent and if only the former are evidentially relevant for identifying the correct theory of reference, then a substantial reform of the usual practices in the philosophy of language is required, since actual or possible cases written so as to elicit metalinguistic intuitions about reference are extremely common in this field."

Justin Sytsma

[Edit: I shouldn't have put the name "John" into Kripke's mouth; Kripke actually discusses *our ordinary man* using the name "Gödel." Strangely, he goes on to phrase the elicited intuition in terms of who *we* are talking about when we use the name. Of course, as we (being privy to the sordid tale of Gödel and Schmidt) have more information than the ordinary man, we must associate a different description with the name "Gödel"; as such, I take it that he means the *we* relative to our putting ourselves in the ordinary man's shoes.]

Jonathan Ichikawa

Here are three worries:

Worry 1:

Dave's right -- you still haven't done any tests about how subjects use names. And, as has been pointed out a couple of times (I apologize to whoever was first and/or official on this point; I forget), if the subjects understood the probe at all, they can't have been the relevant sort of descriptivists. Kripke's argument requires only the coherence of the suggestion that, e.g., Tsu Ch’ung Chih did not really make this discovery.

Worry 2:

I don't think I agree that "Martí's criticism assumes that people's metalinguistic intuitions about the reference of, say, proper names tend to be incongruent with the way they use these names". I don't this is an assumption; I think it's the most natural interpretation of your data. If some people say that 'Gödel' refers to Schmidt because he's the one who satisfies the description, even while accepting the stipulation that Gödel isn't the one who satisfies the description, this shows that their intuitions about reference are incongruent with their use. So that you didn't find that incongruence repeated in another similar example doesn't refute her.

Worry 3:

Your prompt doesn't tell us whether Tsu Ch’ung Chih was a great astronomer or not. We know that he's a fraud, but it's an open question whether he's also a great astronomer, for at least two reasons. First, it may be that he was a great astronomer, independently of this particular alleged discovery. Second, it may be that it is, or is thought to be by some subjects, sufficient for being a great astronomer that one be an astronomer who is very famous and influential. So I wish that the probe had asked about easier-to-evaluate sentences like "Tsu Ch’ung Chih became famous" or "Tsu Ch’ung Chih was a dishonest person".

Jonathan Ichikawa

(A small correction, lest I fall into the mistake I've just prescribed against. In Worry 1: "...if the subject understood the probe at all, the relevant sort of descriptivism cannot be true of the names in their mouths.")

Edouard Machery

Plenty of food for thoughts. Here is a first set of replies (more to come):

Dave, Jonathan and Colin claim that both cases are metalinguistic. It is certainly true that in both cases "Gödel" is cited, not used.

But, first, the probe we used is similar (though not identical) to the probe Martí recommended using to elicit the relevant kind of intuitions. In the probe she sketches at the end of her 2009 paper, the proper name is also cited.

Second, and more important, it would be a mistake to conclude that the answer to the use case (we call it the linguistic case in the Analysis paper) says nothing about how people use words. After all, typically, if you say that someone's utterance is true, you are disposed to utter it yourself. There are of course exceptions (e.g., sentences involving indexicals), but the sentence examined here does not seem to be one of them.

Here is another way to put the same point. If subjects are willing to say that Ivy's utterance of “Tsu Ch’ung Chih was a great astronomer” is true, they would probably also be willing to answer "I agree" if they had been asked the following question:

Ivy says, “Tsu Ch’ung Chih was a great astronomer.” Having read the above story and accepting that it is true, do you agree with her: "I agree" or "I do not agree".

And if they answer "I agree", they are plausibly disposed to say "I agree that Tsu Ch’ung Chih was a great astronomer."

Edouard

Justin Sytsma

Jonathan,

I am not at all sure why it should be the case that “if the subject understood the probe at all, the relevant sort of descriptivism cannot be true of the names in their mouths.” The Gödel story (or the Tsu Ch’ung Chih story) gives us a case in which divergent descriptions are associated with the name. The narrator (and the subject with her if she is engaging with the pretence of the story) associates a different description with the name “Gödel” than John does. The subject, if they understand the probe, should not think that Gödel is the man who discovered the incompleteness of arithmetic, but the man who stole the manuscript and claimed credit for the work.

It strikes me that it is essential to the example that it be about John (or about Kripke’s “ordinary man”), as the whole idea is to have a situation in which the description associated with the name best matches somebody besides the bearer of the name. (As an aside, I don’t think the Gödel example actually does this, as John presumably also thinks that Gödel underwent an event of being named “Gödel,” which isn’t true of Schmidt.) To set up such a situation (or for us to recognize that such a situation holds), we have to know more than the people whose description is at issue; but this means that we will associate a different description with the name.

Because of this, I have serious doubts about Marti’s suggested experiment as well as the one that MOB carry out. The problem is that subjects should associate a different description with then name than John, they should also know this, and thus they should interpret John’s utterance relative to his ignorance. For MOB’s true/false question this leads to a rather clear speaker’s/semantic reference ambiguity. For Marti’s suggested experiment, however, John has learned something new and thus updated the description he had associated with the name “Gödel” (he might have updated the name differently, saying instead: “Today was a strange day as we learned that Gödel bears the name ‘Schmidt’”). Marti suggests that we give subjects the following sentence and then ask for their reaction: “John exclaims: ‘Today is a sad day: we have found out that Gödel was a thief and a liar.’” But, in this sentence John is now using the name “Gödel” the way the subject should, so it isn’t clear why the subject should find this strange.

Brian Weatherson

Edouard, wouldn't it be a lot better to have these experiments, and ask whether the subjects say "I agree that Tsu Ch’ung Chih was a great astronomer", rather than comparing intuitions about what they would say?

I find it impossible to understand what the subjects who give descripivist answers are doing if not something like responding to the speaker reference, rather than semantic reference, of Ivy's utterance. Now probably that says more about my powers of understanding than the facts of the case. But if there is a speaker-reference/semantic-reference distinction going on here, that's one of the exceptions you note above. So it's hard from the armchair to see why this case wouldn't be an exception to the rules you're using in the last three paragraphs.

Edouard Machery

Justin,

Our project is a bit more ambitious than *just* challenging the evidence meant to support the causal-historical theories of reference over the descriptivist theories. What we want to say is this: The *relevant* intuitions for determining what the correct theory of reference is vary across and within cultures, and this variation has significant philosophical implications (see for example Mallon, Machery, Nichols, and Stich's paper forthcoming in PPR).

Now, it is true that most philosophers of language take our *metalinguistic intuitions* to be the relevant source of evidence. This is very easy to establish, and Martí herself acknowledges it in a footnote of her paper.

But still one might wonder whether these philosophers are right. Perhaps the relevant kind of evidence consists of the kind of judgments Martí is alluding to.

Suppose *for the sake of the argument* that this is the case. Does that entail that metalinguistic intuitions have no evidence value? Clearly no: if metalinguistic intuitions are congruent with the way people use words, then even if we grant the importance of the kind of judgments singled out by Martí, metalinguistic intuitions do still provide relevant evidence for determining what the right theory of reference.

is this clearer?

Justin Sytsma

Edouard,

I appreciate the clarification! I guess I’m not sure that I buy that you were really setting out to show that the relevant intuitions vary across cultures, as that seems to call on a positive theory of what intuitions are relevant to determining the correct theory of reference that didn’t seem to be on offer (although perhaps I just need to reread the PPR article). Rather, it seems that you were testing the intuitions that were called on, granting that they are relevant. It then seems that the claim that these intuitions aren’t relevant isn’t really a direct objection to what you were doing. Now, insofar as your *overall* goal is to show that there is variation in the relevant intuitions (whatever those intuitions happen to be), Marti’s objection sets up a *possible* new target (linguistic intuitions rather than metalinguistic intuitions). It strikes me, though, that Marti should now provide *some* positive evidence that these intuitions accord with the causal-historical account (Kripke, after all, provided his own metalinguistic intuition as evidence). I’m not actually sure what the evidence for linguistic intuitions would look like. But, it does strike me that using a Gödel-type example isn’t the best way to go about collecting it. The Gödel example is based around making a judgment about who John is talking about from a perspective of superior knowledge. As such, it strikes me that it is designed to elicit what is being called a metalinguistic intuition. If it is instead linguistic intuitions that are desired, however, then it seems that you want to test how people use names (how John, or people like him, use the name “Gödel,” for example), not how people who have better descriptions than John think about John’s use of the name. The trick, if I am understanding correctly, will be to come up with a way of testing whether people use names in ways that diverge from the descriptions that they associate with those names at the time.

Edouard Machery

Jonathan,

Worry 1

"you still haven't done any tests about how subjects use names": I explained above why this is mistaken.

As for your second point, I don't understand it, but what I say below may be relevant.

Worry 2

I am not sure to understand what you say there, but let me try this. You write:

"If some people say that 'Gödel' refers to Schmidt because he's the one who satisfies the description, even while accepting the stipulation that Gödel isn't the one who satisfies the description, this shows that their intuitions about reference are incongruent with their use."

This is only one way to describe the situation. Here is a much more natural description: some subjects say that supposing that the thought experiment were true, 'Gödel' *in their idiolect* would then refer to Schmidt because he's the one who satisfies the description, while accepting the stipulation that the man *originally called 'Gödel'* (viz. the man that is the reference of 'Gödel' in people's idiolects in the past) isn't the one who satisfies the description.

Now first, if this is the correct way of understanding the situation, it is perfectly possible to have descriptivist metalinguistic intuitions in this case - contrary to what you seem to say in your first worry (but again I don't really understand what you say there).

Second, the case provides no evidence whatsoever that people's metalinguistic intuitions are incongruent with what they would say under the assumption that the thought experiment is true.

Worry 3
This seems completely ad hoc and desperate.

Edouard Machery

Brian,

You write: "wouldn't it be a lot better to have these experiments, and ask whether the subjects say "I agree that Tsu Ch’ung Chih was a great astronomer", rather than comparing intuitions about what they would say?"

Fair enough! I also agree that if it is the case that when some subjects say that "TCC was a great astronomer" is true, they say so because they focus on the speaker's reference, they would then be *unwilling* to assert "TCC was a great astronomer".

But note that these subjects should also be willing to say that literally this sentence is false. I have no evidence that people were tempted to say so. I need to check with my coauthors, but I do not remember any subject writing anything about this in the comment section.

Furthermore, it would still be interesting that judgments about truth-values track judgments about reference. When I presented this work to philosophers of language a couple of years ago, many expressed doubts that people would be willing to endorse the claim that Gödel was a F, where this endorsement makes sense if "Gödel" refers to Schmidt. Some philosophers seemed to conjecture that asking about the truth-value of a sentence would, so to speak, prevent people to focus on speaker's reference or, to put it differently, that it would clarify that the question was not about what the speaker intended to talk about.

You also note:

"I find it impossible to understand what the subjects who give descriptivist answers are doing if not something like responding to the speaker reference, rather than semantic reference, of Ivy's utterance."

More on this in the coming months. Max Deutsch, Cathy Carroll, justin Sytsma, and I have some new data showing that our original findings (for memory, people in Hong-Kong tend to have apparently descriptivist intuitions) cannot be explained by means of the speaker's/semantic reference distinction.

Still, I find your reaction (ad JI's too) a bit puzzling. When Kripke presents the Gödel case in N&N, his claim is not that the reference of "Gödel" could not be fixed descriptively, but that, as a matter of fact, it is not. This suggests that it should at least be possible to have coherent descriptivist intuitions about semantic reference in this case.

Jonathan Ichikawa

I've just read the paper. (My first comment was made just on the basis of this post.) I realized that I'd been assuming something that wasn't true: that the new data confirmed the earlier surprising data that some people responded in some sense descriptively, and that cultural differences predict these differential responses. But in fact, all three groups surveyed reported, on the whole, Kripkean intuitions as opposed to discriptivist ones, and there wasn't any discovered cultural disagreement between them. So the data really is just that there's no discovered difference between intuitions about reference and intuitions about truth-value. Which is all you claimed in the post -- I was just assuming things that weren't there.

My 'Worry 3' above was intended to address the hypothetical data that a surprising number of people responded descriptively. In fact, that didn't happen; so I agree, my proposed interpretation there isn't very well motivated. (Although I'm still pretty confident that if you'd used the prompts I suggested, the rate of the Kripkean intuition would be even higher, for the sorts of reasons I raised.)

But Brian's worry now seems extremely important. The Kripkean is concerned with explaining away the particular surprising intuitions cited in the original paper. What's going on with people who report the descriptivist metasemantic intuition about reference? An obvious candidate explanation is, they're responding to the speaker's intended reference, not to the semantic reference of the term. This is meant as an explanation of why these particular subjects are saying this particular surprising thing.

Now we have new data about different subjects who do not say the surprising thing; they, from the Kripkean point of view, get the answer to the metasemantic question about reference right. The argument from this data to the debunking of the suggestion about speaker reference has to be extremely indirect; it goes through general principles about agreement between intuitions about reference and intuitions about truth values, and then between intuitions about truth values and semantics for names used. There are so many steps here, and so much reason to dispute them in particular cases, that I think the ultimate force of the data is pretty weak (though not zero).

Here's hypothetical data that would have been impressive. We survey the people who gave the surprising response in the initial study -- those who said that when John uses the name 'Gödel', he's 'talking about' the man who really discovered incompleteness. And we discover that most of them give any, many, or all of the following descriptivist responses to probes (given in both 'linguistic' and metalinguistic versions, in your terminology in the paper):

NO to: John says to his friend Mary, "Gödel is famous for discovering incompleteness." Is what John says true?

NO to: Did Gödel become famous for discovering incompleteness?

YES to: John says to his friend Mary, "Gödel deserved credit for the discovery of incompleteness." Is what John says true?

YES to: Did Gödel deserve credit for the discovery of incompleteness?

Data like that would go quite a ways towards convincing me that the people studied did use names descriptively. (It'd still be an open question, I think, how philosophically significant that would be, but it would certainly be interesting and surprising.) But I think the data pointed to in this new paper is much, much weaker than the hypothetical data I've just outlined. And, expressing my own empirical prediction, I'd be willing to wager a significant chunk of cash that data like I've outlined here wouldn't be found, were the relevant experiments conducted.

Edouard Machery

Jonathan,

You write: "What's going on with people who report the descriptivist metasemantic intuition about reference? An obvious candidate explanation is, they're responding to the speaker's intended reference, not to the semantic reference of the term. This is meant as an explanation of why these particular subjects are saying this particular surprising thing."

As I mentioned in the reply to Brian about this, a reply to this point will have to wait. But several unpublished results suggest that this explanation is not correct.

You also say: "Now we have new data about different subjects who do not say the surprising thing; they, from the Kripkean point of view, get the answer to the metasemantic question about reference right."

This is also incorrect - about a third of the subjects have descriptivist intuitions in the three countries examined. And within-culture variation has the same philosophical implications (see the PPR paper).

You also write: "The argument from this data to the debunking of the suggestion about speaker reference has to be extremely indirect;"

I don't follow you here. We do not argue in the paper that our data undermine the speaker's reference objection, and I conceded above that this could be a worry. We only reply to Martí's point.

Still, it is worth repeating that if subjects who give descriptivist answers in the use cases are focusing on the communicative intentions of the speaker, they should be willing to say that the sentence is literally false. But they don't.

Finally, you say that "But I think the data pointed to in this new paper is much, much weaker than the hypothetical data I've just outlined." I don't see why. They are closely related to the sentences you yourself propose.

E

jonathan weinberg

Fwiw, though I generally am on the same wavelength as Edouard in this discussion, I would disagree with this:

"And within-culture variation has the same philosophical implications (see the PPR paper)."

A certain amount of within-culture variation can, I think, be explained away as performance error, subject confusion over the task at hand, etc. Unless we have something that will count as a baseline level of performance of some sort?

Jonathan Ichikawa

Hmm. Maybe I need to read the Martí again. If what you're saying here is right, Edouard, then Martí's point is different from the point I thought she was making. But I do have a tendency to attribute my own views to others, when they're suitably similar.

Justin Fisher

Two concerns:

(1) The prompt seems to beg the question in favor of Kripke. "Now suppose that Tsu Ch’ung Chih did not really make this discovery. He stole it..." This presumes that the name refers to the young thief, not the old astronomer. A neutral prompt would instead need to say something like, "In fact, what happened was that an aging astronomer made the discovery, but died before publicizing it. A young man stole the old man's journal, and claimed the discovery as his own, which is how the world came to associate the discovery with the name 'Tsu Ch'ung Chih'."

Similarly, the metalinguistic prompt says that the young man is "widely believed" to have made the discovery. If (Kripke's caricature of) Descriptivism were right, then contemporary widespread belief would have it that the old man (i.e., TCC) made the discovery, not the young thief. So again, this prompt begs the question in favor of Kripke.

(2) This example (like the Godel/Schmidt example) is useless as a test case for descriptivism. Everyone tacitly associates lots of descriptive information with each name, including that people often pronounced that name while looking at (or for) its referent, and that there has been a chain of communication that passed that name down to us. All of this descriptive information points toward the young thief as being the actual referent. Only the one textbook description (that TCC made the discovery himself) fails to fit the young thief, but even that description is tacitly associated with many descriptions that are true of the young thief (e.g., that TCC claimed to make the discovery himself, and was credited for it). Hence, many Descriptivists would say the name refers to the young thief, just as Kripke does. Since both sides can plausibly give the same verdict about this case, it can't help us choose which side is right.

Edouard Machery

Jonathan,

This is a good point. On the other hand, cross-cultural differences raise various issues (different use of scales across cultures, issues with linguistic comprehension, differences in pragmatic expectations, etc.).

e

Edouard Machery

Justin,

(1) This is entirely correct, but doesn't it make Chinese subjects' judgments even more striking?

(2) This is very controversial. You are assuming that every information associated with a name contribute to reference fixation. This is surely a position hold by some descriptivists, such as plausibly Evans, but not by all descriptivists (think of Searle and Jackson, e.g.). Most important, it is not the version of descriptivism criticized by Kripke

e

Justin Fisher

A late response to Eduard (sorry I didn't check back sooner):

Regarding (1), when subjects offer responses that explicitly contradict a (question-begging) prompt, I do indeed find that striking, but in a way that calls seriously into question how well they understood the prompt, and how seriously they were taking the whole enterprise. It would be much better, I think to not use a question-begging prompt, and hence (at least somewhat) mitigate this concern.

Regarding (2): Of course there are different possible versions of Descriptivism and different versions may take different descriptions to play a role in reference determination. Kripke's example and the example used in this study *at*best* challenge only a small number of possible descriptivist views. Unlike Eduard, I read Jackson as allowing a great deal of deference within reference-determining descriptions. So too for Lewis and Chalmers, whom (along with Jackson) I take to be among the strongest (broadly) descriptivist views going. So, I think there are real worries that Kripke's and Eduard's examples are relevant only to strawman versions of descriptivism.

(BTW, I think all versions of descriptivism are deeply wrong; I just don't think this is a good way of testing the most plausible versions of descriptivism.)

ramen

Very interesting article and especially very interesting discussion following, all of this is great food for though. I think i'll have to read Martí again, some stuff seem to have gone over my head the first time! :)

The comments to this entry are closed.

3QD Prize 2012: Wesley Buckwalter