Blog powered by Typepad

« Stubenberg on neutral monism | Main | Poll: Two belief ascriptions »

February 10, 2005



After having thought about the problem for a while, I think the particular de re knowledge Soames has in mind really matters. That is to say, you can't get quite the same problem off the ground just by pointing to disquotational knowledge.

Here's why. Let's briefly review the problem. According to Stalnaker, what makes for an informative identity claim is that I haven't ruled out that the two words flanking the identity corefer. So when we represent my epistemic situation, there are worlds in the set compatible with what I know in which the two words corefer, and there are worlds in that set in which the words don't. Call this set C1.

Now suppose I have disquotational knowledge; in the usual example, I know that `Hesperus' refers to Hesperus, and `Phosphorus' refers to Phosphorus. As Dave notes, the English sentences " `Hesperus' refers to Hesperus" and " `Phosphorus' refers to Phosphorus" together entail that `Hesperus' and `Phosphorus' corefer. Since these sentences report the contents of my beliefs, no worlds in which the two words fail to corefer are compatible with what I know. Call this smaller set C2, i.e., C1 is a proper superset of C2. The worry is that, just by having this trivial, disquotational knowledge, I narrow down the set of worlds compatible with what I know from C1 to C2.

But that just can't be right. Stalnaker usually begins by asking what some sentence does, and then we can worry about how to fit this behavior into a systematic theory. Well, here's what the disquotational sentence seems to me to do: it seems to me to say no more and no less than that `Hesperus' refers to the thing we call `Hesperus'. At any rate, that seems to be the content of my presupposition when I presuppose that `Hesperus' refers to Hesperus in a situation naturally described as one in which I don't know that Hesperus is Phosphorus. That means that the disquotational belief does not cut down the set of worlds compatible with what I know from C1 to C2, and hence disquotational knowledge does not pose a threat to Stalnaker's account of Assertion.


I feel like this meta-linguistic stuff is only complicating things and is not at the heart of the issue. The general complaint against Stalnaker (which Jackson and Chalmers have the tools to try to avoid) can be recreated without it.
The general problem is that the following happens: A set of presuppositions (taken Stalnaker's way) jointly narrow down the context set. But the assertion made with a sentence (in that context) may be informative though it does not, in fact, narrow down the context set.

We can recreate this without the metalinguistic stuff. Consider the following presuppositions:

P1 Last night Hesperus was the only visible star.
P2 This morning Phosphorus was the only visible star.

These entail:
CLAIM Exactly one star was visible both in the morning and in the evening.

Certainly, CLAIM may be used to assert something informative in this context (Everyone would agree with that). It also seems to me that what is asserted with CLAIM is just the proposition semantically expressed by CLAIM (what else could it be?). But this proposition is already entailed by the presuppositions P1 and P2. So on Stalnaker's view the proposition semantically expressed by CLAIM ought not to be informative. But it is.

On the Chalmers-Jackson view, one can (try to) avoid this type of problem because the presuppositions are taken epistemically. So I tend to agree with Chalmers that for the 2D person, what goes wrong is the way Stalnaker is thinking of presuppositions. The case I bring up (completely inspired by Soames' arguments against possible world semantics) is supposed to support that diagnosis (if one likes the 2D type approaches to begin with) and try to get away from the messy meta-linguistic stuff.


Both of these comments seem reasonable to me. Bernard's comments suggests a general strategy that Stalnaker might use in at least in some of these cases. He can argue that although it might seem at first that these cases involve a presupposition that P, in fact the relevant presupposition has something other than P as its content. This strategy seems especially apt where the presupposition is introduced by a prior utterance of a sentence that would ordinarily express P: in this sort of case Stalnaker might be able to use the 2D strategy to argue that the utterance narrows down the context set according to its diagonal proposition rather than according to P. The strategy can also be used in cases where our main reason to think that the presupposition that P is in place is that an attitude ascription 'S presupposes that P' seems to be true: in such cases, Stalnaker could invoke his 2D semantics for attitude ascriptions to make the case that the ascription could be satisfied by a presupposition whose content is something other than P (e.g., a diagonal proposition in the vicinity). I take it that Bernard's suggestion is that the second version of the strategy applies in the disquotational case, where the datum is arguably just the claim that subjects presuppose that 'Hesperus' refers to Hesperus. I suppose that one could also argue that this move applies in at least some versions of Angel's case, depending on just how the case (and the presuppositions) are described.

It's not easy to see how Stalnaker can use this strategy in the de re case, though. Here the reasons for thinking there is a de re presupposition needn't turn on prior utterances or on attitude ascriptions. What matters here (as Bernard pointed out in conversation) is that the subject stands in a certain sort of causal/cognitive relation to a particular object (Venus), making the case appear to be exactly the sort of case where Stalnaker is already committed to there being substantive de re knowledge, whose content is a proposition that is true in a world depending on the properties of the specified object in that world. The same point applies in Angel's case, if the presuppositions are characterized in the appropriate way. All this tends to confirm that the central issue isn't the diagonalization strategy or anything metalinguistic, but rather is the way that presuppositions constrain the context set.

Bob Stalnaker

Dave invites responses to his discussion of Soames’s criticisms of Stalnaker from “those more familiar with Stalnaker’s program”. Some have questioned whether I really understand that program, but I am at least familiar with it, so let me try to say what I think is going on, and how one ought to respond , from the point of view of that project, to the problem raised. I haven’t yet read Soames’s book, so I am going on Dave’s exposition and development of the objection.

The initial puzzle, as I understood it, was this: an independently motivated semantics tells us that certain statements (most notably Kripke’s paradigms of necessary a posteriori truths) are statements that seem to convey contingent information despite the fact that they are necessary.. To convey contingent information, it seems natural to believe, is to rule out certain possibilities. But how can a statement whose content is a necessary truth, and so which excludes no possibilities, convey contingent information? I approached the problem in two stages: first, try to get clear what contingent information it is that is conveyed, and second, to try to explain how statements with the semantics that the problematic statements seem to have are able to convey that information. The two-dimensional apparatus, and the operation of diagonalization enters the story only at the second stage. To characterize the information that seems to be conveyed, my strategy was to describe a range of possibilities that seemed, intuitively to be the possibilities that were compatible with the prior beliefs and presuppositions of speaker and addressee in a plausible context in which the statement seemed to convey information, and then to consider which of those possible situations seemed to be excluded by the statement. The prior context (the context set) was not determined by first asking what propositions the speaker and addressee believe in common, and then define the context set as the possible situations in which those propositions are true. Rather, the idea was to describe (in a way that is as noncommittal as possible about the semantics for the expressions whose interpretation is in question) the possible situations that seemed to be left open by relevant epistemic situation, and then to define the propositions presupposed in the context as those that were true in all of the relevant possible situations. The possibilities are described from the external perspective - they are the theorist’s characterization of the scenario.

Now in the Hesperus/Phosphorus case, it seems clear that there is a possible world compatible with the addressee’s prior beliefs in which distinct planets, or other astronomical bodies, appear in the morning and the evening in the way that Venus in fact appears. And it seems clear that this is the kind of possible situation that the speaker means to exclude when he says “Hesperus is Phosphorus.” (Still no use of the 2D framework yet - that comes in only when one explains how it is that the sentence “Hesperus is Phosphorus” conveys that information.)

Now Soames’s objection, as I understand it, is to the setting up of the problem, and specifically to the characterization of the prior epistemic situation as one which is compatible with a possible world in which there are two distinct planets involved, one that is the referent of ‘Hesperus’ and the other which is the referent of ‘Phosphorus’. To quote Dave’s exposition, the “objection is that ‘Hesperus is Phosphorus’ can be informative even when the speakers have already made certain presuppositions that eliminate the relevant worlds from the context set.” My first reaction is that this is just wrong. My question was, what is the world like (or what might it be like, since O’Leary isn’t sure), according to O’Leary, who doubts that Hesperus is Phosphorus, but who knows that ‘Hesperus’ denotes that object [pointing to the relevant place in the night sky] and that ‘Phosphorus’ denotes that object [pointing, later, to the relevant place in the morning sky])? We don’t answer this question by using our beliefs about the semantics for the names and demonstratives that are used in this characterization of what O’Leary knows and doubts, since it is the semantics for those expressions that is under discussion. Rather, we try to describe, in noncontentious terms, what the world is like according to him. It seems undeniable that the world as O’Leary takes it to be or the world that is one of the ways O’Leary thinks it might be) is one in which the two names denote different objects (and in which O’Leary is demonstrating different objects on the two different occasions.). After all, O’Leary might say, “For all I know, ‘Hesperus’ and ‘Phosphorus’ are names for different planets.”)

As I said, the 2D apparatus hasn’t entered the story yet, but it will enter in the further dialectic about whether I have given the right characterization of the prior context. In “Assertion”, the diagonalization strategy is applied only to ordinary assertions, but it is also applied (in later papers) to attributions of belief, and other attitudes. The issue about the prior context is an issue about what presupposition attributions are correct, and how they are to be interpreted. I can agree with Soames that it is common knowledge, in the context that is relevant to the H/P example, that ‘Hesperus’ refers to Hesperus, that ‘Hesperus’ refers to that planet, etc. The question is what these attributions say about the proposition that is being presupposed. One can diagonalize these attributions in the same way that one diagonalizes an assertive utterance. In both cases, one is explaining how a form of words (an assertive utterance, or a that-clause) that seems to be expressing (or referring to) a singular proposition is in fact referring instead to a different proposition, obtained in a systematic way (or at least somewhat systematic), from the standard semantics.

Dave suggests that this kind of problem in not a problem for the version of 2D-ism that he and Jackson endorse, and I think his response is similar to the one I am here suggesting. But he says that “this sort of move doesn’t seem to be available to Stalnaker. The reason is that the 2D content (if any) of presuppositions and other mental states plays no role in his framework.” It is right that I assume that the objects of attitudes are (one-dimensional) propositions, but I do want to apply the 2D apparatus to attitude attributions. With both assertions and attitude attributions, the framework comes in, not to characterize the content of either an assertion or a that clause, but to explain how a sentence or a clause with a certain prima facie semantics can be used to express or denote a proposition different from the one that is given by the simple semantics.


Thanks, Bob. That looks like the most promising line of response, and seems closely related to the line discussed in Bernard's and my comments above. I think the residual issue is the worry in the second paragraph of my comment (the one just prior to your comment).

I'm not certain about your views of de re belief, but I took it (maybe wrongly) that you were committed to there being at least some cases where beliefs about an object have a content that is true only in worlds where that object has a given property. Once we allow that there are some such cases, it seems hard to deny that there could be two simultaneous beliefs of this sort about a single object through different routes (e.g. beliefs arising when looking at the same object both directly and via a mirror, without realizing it). If we then allow that two such beliefs can function as presuppositions, we'll be able to generate a version of the original problem. So I presume you hold either (i) that there are no beliefs with this sort of content, or (ii) that subjects can't have two simultaneous beliefs of this sort about the same object via different routes, or (iii) that such pairs of beliefs can't function as presuppositions. I'd be interested to know which.

Ben Blumson

Suppose speakers presuppose de re of Hesperus that ‘Hesperus’ refers to it and de re of Venus that ‘Venus’ refers to it. Stalnaker needs to resist the conclusion that they presuppose de re that ‘Hesperus’ and ‘Phosphorus’ corefer.

Why should it follow that the speakers presuppose that ‘Hesperus’ and ‘Phosphorus’ corefer? The argument goes that if someone believes de re of Venus that ‘Hesperus’ refers to it, then they believe the proposition that is the set of all possible worlds where Venus has the property of being called ‘Venus’.

But there are independent reasons to reject this. Take the case of Kripke’s Pierre puzzle. Pierre seems to believe de re of London that it is ugly, but also de re of London that it is pretty. But Pierre’s does not seem to be inconsistent in holding these believes. If we say that Pierre’s beliefs are de re in the sense that he believes the proposition that is the set of worlds where London has the property of being pretty and the set of worlds where London has the property of being ugly, we are forced to say that Pierre’s beliefs are inconsistent.

If this isn’t the content of the beliefs, in what sense are they still de re beliefs about London or Venus? I think Stalnaker says that Pierre’s belief about London is still de re because what Pierre believes is counterfactually dependent on the way London is under ideal conditions. So Pierre can stand in the right sort of causal relation to London to have a de re belief about it, without Pierre’s belief being the proposition that London has the property of being pretty.

I think that this is a rough summary of what Stalnaker says about belief de re in ‘Belief Attribution and Context’, though I’m not sure I understand the whole of that paper.

It seems he should be able to say a similar thing here. Someone could have the de re belief of Venus that it is called ‘Phosphorus’ and another de re belief of Venus that it is called ‘Hesperus’, but still believe that ‘Hesperus’ and ‘Phosphorus’ do not corefer, without being inconsistent.

Laura Schroeter

Interesting post. It's helpful to get a response from Stalnaker himself about his methodology. It seems clear that Soames is starting with a different methodology, according to which we have a fixed all-purpose specification of prior belief contents, and this specification is then used to directly settle the context set. Stalnaker responds that he’s not assuming a fixed characterization of the subject’s beliefs – instead, he’s approaching the problem by looking for a pre-theoretical characterization of the relevant empirical possibilities that the confused subject is leaving open.

I wonder if the same methodological disconnect is involved in Dave's re-statement of the problem (picking up on Angel’s response). Dave is taking the moral of Stalnaker’s response to be that we need to be careful in attributing a fixed, all-purpose content on the basis of someone’s being willing to assert a de re sentence like ‘Hesperus was the only visible star’: for sometimes the speaker’s understanding is best captured by the diagonal proposition. But insofar as speakers sometimes believe de re propositions, Dave urges, the basic problem will arise again: all-purpose de re belief contents can figure in the context set and prevent an intuitively informative identity claim from having a contingent diagonal intension.

I’m wondering if this isn’t missing some elements of Stalnaker’s position – in particular, the contextual variability of content ascriptions. (Perhaps I should say I’m hoping to get some clarification here from Stalnaker himself on this point.) I would have thought that on Stalnaker’s view, it might sometimes be contextually appropriate to attribute de re beliefs about Hesperus & Phosphorus to O’Leary – e.g. when we’re characterizing his discussion of astronomical laws with Daniels. And in this discussion, the de re beliefs may figure in the context set. Nonetheless, it’s not contextually appropriate to attribute de re beliefs about Hesperus & Phosphorus when determining the context set for O’Leary and Daniels’s conversation about whether Hesperus is Phosphorus. If Stalnaker takes de re belief attributions to be context-dependent in this way, then he need not deny any of (i) – (iii).

Why think belief attributions are context-dependent in this way? Two points seem relevant. First, total belief states are highly complex, and we don’t fully characterize every aspect of a subject’s cognitive state with a particular proposition (set of possible worlds). Choosing different ways of characterizing possible worlds will bring out different aspects of the subject’s cognitive state: e.g. we might focus on objects (res) vs. just properties, or we might include semantic properties of words vs. just non-semantic states of affairs. (Stalnaker, I take it, isn’t committed to Dave’s view that there is some canonical, epistemically basic way of fully specifying every metaphysically possible state of affairs.) Second, when we attribute particular beliefs, we as interpreters are singling out contextually relevant aspects of the subject’s cognitive state. Often, we’ll advert to our own explanatory purposes – for instance, we’ll want to highlight those aspects of O’Leary’s total state of mind that help explain why he’s so interested to hear Daniels say ‘Hesperus is Phosphorus’. So Stalnaker may well deny there is any single context-independent specification of the content of particular beliefs, whether 1D or 2D. Indeed, he might go even further and deny there is any context-independent way of individuating particular beliefs in the first place.

Bob Stalnaker

I agree with Dave that the key issue here concerns the ascription of de re beliefs, or presuppositions, to the speaker and addressee in the Hesperus/Phosphorus example. There are different accounts of how a de re belief ascription should be understood (some of which reject the assumption that it is the ascription of a specific belief), but all agree, I think, that a de re belief ascription is a case where the ascriber refers to an individual, and describes the subject’s belief state as a function of that individual. One can characterize a state of belief (or a context set, representing what is being presupposed) in this way only when one can unambiguously distinguish the possible situations compatible with the relevant belief or presupposition state in terms of the individual in question. As Laura says, whether one can do so will depend on context, since the task will be to distinguish the possible situations compatible with the beliefs from alternatives that might, in the context in question, be compatible with the beliefs. As I have said, I think of a belief (or presupposition) state as given by a set of possible situations - the alternatives left open by the believer, or in the context. The propositions believed or presupposed are those that are true in all those possible situations. If one approaches the situation this way, then, to stay with the H/P case, the relevant context will be one that includes a possible situation in which there are two distinct planets, one referred to by ‘Hesperus’ and one referred to by Phosphorus’. But even if this is clear, it is much less clear how to answer a further question: which of these planets, if either (in the world as it might be, according to O’Leary) is Venus? I don’t think, without more context for the story, that either theory or intuition should yield an answer to this question. It is not that O’Leary’s beliefs or presuppositions should be characterized by a set of possible worlds containing objects of indeterminate identity, but that it is indeterminate, in such cases, exactly what possible worlds are the right ones for characterizing O’Leary’s belief state. Without some basis, either in O’Leary’s cognitive capacities, or in the context in which they are attributed, to identify one or the other of the planets as Venus, the ascriber will not be able to use the reference to Venus, or to a singular proposition about Venus, to characterize O’Leary’s beliefs, and similarly for Pierre, Ralph. Lois Lane, etc.

But in a context in which the identity confusions are not in focus, de re belief ascriptions, and the ascription of singular belief, is (on my view) unproblematic. (And one can’t characterize a state of belief at all without using either some particular things, or some properties, relations, or other materials that the ascriber finds in the world) So Dave is right when he took it that I am “committed to there being at least some cases where beliefs about an object have a content that is true only in worlds where that object has a given property.” But it is not, on my view, that it is the intimacy or vividness of one’s acquaintance with an object that makes it correct to ascribe a singular belief to a person. The issue is whether (in the relevant context) the person’s cognitive capacities and dispositions can be unambiguously represented, or unambiguously distinguished from the relevant alternatives, by a set of possibilities involving a particular individual.

The kind argument that Dave sketched above (“Once we allow that there are some such cases, it seems hard to deny . . “) is what like the one I described as the “divide and conquer strategy” in my paper, “Belief Attribution and Context” that Ben Blumson refers to, and which I there argue can be avoided..

Of the alternatives Dave presents to me, (ii) is closest to what I want to say, but the way I would put it is that, in cases where one has beliefs that are (in some sense) about the same object by two different routes, then this is not a case in which (without further contextual clues) a de re belief attribution will be correct.

I think Laura’s characterization of my position is exactly right. Doesn’t she make it sound plausible?


Thanks, Bob, and also Laura and Ben. Before saying more I think it's useful to distinguish issues about ordinary de re belief ascription, i.e. about the truth of ascriptions such as "X believes of Y that it is F", from issues about singular belief content, i.e. about whether the content of a subject's belief is true only in worlds where a particular object (Y) has a particular property (F). (Note that in this sense, some possible-worlds contents are singular contents.) Ben suggests, in effect, that Bob can hold that de re belief attributions can be true even if the subject doesn't have a belief state with corresponding singular content. That's an interesting line of response, but it looks like Bob doesn't endorse it. Bob's remarks suggest that he holds that de re belief attributions are correct only when the subject has a belief state with an appropriate singular content. In any case, whatever one says about the connection between the two, one can raise the crucial issues wholly in terms of singular belief contents, without mentioning ordinary de re belief ascriptions.

Bob and Laura both appeal to the role of contextual factors in de re belief. Here again, I think one needs to distinguish the claim that the truth of an ordinary de re belief ascription depends on context from the claim that whether a subject's belief has singular content depends on context. The former is plausible enough, but given the distinction above, it's the latter that's most relevant here. One then needs to distinguish two further theses: (A) the thesis that whether a subject's belief has such-and-such singular content depends on the subject's context, and (B) the thesis that the truth of claims of the form 'the subject's belief has such-and-such singular content' depends on the ascriber's context. Some parts of Laura's remarks, appealing to the explanatory purposes of ascribers, suggest (B) -- speaking loosely, a view on which there may be no ascriber-independent fact of the matter about whether a believer has a belief with such-and-such singular content. That view of belief content, while interesting, seems somewhat more deflationary about content than anything suggested in the work of Bob's that I've read. Bob's remarks seem to me to mostly suggest thesis (A), though I'm not 100% certain about this (Bob, please correct me if I'm wrong).

I take it that if one responds to the puzzle cases by embracing thesis (A), things will go roughly as follows. In cases where the subject has only Hesperus-beliefs (i.e. beliefs of the sort that would be expressed by saying 'Hesperus is phi'), and no Phosphorus-beliefs, then the subject's context is not problematic, so these beliefs have singular content involving Venus. Likewise if they have Phosphorus-beliefs and not Hesperus-beliefs. Further: if there were a subject who is intrinsically identical to a subject in our world with both Hesperus-beliefs and Phosphorus-beliefs, but who lives in a context where these beliefs are triggered by different objects (Mars and Jupiter, say), then that subject has beliefs with singular content involving Mars and Jupiter, respectively. But for the subject in our world with both Hesperus-beliefs and Phosphorus-beliefs (or at least for some such subjects), both of which are triggered by Venus, the subject's context is such that it prevents the subject from having singular contents for each of these beliefs. Instead, the subject's Hesperus-beliefs will have non-singular contents -- maybe a content true in worlds where the object called 'Hesperus' by the subject is phi, or where the object visible at a certain place in the evening sky is phi, or something like that. And likewise for their Phosphorus-beliefs. Is that the general idea?

A couple of worries come to mind. One, of course, is a worry about ad-hocness. Another is a worry about just how the relevant non-singular contents are determined. Presumably these will be broadly descriptive contents, such as those above. Those moved by Kripke's anti-descriptivist arguments might be inclined to doubt that such descriptive contents will always be available. Perhaps in linguistic cases, Bob can appeal to metalinguistic contents ("the object called 'Hesperus' is...). But there are plenty of non-linguistic cases to worry about: e.g. perceptual cases, where a subject sees objects on both sides of a symmetrical visual fields and forms beliefs about them, without realizing that they are in fact the same object. How should we understand the non-singular contents in these cases?

Of course, my own view is that one can always find relevant non-singular contents in these cases, but I'm much more sympathetic with broadly descriptivist approaches than Bob is. Whatever one thinks about descriptivism, I think it's reasonably plausible that if broadly descriptive contents can always be associated with a subject's belief states in these cases with dual presentations of a single object, then broadly descriptive contents can also be associated with a subject's belief states in similar cases where there is no such dual presentation. (Perhaps singular contents are also present in these latter cases, but it seems that the descriptive contents should at least be available.) And if so, that seems to undercut at least some lines of anti-descriptivist thought. It would also tend to support a more robust sort of two-dimensionalism than Bob accepts.

Furthermore, one might worry that once these broadly descriptive contents are available, and can be used to explain a subject's behavior in the case of dual presentations of a single object, then similar contents could be used equally well to explain a subject's behavior in the case of dual presentations of distinct objects (the Mars/Jupiter case above). And then it's not a big leap to infer that they can be used to explain (at least many aspects of) a subject's behavior in any case of apparent singular belief.

Note that one can also get similar reasoning going for cases involving dual presentation of a natural kind in the environment, or of a property in the environment. For similar reasons, it looks like Bob will need to appeal to broadly descriptive contents in these cases. And again, it seems that if broadly descriptive contents are available in these cases, then they'll also be available in cases where different kinds or properties are presented, and in cases where there's just a single presentation of a kind or property. And if the descriptive contents can do most of the work that we want contents to do in the dual-presentation case, then they'll also be able to do most of the work that we want contents to do in the single-presentation case. I imagine that this would be an uncomfortable result for Bob. So it would be useful to know where he gets off the bus.

Laura Schroeter

Bob may not be following this thread any more. But I thought the following passage made it pretty clear that he endorsed both types of contextualism, A and B: (Note the either/or in the second sentence)

"It is not that O’Leary’s beliefs or presuppositions should be characterized by a set of possible worlds containing objects of indeterminate identity, but that it is indeterminate, in such cases, exactly what possible worlds are the right ones for characterizing O’Leary’s belief state. Without some basis, either in O’Leary’s cognitive capacities, or in the context in which they are attributed, to identify one or the other of the planets as Venus, the ascriber will not be able to use the reference to Venus, or to a singular proposition about Venus, to characterize O’Leary’s beliefs, and similarly for Pierre, Ralph. Lois Lane, etc. "

Bob Stalnaker

Sorry to have dropped out of this discussion for a while, but a hard disk failure slowed down my electronic life for a while.

First, let me agree with Laura’s response, on my behalf, concerning theses (A) and (B). I want to claim that the correctness of de re belief ascriptions depend on the ascriber’s context, as well as on the subject’s situation. And I also assume, as Dave says, that a de re belief ascription should be understood as the ascription of belief with a singular proposition as its content.

But Dave’s worries need answering, whether or not I accept thesis (B), and I think they get to the heart of the issue.

Before getting to the main issue, a comment on one that is tangential: I don’t want to put a lot of weight on metalinguistic descriptions, such as “the planet called ‘Hesperus’” I agree that “there are plenty of non-linguistic cases to worry about.” Diagonalization (on my account) is performed on the theorist’s or ascriber’s description of the relevant possibilities. So in the perceptual example that Dave gives, the relevant world will be one in which there are different objects in corresponding places on the two sides of the subject’s visual field. Even in cases where the relevant beliefs are linguistically expressed, the subject may speak a different language. When I say that the Babylonians believed that Hesperus was distinct from Phosphorus, I am not talking about their use of the names - I assume they had different names. The ascriber diagonalizes on her own use of the names (relative to a world as the Babylonians took it to be.)

The main challenge, I take it, is this: if “broadly descriptive contents” are “available, and can be used to explain a subject’s behavior in the case of dual presentations of a single object,” then why can’t they be used in all cases of apparent singular belief? (and similarly, for dual presentation, and single presentation, of kinds or properties)

I am skeptical about the category, “broadly descriptive content”. What one falls back on, to describe the content of a belief in the dual presentation cases, is not content that is different in kind from the content that (in such cases) is inapt for characterizing the subject’s state of mind (descriptive content rather than singular content, or content involving some property or natural kind.). Rather, one falls back on content of the same general kind, but that uses descriptive resources that are apt for describing, in an unambiguous way, the world as it is according to the believer. My view is that all content involves reference to objects, properties, relations, events, found in the actual world. Some are more general and abstract than others, but there is no neutral level of description to which we have a priori access. (I am no doubt overstating the point: I don’t want to argue for a universal claim about all concepts, including mathematical and logical concepts. The point is that we can’t give any kind of substantive description of the empirical world without using language that gets its content from our interaction with things, properties and relations that we find in the actual world (and we can’t think substantive thoughts about the empirical world except in virtue of interaction with the things and properties we think about.)

So the descriptive resources we fall back on to describe the state of mind of a person who doesn’t know that Hesperus is Phosphorus, or that Cicero is Tully, might involve singular reference to something else, if a confusion by the subject about that thing is not involved, or not relevant. So O’Leary thinks that Tully is someone different from Cicero. He knows that Cicero denounced Cataline, but thinks that Tully did not. I might characterize the world as O’Leary takes it to be as a world in which there is a person who denounced Cataline, and was named “Cicero” and another who did not, who was named “Tully”. No problem with taking O’Leary to have a singular belief about Cataline (even though he probably knows even less about him than he knows about Cicero). Or to use an example I used in a paper. O’Leary believes that gold is an alloy of tin and copper. This is an impossible state of affairs, so I might describe the world according to O’Leary as a world in which an alloy of tin and copper plays (and has always played) the role that gold plays in the actual world. No problem with the fact that I am describing this possible world in terms of tin and copper - the actual kinds.

So if you can find alternative descriptive resources, in the case of dual presentation, why can’t one also find then in cases of single presentation? One usually can, but in the normal case, they won’t provide any more accurate characterization of the subject’s beliefs, and they won’t ever get one to a purer level of description.

I take the crucial difference between Dave and me to be over whether there exists a level of pure description, a priori accessible, nothing that is “twin-earthable”, that is sufficient for giving a substantive description of the world. If there were, I would agree that the purest description of an informational state would be in these terms. But if one thinks there is no such level of description, then we have to describe the world as it is according to someone with the material we actually have, and where the believer in question is sufficiently confused or ignorant about what the world is like, this will be require some circumlocution. Ad hoc it may be, but I think it is not the adhocness of the theory, but of the way people understand and are able to describe the states of mind of others.

I don’t expect Dave to agree with me about which side to take on the main issue that I am saying divides us, and perhaps he would not describe the issue exactly as I have, but I hope he will agree that this is where the action is.

The comments to this entry are closed.