Similar issues concerning demonstratives are contained in a later paper, "Do Demonstratives Have Senses?" Philosophers' Imprint 2 (2002), http://www.philosophersimprint.org/002002/. It would in some ways be better if we read that paper instead, but it is extremely long: too long for a single session.
The papers by Evans and Campbell we just read primarily concern demonstrative thoughts: They are about how we should think of the senses associated with, e.g., demonstrative beliefs. They do not really speak to the question how we should think of the senses of demonstrative utterances. That question is discussed explicitly in my paper "Do Demonstratives Have Senses?" Philosophers' Imprint 2 (2002), http://www.philosophersimprint.org/002002/, and that, in a way, is the paper we should be reading here But it is extremely long: too long for a single session. So, instead, we are reading this paper, which discusses parallel issues, but in the case of proper names.
The issue here is largely one we discussed earlier, in connection with Frege and Russell: If speakers can associate (in Russell's case) different descriptions with a given name, and if (as he seems to suggest in his discussion of the Bismark case), all that really matters is that these descriptions pick out the same object, then we shouldn't we think of the meaning of the name as just that object? Perhaps every speaker must associate some description with the name, but these look merely idiosyncratic. Why shouldn't we think of them as no more relevant to the meaning of the name than the various ideas we associate with it (to go back to Frege again)?
As I set up the discussion, I first note that there are strong arguments—based upon Frege-style identity puzzles—for the claim that the contents of beliefs are not singular propositions but need to be individuated more finely. And despite their other disagreements, Perry, Kaplan, Evans, and Campbell all agree with this claim. On the other hand, however, Kaplan and Perry (and many others) want to claim that utterances are different: Utterances of "Hesperus is a planet" (or "Thatbow ship is an aircraft carrier") and of "Phosphorous is a planet" (or "Thatstern ship is an aircraft carrier") express singular propositions and so have the same content.
This combination of views is what I call the "Hybrid View". It is natural to think that it must somehow be unstable: that the fact that speakers associate different belief-contents with "Hesperus is a planet" and "Phosphorous is a planet" must somehow be linguistically significant. The goal of the paper is to develop an argument for that conclusion.
I just intimated that there are prima facie reasons to suppose that the Hybrid View is unstable. What such reasons can you think of? Perhaps one of them might have to do with the idea of expressing a belief? Perhaps another might have to dow ith belief-attribtution: the semantics of such sentences as "Bill believes that Hesperus (vs Phosphorous) is a planet"?
Section 2 of the paper argues that the Hybrid View is committed to what I call the "extensionality of understanding": To understand an utterance of "George Orwell wrote 1984", for example, one need not think of George Orwell in any particular way, though one will have to think of him in some way (since belief is intensional). It's enough to realize that the speaker has said, of the person you are thinking of in some particular way—even if the speaker may be thinking of him in a completely different way—that he wrote 1984. So in that sense, reference is all that matters for communication—and so, plausibly, for meaning.
In section 3, I argue that "[t]he transmission of belief from speaker to speaker is a basic purpose of the practice of assertion". The argument proceeds by asking why we should think that understanding even requires preservation of reference. For example, suppose Bob thinks that "Hesperus" refers to the moon. Then, it seems, Bob is definitely going to misunderstand me if I say, "Hesperus is a planet". Why is this so obvious? I suggest that the answer is that understanding must at least guarantee that, if one believes what someone says, then one won't end up with a false belief if the belief they were expressing was true. This is supposed to establish that "[t]he Hybrid View...requires that communication be essentially a means for the transmission...of true beliefs from one speaker to another".
The central point of section 4 is to argue that "questions about the meanings of sentences may be addressed...by asking what is or, in some sense, ought to be common" to the different beliefs different speakers associate with that sentence. This is what will be used in the arguments that follow, so what is most important is simply that you understand the argumentative strategy just elaborated, not the complex argument given here for the usefulness of that strategy. So if you have trouble with section 4, then you might try just skipping it. (I wrote this, though almost 25 years ago, and I'm having trouble.... I'll try to explain the argument in the note that follows.)
I first suggest that we can easily enough understand the idea that a sentence has a certain cognitive value for a particular person: it's the content of the belief that person would form were she to accept that sentence as true. I then ask what we should make of the idea of the cognitive value a sentence has in itself (where I really should be talking about utterances but don't for simplicity). It is clear that people can't just associate any belief at all with a sentence and still count as understanding it. So I suggest that we may think of the cognitive value of a sentence as what is common to the different cognitive values it can have for different speakers who understand it. With that in hand, I then argue that the meaning of a sentence may plausibly be identified with its cognitive value: More precisely, I argue that the argument given in section 3 tacitly depended upon such a claim.
The form of the argument is this: There are arguments in section 3 purport to establish claims of these two forms:
- the reference of a name determines its meaning
- the meaning of a name determines is reference
What the arguments actually establish, however, is:
- the reference of a name determines its cognitive value
- the cognitive value of a name determines is reference
To get (1) and (2) from (i) and (ii) one needs:
- the cognitive value of a name determines is meaning
- the meaning of a name determines its cognitive value
But if we have (a) and (b), then we might as well identify meaning with cognitive value.
In section 5, then, I use the argumentative strategy mentioned to argue against the Hybrid View. I suggest that when we talk about communication as a mean for transferring information, we do not (or should not) mean just "true beliefs" but knowledge (or, at least, justified beliefs). So, as I put it, understanding must enable (that is, make possible) the transmission of knowledge (assuming circumstances are otherwise suitable). If so, then the question we want to ask is: How must the beliefs different speakers associate with a given sentence be related if they are to be able to transfer knowledge to one another?
It's a difficult question, about which I do not say very much, why (or even whether) we should think of knowledge, rather than true belief, as the crucial notion here. What might be said for or against that idea?
I then introduce what I call the Problem of Content. The question it raises is based upon a simple observation. If someone utters "Hesperus is a planet", and if you do not know that Hesperus is Phosphorous, then this might put you in a position to come to know that Hesperus is a planet, but it will not put you in a position to come to know that Phosphorous is a planet.
Why not? Is there more to be said than, "Oh, yes, it seems that way intuitively"?
In section 6, we finally get to the argument against the Hybrid View: The argument is that it cannot solve the Problem of Content. The argument depends upon a series of examples that are supposed to show that "getting the reference right" is not enough to enable to transmission of knowledge. If so, then "getting the reference right" is not enough to guarantee understanding, either.
Fodor once remarked that the problem with the form of argument being given in this section is that you have to have a cat for every mousehole. Are there mouseholes for which I've got no cat? I.e., other reasonable options a Hybrid Theorist might puruse?
In section 7, I explore the suggestion that what is required to enable transmission of knowledge is not just sameness of reference but known sameness of reference. I first motivate this idea as a natural one for a defender of the Hybrid View to suggest. I then argue that, nonetheless, it is not compatible with the Hybrid View but collapses it into a Fregean view---though not into Frege's view. The claim is that the "strict" Fregean view, which would hold that, to understand someone else's use of a name, one must think of the referent in the very same way that they do is not only obviously too strong, but that even Frege recognized this fact. The point is simply that one cannot just think of the referent any old way—in which case reference does not determine cognitive value and so does not determine meaning, either.
As noted in the closing section, none of that gives us a positive theory of sense. And, in some ways, the tenor of that section is a bit despairing. Indeed, in a later paper, "Do Demonstratives Have Senses?" I suggest that there may really be no more to be had: Maybe all there really are are the different cognitive values a sentence has for different speakers and various conversational norms that enforce certain sorts of relations between those cognitive values.