Philosophy 1860

For our purposes, the main lesson of this paper is that certain logico-linguistic issues about proper names are not just issues about language. They are intimately tied up with epistemological issues about the nature of our cognitive relationship to the world. That, Russell argues, is what makes those issues especially important.

The main topic of the paper is "what it is that we know in cases where we know propositions about 'the so-and-so' without knowing who or what the so-and-so is" (p. 108). This is a question about the 'objective content' of our knowledge: about what proposition we know in such cases. So, in that sense, it is a question about the logical structure of cognition. But the question is really one about the nature of our cognitive relation to the world.

Russell contrasts cases in which we know an object only 'by description' with cases in which we are 'acquainted' with the object of our knowledge: cases in which we have what he calls "a direct cognitive relation" (p. 108) to the object. For the moment, we may think of the basic case of acquaintance as perceptual awareness. So, sitting at my desk, I am visually aware of my keyboard, so that would be a case of my being acquainted with it.

It is important to Russell (see p. 109) that acquaintance is a relation between a thinker (a 'subject', in his terminology) and an object that is 'presented' to the thinker. He seems to suggest here, as well, that we can be acquainted with objects we are not currently perceiving, but which we remember.

Is this last claim consistent with what Russell argues later in the paper? Why or why not?

Russell gives an initial example of a case in which we know an object only by description: the candidate who gets the most votes. This involves what he calls a 'definite description', by which he means a phrase of the form 'the F'. Russell says that, in this case, we are not 'acquainted' with the object of our knowledge (or, at least, need not be).

Note that Russell seems here to be exploiting an ambiguity in the English word "know". Other languages (e.g., German and French) have different words for these two notions. But in English, we speak both of knowing facts (Sally knows that snow is white) and also of knowing objects (Sally knows John). It is this latter sort of knowledge that one has "by acquaintance" or "by description". Russell's question then is: How does whether one knows an object by acquaintance or by description affect the sort of knowledge that one can have about that object?

Russell suggests, on pp. 109-12, that there are several sorts of things with which we can be acquainted. The first is 'sense-data', by which he means particular elements of sensory experience, such as color-impressions and sounds. (Note how Russell distinguishes, on p. 109, between "the supposed physical object" of perception and "the direct sensible object".) Second, Russell thinks we are acquainted with ourselves, and he argues on p. 110 that one could not merely have descriptive knowledge of oneself, because any description one might employ would have to make reference to oneself anyway. Third, Russell thinks we are acquainted with 'universals', by which he means, roughly, properties (such as being green) and relations (such as being taller than).

Russell implicitly takes the distinction between acquaintance and description to be exhaustive. (See e.g. p. 110.) I.e., he assumes that, for any given object about which we can think, we must either be acquainted with it or else know it by description. We shall consider at some length later whether this is true.

What reasons might there be to think that the distinction between acquaintance and description is exhaustive? Is it plausible that it is? Or is there some other kind of knowledge of objects one might have?

On the other hand, Russell denies that we are ever acquainted with physical objects (e.g., my keyboard) or with other people's minds, insisting that we can only ever know such things by description (p. 112). This is when we know that there is a unique thing that satisfies the description, but we do not know which thing satisfies the description. Thus, to take a version of Russell's example, we know that someone will receive the most votes in the 2022 election for governor of Rhode Island, but we do not know who.

As always in such cases, it is worth spending a moment to think about other examples of your own creation.

Russell does not actually argue for this claim, but it is clear enough from the text (e.g., p. 114) that Russell thinks we are never really perceptually aware of other people, say, but only with "certain sense-data" that, perhaps, we are caused to have by that person. He thinks this is even clearer when the person lived long ago: We cannot be acquainted with (say) Aristotle, Russell thinks; Aristotle himself cannot be 'presented' to us; we know him only through his writings; we know him only as 'the author of this and that'.

Russell goes on to argue that, if we know an object only by description, then the knowledge we have about that object is, in a sense, not really knowledge about that object at all. So, Russell thinks, we do not really know that Aristotle was Greek. Rather, what we know is really of the form: There was a (unique) person who wrote this and that, and that person was Greek. (Indeed, Russell thinks "Greek" may need further analysis, as well.) And so, in general, if we fully "analyze" what we know in such cases, we find that it is "directly" about things with which we are acquainted. The object we know by description enters only as the value of a variable: the x that is uniquely so-and-so.

Russell is thinking of what we know—the 'proposition' we know—as articulated into various 'constituents'. (Propositions of this kind are nowadays known as 'structured propositions'.) To take some simple examples, the proposition that 2 is prime would contain as constituents the number 2 and the property of being prime; the proposition that Alex loves Tony would contain as constituents Alex, Tony, and the relation of loving. Note that these constituents need to be 'put together' in the right way: that Alex loves Tony is a different proposition than that Tony loves Alex, since (tragically) the one can be true and the other can be false. (The question how the constituents of a proposition 'fit together' is the so-called 'problem of the unity of the proposition'.)

I am being a bit imprecise here, but not in ways that will matter for our purposes. Russell actually thinks of judgement (or knowledge, or belief) not as a relation between a thinker and a proposition but as a many-termed relation between the thinker, judging (say), and the various constituents of the judgement. This theory, which Russell describes on pp. 117-8, is known as the 'multiple relation theory of judgement'. See "On the Nature of Truth and Falsehood" for an argument for it.

Russell says that "proper names are usually really descriptions" (p. 114, punctuation omitted). Exactly what does he mean by this claim? The answer is to be found in the discussion of Bismarck, on pp. 114-7.

In the full paragraph on p. 116, Russell raises a question about communication involving proper names. What problem is being raised here? To what extent is it a problem for Russell? What is his solution to it? This is a useful question to ponder, as answering it requires putting several aspects of Russell's view together.

This discussion motivates what Russell calls "[t]he fundamental epistemological principle in the analysis of propositions containing descriptions", which he states as: "Every proposition which we can understand must be composed wholly of constituents with which we are acquainted" (p. 117).

Russell says that it is "plain why [he] advocate[s] this principle". So this should be an easy question: Why does he? If there's an argument for it, it would seem to be given on pp. 117-8. (Is it 'plain' that Julius Caesar is not a constituent of any proposition we can know about him? Why does Russell think it is?)

Russell says that his "fundamental principle" just amounts to saying that "we cannot make a judgment or a supposition without knowing what it is that we are making our judgment or supposition about" (p. 118). Is that right? What unstated premises might be in play here? (What is involved in 'knowing what it is that we are making a judgement about'?)

On pp. 119-20, Russell argues against the view that our judgements really concern our 'ideas' about objects, thus distinguishing his view from such an 'idealist' view. Russell insists that, when we are acquainted with something, there is a 'direct' relation between our mind and that thing, even if the thing is outside the mind; in particular, there is no need for any 'idea' (mental object) to intervene between us and the object. (We'll soon see Frege making almost the same point. It would be interesting to compare his treatment of this issue with Russell's.)

What does this discussion imply about Russell's view of 'sense-data'?

It's important to distinguish two aspects of Russell's view here: His overall distinction between knowledge by acquaintance and knowledge by description, and his more specific (and extremely narrow) view about what we can be acquainted with. One could think that the distinction between acquaintance and description was important but draw the boundaries of acquaintance differently from how Russell does. Indeed, in many ways, a recurring theme in many of our later readings will be what sorts of cognitive and perceptual relations to things support knowledge about them that is "direct" in the sense that Russell thinks knowledge by acquaintance is "direct".

There might be different notions of directness and indirectness in play here. Can you identify some of the options?

The central purpose of this paper is to establish a distinction between the reference of an expression—primarily, a proper name—and what Frege calls its sense. The reference of a name is the name's bearer: the thing it is a name of. So the reference of the name "Gottlob Frege" is Gottlob Frege, that very person. It is not so easy to say what the sense of the name is, and Frege does not seem to tell us very much about what it is. Rather, as I said, his purpose here is to argue that names do have sense, and that their sense is different from their reference. In particular, the claim is, it is possible for two names to have the same reference but to have different senses.

The argument for this claim is contained in the first pargraph of the paper. It is not an easy argument to understand. Trust me on this.

Frege begins by mentioning a puzzle about identity statements, such as "Hesperus (the evening star) is the same thing as Phosphorous (the morning star)". As it happens, this is true: Hesperus and Phosphorous are both Venus. The puzzle is generated by the fact that such a statement can be informative and, in particular, that such statements need not be analytic or a priori but, as Frege puts it, may "often contain very valuable extensions of our knowledge" (p. 56).

Frege mentions different astronomical examples here. Can you elaborate what he has in mind?

Other examples involve pen names. For example, the author known as "George Eliot" was actually Mary Ann Evans, who used a man's name to avoid the stereotypes associated with women writers. Other examples you often see are Mark Twain (Samuel Clemens) and George Orwell (Eric Blair). Musicians and actors provide other examples, as do (in fiction) superheroes.

Frege then raises a problem for the claim that identity is a relation between objects: If it is, then it looks as if "Hesperus is Phosophorous" and "Hesperus is Hesperus" assert that exactly the same relation obtains, namely, a relation between Venus and itself. But then how can the former be informative and the latter a mere instance of the law of self-identity? How, as it is put, can the one have a different "cognitive value" (in German "Erkenntniswert", literally: value for knowledge) from the other?

Frege mentions that, for such reasons, he once held himself (in his first book Begriffsschrift, or Conceptual Notation) that identity was not a relation between objects, but was actually a relation between names. Thus, "Hesperus is Phosphorous" was supposed to mean something like: The name "Hesperus" has the same denotation as the name "Phosphorous".

It's important to see what Frege means by the claim that identity is a relation between objects. Consider, for example, the relation "taller than". This is a relation between objects, such as people: A sentence of the form "A is taller than B" is true if the person A is taller than the person B (to focus on that case); the names involved have nothing to do with this relation. And that is how we would treat it in logic: The extension of "(1) is taller than (2)" is the set of ordered pairs <x,x> such that x is taller than y. Similarly, the extension of "(1) = (2)" is the set of pairs <x,x>, so in logic we do treat identity as a relation between objects.

In the middle of the paragraph, beginning with the words, "But this relation would hold...", Frege argues against this view. To be kind, it is not obvious what argument Frege is giving here. The key seems to be Frege's observation that the relation between a name and its bearer is arbitrary or, perhaps, conventional. But his underlying thought seems to be that the claim that "Hesperus" and "Phosphorous" have the same denotation is a claim about linguistic practice, whereas "Hesperus is Phosphorous" was meant to be a claim about celestial bodies. (Note the remark that the "signs themselves would be under discussion" (p. 56) were identity a relation between names.) So the 'name view' of identity gets the "subject matter" (p. 57) of such statements wrong.

There have been writers who questioned this 'orthodox' reading of the paper. See the related paper by Thau and Caplan, and my reply.

Frege then transitions, without explicit mention that he is doing so, into the development of his own view. He notes that the mere difference of shape (spelling, pronunciation, whatever) between "Hesperus" and "Phosphorous" cannot be what accounts for the difference in cognitive value that we are trying to explain. Rather, there will be a difference in cognitive value only if there is a difference in what Frege calls the "mode of presentation", which he illustrates using a geometrical example.

Can you give an example to show that the mere difference of shape (spelling, pronunciation, whatever) between names cannot be what accounts for the difference in cognitive value that we are trying to explain? Can you give an example where there is no difference of shape (etc) but there is still a difference of cognitive value, even though the reference is the same?

Frege then says that each name has a "sense" that "contains" a "mode of presentation" that is associated with that name, but most commentators have simply supposed that the sense is the mode of presentation. In any event, the implication (which Frege makes explicit elsewhere) is that the fact that "Hesperus is Hesperus" and "Hesperus is Phosphorous" differ in cognitive value is supposed to be explained by the fact that "Hesperus" and "Phosphorous" have different senses.

The puzzle here, though particularly stark in connection with identity-statements, is not essentially about identity. The sentences "Hesperus is a planet" and "Phosphorous is a planet" also have different 'cognitive values'. In particular, someone might believe (or even know) that the one is true without believing (or knowing) that the other is true. Frege also holds that this difference is explained by the fact that "Hesperus" and "Phosphorous" have different senses. (Frege does not mention this version of the puzzle here but does so, again, in other places.)

How exactly is the fact that "Herperus" and "Phosphorous" have different senses supposed to explain the difference in cognitive value between "Hesperus is a planet" and "Phosphorous is a planet"? What assumptions about sense and cognitive value must such an explanation make?

In the next few paragraphs (pp. 57-9), Frege states several theses about sense and its relation to reference:

• The sense of a name is a linguistic feature of it (an aspect of its meaning, as it were), and anyone who understands the name must know what its sense is.
• Names with the same sense must have the same reference (sense determines reference), but names with the same reference may have different senses.
• It is possible for a name to have sense without having a reference. (Better examples involve names like "Homer", which some scholars think does not denote any single person but became attached to the Greek epics through some confusion. If one supposes that such scholars are correct, then the name gives us a good example.)
• Ordinarily, when one uses a name, one uses it to 'talk about' its reference.
• But when one uses words in 'indirect speech', one uses them to talk about their sense. Thus, if one says, "Lois said that Superman can fly", then one is talking about the sense of Lois's remarks, as is clear from the fact that it is one thing to say that Superman can fly and another to say that Clark can fly. The same is true of such constructs as "Lois believes that Superman can fly". Here it looks as if what one is saying Lois believes is determined by the sense of the name "Superman", since Lois does believe that Superman can fly, but not that Clark can.

Over the next several pargaraphs (pp. 59-62), Frege argues that the sense one associates with a name must "be distinguished from the associated idea" (p. 59), by which he means something like a mental image. (We have already seen Russell make a similar argument.) The larger point at issue here, though, is whether sense is subjective. The really crucial claim for Frege is that senses can be shared, in the sense that two people can attach the same sense to a given name. As he sees it, this is essential for successful communication: "One can hardly deny that mankind has a common store of thoughts which is transmitted from one generation to another" (p. 59). But see the footnote on p. 58 for a caveat.

What is Frege's argument that senses are not "ideas"? How much of the argument turns on special features of ideas as opposed to something else subjective that sense might be? That is: To what extent does the argument show that senses are not only not ideas but are not subjective at all? Perhaps more importantly: What does the way Frege argues here tell us about how he is thinking about 'sense'?

Frege then turns to the question what we should regard as the sense and reference, not of a name, but of a whole (declarative) sentence. This question may sound peculiar, and if one is thinking of the reference of a name strictly as its bearer, it can't but sound peculiar. So the very fact that Frege raises this question indicates that he is not thinking of reference just in that way. Rather, he has in mind a more general notion that Dummett calls 'semantic value'. (See the related reading.) One might think of this (as Dummett does) in terms of logic: What is the central fact about a sentence from the point of view of logic?

Frege says first that "Such a [declarative] sentence contains a thought", by which he means a certain "objective content, which is capable of being the common property of several thinkers" (p. 62). (So what Frege means by a 'thought' is something like a proposition.) Frege then argues that the thought "contained" in a sentence cannot be its reference. He then concludes that the thought must instead be the sentence's sense and goes on to argue that the reference of a sentence is just its truth-value. For our purposes, we need not consider the details of this argument, but my own view, for what it is worth, is that what lies behind this discussion is the role of truth-functional connectives in logic. (See the related paper by Heck and May.)

What is Frege's argument that the thought "contained" in a sentence cannot be its reference? To what assumptions about thoughts and references does it appeal?

Oddly, Frege does not seem actually to give any argument for the claim that the thought "contained" in a sentence is its sense. Some people therefore have suggested that by a "thought" Frege just means: sense of a sentence, so that "thought" for him is just defined as "sense of a sentence". Does that seem right? What alternative might there be? One might find inspiration for such an alternative in what Frege said earlier about indirect speech.

Russell is generally interested here in what he calls denoting phrases. The examples he gives are of two types: quantifier phrases, such as "a man" or "every man", and descriptive phrases, such as "the man with the yellow hat". The latter are the main focus of the paper, but it is important to understand what Russell is saying about the former in order to understand what he is saying about the latter.

Russell's central idea is "that denoting phrases never have any meaning in themselves, but that every proposition in whose verbal expression they occur has a meaning" (p. 480). Russell first explains this point in connection with quantifier phrases, and he claims that "Everything is F" means: F(x) is always true (i.e., is true for all values of x). The contrast here is with a view, which Russell seems to have held earlier, in The Principles of Mathematics, that "Everything" names a sort of variable entity. (He alludes to this view in the first paragraph.) So what Russell is saying is that quantifiers are not names of things. That is right, by our current lights, and much of the discussion on pp. 480-1 explains what is now standard fare in basic logic.

Here's a simple exercise: Translate Russell's verbal statements of what such sentences as 'C(all men)' mean into modern logical notation.

The way Russell puts this point might well be regarded as misleading, however, or even false. Some contemporary theories of quantification (which descend from Frege's) regard quantifiers like "every" as having meanings of their own without regarding them as being names of anything. I will explain this in class by presenting some of the basic ideas behind so-called generalized quantifers. For our purposes, however, the important point is that Russell is claiming that quantifiers behave very differently, from a logical point of view, from how names of things do.

Some authors (perhaps most famously Richard Montague) have suggested that names are also quantifiers. So we might write: (Socrates x)(x is tall), where "Socrates x" is a quantifier, and the sentence is true iff "x is tall" is true when 'x' has as value Socrates. So this is equivalent, in effect, to ∃x(Socrates(x) & x is tall), where 'Socrates(x)' is a predicate true of Socrates and no one else. In some ways, Russell's view is similar, in so far as he regards names as 'disguised descriptions' and holds that descriptions are quantifiers.

Russell turns at the bottom of p. 481 to descriptive phrases. His view, as he states it, but modernizing the notation, is that a sentence of the form "The F is G" means:

∃x[Fx ∧ ∀y(Fy → x = y) ∧ Gx]

I.e.: There is one, and only one, F, and it is G. As Russell notes, this incorporates an assumption that "The F" always involves an assertion of uniqueness. This can be questioned, and not just in cases of so-called plural descriptions (such as "the books on the table"). We will see Strawson raise questions along these lines later.

Russell's main point, however, does not concern the specific analysis displayed above (which, as Gareth Evans once quipped, simply massacres the surface structure of the sentence). The main point, rather, is that descriptions are a sort of quantifier, in this case, one might say, a "complex" quantifier. I'll explain in class, again, how this insight can be expressed in terms of generalized quantifier theory.

Important Terminological Note: You will see in our readings mentions both of the 'Theory of Descriptions' and of the 'Description Theory'. The former is the view Russell is expounding here: that descriptions (expressions of the form 'the F') are quantifiers. The latter is the view that Russell mentions in "Knowledge by Description": that proper names (like "Bismarck") are really 'disguised' or 'abbreviated' descriptions (so "Bismarck" really means: the first chancellor of the German empire). Although Russell accepts both of these views, they are independent: One could hold either without holding the other, and many people actually do accept the Theory of Descriptions without accepting the Description Theory. Strawson may have held the opposite combination of views: Accepting the Description Theory without accepting the Theory of Descriptions.

Russell sees the main alternative to his view as being a version of Frege's, who did indeed hold that descriptive phrases are names. So Frege thought that "the smallest prime number" referred to the number 2, but that this phrase also has a sense different, say, from "two".

In fn. 1 on p. 480, Russell remarks that the view he had held earlier, in The Principles of Mathematics, "is very nearly the same as Frege's". Most scholars nowadays think that is not true and that Frege's view was quite different. It's a nice question, therefore, whether Russell's criticisms of 'Frege' are really criticisms of Frege or rather of his own prior self. But we'll not concern ourselves too much with such historical issues.

Russell's first objection to Frege's theory is that, if a descriptive phrase has no denotation, then sentences containing that phrase should be meaningless. And Frege does indeed accept this consequence, to a point: Frege thinks, e.g., that if Homer never existed, then "Homer wrote the Iliad" is not false but without truth-value (i.e., it has no reference). But, Russell says, "The King of France is bald" is not meaningless but "plainly false".

Why does Russell think "The King of France is bald" is "plainly false"? Is he right? How can one tell? What is the bearing on this question of examples like "If there is a greatest prime number, then the greatest prime number is odd"? (Russell considers a convoluted version of this kind of example: the bit about "the u" on p. 484.) What should Frege say about that kind of case?

Russell gives another argument against Frege's account on pp. 485-8. This is known as the Gray's Elegy argument, due to an example Russell uses. The argument is extremely confusing, and I would not claim to understand it. There is some good work on this, however: See, for example, this paper by William Demopoulos or this one by Berit Brogaard. But this is more an historical issue, so I do not propose to spend much time on the Gray's Elegy argument, though you should at least skim it.

If you have any ideas about how the Gray's Elegy argument should be understood, feel free to express them!

Russell's main defense of his own theory consists in his arguing that it does a good job, and a better job than Frege's (or Meinong's), of solving three puzzles that he describes on p. 485. The first of these is the puzzle about identity that we have already seen in Frege. The second is a puzzle about excluded middle: Mustn't it be the case that either "The King of France is bald" or "The King of France is not bald" is true? The third puzzle, which Russell introduces using a very confusing example, seems mostly to concern claims of non-existence. So one might put it by asking how it could be possible to say, truly, e.g., "The greatest prime number does not exist". On Frege's theory, it again looks as if this ought to be without truth-value, since "The greatest prime number" does not denote anything.

In case you are curious, and don't remember, here's the simplest proof that there is no greatest prime. Suppose there is; call it p. Then consider p! + 1. This is not divisible by any number k such that 1 ≤ k ≤ p, since each of these divides p! and does not divide 1. So either p! + 1 is itself prime or it is divisible by some prime > p.

Russell explains on pp. 488-91 how his view resolves the three puzzles.

In the case of identity, the key move is to deny that "Scott is the author of Waverly" really is an identity-statment at all. Rather, that sentence means:

∃x(W(x) & ∀y(Wy → x = y) & x = Scott)

where W(x) means: x wrote Waverly. So Russell does not think that the sentence is really an identity statement at all.

What, then, should Russell do about, say, "Twain is Clemens"?

In the case of excluded middle, the solution rests upon a distinction between "primary" and "secondary" occurrences of a description, which Russell explains on p. 489. This is really an observation about scope. A 'primary' occurrence is an occurrence with the widest possible scope, whereas a 'secondary' occurrence is one with narrower scope. So, in particular, Russell regards a sentence like "The King of France is not bald" as ambiguous between these two readings:

(1) ∃x(KF(x) & ∀y(KF(y) → x = y) & ¬B(x))
(2) ¬∃x(KF(x) & ∀y(KF(y) → x = y) & B(x))

Note how, in the former, the negation is within the scope of the existential but, in the latter, it is outside it.

As Russell notes, (2) is really the negation of "The King of France is bald", not (1). So excluded middle does hold for that sentence; it's just that we have to be careful about which statement is its negation.

Explain why (1) is false and (2) is true.

The same distinction allows Russell to explain how "The greatest prime number does not exist" can be both meaningful and true. How so? And what should Russell do about "Homer did not exist"?

Russell concludes the paper by gesturing at epistemological implications of his theory, which we have already encountered in "Knowledge by Acquaintance and Knowledge by Description". But you can now see what Russell meant, in that paper, when he talked about the 'logical' considerations that also support his view.

Strawson is out to argue that Russell's Theory of Descriptions "embodies some fundamental mistakes" (p. 321). Strawson is particularly concerned to reject Russell's argument for the existence of "logically proper names" and, with it, his argument that 'ordinary' proper names are really descriptions (which, recall, is the Description Theory of Names).

As Strawson sees it, these arguments rest upon the claim that "if there are any sentences which are genuinely of the subject-predicate form, then the very fact of their being significant...guarantees that there is something referred to by the logical (and grammatical) subject" of those sentences (p. 323). So he will argue, by contrast, that a sentence can perfectly well be significant even if its "logical subject" fails to refer to anything.

One might worry here that Strawson is not distinguishing clearly enough between the Theory of Descriptions and the Description Theory: The former does not imply the latter. If we do distinguish these, however, then at which target are Strawson's arguments really directed? Or should they be understood as directed not at these theses themselves, but rather at arguments Russell gives for them?

On pp. 321-2, Strawson outlines two "obviously bad arguments" for the claim that the King of France must, in some sense, exist. Which premises of these arguments does he reject? And are these arguments "obviously bad"?

Strawson begins his argument against Russell, in section II, by making some distinctions, between:

• Sentences
• Uses of sentences
• Utterances of sentences

And similarly for other expressions, e.g., names or descriptions. The first and last should be fairly clear—though note that an utterance is an action that someone performs, one that involves the intentional 'production' of a sentence, which may be spoken, written, signed, etc. It is not so clear what Strawson means by a 'use'. Fortunately, this does not seem to be a particularly important notion for Strawson: His various claims can, so far as I can see, be stated just in terms of sentences and utterances.

It is common nowadays to distinguish sentences types from sentence tokens. To understand the distinction, consider the list above. How many words are there in that list? In the type sense, there are four words: "sentences", "uses", "of", and "utterances". In the token sense, there are seven words. What Strawson means by "sentences" are thus types; what he means by "utterances" corresponds to tokens. As for 'uses', I think we should say that people who make two utterances of the same (type) sentence make the same use of it if those utterances have the same meaning as so used, or if those uses express the same proposition (see p. 327).

Perhaps the most important thing in this paper is a distinction between two senses of the verb "refer", which emerges at p. 326. So far, we have been speaking of reference as a relation between words and things: "Hesperus" refers to Venus. But one can also speak of reference as an act: To whom were you referring? Strawson seems to want to insist that the latter is really more fundamental: "'Mentioning', or 'referring', is not something an expression does; it is something that one can use an expression to do", he says (p. 326). By contrast, meaning or significance is something an expression (e.g., a sentence) can have (p. 327).

Overall, then, Strawson thinks Russell conflates the question whether, say, "The King of France is bald" is meaningful with the question whether the subject-phrase, on some particular occasion of utterance, refers to anything: The latter question cannot sensibly be asked about the expression (type), but only about utterances of it (tokens). What the sentence means is supposed to be revealed by the "general directions" regarding its use, on various occasions, to say different things about, potentially, different objects (p. 327).

Strawson uses a variety of examples to press this point, one of which involves "I". Another involves what have come to be called "incomplete descriptions", such as: The table is covered with books. These pose a challenge to Russell's theory because it is obvious that there is not a unique table.

Why exactly do incomplete descriptions pose a problem for Russell?

In Russell's reply (which you do not have to read), he insists that Strawson is confusing different issues. Russell never meant his theory to address issues concerning what he calls `egocentric' words, of which "I" is the best example. (Russell means words whose reference can change with the occasion of use, words that are 'context dependent', as it is put nowadays.) We can get away from such issues by considering, e.g., "The King of France in 1905 is bald". Do Strawson's arguments still work as applied to such a case?

Strawson goes on to insist, however, that if someone does seriously use a sentence containing a descriptive phrase without thereby referring to anything, then they are "not making either a true or a false assertion" (p. 329); that is, they are not saying anything at all. In that sense, Strawson is accepting something Russell seemed to find absurd: Russell insisted that, if someone were seriously to claim, right now, that the King of France is bald, then they would have said something that was not meaningless but "plainly false". Strawson is saying that, yes, the sentence they uttered was not meaningless, but their utterance, in effect, was meaningless: They made no significant assertion.

This seems to return us to a question I asked earlier about Russell: Why does Russell claim that "The King of France is bald" is plainly false? But we now face a more general question. Forget about whether Russell or Strawson is correct. What kinds of considerations might possibly be brought to bear to help us decide this kind of issue?

In section III, Strawson goes on to explain his view that (contemporary) utterances of "The King of France is bald" somehow 'misfire'. Strawson claims that someone who uttered this sentence would thereby "imply" that there was a (unique) King of France. But he notes that "The King of France is not bald" has the same 'implication', and that suggests that this is "a very special and odd sense of" implication (p. 330). It is, in fact, a version of what has come to be called presupposition.

Thus, if I were to ask you, "Have you stopped smoking crack?" you might reasonably want to refuse to answer either "yes" or "no" because, as a lawyer might say, the question presupposes facts not in evidence. In particular, it presupposes that you have smoked crack in the past. But if both "You have stopped A-ing" and "You have not stopped A-ing" presuppose that you used to A, then this cannot be an ordinary logical implication: If both P and ~P imply Q, then Q is itself a logical truth. And surely it is not a logical truth that you used to smoke crack!

Can you think of other natural examples of words, like "stop", that carry presuppositions?

Similarly, Strawson wants to say that, if one utters "The King of France is bald", one does not assert but only presupposes that there is a (unique) King of France. So it is no part of what one is claiming when making such an assertion that there is a (unique) King of France, and the non-existence of such a monarch does not imply that your utterance is false. Rather, in that case, as Strawson puts it, the question whether the King of France is bald simply does not arise.

How satisfying is this suggestion that some assertions involving descriptions 'misfire'?

Strawson goes on to illustrate this point with what has become a very famous example: Someone says, "This is a fine red one", when there is nothing to which "this" might plausibly refer. In this case, Strawon suggests, one might reasonably want to say that nothing has really been said by the person who uttered this sentence.

Is this example really of the same sort as the one involving "the King of France"? In what ways might they be similar or different?

As mentioned earlier, it's critical to distinguish the Theory of Descriptions from the Description Theory of Names, and in many ways Strawson is focused on the latter more than the former. As it happens, there are other arguments in favor of the Theory of Descriptions, namely, that descriptions can take scope with respect to other operators, such as tenses. Consider, for example, the sentence: The president of the United States will one day be a woman. I contend that this sentence is ambiguous. How so? In what sense is this a scope difference?

Donnellan's main claim in this paper is that there is something wrong with both Russell's and Strawson's account of descriptions, because both of them account for at most one of the two uses of descriptions that Donnellan distinguishes. That said, in many ways one might think of Donnellan as trying to build on Strawson's central point about the importance of referring as an act.

The central contribution of this paper is the distinction between attributive and referential uses of descriptions. One uses a description attributively when one wishes to speak about whomever, or whatever, satisfies it; in this case, one may not know which thing satisfies it. By contrast, one uses a description referentially when there is a specific object about which one wishes to speak, and one only uses the description in order to help one's audience identify the object in question. Only in this case would it make sense to ask, "Who do you mean?" or "Who are you talking about?"

Make absolutely sure you understand this distinction. One good way to do that is to come up with some examples of your own to illustrate it. You might also pay attention as you read non-philosophical material and see if you can find some good, independent examples of each use. If you're having trouble, ask about it on Canvas.

As Donnellan is at pains to emphasize (pp. 285ff), this is a feature of uses of descriptions, not a matter (as Strawson seems to imply) of where they occur in a sentence (as subject or as predicate). Thus, Donnellan suggests that someone who uttered "Smith's murderer is insane" could be using the description either way, depending upon what sort of thing they were trying to say.

Obviously, "Smith's murderer" is not a definite description in the strict sense of a phrase of the form "the F". But it is common to treat it as a description nonetheless, having the same meaning as "The murderer of Smith". Exactly what the relation between these two expressions might be is a matter for linguistic theory and not one we shall pursue.

Donnellan uses a variety of locutions in trying to capture what is special about the referential use: He talks of the thing that is "meant"; of something that the speaker has "in mind". Are these helpful? Why or why not? Or better: In what ways are they helpful and in what ways are they not?

Perhaps the most important difference between these uses, however, appears when we assume that the description is "improper", i.e., that there is no object that satisfies it (pp. 286ff). Donnellan claims that, if we utter "The F is G" using the description attributively, and it is improper, then there is no sense in which anything has been said to be G. But if the description is used referentially, then one might still have managed to refer to something and to say of it that it is G, even if that thing is not F at all. This is a phenomenon that neither Strawson nor Russell seems to have envisaged.

It's common to distinguish two ways in which a description can be improper: It can be satisfied by nothing, or it can be satisfied by more than one thing. Strawson does discuss cases of the latter sort (so-called 'incomplete' descriptions). Could what he says about those cases be extended to cases of 'empty' descriptions (as we might call them)? If so, how different is Donnellan's view really from Strawson's, in the end?

Donnellan notes (pp. 288-9) that, although there is, in both of these cases, some sort of implication or presupposition that something fits the description, what is implied or presupposed depends upon the kind of use being made. In the attributive case, the implication or presupposition is general: One is implying or presupposing that something fits the description uniquely. In the referential case, by contrast, one is implying or presupposing that some particular thing—namely, the thing about which one wishes to speak—fits the description uniquely.

In section IV, Donnellan extends these claims in two directions. First, he argues that the question whether a description is used referentially or attributively cannot be reduced to the question whether the speaker knows of some particular thing that it fits the description uniquely (p. 289). Second, he argues that it is possible to use a description referentially even if one believes that the object to which one wants to refer does not fit the description, or even that nothing does (pp. 289-90). As Donnellan notes in section V, this latter point qualifies the earlier one about presuppositions (p. 291): In the referential case, it would seem, there is only a "presumption", which can be over-ridden, that the object about which one wishes to speak fits the description.

In section VI, Donnellan argues that his results are inconsistent both with Russell's theory and with Strawson's. Can you briefly summarize why? Which of the two theories would you suppose would have a harder time accomomdating the referential–attributive distinction?

In the last couple sections of the paper, Donnellan attempts to say something about the nature of the distinction he has isolated. In section VII, he denies that it is any sort of ambiguity, either syntactic (or 'structural', e.g., a scope difference, as with "Someone loves everyone") or lexical (which Donnellan calls 'semantic', meaning a difference in the meanings of the words, as with "bank"). He suggests we might think of the ambiguity as "pragmatic", a matter of the speaker's intentions, but it is not clear what this might mean.

The issue here is a very general one. Both Russell and Frege, one might say, tie the proposition the speaker expresses very tightly to the meaning of the words the speaker uses. Strawson to some extent tries to open up some space between these, for example, in the case of incomplete descriptions, such as "The table is covered with books". One might suggest that Donnellan is going even further in this direction. How so?

In section VIII, Donnellan suggests that we might understand the referential use in terms of a notion of "say[ing] something true about someone" (p. 298). Thus, even if Jones did not murder Smith, we might say, in the original example, that we referred to Jones and said of him that he was insane. And it seems as if Donnellan thinks this is the best we can do. He seems to want to deny that any "statement" is made in such cases (i.e., that any definite proposition was expressed), but seems to want to insist that the best we can do is to use the said-of locution. This sort of idea is developed yet further in the final section, in which Donnellan suggests that there is something fundamentally right about Russell's idea that, when we genuinely refer to something, we do so without thereby ascribing any specific properties to it. (Of course, having referred to it, we presumably will then ascribe some properties to it.)

Suppose that the best we can do, in cases of referential use, is to say that the speaker said of a particular object that it is G. Suppose, moreover, that in characterizing what the speaker said, we ourselves can refer to the object to which the speaker was referring using any tool at our disposal: any name of the object, e.g. It would be natural to understand this as a challenge to Frege. Why? (Hint: Consider a case in which, say, "the morning star" is used referentially.)

Kripke aims to argue "that the considerations in Donnellan's paper, by themselves, do not refute Russell's theory" of descriptions (p. 100). He remains officially neutral on whether Russell's theory is correct (though he tentatively suggests that it probably isn't).

In the first section, Kripke summarizes Donnellan's view and the arguments he gave for it. In the second, he considers some applications other authors have wanted to make of Donnellan's distinction. As mentioned above, you can skip that section (since it presupposes material we have not yet read).

In section 2, Kripke uses some notation from modal logic. We will discuss that later; for now, you needn't worry about it (since you can skip this section). If you're curious, the diamond means "it is possible that" and the box means "it is necessary that".

Kripke's early discussion, in section 3, presses on Donnellan's unclarity, towards the end of the paper, about (i) what "statement" someone makes when they use a description referentially and (ii) exactly what kind of distinction the referential–attributive distinction is. Kripke insists that if the distinction is merely "pragmatic", and is not any sort of syntactic or semantic (lexical) ambiguity, then it is hard to see why Russell should be concerned. Russell meant to be analyzing the meaning of descriptive statements, not the different "uses" one might make of them.

If there are parts of the discussion in 3(a) that you do not fully understand, you should of course feel free to ask about them. But the real core of the paper is what follows, and that will be our focus.

The key to Kripke's analysis of the phenomenon Donnellan uses to motive the referential–attributive distinction (henceforth, the "D-phenomenon") is the distinction he draws on p. 111 between speaker's reference and semantic reference. The distinction is inspired by, and is arguably a special case of, H.P. Grice's distinction between (i) what one says when one makes a given utterance (i.e., what the words one used mean) and (ii) what one means, which is connected to one's broader purposes in 'saying' what one did. The semantic reference of an expression is fixed by linguistic facts or conventions and (ignoring things like tense and demonstratives, for the moment) does not vary from occasion to occasion. The speaker's reference, on the other hand, can vary: It is "that object which the speaker wishes to talk about, on a given occasion, and believes fulfills the conditions for being the semantic referent of the designator" (p. 111).

An 'idiolect' is like a dialect, except that it is the language of one person (e.g., English as I personally speak it). There are large issues in philosophy of language, to which Kripke briefly refers, about the relation between idiolects and 'public' languages, such as English. Those issues will not concern us, and it will not hurt just to replace 'idiolect', for the most part, with 'language'.

It is tempting to borrow from Grice and say that the speaker's referent is the object one means. It would be worth exploring whether some such equation can be made to work, i.e., to try to show as precisely as possible how Kripke's distinction can be seen as a special case of Grice's. (This suggestion is only intended for students who have some familiarity with Grice's distinction.)

Kripke suggests further that the reason speaker's reference can come apart from semantic reference is because one can use a name with two sorts of intentions. On the one hand, one can use the name with the sole intention of referring to its semantic referent. In that case, the speaker's reference is guaranteed to be the same as the semantic referent. On the other hand, however, one can intend to refer to some person, say, that one can see, and thinking that this person is NN, one might then go on to use "NN" in saying something about them. If one is wrong about who that person is, then the speaker's reference will not be the same as the semantic reference, but will be the person one can see.

One of Grice's "conversational maxims" is the so-called "maxim of quality", which says that one should only say what one has reason to believe is true. Let's accept this as a general (but not strict) rule of conversation. Then if someone says "The F is G", we may ask what sorts of reasons they might have to believe what they are saying. Broadly speaking, those reasons divide into two sorts. One might have general reasons to believe that the F is G: One might think anything that is F is G and also think that there is exactly one F. Or one might have particular reasons: One might think that some particular thing x is G and also think that x is the one and only F. How might this observation be used to reinforce Kripke's point?

Perhaps one of the best known features of this paper is the "test" that Kripke describes on p. 113:

If someone alleges that a certain linguistic phenomenon in English is a counterexample to a given analysis, consider a hypothetical language which (as much as possible) is like English except that the analysis is stipulated to be correct. Imagine such a hypothetical language introduced into a community and spoken by it. If the phenomenon in question would still arise in a community that spoke such a hypothetical language (which may not be English), then the fact that it arises in English cannot disprove the hypothesis that the analysis is correct for English.

Kripke then introduces three "Russell languages", in which various stipulations are made about how descriptions behave semantically (pp. 113-4). He goes on to argue that the D-phenomenon would arise in all three of those languages.

Make sure you understand why Kripke's test is supposed to be a good one and how it is supposed to apply to the referential–attributive distinction.

Maybe the most striking of these arguments is one Kripke makes about quantifiers: Even what are obviously quantifier phrases can be used to illustrate the D-phenomenon. Thus, one might say (in the party case) "Someone over there is drinking champagne, and he's having a great time". Kripke's claim is that, in such a case, one could still 'refer' to a particular person who was not drinking champagne, and say of them that they were having a great time. If so, then the D-phenomenon cannot show that descriptions are not quantifiers, since the phenomenon occurs with quantifers such as "Some".

Kripke concludes this part of the discussion with the remark: "Since the phenomenon Donnellan cites would arise in all the Russell languages, if they were spoken, the fact that they do arise in English, as actually spoken, can be no argument that English is not a Russell language" (p. 115). Explain exactly what Kripke means.

Kripke allows that it is possible that descriptions behave ambiguously, and he introduces the D-languages to illustrate what this hypothesis would involve. He then argues on very general, methodological grounds that we should prefer the hypothesis that English is a Russell language.

In some ways, Kripke's entire case against Donnellan is contained in the example he gives on p. 111 about Smith and Jones. What this example purports to show is that the D-phenomenon has nothing special to do with descriptions, since it can also arise with proper names. Why is it such a problem for Donnellan if that is true?

Kripke does not apply his "test" to what he calls the "unambiguous D-language". If we did, I claim, we would find that the D-phenomenon arises within the unambiguous D-language itself. I claim further that this is a more powerful reason than the ones Kripke gives to think that English is not a D-language. See if you can spell out this argument. (This may be a version of the considerations that Kripke gives on pp. 118-9.)

Kripke also gives a complex argument against Donnellan involving pronouns (pp. 121-2). If you have some experience with linguistics, you might benefit from working through this argument; if not, feel free to skip it.

Quine is interested in the question to what extent we can make proper sense of the notion of necessity, and what philosophical commitments we must make to do so. And this question is tied up, for him, with the question what our logic of necessity should be like.

There are, Quine says, three ways we might treat necessity logically.

1. We may regard it as a property of sentences.
2. We may regard it as a "statement operator", attaching to statements to make other statements.
3. We may regard it as a "sentence operator", attaching, like negation, to a formula to make another formula.

The difference between the second and third is that, in the latter case, the formula may contain free variables, which may later be bound by quantifiers. Thus, at the third grade, one can write things like: ∃x(Fx ∧ &neg;□Fx). This is not possible at the second grade, since the box can only be attached to a 'statement' (meaning a formula without free variables).

What Quine calls a 'sentence operator might better be called a 'formula operator', since the term 'sentence' is usually used to mean closed formula, i.e., formula without free variables.

Here's a little exercise to check your understanding. On p. 157, Quine lists three examples of the use of a 'sematical predicate', (1)-(3), but then only two examples of the use of a 'statement operator', (4)-(5). What happened to the other one?

Quine's overall point may be summarized this way. Step (1) is relatively harmless. We do this in logic anyway when we speak of a sentence as being valid. Step (2) is mostly harmless, and it is completely harmless if understood as a notational variant of step (1). But it is dangerous in so far as it encourages one to move to step (3), which Quine regards as philosophically suspect.

In section I, Quine introduces the notions of purely referential occurrence, referential opacity, and truth-functionality.

• An occurrence of a term is purely referential if one can substitute any other name of the same object without change of truth-value (salva veritate).
• A 'context' in which a sentence might occur is opaque if occurrences of terms in that context, which otherwise would have been purely referential, become not purely referential. Quine notes that "Quotation is the referentially opaque context par excellence" (p. 159).
• A context is truth-functional if substitution of sentences with the same truth-value preserves the truth-value of the whole.

Examples:

• Superman can fly.
  The occurrence of the italicized term is 'purely referential', since replacing it with "Clark" does not change the truth-value.
• Lois believes that Superman can fly.
  This occurrence of "Superman" is not purely referential, since replacing it with "Clark" does change the truth-value.
• Superman can fly and _____
  This is a 'transparent' context, since substitution of co-referential terms will be fine in whatever replaces the blank. (Unless, of course, it was already not fine. But Quine is careful to state the condition so as to allow for that.) It is also a truth-functional context.
• Lois believes that _____
  This is an opaque context, since if we fill the blank with "Superman can fly", we cannot replace "Superman" with "Clark" without a change of truth-value. So a referential occurrence has become non-referential by being embedded in this context. This is also not a truth-functional context.

Can you give an example of a sentence containing two occurrences of the same name, where one occurrence is purely referential and the other occurrence is not? (This possibility explains why we need to talk about referential occurrences and not just whether a name occurs referentially.)

In the examples just given, the referentially opaque contexts are also non-truth-functional contexts. So one might wonder: Are there any contexts in which you can substitute equals salva veritate but that are not truth-functional? On pp. 161-2, Quine presents an argument known as the Slingshot, which purports to show that there are not. (The argument has antecedents in Frege but probably first appears in Alonzo Church.) More precisely, the Slingshot is supposed to show that any context that (i) permits substitution of equals salva veritate (i.e., is not referentially opaque) and (ii) also permits substitution of logical equivalents, salva veritate, is truth-functional.

The argument is extremely simple, but Quine's notation may be confusing, so here it is a slightly different version. Suppose that p and q have the same truth-value. Observe that:

1. The statement "The x such that [(x = Venus & p) or (x = Mars & not-p)] = Venus" is logically equivalent to p.
   If p is true, then the left-hand side refers to Venus; if it is false, then to Mars. So the statement will have the saem truth-value as p, either way.
2. The statement "The x such that [(x = Venus & q) or (x = Mars & not-q)] = Venus" is logically equivalent to q.
3. If p and q have the same truth-value, then the two terms "The x such that [(x = Venus & p) or (x = Mars & not-p)]" and "The x such that [(x = Venus & q) or (x = Mars & not-q)]" refer to the same thing: Venus, if p and q are both true; Mars if p and q are both false.

Quine's own argument could be re-written using set-theoretic notation: {x: x = ∅ ∧ p} and {∅}. These will be equal if p is true and not equal if p is false. But this would not yield the logical equivalence of p with "{x: x = ∅ ∧ p} = {∅}", since that involves at least some facts about sets (e.g., extensionality). Indeed, one might question whether Quine's presentation, which proceeds in terms of 'classes', really yields logical equivalence.

Now suppose that F(p) is a context satisfying assumptions (i) and (ii). Then the following must all have the same truth-value:

1. F(p)
2. F(the x such that (x = Venus & p) or (x = Mars & not-p) = Venus)
3. F(the x such that (x = Venus & q) or (x = Mars & not-q) = Venus)
4. F(q)

(i) and (ii) are equivalent by substitution of logical equivalents, justified by (a); (ii) and (iii) by substitution of equals, justified by (c); (iii) and (iv) by are equivalent by substitution of logical equivalents, again, justified by (b).

What do you think Russell might say about the Slingshot?

In the context of the paper, the primary lesson of this argument is supposed to be that "quantifying into" referentially opaque contexts is a troublesome matter. Quine will turn to that argument shortly.

In section II, Quine notes that we need necessity-like notions in logic, namely, validity and implication, but insists that these are properly understood as properties of or relations between sentences. He then argues that, with sufficient care taken about quotation, the use of necessity as a statement operator (attaching only to sentences, i.e., to formulae with no free variables) can be explained in terms of the 'semantical' use: "nec(9 > 5)"—or □(9 > 5)—just means: Nec('9 > 5').

If you took PHIL 0640, much of what Quine says here may sound familiar. That is becuase the book we use, Goldfarbs's Deductive Logic, was written as an updated version of Quine's textbook Methods of Logic and shares much of Quine's philosophical orientation.

By the way, the 'alternation' of two formulas is their disjunction; and Quine uses (x) for the universal quantifier, rather than ∀.

On the other hand, however, Quine notes that iterated modalities, which are characteristic of modal logic, are not easy to make sense of on this approach. The translation works, but it is not clear what the translated sentences really mean. Quine thinks, then, that, if we explain necessity as logical necessity, that does not really suffice to explain the use of "nec" as a statement operator. In that regard, then, he thinks that even the second grade of modal involvement may not be well-motivated.

Exercise: Explain in your own words why Quine thinks it just isn't clear what it would even mean to say that
(36) Nec'(x)(x is red → x is red)'
is valid. (The examples leading to (42) are even more complicated.)

In section III, then, Quine turns to the use of "nec" as a sentence operator. With this, we allow such constructs as "(∃x)(nec(x > 5))", meaning: there is a thing that is necessarily greater than 5.

Quine notes first that, with this move, we make the translation of "nec" to "Nec" impossible, since the variable would have to appear within quotes.

Make sure you understand why 'quantifying into' quotation marks is so problematic.

See Quine's paper "Quantifiers and Propositional Attitudes" for the classic discussion of quantifying in.

A more serious issue, however, is that the context "nec(...)" appears to be referentially opaque, since "nec(9 > 5)" looks true and "nec(the number of planets > 5)" looks false. If so, however, then, as Quine says, "nec(x > 5)" simply isn't true or false of an object at all. (As he tends to ask elsewhere: Which thing is it that is necessarily greater than 5? Is it 9? I.e., the number of planets?) If so, then the value assigned to "x" can't simply be an object.

Articulate this point of Quine's a bit: If we assign "x" the value 9, then is "nec(x > 5)" true or false? Why is the question difficult to answer? Or, at least, why does Quine think the question is difficult to answer?

The response Quine envisages is that "nec(the number of planets > 5)" is ambiguous as to the scope of the description: See (49) and (50).

Explain Quine's point using Russell's notions of primary and secondary occurrence. What would this imply about substitution?

Quine notes on p. 173 that, if we are going to quantify into "nec", then we need to be careful about how we apply universal instantiation. What restriction do we need? (Hint: How do we have to read (53) to make it true?) Can we formulate the required restriction using Russell's notions of primary and secondary occurrence? Why or why not?

Quine concludes by arguing that, if we accept quantification into "nec", then we are committed to "Aristotelian essentialism", i.e., to the doctrine that there is a distinction to be drawn between properties of an object that are essential to it and those that are merely accidental. In particular, Quine notes, it looks as if we are committed to the claim that "being greater than 5" is a property that 9 has essentially, but "being the number of planets" is a property it has only accidentally.

Quine does not say why he regards Aristotelian essentialism as problematic. What sorts of worries do you think he might have? (See the last paragraph for some hints.)

If we decide to quantify into "nec", then "nec(...)" is a context that is not referentially opaque and, presumably, allows substitution of logical equivalents salva veritate. Why doesn't the Slingshot then imply that "nec(...)" has to be truth-functional? (Hint: Russell again.)

There are four crucial things to understand in this lecture:

1. The distinction between necessity and a priority
2. The distinction between rigid and non-rigid designators
3. The distinction between a description's "giving the meaning" of a name and its merely "fixing the reference" of that name
4. The argument Kripke gives against the description theory of names, where that means the theory that a description 'gives the meaning' of a name

The first two of these we'll discuss this time; the last two, we'll discuss next time.

I'd suggest writing, just for yourself, at least a couple sentences about each of these, as a way of checking your own understanding.

On pp. 22-34, Kripke quickly summarizes a lot of the history we have just studied. Is there anything about this description that strikes you as wrong? or perhaps as biased? (I'm not necessarily saying there is anything wrong or biased. Then again, I'm not saying there isn't.)

On pp. 34-9, Kripke introduces the important distinction—which first appears in these lectures—between what is a priori and what is necessary. As he notes, these had largely been conflated. The key point here, really, is that the two notions apply to two different things. The term "a priori" should be regarded as an epistemic adverb: It denotes a way of believing or knowing something. So we can say that someone knows a priori that 2+2=4, say. What is necessary, by contrast, is a proposition (or something along those lines), and whether a proposition is necessary does not depend upon what anyone knows.

To tease these apart, Kripke gives a few examples. Kripke remarks that one could know a certain mathematical fact because it has been proven by a computer. But that knowledge would not, he claims, be a priori, though the mathematical fact, if true, would of course be necessary, like any mathematical fact. More strongly, if Goldbach's conjecture is true, it is necessary; but perhaps there is no possible proof of it, and it is not even possible (for a person at least) to know it a priori.

It is not obvious that one does not know mathematical facts that are proven by computers a priori, and Tyler Burge has argued the opposite, in "Computer Proof, A Priori Knowledge, and Other Minds" (PhilPapers). But there are better examples. Can you come up with a case in which someone might have empirical verification even for what turns out to be a logical truth?

As Kripke notes, this does not by itself show that there isn't some connection between a priority and necessity. It does show that these are different notions and that, if there is such a connection, it needs to be established by argument, not just assumed.

Kripke then turns to the sorts of issues raised by Quine about de re and de dicto necessity, or essentialism. He first argues that the notion of an essential property (a property an object necessarily has) is well-grounded intuitively (pp. 41-2). For example, it seems obvious that Barack Obama might not ever have been the US President—that, as Kripke keeps saying, he might not have been—even though, if we describe him as "the first Black President", then it's equally obvious that it's a necessary truth that the first Black President was at some time the President.

The sentence "The first Black President might not have been the President" exhibits a scope ambiguity. How could it be represented? Why does pointing this out not help Quine?

Kripke then turns, on pp. 42-53, to a set of questions about the nature of possible worlds and our cognitive access to them. (These issues will not be our focus, but it is important to understand what is going on here.) Kripke's main target is David Lewis, who holds an extreme form of "modal realism" (see his later book On the Plurality of Worlds) according to which (i) possible worlds are something like "alternate universes", containing real people, e.g., just like you and me, and (ii) for every way things might have been, there is a possible world (an alternative universe) in which things are that way. So, according to Lewis, there is a possible world in which Mitt Romney won the 2012 US election and became President of the US, and another in which he won but was assassinated by a crazed former president who went insane after his affair with an intern was revealed. But note that, in these worlds, it isn't really Mitt Romney who wins. That's because Mitt lives in our universe, not in some alternate universe, and that raises the question of "transworld identity": How do we know which of these other people in this other universe is Mitt's "counterpart"? It is hard to see what other sort of answer might be given than one that involves some sort of qualitative similarity.

If transworld identity has to be explained in terms of qualitative similarity, then that might suggest that every proper name must be associated with "purely qualitative necessary and sufficient conditions for being" its referent (p. 46). And that seems, at least, to be very friendly to the description theory of names. Why?

It's against this view that Kripke argues. His central claim is that we should not think of possible worlds as "alternate universes", so that we have to figure out who is playing the Mitt Romney character in one. Rather, we should think of them simply as situations we imagine and that include Mitt Romney by stipulation. (It might be better if we called these possible situations, since they are usually just fragments of complete worlds. And whether imagination is the right faculty is not obvious, but it helps to fix ideas.) So, Kripke claims we can simply consider Mitt himself and ask: Could he have won the 2012 election? Note that this question is not supposed to be epistemic: Could it turn out that Romney actually won? We're assuming he lost and asking if it's possible that he could have won.

Kripke suggests that Nixon could not have been an inanimate object and, perhaps, might necessarily have been human: There is no possible world in which Nixon is (say) a wooden table; perhaps there is no possible world in which he is a robot, or a gorilla. Explain in what sense this is a different claim from the claim that Nixon could not turn out to have been a robot or a gorilla.

Kripke then introduces the second really important distinction in this lecture: between rigid and non-rigid designators. A rigid designator, he says, is one that always denotes the same object in every possible world. So, he says, "9" always picks out the same thing in every possible situation, namely, the number 9, whereas "the number of planets" could pick out different numbers in different situations. Note that the claim here is not that "9" couldn't have meant something different and that, as it would have been used in some different situation, it couldn't have referred to something else. Rather, Kripke's claim is that when we use "9" to describe a possible situation, it always refers to 9; whereas we can use "the number of planets" in describing a possible situation and use it to refer to the number of planets in that situation, which need not be 9.

(There is then a brief return to questions about transworld identity, which elaborates some of the earlier themes. You should read these parts, but they will not be our main focus.)

On p. 53, Kripke turns to criticisms of the description theory. First, however, he introduces his distinction between using a description to "give the meaning" of a name and using it only to "fix the reference". Kripke never really explains this distinction very clearly, simply illustrating it with examples. But what he seems to have in mind is that, if one uses a description to "give the meaning", then the name is synonymous with the description and functions largely as an abbreviation for that description (see the top of p. 32). If, by contrast, one uses the description just to "fix the reference", then the name is not an abbreviation for the description.

Kripke explores this distinction (on pp. 54-7) using the example of the standard meter stick, but there are real-life examples that can be used instead. One, noted by Sir Michael Dummett, is the name "St Anne", which was stipulatively introduced as a name for the mother of the Virgin Mary. But, Kripke says, we do not use "St Anne" as an abbreviation for "the mother of Mary". If we did, then "St Anne had a child" would be equivalent to "The mother of Mary had a child" and so would be a necessary truth. But St Anne, like anyone, might have died in infancy, succumbing to a childhood illness, and then she wouldn't have had a child. So "St Anne had a child" is no necessary truth, and the description is only being used to fix the reference, not to give the meaning. Note, on the other hand, that "St Anne had a child" is arguably a priori: Anyone who understands the name "St Anne" knows that she had a child (assuming that Mary was a real person, etc). So this is an example of the sort of statement that Kripke claims is contingent even though a priori.

Although Kripke is mostly interested in 'simple' modal statements like "St Anne might have died in infancy", counterfactual conditionals also play a role in his argument. These, in English, take the form "If so-and-so had happened, then such-and-such would have happened". (Thus, they are sometimes also called 'subjunctive' conditionals.) These are not truth-functional. A standard contrast is between "If Oswald did not kill Kennedy, then someone else did" (true: someone killed him) and "If Oswald had not killed Kennedy, then someone else would have" (not obviously true). Thus, Kripke remarks that something like "If heat had been applied to stick S at time t, then it would not have been one meter long" seems true: If so, then stick S could have been other than one meter long (and would have been, had heat been applied to it), so it is not necessary that it is one meter long.

Note here again that Kripke's point is that when we use the name "St Anne" to describe a possible situation, we always use the name to refer to (our) St Anne. He is not saying that, in other worlds, people could not use the name "St Anne" to refer to a different woman, and if they fixed its reference the same way we do, then they might. (This particular example is complicated by some claims Kripke wants to make about essence, but those claims are irrelevant to the example.)

Kripke does not deny, then, that there are at least some cases in which a description is used to fix the referent of a name. But he does deny that a description ever gives the meaning of a name. The argument is just a generalization of the one just given. What description might we associate with the name "Otto von Bismark"? Perhaps "the first chancellor of the 19th century German Empire". But, obviously, "Bismark was a politician" is no necessary truth (he too might have died in infancy), whereas "The first chancellor of the 19th century German Empire was a politician" looks to be a necessary truth. And, Kripke suggests, the same sort of thing will happen with any description one might choose, unless one just happened to choose one that expressed an essential property of the object. (Indeed, it is not obvious that objects have enough essential properties that there will always be a description, given entirely in terms of essential properties, that picks it out uniquely.)

Kripke remarks on p. 58: "'Hesperus' rigidly designates a certainly heavenly body and 'the body in yonder position' does not—a different body, or no body might have been in that position, but no other body might have been Hesperus (though another body, not Hesperus, might have been called 'Hesperus')." Explain the contrast Kripke is drawing in the parenthetical at the end.

Kripke writes: "Frege and Russell certainly seem to have the full blown theory according to which a proper name is not a rigid designator and is synonymous with the description which replaced it" (p. 58). Historical issues are not our central focus, but one might wonder nonetheless whether this is true. Is it? What evidence does Kripke give for this claim? (The considerations in the rest of the paragraph might be thought to be such evidence. Are they?)

Kripke takes these considerations, then, to refute the "ordinary" description theory, construed as a theory of the meanings of names (and not just as a theory of what fixes reference). He then turns, on p. 60, to the cluster theory and argues that it succumbs to similar objections.

Kripke takes the cluster theory to consist of six theses:

1. The "cluster" that a speaker S associates with a name 'N' is the family of properties φ such that S believes 'φ(N)'.
2. One of the properties, or some conjointly, are believed by S to pick out some individual uniquely.
3. If most, or a weighted most, of the φ's are satisfied by one unique object y, then y is the referent of 'N'.
4. If the vote yields no unique object, 'N' does not refer.
5. The statement, 'If N exists, then N has most of the φs' is known a priori by the speaker.
6. The statement, 'If N exists, then N has most of the φs' expresses a necessary truth (in S's language).

Note, as Kripke says, that (6) will be part of the theory only if one thinks that the cluster "gives the meaning" of the name. There is also an additional condition, (C), that specifies that the account must not be circular.

Formulate Kripke's argument against thesis (6) of the cluster theory in your own terms.

There is a response to this argument that is nowadays fairly common: Instead of 'ordinary' descriptions like "the φ", we can use so-called 'actualized' or 'rigidified' descriptions of the form "the actual φ". So, in the case of "St Anne", the associated description would be "the actual mother of Mary" or, perhaps, "the woman who was Mary's mother in the actual world".

If "St Anne" abbreviates "the actual mother of Mary", then "St Anne had a child" abbreviates "The actual mother of Mary had a child". The former, as we saw, is contingent. Is the latter contingent or necessary? It may help to consider "The woman who was the mother of Mary in the actual world might have died in infancy". (Be careful about necessity vs a priority!!)

My own view is that the real purpose of Lecture I, overall, is to introduce the distinction between necessity and a priority and, with it in place, to eliminate the "strong" form of the description theory that takes descriptions to "give the meaning" of names. The real action is then in Lecture II, to which we turn next.

I am inclined to think that almost none of the authors Kripke cites ever held the strong form of the description theory. (An exception might be Lewis, whose motivations are very different from those of Russell and Searle.) Have a close look at the passage from Searle that Kripke quotes on p. 61 (and the surrounding remarks). If we assume that Searle (like everyone else at that time) is conflating necessity and a priority, then we need to ask what exactly Searle means here: We need to disambiguate his use of "necessary". What do you think he means?

In my view, Lecture II contains what are by far the most important of Kripke's arguments.

Recall that the cluster theory consists of six theses:

1. The "cluster" that a speaker S associates with a name 'N' is the family of properties φ such that S believes 'φ(N)'.
2. One of the properties, or some conjointly, are believed by S to pick out some individual uniquely.
3. If most, or a weighted most, of the φ's are satisfied by one unique object y, then y is the referent of 'N'.
4. If the vote yields no unique object, 'N' does not refer.
5. The statement, 'If N exists, then N has most of the φs' is known a priori by the speaker.
6. The statement, 'If N exists, then N has most of the φs' expresses a necessary truth (in S's language).

Thesis (6) is part of the theory only if one thinks that the cluster "gives the meaning" of the name.

Kripke begins Lecture II by re-emphasizing the importance of the additional condition, (C), that specifies that the account must not be circular.

As it happens, some philosophers have raised questions about this part of Kripke's discussion. If the theory in question were, "'Socrates' refers to the person to whom 'Socrates' refers", it would obviously be circular. But the theory is "'Socrates' refers to the person called 'Socrates'", and that is not obviously circular: Perhaps being called by a name isn't the same as being referred to by the name. Of course, one then wants to be told what 'being called by a name' is. This will not be on our line of inquiry, though.

He then summarizes the arguments given against thesis (6) in the first lecture, remarking that Searle's claim that it is necessary that Aristotle has some of the properties commonly attributed to him is false. This is clearly too strong, since being human is such a property, and Kripke himself regards this as an essential property. (This is not argued until Lecture III.) But this does not really matter: If thesis (6) is to be true, then it has to be necessary that Aristotle has enough properties to individuate him uniquely, and properties like being human are not likely to do that.

That said, recall that it is not obvious that anyone (with the possible exception of Russell) ever held the description theory in the strong form that includes thesis (6).

With thesis (6) disposed of, then, Kripke turns to the remaining theses, which constitute the cluster theory of reference-fixing. And Kripke allows that, for some names (Neptune, Jack the Ripper, St Anne), it may be a good account. But he claims that it does not give a good account of how reference is fixed for most names.

One important point to note, right away, is that these arguments do depend, whereas the arguments against thesis (6) do not depend, upon what the decriptions associated with a given name actually are. The modal argument shows (or purports to show) that no name is ever synonymous with a description, for any speaker, even names like "St Anne", where it is uncontroversial what description fixes its reference. None of the arguments in this lecture would or could refute that claim. They are directed more at 'ordinary' names, like "Aristotle" or "Gödel", where it is just not obvious what descriptions (if any) fix its reference. So the arguments will necessarily require some idealization or guesswork, and the option will always exist of replying: Oh, it's not those descriptions that fix the reference but these other ones.

Kripke gives two main arguments against the cluster theory. So far as I know, there are no standard names for these, but I call them the argument from ignorance and the argument from error. (Collectively, these are sometimes called the epistemic arguments, to distinguish them from the modal arguments given in Lecture I.)

In the argument from ignorance (pp. 80-2), Kripke claims that people can refer to an object even if they do not have enough information about that object to individuate it. Kripke uses the example of "Richard Feynmann", noting that many people who understand this name know little more about Feynmann than that he is a physicist, and do not know enough about him to distinguish him from many other physicists. And, indeed, this seems quite common.

Can you give some other examples like Kripke's "Feynmann" example?

Kripke presents this as an argument against thesis (2): that speakers generally believe that they have enough information to individuate an object. In fact, however, it seems better directed at thesis (4): that if the information a speaker has does not pick out a unique object, the name does not refer. What speakers believe about this sort of thing does not seem terribly interesting, let alone crucial.

In the argument from error (pp. 82-5), Kripke claims that people can refer to an object even if the information they have about that object fails to apply to anyone, or applies to someone else. Here, the central example is the Gödel–Schmidt example. Most people, Kripke says, probably believe no more about Gödel than that he proved the incompleteness theorem. But, he insists, even if that were (shockingly) to turn out not to be true true—if someone else proved the theorem and Gödel stole it from them—people who had 'heard' nothing else about Gödel than that he proved the incompleteness theorem would still use the name "Gödel" to refer to Gödel. So the reference, for a given speaker X, is not fixed by the information that X associates with the name.

Kripke also mentions some real-life examples, involving Peano (who did not discover the so-called Peano axioms); Columbus (who was not the first European to visit the Americas); and Einstein (who did not invent the atomic bomb). Can you think of other such examples?

When we make the judgement that someone who knew nothing else about Gödel than that he proved the incompleteness theorem could still refer to Gödel in the circumstances described, what kind of judgement are we making? On what basis do we make it? Is this a case where we are relying upon "intuition"? Do the real-life examples seem somehow different, in this respect, from the invented example about Gödel and Schmidt? Or are our reasons in these cases the same?

This argument is directed primarily against thesis (3): The information the speaker has does pick out an object uniquely, but it is the wrong one. It also constitutes an argument against thesis (5): Even if it is true that Gödel proved the incompleteness theorem, it is not a priori that he did so; and, Kripke claims, that is true even for a speaker who knows nothing more about Gödel than that he proved the incompleteness theorem.

It's important to note that Kripke is not claiming that, according to the cluster theory, the name "Gödel" would refer to Schmidt in the hypothetical situation for all of us, but only for those who 'know' only that Gödel proved the incompleteness theorem. Why not? In what way might this fact point toward a better view?

Note that this is a stronger claim than the one made in Lecture I: Kripke is not just saying that it is not a necessary truth that Gödel proved the incompleteness theorem, i.e., that there are other possible worlds in which he did not prove it; he is claiming that we can at least make sense of the suggestion that Gödel did not in fact prove the incompleteness theorem in the actual world, and that is so even if the only thing we 'know' about Gö is that he proved the incompleteness theorem.

By contrast, while of course Jack the Ripper might never have killed anyone (he might have died in infancy), we cannot make sense of the suggestion that Jack the Ripper did not actually commit the grisly murders. Why not?

Can you think of real-life examples where you yourself have very limited information about some historical figure and yet you can make sense of the idea that the information you have might be wrong?

One of the things that makes these arguments difficult to evaluate is that they depend, in a way the "modal" argument does not, upon exactly what sorts of descriptions one thinks ought to form part of the "cluster". The sort of view Kripke spends most of his time discussing is sometimes called "famous deeds" descriptivism, since the descriptions all seem to concern the "famous deeds" of the referent. On pp. 87-90, then, Kripke considers some other suggestions about what the descriptions ought to be. Do any of these seem promising? What other options might there be?

At the bottom of p. 90, Kripke turns to his positive account of how the reference of a name is fixed. The view he describes is an antecedent of the so-called "Causal Theory of Reference", though I'll register some caveats about that attribution below.

Kripke divides the question how a name refers to its bearer into two parts.

• First, there is an initial "baptism" or "dubbing", during which the reference of the name is originally fixed. Thus, the name "Richard Feynmann" originally referred to Richard Feynmann because it was bestowed upon him by his parents.
• Second, as the name is used in communciation, it is "passed from link to link", and the name "Richard Feynmann" refers to Richard Feynmann for those of us who were not present at the initial dubbing because of the complex chain connecting us to the original use.

Kripke offers this only as a "rough statement of a theory", and he is particularly cagey about the second condition, which needs careful statement.

It turns out, then, perhaps surprisingly, that Kripke's primary target is the view now known as "individualism": the view that what someone's words refer to is wholly determined by facts about that person as they are in themselves and, in particular, do not depend upon their social environment. Certainly Russell's descriptivism takes that form. But does descriptivism have to take that form?

Kripke mentions a number of problems that affect the second condition, the key examples being Napoleon the aardvark and the neighbor George Smith. Explain how these examples motivate the suggestion that, when one picks up a name from someone else, one will inherit the reference of that name only if one intends to use the name with the same reference as the original speaker. How well does this condition deal with the original problem?

Gareth Evans, in one of the related readings, offers the following counterexample to Kripke's analysis. It turns out that the name "Madagascar", as used to refer to an island off the coast of Africa, originally referred to a portion of the African mainland in Somalia, near Mogadishu. Even though everyone intended to be preserving the reference of the name, some confusion somewhere along the way led to its becoming a name of the island. Can you spell out in detail why this is a problem for Kripke?

Kripke remarks at one point that "A certain passage of communication reaching ultimately to [Feynmann] himself does reach the speaker" (p. 91). This suggests that, for Kripke, reference is sustained by a causal relationship between a speaker and the object of reference. In fact, however, this is misleading. Kripke allows that the reference of the name may originally be fixed by description, and he again gives the example of "Neptune" (p. 96, fn 42). The causal connection, then, is supposed to be not between the speaker and the object but between a speaker not present at the baptism and those who were. Devitt, in the related reading, is quite clear about this: Kripke's theory is a theory of reference transfer.

As Kripke notes (p. 92), the theory he proposes is somewhat similar to one suggested by Strawson. The difference is supposed to be that, on Strawson's account, you have to remember from whom you learned the name, so that "Gödel" might mean: the man Jones calls "Gödel". On Kripke's account, by contrast, you don't need to know from whom you learned the name. How significant a difference is this? (What is doing the work of securing refernce on Kripke's view?)

In fact, Kripke emphasizes that he is not giving an "analysis" or a "reductive account" of reference. This is in part because the second condition itself involves the notion of an intention to preserve reference.

On p. 97, Kripke begins a discussion of the modal status of identity statements, such as "Hesperus is Phosphorous". Kripke of course agrees that if the statements involve descriptions, then they can be contingent. If an identity statement involves names, however, then, Kripke argues, it is necessarily true if true at all: There is no possible world in which Hesperus is not Phosphorous. So Kripke endorses the "necessity of identity".

Ruth Barcan Marcus had claimed before Kripke that true identities are necessary, but she had also claimed that, as Kripke puts it, "...if you really have names, a good dictionary should be able to tell you whether they have the same reference (p. 101), so that it cannot have been an empirical discovery that Hesperus is Phosphorous. Why isn't Kripke committed to the same combination of views? (The remark Kripke cites in fn 47 is on p. 142 of the "Discussion" listed as a related reading for Quine.)

It's very important to see exactly what Kripke is and is not claiming here. He is not claiming that we could not find out that, shocking as it may seem, Hesperus is not Phosphorous after all. Nor is he claiming that there is not a possible situation in which "Hesperus" and "Phosphorous" are used much like we use them, but in which these names refer to different objects; so that, in that situation, the sentence "Hesperus is Phosphorous" would be false. What Kripke is claiming, rather, is that there is no possible situation that we could correctly describe as one in which Hesperus is not Phosphorous. This is because names are rigid designators: "Hesperus" always refers to Hesperus (i.e., Venus) when we use it, and "Phosphorous" always refers to Phosphorous (i.e., Venus). So it is not possible to use these names to describe a situation in which Hesperus would not be Phosphorous. To do so would be to describe a situation in which Venus was not Venus, which is surely impossible.

Early in this lecture (p. 77), Kripke mentions that the fact that the sentence "2 + 2 = 4" could have been false, and would have been false had the words been used differently, does not show that it is not a necessary truth that 2 + 2 = 4. Explain how this bears upon his claim that true identities are necessary.

Some philosophers have also endorsed the "necessity of diversity": the claim that every false identity statement is necessarily false. What kind of argument might one give for this claim? How does it compare to Kripke's argument for the necessity of identity?

Here's a case to think about. Mario and Aldo Andretti (famous race car drivers) are identical twins. Presumably it is possible that the zygote from which they both developed should not have twinned and so should have produced only one child. Consider, then, such a situation, and call the single child born in that situation "Ennio". Could one correctly describe this situation as one in which both Mario and Aldo were both Ennio and so were the same person? If so, then diversity is not necessary.

This paper has two main goals. The first is to raise a series of methodological questions about the role (allegedly) played by 'intuition' in arguments about reference. The specific focus, ultimately, is on Kripke's use of the Gödel–Schmidt case in his arguments against the description theory, but the worries MMN&S express are supposed to generalize far beyond that particular case. The second goal is thereby to undermine what MMN&S call 'arguments from reference': arguments given in areas of philosophy quite distant from philosophy of language that appeal to the sorts of conclusions for which Kripke argues (and also Putnam and Burge, whom we shall read shortly).

In some ways, the point of our reading this paper now is to prepare us to read Putnam and Burge: We'll want to pay attention to the question whether and how 'intuition' figures in their arguments.

In section 1 of the paper, MMN&S explain what they mean by 'arguments from reference' and give a series of examples. The key feature of such arguments, as already said, is that they use quite general claims about reference to derive metaphysical or epistemological conclusions. These claims about reference may be different in different cases. Concerning the case discussed in §1.1, for example, eliminativist arguments (including those offered by one of the authors of this paper) often seem to depend upon the claim that, e.g., the reference of "belief" is fixed descriptively. So responses to eliminativism have often rested upon a rejection of descriptivism. The overall moral is supposed to be that one's views about the nature of reference seem liable to affect one's views on quite distant topics: such as whether 'beliefs' or 'races' exist.

The remarks made about the debate over race in §1.2 seem to me to be not quite accurate (though Mallon has been an important contributor to that debate). In particular, Appiah's argument that there are no races is, in the first instance, that there are no natural kinds that correspond to what we call 'races'. Indeed, he argues in some detail that arguments for racism given by such authors as Thomas Jefferson crucially depend upon the assumption that races are natural kinds. But since our focus is not on the metaphysics of race, we can set this issue aside.

In section 2, MMN&S raise the question how we are supposed to decide, in the first place, what the correct theory of reference is. Their claim is that a crucial role is played by what they call the "method of cases": the provision of 'cases', such as the Gödel-Schmidt case, and the extraction of 'intuitions' about those 'cases'. Indeed, it would seem that what MMN&S actually think is that the 'method of cases' plays an essential role in the debate over theories of reference: They seem to think that, without using that method, it would be impossible to decide among theories of reference.

Just to be clear, Evans (whom MMN&S cite on p. 339) was in no sense a descriptivist.

Note that the claim is not just that examples play an important, or even essential, role. The emphasis, rather, is on the role played by our intuitions about those examples. This raises the question what 'intuition' is supposed to be, but, unfortunately, MMN&S do not ever seem to tell us. The usual view, though, seems to be that 'intuitions' are 'fast' responses that are relatively unmediated by explicit reasoning. In particular, no special training is needed for someone to have an 'intuition' about the Gödel–Schmidt case. You could, for example, ask a bunch of your friends about it and collect their 'intuitions'.

In §2.2, then, MMN&S describe an experiment (first reported in the related reading mentioned above) that they did along these very lines. They presented undergraduates at Rutgers and Hong Kong with a version of the Gödel–Schmidt case and collected their responses. What they found was that the Rutgers students (who were limited to North Americans) were more likely to agree with Kripke's verdict about the case than were the Hong Kong students (who were limited to East Asians). This seems to suggest that people's 'intuitions' about the Gödel–Schmidt case vary with culture—and, as they note, other experiments suggest that intuitions also vary within cultures.

In §3, MMN&S consider the question how we might respond to this fact, and do so while still preserving 'arguments from reference'.

• The first option, mentioned in §3.1, is to give up on the idea that there is any such things as the 'correct' theory of reference (or at least that we could ever find it). If so, however, then of course arguments that appeal to claims about what the correct theory is will be impossible.
• A second option, mentioned in §3.2, is to suggest that the 'method of cases' plays less of a role than MMN&S suppose, and that we can decide among theories of reference by deciding among their consequences for other areas of philosophy. But then we cannot use claims about which theory of reference is correct to argue for claims in other areas; we're actually arguing in reverse now: From philosophical conclusions in other areas to the correct theory of reference.

Is it obvious that, if we use consequences for other areas of philosophy to decide among theories of reference, we cannot use theories of reference to argue for conclusions in other areas of philosophy? Why might it not be?

In the final paragraph of §3.2, MMN&S allow that someone might argue that "some other considerations might be used to justify" theories of reference. They confess, however, that they "have no idea what other considerations philosophers of language might appeal to" (p. 343). Do you? Or, to put the question more directly: How central a role do 'intuitions' about 'cases' play in Kripke's arguments? (You might want to reflect here on some of the real-life analogues of the Gödel-Schmidt case that Kripke mentions, e.g., the Peano and Einstein examples.)

The option MMN&S seem to take most seriously—they devote eight pages to it, as opposed to one page for the first two options combined—is what they call 'referential pluralism' and discuss in §3.3. This is the idea that maybe different theories of reference are true for different groups of speakers. (So, e.g., maybe descriptivism is true for native speakers of Chinese, but not for native speakers of English.) MMN&S argue, however, that referential pluralism is hopeless. MMN&S's argument here is in a way very different from their previous arguments, since it seems to depend upon intra-cultural variation in 'intuitions' rather than on inter-cultural variation.

It's important to distinguish between the relation of reference, and what is true of it, and the way words like "refer" and "talk about" are used in English. There is no particular reason to think that the way ordinary people use the word "refer" closely tracks the relation of reference in which we, as theorists of language, are interested. (Compare: Questions about the nature of time are one thing (studied by physicists); questions about how tense works in English are quite different (and are studied by linguists and philosophers). If we do make this distinction, which of these things do experiments like the one MMN&S performed concern? Which of these things would a 'pluralist' theory of reference concern? (You might want to reflect here on the quote from Lycan on p. 335.)

Deutsch's central goal in this paper is to argue that the cross-cultural variation in 'intuitions' reported by MMN&S poses no threat to the viability of the theory of reference. He makes two main claims: First, MMN&S over-state the role played by 'intuition' in Kripke's arguments; Second, their experiments do not show that there actually is any such variation.

Deutsch first points out, in §2, that the 'method of cases', as MMN&S call it, seems at first blush simply to be the method of counterexample: Kripke aims to undermine a general claim implied by (actually, constitutive of) the description theory of names by showing that it has a false instance. If there is something different about Kripke's argument, it must be his (alleged) reliance upon intuitions about the Gödel–Schmidt case. Deutsch proceeds to argue, however, in §3, that it is far from clear what role 'intuitions' are supposed to play. What matters for Kripke's argument is to whom the speaker in the Gödel–Schmidt case would be referring. That has nothing to do with anyone's 'intuitions'. As Deutsch puts it: "...[T]he predictions of a theory of reference concern terms and their referents, not competent speakers and their intuitions" (p. 448).

Deutsch also mentions an example due to Gareth Evans, in "The Casual Theory of Names" (pp. 195-6, PhilPapers), which Evans uses to argue against Kripke's own 'causal theory' of reference. That theory says that reference is inherited along a causal chain so long as each speaker intends to preserve the reference of the person they learned the name from. But, Evans says (drawing upon a book on geographical nomenclature), the name "Madagascar" originally referred to a part of Somalia near Mogadishu. There was some sort of confusion, and it ended up referring, as it does today, to an island off the southeast coast of Africa. And it did so despite the fact that no-one ever meant to change its reference. So that looks like a counterexample to Kripke's theory. Deutsch emphasizes that, contrary to what MMN&S say, Evans's claim that "Madagascar" refers to the island was not one about anyone's 'intuitions', but a claim that can just be checked on Wikipedia.

As I mentioned in the notes on MMN&S, we need to distinguish claims about the nature of reference from claims about how such expresssions as "refer" and "talk about" work in English. Can you use this distinction to elaborate Deutsch's point? (See especially the end of the first full paragraph on p. 449, and see fn. 6 on p. 453 for related remarks.)

Deutsch gives a couple examples of cases in which proponents of a particular theory are well aware that their view contradicts 'intuition'. One of these should be familiar to you. Which? Can you develop Deutsch's treatment of this example a bit further?

In §4, Deutsch considers an interpretation of MMN&S that would allow them to accept these points. On this view, 'intuitions' about cases function as evidence for claims about (say) what the reference of "Gödel" is in those cases. Deutsch allows that some philosophers may sometimes treat 'intuitions' as evidence in this sense. But he insists that this is unnecessary, and that neither Kripke nor Evans need do so.

Deutsch writes: "The causal source of Kripke's judgment about the Gödel case is intuition. Kripke does not literally see or otherwise perceive that 'Gödel' does not refer to Schmidt in his fiction...". But we do not "literally see or otherwise perceive" that, say, "Charles Larmore" refers to Charles Larmore, either. (Maybe if he were wearing a name tag, one could see that.) So what does Deutsch seem to mean by this claim?

In § 5, Deutsch turns his attention to MMN&S's experimental results. Concerning their version of the Gödel-Schmidt case, Deutsch notes that the question they ask at the end seems to be ambiguous between a "speaker's reference" reading and a "semantic reference" reading. Worse, the former reading may in fact be the more likely one (see fn 7).

Explain in your own words why, if Deutsch is correct about this, this poses a problem for MMN&S.

After some introductory remarks, Putnam begins by sketching and then criticizing the traditional distinction between intension and extension. As he notes, intensions were often taken to be "concepts" and therefore to be a sort of mental entity. Frege—think here of his remarks about ideas—rejected this view, insisting that senses are objective. But, as Putnam notes, Frege nonetheless regards grasping a sense as a psychological act or state. Frege of course also held that sense determines reference, or that intension determines extension (if we identify that distinction with Frege's).

Thus, Putnam claims, traditional views about meaning are committed to two claims:

1. "...[K]nowing the meaning of a term is just a matter of being in a certain pyschological state..." (p. 135).
2. "...[T]he [intension] of a term...determines its extension..." (p. 136).

Putnam's central claim is that these two conditions are incompatible, and that is because they jointly entail:

• 3. The extension that a term has for a given speaker is determined by that speaker's psychological state.

The argument for (3) is that (1) entails that what intension a term has for the speaker is determined by her psychological state; the intension then determines the extension, by (2). (See pp. 137-8.)

One of the most important issues here is exactly what a "psychological state" is supposed to be, and Putnam has in mind a very particular conception of such a state. He explains on pp. 136-9 that he takes "traditional" philosophy to be committed to a doctrine he calls methodological solipsism: whether someone is in a given psychological state depends upon nothing "outside" that particular individual; it does not depend upon the existence of any other individual, nor even upon the existence of the external world. To know what psychological states some person is in, you need only look at that person, and nothing else. Nowadays, this view is often called "Cartesianism", since it reflects Descartes's method of reflecting inwardly when investigating the mind. (Methodological solipsism is not, however, committed to the view that all psychological states are introspectible.)

Putnam calls psychological states, so understood, psychological states "in the narrow sense". (The wide–narrow terminology has stuck, so remember those terms.)

To be kind, it isn't entirely clear what Putnam means by a psychological state in the narrow sense. In fact, Putnam expresses a great deal of skepticism about the very notion. (See the remarks on jealousy, etc.) But one way to try to make it more precise would be to say that, at the very least, if two people have all their intrinsic physical properties in common, then they must have all their (narrow) psychological properties in common. Explain why that helps. Or not. (Hint: This would assume a weak form of mental–physical supervenience.)

We can understand this as a sort of supervenience claim: What psychological states one is in depends only upon (i.e., supervenes upon) one's "intrinsic" features, i.e., upon how things are "from the skin in".

Supervenience is a relation between two types of 'facts'. Facts of type A are said to supervene on facts of type B if any difference in A-type facts implies a difference in B-type facts. So "supervenes on" is the converse of "determined by": The A-facts supervene on the B-facts if the B-facts determine the A-facts—if, once the B-facts are fixed, the A-facts are fixed as well. (See the SEP article for more detail.) So one way to understand psychological states in the narrow sense is that they are meant to supervene on the intrinsic physical properties of their subject.

So Putnam therefore says that he will understand (1) above as:

• (1') Knowing the meaning of a term is a matter of being in a certain pyschological state in the narrow sense.

We can therefore also strengthen (3) to:

• (3') The extension that a term has for a given speaker is determined by that speaker's psychological state in the narrow sense.
• (4) Hence, if two speakers are in the same psychological state in the narrow sense, then each term they understand must have the same extension for them.

It is (4) that is the target of the examples that follow.

On pp. 138-9, Putnam argues that Frege's insistence that senses are public and objective is really irrelevant to the central issues. Why? And does that seem right?

What follows, on pp. 139-42, is the now famous "Twin Earth" thought experiment. Putnam has us imagine a distant planet (or, perhaps, a counterfactual world) that is exactly like Earth except that what's in the lakes, etc, isn't H₂O but something else: XYZ. The argument then proceeds in a few steps:

1. If astronauts today were to visit Twin Earth, they'd report back that there was no water in the lakes there, but some other substance, XYZ, and that the Twin Earth term "water" didn't mean water at all, but something else, which we may call 'twater' (which is XYZ).
2. Even before the advent of chemistry, it was still true that, on Earth, "water" was true only of H₂O and, by parity, on Twin Earth, only of XYZ.
3. Now let Twin Oscar be Oscar's physical duplicate on Twin Earth in 1750. Then Oscar and his Twin are in the same psychological state in the narrow sense, but the term "water" has different extensions for them.

The crucial premise in this argument is obviously (ii). Premise (i) should be uncontroversial: We know enough chemistry (as do the Twin Earthers) that the descriptions we associate with "water" are enough to pick out H₂O as its extension. (We believe things about 'water' that they do not, so we are not in the same psychological state in the narrow sense.) And (iii) is just a consequence of what 'psychological states in the narrow sense' are. Putnam's first argument for (ii) depends upon the fact that "water" is a "natural kind term": It is, he says, a presupposition of our use of it that we use it to pick out a certain kind of stuff; water is whatever is the same kind of stuff as typical examples of water. And what "same kind of stuff" means will not necessarily be obvious but may require empirical investigation.

When the Twin Earthers say, "There's water in the lakes", do they speak truly or falsely?

It is often pointed out that, since people are mostly water, there is no way that Twin Oscar can be a physical duplicate of Oscar. How serious is this worry?

Putnam gives another argument for premise (ii) in the section entitled "Let's Be Realistic", on pp. 153-7. What is that argument? How does it compare to the argument about kind terms on pp. 141-2?

The conclusion we are meant to draw from this example is that the extension that a word has for a particular speaker can depend upon facts about the physical environment in which they live. More precisely, much as with Kripke, the thought is that what extension a term has for someone is in part a matter of causal relationships between them and their physical environment. This view is nowadays known as (semantic) externalism, in contrast to internalism, which is the view Putnam is arguing against.

On p. 143, Putnam introduces another now-famous example, which he takes to be of roughly the same form. Putnam says that he cannot personally tell elms from beeches (though let's assume he can tell them from other sorts of trees), then claims that his "concept" of a beech is the same as his "concept" of an elm. (I take it that he means he has the same beliefs about each, so that the "cluster of descriptions" he associates with each is the same.) Nonetheless, Putnam claims, his word "elm" applies only to elms and his word "beech" applies only to beeches: If he points at a beech and says "That's an elm", he speaks falsely. But there is a different Twin Earth on which the words "elm" and "beech" are switched, so that Twin Putnam's word "elm" applies to beeches.

Putnam takes this example to show "that there is a division of linguistic labor" (p. 144): People who do not know how to identify elms and beeches (or whatever) can depend upon people who do, and the extension the word has for me can be determined by the extension it has for the "experts". Thus, as in the original Twin Earth case, the lesson is supposed to be that a person's "individual psychological state...does not fix [the] extension" of a term that is subject to a division of linguistic labor (p. 146).

So the conclusion we are meant to draw from this sort of example is that the extension a word has for a particular speaker can depend upon the social environment in which she lives. This view is nowadays known as anti-individualism, in contrast to individualism, which is the view Putnam is arguing against.

Like the H₂O–XYZ example, the one Putnam uses to argue for anti-indvidualism also involves "natural kind" terms: "elm" and "beech", which are of course names of types of trees. Can you come up with a similar example that doesn't involve natural kind terms? Can you come up with an example that might be used to argue for externalism that doesn't involve natural kind terms?

Suppose Putnam is correct that (1) and (2) above are incompatible. Which should be jettisoned? Should we conclude that knowing the meaning of a term is not just a matter of being in a certain psychological state in the narrow sense? I.e., that Oscar and Twin Oscar disagree about what "water" means? Or should we conclude that intension does not determine extension? I.e., that "water" means the same for them, but applies to different things? For that matter, is either option comfortable?

The central target of this paper is 'individualism': the view that the contents of a person's mental states—for example, what they believe—are wholly determined by facts about that person as they are in themselves and, in particular, do not depend upon their social environment. As Burge notes in §IV (which you do not have to read), this sort of view is familiar from Descartes and, more relevantly for us, from Russell. On Russell's view, remember, we can only really think about that with which we are acquainted, and that does not depend upon our social environment. So Burge's arguments extend those of Kripke.

In §I, Burge explains some of the terminology he will use. First, Burge emphasizes that he will be focused on "propositional attitudes". These are mental state that are typically reported using a "that-clause", e.g., "John believes (knows, hopes, fears, etc) that roses are red". The 'attitude' is belief (or knowledge, or hope), and the 'that-clause' gives the proposition that is said to be the content of that attitude. Burge calls that-clauses "content clauses". Burge emphasizes that he does not need to make substantial assumptions about when contents are different: His arguments will turn on differences in extension (or reference), and anyone would count "a is F" and "b is F" as reporting different contents if a≠b.

Burge also notes that he will need to be able to talk about the 'contents' of sub-sentential experessions, like predicates. He uses the term "notion" to do so. So, e.g., he talks about the 'notion' of 'arthritis'. Another term often used here is 'concept': So we talk about someone as having (or not having) the concept of arthritis. Burge uses 'notion' because it brings with it less baggage.

Another important term is that of an oblique occurrence. This is essentially the same as Quine's notion of a 'referentially opaque' occurrence: An occurrence is oblique if it cannot be substituted salva veritate (without change of truth-value) by other terms with the same extension. This is typically the case when a term occurs in a that-clause, but, as Burge notes, in ordinary speech we are often not terribly strict about this. But, even in ordinary language, such differences can make a difference, and that's what matters here: We need not always be concerned with specifying the content of someone's beliefs precisely, but sometimes we are, and ordinary language provides us with the resources to do that. The important point here is that Burge's arguments concern oblique occurrences of terms like "water".

One other term Burge uses that may be unfamiliar is "idiolect". An 'idioelct' is like a dialect, but it is the language of a single person, as they understand it, as opposed to that of a (relatively small) community.

Burge summarizes the upshot of all of this when he writes: "The crucial point [is] that obliquely occurring expressions in content clauses are a primary means of identifying a person's intentional mental states or events" (p. 104). Burge is really interested in questions about people's beliefs (and other mental states). But he intends to get at those questions by looking at how we report people's beliefs. So, e.g., if it's true to say, in certain circumstances, that Bert believes that water is wet (a fact about language), then Burge will take that to imply that Bert does believe that water is wet (a fact about mental states).

This is one of the ways in which Burge's discussion differs from Putnam's: Whereas Putnam was primarily concerned with questions about language (what "water" refers to, etc), Burge is primarily interested in questions about cognition, that is, about mind: what someone's beliefs are about, etc. Putnam sometimes says things like, "People on twin earth have the same beliefs about water that people on earth do": The contents (intensions) of the beliefs are the same, but the extensions differ (so intension does not determine extension). So it often seems as if Putnam regards beliefs as 'psychological states in the narrow sense'.¹ One way to think of what Burge is arguing here is that this is wrong: Beliefs are not 'narrow' in Putnam's sense; Twin Earth-like arguments do not just concern language. (See the related paper by McGinn for an early argument for this claim.)

The names of the characters here, Bertrand and Alfred, are those of Russell and his collaborator on the monumental Principia Mathematica, Whitehead.

The central thought experiment on which Burge bases his arguments in presented in §IIa. The structure is as follows. We consider a person, Bert, who has a variety of ordinary attitudes he expresses using the term "arthritis", e.g., that his father had arthritis, that older people are more likely to have arthritis, and the like. But there is also something non-standard about the way Bert uses the word: Bert will say, e.g., "I have arthritis in my thigh", although that is, by definition, impossible. (The word "arthritis" just means: inflammation affecting the joints.) So it seems as if Bert can perfectly well have a variety of beliefs about arthritis, although he has one belief that is a bit odd.

Now consider another possible situation (a 'Twin Earth') in which the only difference is that the word "arthritis" is used, in Bert's community, with a different meaning: perhaps it is also applied to certain kinds of inflammation affecting muscles. In that situation, Burge claims, Bert would not have any beliefs about arthritis but would instead have beliefs about 'tharthritis' (i.e., whatever "arthritis" means in the language spoken on Twin Earth). If that is right, then it would seem that individualism must be false, since Bert and Twin Bert are physical duplicates.

What reasons does Burge give for the claim that Twin Bert has no beliefs about arthritis? Or does his argument just depend upon our having the 'intuition' that he doesn't?

Compare and contrast Burge's example with Putnam's elm-beech example. Is Burge just giving additional arguments for the 'division of linguistic labor'? Or does his conclusion differ in some way? (It might help here to study the long footnote on pp. 107-8.)

It's worth distinguishing two kinds of view, then. Putnam's original twin earth example does not really depend upon differences in a person's social environment: It's a difference in the physical environment that matters. The view that those kinds of examples is meant to support is usually called (semantic) 'externalism'. The elm-beech example, and Burge's examples, and Kripke's arguments, by contrast, do concern a thinker's social environment. The view those arguments support is what Burge calls 'anti-individualism'. What's the relation between these two views? Could one accept one but reject the other?

One feature of Burge's example is that it suggests that it can be coherent to doubt truths that would typically be classed as 'analytic': It's true by definition that one cannot have arthritis in one's thigh. (In many ways, this kind of point goes back to Benson Mates, in his paper "Analytic Sentences" (PhilPapers).) Can a Burge-like example be developed to show that it can be coherent to doubt, say, that all bachelors are unmarried? Can it really be coherent to doubt analytic truths?

In §IIb, Burge argues that this same sort of example can be given concerning a wide range of other types of expressions. Part of Burge's point here, much as we saw with Kripke, is that these kinds of cases are not only common but normal.

To emphasize: One important way in which this example differs from Putnam's is that it does not involve a natural kind term (like "water"). Indeed, Burge suggests that similar examples could be given for almost any word, not just 'common nouns' but many adjectives and verbs. Here, then, is an exercise to help check your understanding: Develop, in the same sort of detail Burge develops his original example, one that concerns some word other than a common noun. And here's a question: Are there words that are not 'twin-earthable', as people sometimes say? What distinguishes those cases?

Can you think of any natural examples of this sort from your own experience?

In §IIc, Burge considers the question what makes such examples possible. His first suggestion, made in §IIb, is that "The argument can get under way in any case where it is...possible to attribute a mental state or event whose content involves a notion that the subject incompletely understands" (p. 107). In particular, there is no need for the subject to make the sort of 'conceptual error' that Bert does in the original thought experiment. It is not required, that is, that Bert actually have the belief that it is possible to have arthritis in the thigh, but only that he does not know the definition of "arthritis" and so do not know whether it is possible to have arthritis in the thigh. In that case, too, Burge claims, we will be able to develop a version of his example.

There seem, then, to be two sorts of phenomena at issue in Burge's thought experiments: incomplete understanding and conceptual error. How might these be related to the arguments from ignorance and error in Lecture II of Naming and Necessity?

Suppose that Bert hard-headedly responds to the doctor, "Well, I don't care what you fancy doctors say! I know I've got arthritis in my thigh, because it killed my Daddy!" What should we say about this case? Does Bert believe that arthritis killed his father? Perhaps better: What does this example show us about the presuppositions of Burge's example? (There's some relevant discussion on pp. 111-2.)

Burge argues further that, although incomplete understanding must be present at some stage of the thought experiment, it might be present only in the counterfactual situation. If so, then "...even those propositional attitudes not infected by incomplete understanding depend for their content on social factors that are independent of the individual..." (p. 113, emphasis removed). That is: "...[C]ommunal practice is a factor...in fixing the contents of my attitudes, even in cases where I fully understand the content" (p. 114, my emphasis).

There is a detailed argument for this claim in the last paragraph of §IIc, on p. 113-4. Can you restate that argument in your own words?

---

¹Other times, Putnam speaks differently; he just isn't consistent about this. In some ways, "The Meaning of 'Meaning'" is really a mess.

In §III, Burge considers various attempts to avoid his anti-individualist conclusions by 're-interpreting' the various steps of the thought experiment. For what it's worth, it seems to me that there is something (at least potentially) misleading about the way he describes the dialectic here. He says that he will focus on the 'first stage' of the thought experiment and objections to how he describes it. But, for the most part (though there may be exceptions, as Burge mentions), the first and second stages (real-life and the counterfactual situation) are symmetrical. So whatever someone is inclined to say about the one, they will likely also be inclined to say about the other.

In section IIIa, Burge insists that his interest is in 'commonsense mentalistic notions', meaning, I take it, the notions of belief, etc, with which ordinary people operate in their everyday lives. As we'll see, this is a point of vulnerability. Can you see why?

The first strategy Burge considers, in section IIIb, is to "deny that a subject could have any attitudes whose contents he incompletely understands" (p. 118). So, on this view, Bert does not have the 'concept' of arthritis and does not have any beliefs about arthritis. Burge concedes that, if the subject's understanding is incomplete enough, then this will be true, and he agrees that we would like to have a better sense than we do of what distinguishes the cases in which he is interested from, say, that of a child who has memorized "e = mc²". But he insists that there are plenty such differences, including "A person's overall linguistic competence, his allegiance and responsibility to communal standards, the degree, source, and type of misunderstanding...", and the like (pp. 91-2). But Burge does not really complete the argument against this sort of view here. All he needs is that there are some cases, like the arthritis case, in which incomplete understanding is compatible with the subject's having attitudes with the usual sort of content. It's his opponent, not Burge, who needs to explain exactly when re-interpretation is appropriate and when it is not.

The key to the argument here seems to be this sort of claim: "People are frequently held, and hold themselves, to the standards of their community when misuse or misunderstanding are at issue" (p. 119). What does Burge mean by that? How does he use this claim to oppose the 're-interpretation' strategy?

On p. 120, Burge argues, in effect, that the re-interpretation strategy depends upon the analytic-synthetic distinction. How so? (This question is specifically for people who have some prior exposure to discussions of the analytic-synthetic distinction.)

In §IIIc, Burge considers a variety of proposals about what the content of Bert's 'arthritis' beliefs really is (if they are not about arthritis). The first (on pp. 121-2) you can ignore.

Burge opens the section with the following remark: "I now want to criticize attempts to argue that even in cases where we ordinarily do ascribe content clauses despite the subject's incomplete understanding of expressions in those clauses, such ascriptions should not be taken literally" (p. 121). One might think that had already been argued. Why hasn't it?

The second is that the content is 'indefinite': So Bert's beliefs would be, roughly speaking, ambiguous between arthritis and tharthritis and all the other things that "arthritis" could mean in different possible situations. Burge argues that this is ad hoc: Absent some antecedent commitment to individualism, there is no reason to accept that such beliefs have no definite content.

Burge also claims that Bert's beliefs "admit of unproblematic testing for truth-value, despite [his] partial understanding" and that "The subject and his fellows typically know and agree on precisely how to confirm or infirm his beliefs" (p. 93). What does he mean and why is that relevant?

Burge gives more attention to the other two strategies. The first of these (the third overall) is to take Bert's beliefs to involve a concept that accords with his misconception. So his beliefs would be about tharthritis. Burge makes a number of objections to this strategy:

1. It is far from clear what this other concept is supposed to be (so perhaps this view collapses into the indefiniteness view).
2. This move would prevent Bert from sharing any beliefs about arthritis with others (e.g., that his father had arthritis).
3. This view leads to an implausible account of how Bert reacts to his doctor's telling him that it is not possible to have arthritis in one's thigh.

Suppose Bert is talking to his doctor and says "My father had arthritis". (That would be an important component of his medical history.) One would ordinarily suppose that, as a result, the doctor could come to know that Bert's father had arthritis. Is that possible if Bert does not mean arthritis by "arthritis"? What is the best way for an individualist to handle this kind of issue?

The last strategy Burge considers, which he sees as operating in conjunction with the third, is 'meta-linguistic': It takes Bert's error to concern what the word "arthritis" means; e.g., he wrongly thinks that "arthritis" can correctly be applied to what is wrong with his thigh. Burge makes several objections to this view, too, but the most impressive of these is probably that is it overly sophisticated: The capacity to have beliefs about language does not seem essential to the kind of example Burge gives. More particularly, one might think that children very often have an incomplete understanding of concepts such as arthritis, and yet we do not hesitate to ascribe beliefs about arthritis to them (e.g., that Granny has arthritis in her hand and that arthritis hurts). Yet it is far from obvious that such children are capable of the sorts of meta-linguistic beliefs this view requires them to have.

Burge also brings in here a very different sort of consideration: that the 'ordinary' attributions are explanatorily more powerful than the re-interpretations. What are his reasons for that claim? Even if it is accepted, why is that reason to reject the re-interpretations?

In this case, it is particularly important to keep two key points in mind. First, it is not enough to show that there are attitudes that Bert has that (in some way or other) incorporate his misconceptions. One has also to show that Bert does not have any attitudes in which the notion (concept) of arthritis figures. Second, it is not just the 'unusual' beliefs (such as Bert's belief that he has arthritis in his thigh) for which we need to account. It is also the 'ordinary' beleifs (such as that arthritis is more likely to afflict older people).

We have seen that Nozick's Gambit may always be used to 'internalize the external factors'. How could Nozick's Gambit be deployed in response to Burge?

In section IIId, Burge considers the underlying philosophical motivations that tend to lead people to want to re-interpret his examples. He discusses three. The most interesting of these is the last: that the speaker's deviant beliefs imply that he means something different by his words than we do, so that he should be re-interpreted. Burge again makes various counter-arguments. The one on which he seems to put the most weight is that "The subject's willingness to submit his statement and belief to the arbitration of an authority suggests a willingness to have his words taken in the normal way.... When the verdict goes against him, he will not usually plead that we have simply misunderstood his views" (p. 131).

Throughout this discussion, Burge returns, time and again, to the question whether Bert shares (or even can share) certain attitudes with other people. That seems also to be what he is getting at here: Bert himself acts as if what he means by 'arthritis' is what others mean. Why is it so important to Burge that Bert should share certain attitudes with other speakers? What does Burge's emphasis on this point show us about the underlying assumptions that are fueling his argument?

As we've seen, although Putnam's arguments primarily concern language, they naturally extend to belief. Burge, for example, argues that, not only do Oscar and Twin Oscar mean different things by the word "water", but the beliefs they would express as "Water is wet" have different contents. Oscar believes (as we would put it) that water is wet, but Twin Oscar has no such belief; he, rather, believes that twater is wet. If so, then ordinary psychological states are not "psychological states in the narrow sense".

It is against this sort of view that Loar argues here. He wants to concede that Putnam and Burge have a point as far as language is concerned, but deny that the point extends to what he calls "psychological content". More precisely, he takes Putnam and Burge to have a point about "attitude ascription": the way we report or describe the mental states (e.g., beliefs) of other people. As a matter of correct linguistic usage, Loar is claiming, it would be wrong for us to say "Twin Oscar believes that water is wet". But Loar denies that we should conclude that the beliefs Twin Oscar expresses using the word "water" have a different psychological content from Oscar's. The 'that-clause' that water is wet simply does not, according to Loar, pick out what Twin Oscar actually believes; the language we use to ascribe beliefs does not track the actual facts about belief. ("I shall argue that psychological content is not in general identical with what is captured by oblique that-clauses..." (p. 102).)

One of the key notions in this paper is the notion of 'psychological content': the contents of mental states like beliefs, which is typically taken to mean: how the belief represents the world as being, or how the world would have to be for the belief to be true. But what's most important to Loar is the theoretical role that psychological content is meant to play:

By psychological content I shall mean whatever individuates beliefs and other propositional attitudes in commonsense psychological explanation, so that they explanatorily interact with each other and with other factors such as perception in familiar ways. (p. 99)

The key point here is that claims about psychological content are supposed to be responsive to the needs of psychological explanation. What's meant by 'psychological explanation'? Generally speaking, explanations are answers to 'why' questions, and a psychological explanation is an answer to a question like "Why did Bill walk over to the refrigerator?" that mentions Bill's mental states, e.g.: Because Bill wants a beer, and he believes that there are beers in the fridge. Obviously, this is a pervasive feature of our thought about other people and an important aspect of our own psychology: that we can think about other people's mental states and not just explain but predict their actions.

Loar insists, on several occasions, that his focus is on "commonsense" psychological explanation, by which he means the sort of explanation of people's actions in which we engage on an everyday basis. A similar but different thesis would focus on the sort of explanation that might be given in scientific psychology. For that sort of view, see the related paper by Jerry Fodor.

So Loar's thought is: We should say that two beliefs have different contents when they play different roles in explaining people's behavior; we should say that two beliefs have the same content when they play the same role.

This condition on the individuation of contents plays a significant role in contemporary discussions of Frege cases. Can you see how? (Hint: What makes us so sure that the belief that Hesperus is a planet has a different content from the belief that Phosphorous is a planet?)

Loar sketches three sorts of arguments against particular theses about psychological content, the second of which comes from Burge. Loar sees such arguments as being directed against the thesis that psychological content is determined by "conceptual or cognitive-functional roles". It does not matter for our purposes exactly what that means. Roughly, it means that its determined by the kinds of inferences a person is inclined to draw. (If you want to know more, read section 4.2 of the SEP article on narrow content.) The only thing that really matters for our purposes is that this view is a form of internalism and individualism. So Loar is out to defend those two theses, as views about psychological content.

In the third argument Loar discusses, Nozick's Gambit is employed. How?

Loar takes these sorts of arguments to rest upon two sorts of assumptions:

1. Sameness of de dicto or oblique ascription implies sameness of psychological content.
2. Difference of de dicto or oblique ascription implies difference of psychological content.

Here, "de dicto" or "oblique" ascriptions are, as usual, ones that resist substitution of identicals.

One of the two theses (A) and (B) is implausible if the sort of ascription at issue is de re or relational. Which? In the case of the other thesis, one might actually think that Putnam and Burge only need the de re version. Why?

Loar begins his counter-argument by recounting an example due to Kripke (in "A Puzzle about Belief" (PhilPapers)). The example concerns a character Pierre who does not realize that 'London' and 'Londres' are the same place. It seems difficult for us to describe Pierre's beliefs: It seems right to say that he believes London is pretty, and also to say that he believes London is not pretty. But, Loar insists, it just seems obvious that Pierre's beliefs are not really contradictory: It is (or need not be) not irrational for him to hold both beliefs.

Indeed, it looks as if Pierre could believe (as he would put it) that London is a big city and that Londres is a big city, and he would thus have two beliefs not just one, because those beliefs "interact differently with other beliefs in ordinary psychological explanation" (p. 103). I.e., and roughly, we have the same reasons to think that Pierre's London-beliefs are different from his Londres-beliefs that we would have to think that his Hesperus-beliefs are different from his Phosphorous-beliefs. And that is true even though we would report both beliefs by saying: Pierre believes that London is pretty. Thus, Loar thinks, the facts about Pierre's beliefs and the facts about how one might correctly report them come apart.

Loar gives one example, on p. 103, to illustrate what he means by the claim that Pierre's two beliefs "interact differently with other beliefs in ordinary psychological explanation". Give another such example. What work if any is being done by the phrase "in ordinary psychological explanation"?

Loar goes on to say that Pierre not only has two beliefs, but that "it...seems quite appropriate to regard them as distinct in content". Does he have an argument for this claim? If so, what is it? (See also p. 105.)

On pp. 102-5, Loar gives several arguments against thesis (A). Our main interest will be in (B), however, since that is what plays a role in Putnam and Burge. Loar's discussion of (B) begins at the bottom of p. 105 and is focused on Burge's arthritis example. Against Burge, Loar claims that the contents of Bert's and Twin Bert's beliefs about what they each call "arthritis" are the same, "because they have the same potential for explanatory interaction with other beliefs" (p. 106): The way Bert will behave if he believes something about 'arthritis' is the same as the way that Twin Bert will behave if he believes the same thing about 'arthritis'; e.g., they behave the same way if they believe they have 'arthritis' in their thigh. Loar makes the same claim about Oscar and Twin Oscar's beliefs about water. Loar claims to find this account "intuitive", but he also argues for it using two sorts of examples.

A real-life version of the first (mentioned by Gabriel Segal) concerns the word "pie". In Britain, something is only rightly called a "pie" if it has a crust on top. So, in Britain, a Key Lime Pie is not called a "pie" but rather a "tart". Suppose Bert reguarly travels between Britain and America—perhaps he lives in both, as well—and is unaware of this difference. Loar would claim here that, although the word "pie" seems to be ambiguous in Bert's mouth, he has just one belief that, say, pies are tasty.

Suppose Bert wrongly believes that the word "arthritis" is ambiguous: He thinks there are two different sorts of diseases, both called "arthritis". What would Loar have us say about Bert's beliefs in this case? How might such a case be thought to put pressure on Loar's claim that Bert and Twin Bert have the same beliefs?

The second sort of example concerns a case in which one is given a diary written by someone who is either from Earth or Twin Earth, but one does not know which. In it, we find: "I think I might have arthritis". What does Loar want to say about this case? How is that supposed to answer Burge and Putnam?

One obvious question that Loar does not discuss is whether his response generalizes in the right way. Grant, for the sake of argument, that we can explain why the unknown diarist booked an appointment without knowing where he lives. But what Loar needs to show is that there is no explanation of Bert's behavior that might be sensitive to this difference. Can you think of any psychological explanations that would be sensitive to the difference? What reasons might there be to believe that are none?

Loar considers two sorts of objections to his view:

1. Oscar's belief is true in different circumstances from those in which Twin Oscar's is true. So the "narrow content" that Oscar and Twin Oscar share is not "intentional" or "representational".
2. There is no sensible way to say what the common content of Oscar's and Twin Oscar's common belief is.

On pp. 107-8, Loar sketches but does not develop a response to the first objection. The locus classicus for the sort of view he mentions, which is known as a "two-factor theory", is probably the related paper by Block.

Loar's main response to objection (1), given on pp. 108-9, is that, even if we do not know whether "The water is cold today" was written on Earth or Twin Earth, we still know how the world would need to be for the description to be true. So there is a set of possible worlds in which it would be true, which Loar calls the realization condition for the belief. Which worlds are those worlds? To put it differently, Loar claims that we do how know the world would need to be for that statement to be true. How would it need to be? (Remember that the answers here need to work for both Oscar and Twin Oscar.)

Loar's response to objection (2), given on pp. 109-10, is that, even if that-clauses do not specify realization conditions—which are, he is claiming, all that are needed for ordinary psychological explanation—we have other methods by which we can express them. What are these? The obvious objection here is that ordinary psychological explanation certainly seems to proceed in terms of attributions of beliefs, etc, made using ordinary that-clauses. (Just think about how we actually explain other people's behavior.) How worrying should that objection be?

Stalnaker aims in this paper to explore the phemonena that lie behind the judgements about the thought experiments on which Putnam and Burge base their arguments.

Stalnaker begins by recalling what was central to the description theory—which was, as he sees it, a division of the question what fixes reference into two parts:

1. A relation between a name and a "purely general concept" that can be "grasped by the mind".
2. A relation of "fit", or satisfaction, between the purely general concept and an object in the word.

The key point is that the relation mentioned in (1) is purely mental and so can, one might think, can be effected by the mind itself. E.g., I can choose to associate a particular name with a given concept, because that concept is, so to speak, present to me in thought. As Russell would say: I am acquainted with it.

Putnam and Burge's thought experiments of course aim to challenge this view. But one response to them (in some ways Putnam's own) sees the lesson of those considerations to be limited to language. As we have seen, Burge contested this interpretation, but Loar sought to restore it. And, however convincing the thought experiments might seem, the Cartesian intuition remains: The way the world seems to Bert is the same as the way it seems to Twin Bert; more generally, the way the world seems to me, the way I believe it to be, is independent of my environment, because it is a matter of how things 'seem from the inside'.

The first few pages thus all build to a series of remarks that Stalnaker makes on p. 301:

...[T]he externalist case...was made with examples and thought experiments.... The case is not based on any theoretical explanation for the phenomena.... It is thus open to the internalists...to accept the phenomena—the intuitive judgments about the examples—but argue that they rest on superficial facts about the way we happen to talk, and the way we happen to describe our mental states.

This is more or less Loar's position: The language we use when we attribute mental states reflects "broad" (or "wide") content, which depends upon the relations in which a person stands to their environment; but what really explains behavior are mental states individuated more narrowly, in terms of their "narrow" content. And this is also what captures "how the world seems" to that person, so we can understand why the world 'seems the same' to Bert and Twin Bert.

As Stalnaker mentions, it is often supposed that 'narrow' content "determines a function that takes the external environment into the kind of content to which the externalist's examples and arguments apply" (p. 301). What he means is that we can think of the narrow content of a word (or mental state) as associating with each possible environment what that word would refer to if the person lived in that environment. So, for example, the narrow content of "water" would map Earth to water; Twin Earth, to twater; etc. This point, however, does not play a significant role in Stalnaker's discussions, though it does help make it clear what 'narrow content' might be. (This is more or less the account given in Block's "Advertisement", which is listed as a related reading for Loar.)

One of Stalnaker's central points here is that this response cannot just be met with more examples: Loar's view is meant to accommodate the examples; he does not disagree with Putnam's and Burge's claim that, in some sense, Bert and Twin Bert have different beliefs (specifically, that we would, and should, report their beliefs differently). What is required, rather, is "a theoretical account of intentionality [i.e., of representational content] that explains the externalist phenomena, and that justifies the claim that the phenomena show something about the nature of intentionality, and do not just reflect an accidental fact about the way we happen to talk about speech and thought" (p. 301). The remainder of the paper is devoted to sketching such an account (which had been already been developed and defended by Stalnaker and others, as is mentioned in footnote 5).

Stalnaker seems to be insisting here that 'intuitions about cases' cannot decide the dispute between internalism and externalism. Indeed, one might draw an every stronger conclusion: That even the facts about what refers to what cannot decide the issue. What lessons about philosophical methodology might one draw from that fact?

The specific account of intentionality that Stalnaker mentions is less important, for our purposes, than its overall shape: that it essentially invokes relations to the environment. Stalnaker's own account is a "teleological" theory. For more on these, see the entry on "Teleological Theories of Mental Content" at the Stanford Encyclopedia of Philosophy. For an overview of another sort of account, see the entry on "Causal Theories of Mental Content".

Stalnaker takes the basic notion to be that of a state's carrying information about the world. One example often given, to fix ideas, is that of a thermometer: The height of the mercury in the thermometer (i.e., the 'state' of the thermometer) carries information about temperature. It does so because the height of the mercury varies systematically with (that is, depends upon) the temperature. Of course, it does not do so invariably: Other factors might intervene. But in 'normal' conditions, the height of the mercury reliably indicates the temperature.

Can you think of other familiar cases in which some natural phenomenon carries information in Stalnaker's sense? (One other common example is tree rings, which carry information about the age of the tree.)

Stalnaker proposes that we should think of beliefs as carrying information about the environment in a similar way. As he emphasizes, we need not think of this account as attempting to "reduce" content to information; and we need not suppose that it offers us a complete account of what content is (see pp. 303-4). The important point, for his purposes, is that we have at least some reason to suppose that it is a necessary condition on a belief's having the content that p that it should tend to carry the information that p. Call that the 'informational condition'.

For our purposes, the discussion of what distinguishes beliefs from desires and intentions, and how the content of the latter might be explained (see pp. 302-3), is not critical.

Stalnaker goes on to suggest that the informational condition provides a theoretical basis for the usual judgements about twin earth cases. He claims, for example, that, on Earth, O'Leary's beliefs about 'water' do tend to carry information about water, whereas, on Twin Earth, Twin O'Leary's beliefs tend to carry information about twater. And he makes similar claims about Burge-like examples. More generally, Stalnaker remarks that, if the informational condition is correct, then content must depend upon the external environment. This is because what information a state tends to carry depends, almost by definition, upon relations between it and the environment.

That said, what information a given state carries in any particular case will always be debatable. This is in part because that depends upon what we regard as 'normal' circumstances. However, one of Stalnaker's most interesting points is that this does not undermine the externalist conclusion: Even if O'Leary's water-beliefs only carry information about watery-looking stuff (stuff meeting some description given in terms of observable properties), that will still be because of relations between O'Leary and his environment.

This seems, in fact, to be the crucial issue. An internalist could respond that while O'Leary's mental states do carry information, they carry information about 'watery-looking stuff', so that Twin O'Leary's states will carry the same information. Can you develop the argument Stalnaker is giving here?

On the other hand, this does raise the question, rather sharply, how "normal" conditions are to be characterized. If there is no restriction on what could count as 'normal', then O'Leary's water-beliefs seem not to carry any very specific information at all: If he were a brain in a vat, they might only carry information about the internal workings of a complex computer or the paranoid fantasies of an evil demon (see p. 306). So the claim has to be, e.g., that Twin Earth is abnormal. But it is so, Stalnaker emphasizes, only from the perspective of Earth (see p. 307). Or better: It is abnormal for O'Leary, but not for Twin O'Leary; for him, the situation on Earth would be abnormal.

Although this clarification is helpful, it does not really explain what 'normal' circumstances are. One might argue, nonetheless, that externalism can only really be avoided if there is no restriction on what counts as 'normal'. Why would that be? (Hint: Consider Stalnaker's earlier remark about the inevitability of externalism on the informational conception.)

Stalnaker emphasizes that he is not trying to argue, here, for the informational account but only to show how, given that account, it is possible to explain why the twin earth examples show something important about content itself and not just 'how we happen to talk'. Can you summarize how the informational account does that? What features of the account play a role here?

This paper amounts to a comprehensive reply to Kripke, Putnam, and Burge. Chalmers here outlines what has come to be called 'two-dimensional semantics', which has been much discussed in the years since. The key idea is to attempt to reconstruct Frege's notion of 'sense' in terms of what Chalmers calls 'epistemic intensions'.

In section 2, Chalmers reminds us of some of the reasons there are to think that extension (reference) cannot be all there is to meaning. These should be familiar. He then lists and discusses several features that Frege's notion of sense was supposed to have:

1. Every expression that has an extension has a sense.
   The point here is that the difference between mass terms like "water" and "H₂O" is also to be explained in terms of a difference of sense. (Frege is explicit about this point in "Comments on Sense and Meaning", a paper that was not published in his lifetime.)
2. Sense reflects cognitive significance.
   The point here is that differences of sense are supposed to go with the sorts of cognitive (or psychological) differences on which Frege's example was focused. As Chalmers puts it, "We can think of the sense of an expression as mirroring the expression's role in reason and cognition" (p. 139).
3. The sense of a complex expression depends on the senses of its parts.
4. Sense determines extension.
   As Chalmers notes, it is not easy to say exactly what this means. See below for his account.
5. In indirect contexts, expressions refer to their customary senses.
   This is not something on which we've focused, but the idea is that, when one says, "John believes that Hesperus is Phosphorous", what John is said to believe is the sense of the sentence "Hesperus is Phosphorous", that is, the 'thought' that Hesperus is Phosphorous, as Frege puts it. Chalmers does not discuss this issue in the parts we are reading.
6. The sense of a sentence has an absolute truth-value.
7. The sense of an expression can vary between occasions of use.
   These two theses specifically concern the senses of 'context-dependent' expressions like "this", "now", and "I". We'll look at this sort of issue more closely soon, when we read some of the literature on demonstratives. For now, you needn't worry too much about it (though it is, in some ways, a key motivation for Chalmers's account).

What do the words "intuitive" and "intuitively" mean as Chalmers uses them in section 2?

As we saw earlier, Frege's notion of sense is largely 'programmatic': He introduces it to do certain work (that mentioned in thesis (2)), and he says various other things about sense, but he never tells us exactly what sense is. That is what Chalmers wants to do here. Chalmers's suggestion is that the sense of a name is a condition on reference: It's a condition an object must meet to be the referent of the name. (Indeed, as he says, that's more or less the core idea behind the description theory, to which his view is therefore an heir.) Chalmers argues, however, that we need to understand the notion of a 'condition' here in epistemic terms: The sense of a name is something like the condition that an object in the actual world would have to meet to be the referent of the name; indeed, Chalmers regards the sense as refecting not just 'what determines' the reference, in an objective sense, but how we might go about determining the reference. (This ambiguity, between 'metaphysical' and 'epistemic' senses of "determine" often causes problems.)

As Chalmers puts it:

If a subject uses an expression, then given sufficient information about the world, the subject will be in a position to know the extension of the expression. Furthermore, something like this will be the case however the world turns out: for any scenario, given sufficient information about that scenario, the subject will be in a position to determine what the extension of the expression will be if that scenario is actual. (p 144)

So, roughly: When you understand a name, you do not have to know what it refers to; that might take work to figure out. But what you do have to know is what you would need to figure out to know what the name refers to: You need to know what it would take for a particular object to be the referent of the name. This could involve knowing a descriptive condition, but, as we shall see, it need not.

So Chalmers is claiming that someone who understands a given name must know how to determine what that name refers to. Is it plausible that all speakers who can use a name competently have such knowledge? If not, how should Chalmers reply?

In section 4, Chalmers generalizes this idea, yielding the notion of an 'epistemic intension'. Generally, intensions are functions from possibilities (possible worlds) to extensions. But the 'possibilities' here are supposed to be epistemic possibilities, not metaphysical ones; they are, roughly, ways the world might be that we cannot rule out a priori. A different way to think of this is that epistemic possibilities are possible worlds 'seen from the inside' (or 'considered as actual', as Chalmers puts it). They represent ways things could turn out to be, not just alternative ways things might have been.

This allows Chalmers to accept the thesis that names are rigid without having this imply that "Hesperus" and "Phosphorous" have the same intension. (See section 7 for detailed discussion.) These names do have the same metaphysical (or 'subjunctive') intension. This is because the metaphysical intension of a proper name is a constant function: It's always the same object, since names are rigid; "Hepserus" denotes the same thing with respect to every counterfactual situation. But the epistemic intensions differ: There is a possible world such that, in that world, "Hesperus" and "Phosphorous" have different references (though, when we describe that world from here, so to speak, we say that at least one of those names does not refer to Hesperus, i.e., Phosphorous, i.e., Venus).

Chalmers emphasizes that, in many cases, there will be no description that captures the epistemic intension. The epistemic intension reflects, rather, what speakers would say the referent was in different sitations: It captures a pattern in their dispositions to identify different things as the referent in different such situations. In that sense, the epistemic intension is tightly tied to the reasoning of actual speakers.

Why is it essential that epistemic intensions be tied to the actual linguistic capacities of actual speakers? (One answer is found in Chalmers's discussion of thesis (2), in the next section.)

On pp. 146-8, which you can skim, Chalmers outlines a detailed account of what epistemic possibilities (and so epistemic intensions) are, but we should be able to work with the general idea. (This approach is developed in more detail in Chalmers's book Constructing the World and in the related paper "The Nature of Epistemic Space".)

In section 5, Chalmers considers how well epistemic intensions can perform the role that sense was meant to play. He argues that epistemic intensions plausibly can play that role. Much of the section is concerned with questions about whether epistemic intensions can really help explain the informativity of identity-statements. As Chalmers suggests, in the most usual cases, this is pretty clear: There is an epistemic possibility in which what we would, in that situation, identify as the referent of "Hesperus", is not the same as what we would, in that situation, identify as the referent of "Phosphorous".

On Chalmers's account, any two names for which it's a priori that they refer to the same thing will have the same epistemic intension. Are there such names for which the identity statement connecting them will nonetheless be informative? If so, then Chalmers's account will not be able to explain the informativity of that statement. Why not? And how should Chalmers respond?

Section 6 discusses the senses of demonstratives and indexicals; we've not yet considered those, but you may want to come back to this section after we do. This section does, though, explain some of the motivation for Chalmers's approach. Section 7 dicusses the modal argument, Chalmers's response to which was sketched above.

In section 8, Chalmers discusses Kripke's central objections to descriptivism: the arguments from ignorance and error. Chalmers's main response is that Kripke's arguments help to illuminate what the epistemic intension of "Gödel" is; they do not show that the name has no such intension. When we look at the Gödel-Schmidt case, for example, this shows us that the epistemic intension of "Gödel" is not the same as that of "the person who proved the incompleteness theorem". So what is the epistemic intension? It is, roughly: the person called "Gödel" by those from whom I learned the name (p. 170). So Chalmers's strategy is a version of 'Nozick's Gambit': Effectively, he builds Kripke's theory (or whatever the right theory is) into the epistemic intension.

Chalmers considers two objections to this view. Which of these is most worrying? How good is his response?

Probably the best argument for internalism is what we might call the 'argument from psychological explanation': The differences the externalist claims to find between the beliefs of Fred and Twin Fred seem to make no psychological difference. It's this argument that Braun aims to counter in this paper.

Braun phrases the argument in terms of the question whether any kind of content needs to be ascribed to mental states for the purposes of psychological explanation. Some people have held this stronger view. But, as he mentions, nothing in the argument turns on that choice. Braun is out to defend externalism and to show that externalist content is needed.

In section II, Braun rehearses the sort of argument with which he'll be concerned. The 'Naive Argument', as he calls it, is that Fred and Twin Fred behave in the same ways, so their behavior should be explained in the same way. This argument generates an equally Naive Counterargument: There are natural ways of describing this behavior so that it is not the same. When Fred is thirsty, he drinks a glass of water; when Twin Fred is thirsty, he drinks a glass of twater; etc. As Braun mentions, these two arguments are too naive to be convincing. A better form of the first argument holds that, while Fred's and Twin Fred's beliefs lead to different behavior (relationally described) in their different 'contexts', it's still true that their beliefs would lead them to behave the same way were they in the same context. E.g., if you swapped Fred with his twin, then they'd each behave like the other.

In section III, Braun discusses an argument due to Fodor for the conclusion that broad content is explanatorily idle. That argument is more or less the less naive one just described: Twins will behave the same in any given context. Braun's response is borrowed from the paper of Burge's listed as related: There are contexts in which Fred and Twin Fred's mental states will have different effects. More generally, relational properties of objects (as opposed to their intrinsic properties) can clearly cause those objects to have different causal effects, even when the relational properties are extremely complex. Still, it's easy to think that there must be some way Fodor could restrict the range of appropriate contexts, so more needs to be said.

The responses Braun gives in this section might seem to be closely related to the Naive Counterargument discussed in the previous section. Are they? How are they different?

In section IV, Braun offers a positive reason to think that the broad (or wide) content of mental states is relevant to psychological explanation. In particular, he claims that "There are true law-like causal generalizations using broad content attributions" (p. 380). The most important of these are the ones that explain or predict behavior, itself described in relational terms, such as (C). The point is that the mental states mentioned in these generalizations are characterized in terms of their wide content: Replacing "pencil" in (C) with some 'twinned' verison will not preserve the truth of the generalization. It's also important that these generalizations are 'law-like', meaning: They support counterfactuals. So (C) is not just true of the actual cases but also applies to counterfactual cases. (E.g., if S had believed... or if S had wanted....) Thus, (C) is meant to illustrate the quoted claim.

Exercise: Produce corresponding versions of (C) for "water" and "arthritis", as well as for their Twin Earth counterparts.

Here again, one might wonder exactly how this argument differs from the Naive Counterargument considered initially. How does it? And how is it similar? Would it be right to say that Braun's argument is basically a less naive version of the Naive Counterargument? (Hint: How should Braun deal with the Disneyland case?)

In section V, Braun considers a restatement of Fodor's argument: The non-relational properties of a thing are what determine its effects, so only non-relational (narrow) properties of mental states can be relevant to what laws apply to them. The response is that this is just false: The non-relational properties of a thing will only determine its effects non-relationally characterized; but there is no reason that we cannot consider effects that are relationally characterized and seek law-like generalizations about those.

How is this response related to the earlier response to Naive Argument?

Braun gives a simple example, involving muscles, to illustrate this point. Can you think of more realistic examples from the sciences? Biology might be one source of examples. The social sciences might be another.

In section VI, Braun considers an objection based on 'visiting cases': Suppose Fred visits Twin Earth. Then his cat beliefs will cause behavior directed towards robocats, not cats. Doesn't that show that the generalizations involving cat-beliefs and cat-behavior are false? Braun's answer is that such cases 'trigger the ceteris paribus clause', as it's sometimes put: They are cases where other things are not equal. So they do not contradict the generalization; they are just exceptions to it.

In section VII, Braun argues that the defender of narrow content must make similar moves in order to defend the claim that there are true generalizations involving narrow content. The strategy is to produce a case in which the same narrow content (assumed to supervene on brain states) produces different behavioral effects, because of differences in the creature's body. To avoid the conclusion that even bodily effects cannot be explained in terms of narrow content, one has to regard these cases as exceptions. So, the thought is, Fodor is in no position to object to the corresponding move when made by Braun.

There's a general worry here that this strategy works too well: If every counter-instance can just be declared an exception, then no law can be disconfirmed. So what can be said in favor of the claim that visiting cases really do yield exceptions? One might also wonder whether the case of switching Fred's brain into Martian Fred's body is more exceptional, so to speak, than the case where Fred visits Twin Earth. Can these cases be pulled apart?

That said, one might wonder what true generalizations about narrowly described mental states would have to be like. How should such a generalization be formulated that applied both to Fred and to Twin Fred, and to all their other twins? Would such a generalization also have to apply to brain-in-vat cases? If so, what conclusion might one draw? (See the related paper by Buller for a development of this line of thought.)

At the beginning of the paper, Kaplan recounts some of the developments in what he dubs the "Golden Age of Semantics". In particular, he explains why proper names came to be treated as descriptions: largely for Russell's reasons. But, he notes, the development of intensional logic put pressure on this view, largely as a result of problems surrounding quantification into intensional contexts. The issue—which we have encountered earlier, in Quine—is, as Kaplan puts it:

Under what circumstances does a given individual, taken as value of `x, satisfy th[e] formula "John asserted that x is a spy"? Answer: If the appropriate singular proposition was the content of John's assertive utterance. (pp. 225-6)

Kaplan goes on to explain his notion of singular proposition. The presentation is a bit idiosyncratic.

Some spy is smart.	<<'Some', spyhood>, smartness>
Every spy is smart.	<<'Every', spyhood>, smartness>
The spy is smart.	<<'The', spyhood>, smartness>
John is smart.	<John, smartness>

The details of how Kaplan is representing quantifier phrases do not much matter. (The idea is that the quantifier words are syncategorematic.) What does matter is that the last sentence, "John is smart", is being treated very differently from the others: as expressing a proposition that contains John himself as a constituent.

The difference manifests itself when when we evaluate these propositions at different possible worlds. If we ask whether <<'Some', spyhood>, smartness> is true in some given world, we are asking a question about the things in that world that are spies and are smart, namely: Is there anything that is both? By contrast, if we evaluate the proposition <John, smartness> at a given possible world, then Kaplan thinks this amounts to asking, quite simply, whether John is smart in that world.

One might have thought that we could also ask whether the person who is John in that world is smart in that world. Is that different? Why or why not?

One might well say that the notion of a singular proposition has rigidity 'built into it'. Why? And in what sense?

This leads Kaplan to the statement of the view he wants to defend:

...[S]ome or all of the denoting phrases used in an utterance should not be considered part of the content of what is said but should rather be thought of as contextual factors which help us to interpret the actual physical utterance as having a certain content. (p. 228)

The main payoff of making this move is what it allows us to say about the semantics of demonstratives ("this", "that") and indexicals ("I", "here", "now").

My own view—and, I think, the general view—is that the examples Kaplan initially uses to motivate this idea, on pp. 228-9, are poorly chosen. The examples primarily concern various sorts of ambiguity, which has nothing to do with context-sensitivity. But this point will not matter to us, and there is a way in which the examples can be help one at least get an initial idea what sort of distinction Kaplan wants to make.

At the bottom of p. 229, then, Kaplan turns to the case of demonstratives, such as "that book" or "this man". There is a pronominal use of such expressions—as in "If you see someone who needs help, then give that man a hand"—which Kaplan sets aside. He is interested, rather, in what we might call the "referring" use of demonstratives and which he calls the "demonstrative" use, which (he says) is typically associated with a "demonstration" that serves somehow to pick out the referent of the uttered demonstrative. The central question in which Kaplan is interested can then be put as follows: What sort of contribution to the proposition expressed does the demonstration (e.g., pointing) make?

There are two possible views.

1. We can take the demonstration to be a contextual factor that helps to determine the content of the utterance. The content itself is then a singular proposition.
2. We can make the demonstration part of the content of the utterance, so that the content is a general proposition, something like: The thing at which I am pointing is suspicious.

Kaplan suggests that the former is the correct analysis, and he outlines (on p. 232) some reasons to think (2) is wrong. But, in the end, he sets the issue aside and adopts a strategy not unlike the one Kripke uses in "Speaker's Reference and Semantic Reference": He stipulates that the word "dthat", in his mouth, will have the semantics stated by (1), and then explores what it would be like to have such an expression in our language. So a sentence of the form "Dthat['the F'] is G" will express the singular proposition <a, G>, where a the object that is the unique F.

At the bottom of p. 231, Kaplan gives a somewhat artificial example of an utterance of "That is the same as that" that poses a 'Frege case': It's true but potentially informative. Can you give a more realistic example? (It might help to use 'complex demonstratives', i.e., expressions like "that woman" or "that rock".)

Tough question: How does "dthat['the F']" compare to the actualized description "the actual F"? (There is some debate about this.)

There follows, on pp. 235-8, an extremely convoluted, and therefore difficult, discussion of the relation between an utterance, its meaning, and its content. (It's worth the effort to try to understand Kaplan's discussion, but if it's too difficult just skip to the top of p. 238.) The distinction towards which Kaplan is moving here is now known as the distinction between the context (or world) of utterance and the circumstance (or world) of evaluation. Here's the gist of the argument.

First, consider an "eternal" sentence like "The person who was President of the United States on 10 October 2022 is a woman". The truth-value of this sentence does not depend upon the time at which it is uttered. Its truth-value does, of course, depend upon the facts: There are some ways the world could be that would make it true and others that would make it false. So one might think of its meaning as being its "truth-condition", which we can represent as a function from possible worlds to truth-values (the truth-value the sentence has in that world).

Now consider a "fugitive" sentence such as "The person who is [now] President of the United States is a woman". The truth-value of this sentence will vary with the time at which it is uttered (who the President is then), as well as with other facts (what that person's gender is). So in this case one might think of the meaning of the sentence as a function from worlds and times to truth-values.

Kaplan thinks this idea (which traces to Richard Montague) is right, in a way, but that it misses an important subtlety. Suppose I say now, "The President of the United States is a woman". Then what I say is false. Might what I said have been true? One might think so: Kamala Harris might have been the President instead of Joe Biden. But, and here's the point: It does not show that what I said might have been true that Kamala Harris might be the President in 2025, so that I might truly utter the sentence "The President of the United States is a woman" three years from now. Whether what I said might have been true depends upon who might have been the president now. The utterance of the sentence that I make now fixes its "temporal co-ordinate", so to speak, so that what I say when I utter the sentence now is not "fugitive" or "time-relative" but "eternal", almost as if I'd explicitly added "in 2022".

One way to think of this is as concerning a certain sort of dialogue:

• Alex: The President of the United States is a woman.
• Toni: That's not true, but it might have been.

The question, in effect, is what it takes for the second of these remarks to be true. (How could one continue it: "...and it would have been true had..."?) That is tied up (one might reasonably suppose) with what "That" and "it" refer to in the second remark (what proposition they denote).

So possible circumstances are playing two roles: The sentence can be uttered in different circumstances, and those circumstances (the 'context of utterance') help to fix what has been said; then we can take that proposition (what Kaplan calls the 'content' of the utterance) and evaluate it in different circumstances (the 'circumstances of evaluation').

The best example of this sort of phenomenon is a sentence like "I am here now", which Kaplan mentions on p. 237. This sentence is context-dependent along even more dimensions: not just time of utterance, but also location (for "here") and speaker (for "I"). But any utterance of this sentence will be true. So, if one thinks of its meaning as a function from speakers, times, locations, and worlds to truth-values, then its meaning is the "constant function" that is always true. But that, again, seems to miss something important. Suppose I now utter "I am here now". Then if we ask whether what I said might have been false, the answer seems to be "Yes, of course". I might have been somewhere else. So, Kaplan would suggest, we should think of the context in which I make my utterance as fixing its various contextual parameters—speaker, time, location—so that the content of my utterance ends up being something like: RKH is in their study at 2:21pm EDT on 22 October 2022. The content is not the same as that of: The person making this utterance is where this utterance is being made at the time this utterance is being made.

Actually, there seem to be utterances of "I am here now" that are not true. Can you think of one? Alternatively, can you think of utterances of "I am not here now" that might be true? (The bearing of this phenomenon on Kaplan's theory is not entirely clear, however, so ignore it henceforth.)

Thus, Kaplan thinks that there is a sense in which "I" means: the person making this utterance. But that description only fixes the reference of "I" on a given occasion; it does not "give the meaning", in Kripke's sense.

This is why Kaplan says at the bottom of p. 238 that it would be better to think of the meaning of a sentence as a function from "contexts" to "contents", that is: from the circumstances in which an utterance might be made to the proposition an utterance of that sentence would express if made in those circumstances. The proposition itself can then be regarded as a function from worlds to truth-values, just as with eternal sentences.

In his later writings, Kaplan distinguises between content and what he calls character. The latter is supposed to capture the notion of "the meaning of a word". So the content of any given utterance of "I" will just be a person, but the character of "I" is a function from contexts of utterance to people: from contexts to the person uttering "I". Similarly, the character of "now" is a function from contexts to times (the time "now" is being uttered); the character of "yesterday" is a function from contexts to days (the day before it is being uttered); and so forth.

All of this is supposed to help us better understand the difference between:

1. The President in 2022 is a woman.
2. Dthat['the President in 2022'] is a woman.

What's puzzling is that, on the one hand, according to Kaplan, these sentences express different sorts of propositions (general vs singular). But, on the other hand, in any given world, an utterance of (i) will be true if, and only if, the corresponding utterance of (ii) would be true. If we think of the meaning of a sentence S, then, as a function from worlds to truth values that gives you the truth-value a sentence would have if uttered in that world, then (i) and (ii) have the same meaning. But, Kaplan says, that is a mistake: The meaning of a sentence should be thought of as a function from worlds to contents, and the content of an utterance of (i) is different from the content of an utterance of (ii) made in the same situation, even though those utterances will always have the same truth-value. That comes out in the fact that what's expressed by (i) and (ii) is true in different worlds.

Kaplan suggests that his machinery helps us make proper sense of Donnellan's distinction between referential and attributive uses. How so? And to what extent? (Hint: Does Kaplan agree with Donnellan that "the man drinking champagne" can be used to refer to someone who is not drinking champagne?)

Finally, Kaplan discusses a question that has been in the background of much of his discussion, namely: How exactly is the referent of an uttered demonstrative, such as "that picture", determined? There are three options:

1. The demonstrative refers to the object to which the speaker intends to refer.
2. The demonstrative refers to the object that various 'contextual cues' would lead a "linguistically competent public observer" to pick out.
3. Kaplan's own view: (2), but with intentions playing a limited "disambiguating" role if the demonstration itself is too vague.

We'll spend some time discussing this issue when we read Marga Reimer's paper "Demonstratives, Demonstrations, and Demonstrata", so do not worry too much about it for now.

This paper builds in many ways upon ideas in Kaplan's paper "Dthat". The fundamental question is whether Frege can give a reasonable account of demonstrative reference.

Perry calls all of the expressions in which he is interested "demonstratives", but it is standard nowadays to reserve that term for expressions like "that" and "this dog". Words like "I", "here", and "yesterday" are instead known as indexicals.

Perry begins by explaining certain key doctrines to which he takes Frege to have been committed:

1. If A understands two sentences S and T, and accepts S as true while not accepting T as true, then S and T have different senses. (This is more or less Frege's test for when expressions have different senses.)
2. If S is true and T is not, then S and T express different thoughts. (Because sense determines reference.)
3. S and T have different senses iff they express different thoughts (because the sense of a sentence just is the thought it expresses).

What Perry will argue is that the identification made at (3) cannot hold unless one of (1) and (2) is rejected.

Perry also mentions another thesis of Frege's: If "A believes that S" is true, and "A believes that T" is not, then S and T do not have the same 'indirect reference'. This thesis plays no significant role in the paper, so don't worry if you don't understand it. (That said, I'm happy to explain it to anyone who's curious about it. The idea, roughly, is that complement clauses, like "that S", denote the thought expressed by the sentence S, so "Sue believes that ice is cold" affirms a relation (that of believing) between the reference of "Sue" and the reference of "that ice is cold", i.e., between Sue and the thought expressed by "ice is cold", i.e., between Sue and the thought that ice is cold; so the sentence says that Sue believes that thought. Frege says that, when occuring this way (with 'that'), sentences (and their parts) do not have their 'ordinary' reference but their 'indirect' reference, which he then claims is just their sense.)

On pp. 477-8, Perry notes (and this is a point Frege makes himself) that sentences do not always express thoughts. Some sentences are "incomplete" in themselves, and what thought they express will vary from case to case. The example Perry gives is "Russia and Canada quarreled", which is 'incomplete' as to the time at which the quarrel is alleged to have occurred. The sense the sentence has in its own right, then, is similarly 'incomplete', and the sense an utterance of it expresses on any given occasion will depend upon how that sense is 'completed'.

This terminology comes from another doctrine of Frege's. (See his paper "Concept and Object".) He thought of predicates as being 'incomplete', in the sense that they have a 'gap' in them that awaits fulfillment by a name (say), from which a sentence is constructed. So Frege did not think of "is white" as a predicate, but of "( ) is white" as a predicate, where the empty parentheses indicate the 'gap'. Frege also thought of the senses and references of predicates as being similarly incomplete, but this is one of the most puzzling of his doctrines. (I have written about this with Robert May, in "The Composition of Thoughts" (PhilPapers).)

Frege, Perry says, thinks of context-dependent utterances in the same way: the sentence uttered—e.g., "Russia and Canada quarrelled today"—does not itself express a thought, but only does so in light of the conditions under which the utterance is made. So "today", together with the circumstances of utterance, must provide a "completing sense". And the problem, as Perry sees it, is that this does not seem possible, given Frege's other views about sense.

If "today" always had the same sense, then different utterances of "Russia and Canada quarrelled today" would always have the same sense and so express the same thought. But they can't, since different utterances of that sentence can have different truth-values and so, by (2) above, must express different thoughts. Hence, it must be that "today" expresses different senses when uttered on different days.

Now something about "today" does change from day to day, namely, which day an utterance of it would refer to: It is a feature of what "today" means that any given utterance of it will refer to the day on which it is uttered; similarly, it is a feature of what "I" means that utterances of it refer to the person who utters it. Perry calls this rule mapping utterances to referents the role of the word. (This is more or less how Kaplan thinks of 'character', as well.) The important point here is that the role only gives us a reference. But Frege needs it to do more—or, at least, he needs something to do more: Not just a reference but also a sense must somehow be provided on any given occasion of utterance. The question is how this is supposed to happen.

To restate the problem: Utterances of "Russia and Canada quarrelled today" made on different days must have different senses, since different such utterances can have different truth-values. But where do these different senses come from? Perry suggests that three options are open to Frege.

The first is to take the sense associated with a given utterance of "today" just to be its role. The problem with this proposal is that the sense of a given utterance of "Russia and Canada quarrelled today" will then always be the same, and that violates (2) above, as said already.

What is new in this discussion is the idea that sentences can have roles, not just indexicals. And whether that will help Frege or not, one might wonder if it's a useful extension of the notion. Perry's presentation of the idea is a bit compressed. Can you spell it out in more detail?

The second option is to introduce a somewhat coarser-grained notion of a thought. Perry explains this notion in terms of what he calls "informational equivalence", but one might better think of such thoughts as a hybrid of Fregean and Russellian ideas. Consider the sentence "Tony loves Alex". Frege would have us think of the thought expressed by such a sentence as something like:

<sense of 'Tony', sense of 'loves', sense of 'Alex'>

Russell would think of the proposition expressed as:

<Tony, loving, Alex>

Roughly: Frege thinks of propositions as composed of senses, whereas Russell thinks of them as composed of references.

The proposal under consideration is that Frege might take the sense of "I love Alex", as uttered by me, to be the hybrid:

<RKH, sense of 'loves', sense of 'Alex'>

The role of "I", though it doesn't get us a sense for my utterance of it, does at least get us a referent, and so will get us one of these hybrids.

The main objection Perry raises against this suggestion is that it violates principle (1) above. The example he gives is the now famous Enterprise example.

Painting of Frege by Renee Jorgensen.
(You can also get it on a mug, or tote bag, or clock, or...,
and there are many other philosophers to choose from, as well!)

The corresponding view in this case would have both utterances of "That is an aircraft carrier" express

<Enterprise, sense of 'aircraft carrier'>

But, as Perry notes, one could believe one and not the other, so the utterances need to have different senses.

Perry gives an additional example, involving "today", but (to my mind) it is not terribly convincing. Can you think of better examples involving other indexicals, e.g., "here", "now", "you", "there"? Is it possible to come up with such an example for "I"?

The third option, then, is to try to find some way to let the sense expressed by an indexical on a given occasion vary. Perry supposes that, whatever the sense is on a given occasion, it has to be expressible by some description that picks out the right day. The central objection to this move is that there simply isn't any non-indexical description that will have the same sense as a given utterance of "today". For any non-indexical description of a day "the F", it seems one could coherently doubt that today is the one and only one day that is F. If so, then "the F" and that utterance of "today" must have different senses. Similar problems arise with other indexicals, such as "I": For any non-indexical description "the F" of me, I could coherently doubt whether I am the F. In this sense, the indexical is 'essential': It cannot be replaced by any description (or, indeed, by any non-indexical expression, such as "Riki Heck", since I could coherently doubt that I am Riki Heck).

As Perry notes, this sort of point is originally due to Hector-Neri Castañeda. See the references in note 5. Perry himself develops it in more detail in the paper "The Problem of the Essential Indexical", listed as a related reading.

The objections Perry brings against this view should remind you of objections Kripke brought against the description theory of reference-fixing. Which ones?

Perry goes on to suggest, in section II, that Frege was led by these problems to the view, which he expresses in "The Thought", that the sense each of us associates with our own utterances of "I" is primitive, idiosyncratic, and incommunicable.

Perry expresses some consternation about how this might be true, largely because he continues to insist that, whatever this sense might be, it must involve some description I have of myself. As we'll see when we read Evans, that is questionable. That does not of course mean that Frege's suggestion is easy to understand. Again, Evans will try to make some sense of it.

Perry's main worry about this proposal is that it's hard to see how to extend this idea to other cases, such as "now". Or, rather: It's clear enough how to do so, but it implies that "Frege will have to have, for each time, a primitive and particular way in which it is presented to us at that time, which gives rise to thoughts accessible only at that time, and expressible, at it, with 'now'" (p. 491). And Perry thinks that isn't plausible at all.

Perry doesn't say much about why he thinks that claim is implausible. Why might one?

In section III, Perry sketches an alternative to Frege's treatment. The sense of a sentence, as Perry wants us to think of it, is not the thought the sentence expresses; rather, it helps to determine that thought, which is not a sense but a 'hybrid' of the sort mentioned above. So Hume can 'entertain the sense' of the sentence "I am British" and thereby 'apprehend' the thought:

<Hume, sense of "is British">

Heimson, on the other hand, cannot have a belief with this same content by entertaining that same sense, but only by entertaining some other sense (e.g., that of "Hume is British"). Senses, for Perry, are thus roles, as suggested in the first of the options he offered Frege; and thoughts are the hybrids presented in the second option. But Perry doesn't have the same problems Frege does, because he is not committed to (1)–(3) above.

Perry is clear that he wants to reject (3): the identification of senses with thoughts. What should he say about (1) and (2)?

On p. 494, Perry briefly discusses the way in which senses and thoughts, as he understands them, do and do not figure in the explanation of people's behavior. How do the views he expresses here relate to those in Loar's paper?

On pp. 495-6, Perry considers how his view might handle the sort of case Frege uses to motivate the distinction between sense and reference. In his example, Mary says in the morning "I believe that is the Morning Star" and in the evening "I believe that is not the Morning Star". How does Perry's view do with a slightly different example, in which Mary says in the evening, "I do not believe that is the Morning Star"?

How would Perry account for the Enterprise case?

In this paper, Evans responds to Perry's criticisms of Frege and develops a Fregean account of the senses of demonstratives and indexicals.

Evans first recounts the arguments in Perry's paper, noting that the crucial assumption is "that a Fregean sense of any singular term must be either the sense of a definite description or 'intimately related' to such a sense" (p. 292). Given that assumption, Perry's arguments succeed. Evans claims, however, that the assumption is unjustified. In the second section, he elaborates a conception of sense that he claims has no such commitment.

Evans's discussion is extremely abstract, but the central idea is fairly easy to understand. Consider the terms "Hesperus" and "Phosphorous". To understand such a term, Evans says, one must know what it refers to: "an understanding of the language must be capable of being regarded as involving knowledge of the semantic values of expressions" (p. 293). If so, however, then one might worry that understanding these two terms would involve the very same thing: knowing that they refer to Venus. But, Evans says, "Frege's idea was that to understand an expression, one must not merely think of the reference that it is the reference, but that one must, in so thinking, think of the reference in a particular way"; the sense of the expression is the particular way in which one must think of its referent to understand that expression (p. 294). So there is a particular way one must think of Hepersus (i.e., Venus) in order to understand the expression "Hesperus", and a different way in which one must think of it to understand the expression "Phosphorous".

To put it slightly differently, consider the following three statements:

1. "Hesperus" refers to Hesperus
2. "Hesperus" refers to Phosphorous
3. "Hesperus" refers to Venus

If one can know that Hesperus is a planet without knowing that Phosphorous is a planet, then one can equally know (1) without knowing (2) or (3). Evans's idea, then, is that understanding "Hesperus" involves knowing (1). Not only is one not required to know (2) or (3), but if one knew only (2) and so did not know (1), then, Evans wants to say, one would not understand "Hesperus" (or at least, would not understand it as other speakers do).

Can you now explain what Evans means when he says that, to understand "Hesperus", "one must not merely think of the reference that it is the reference, but that one must, in so thinking, think of the reference in a particular way" (p. 294)?

Evans says that, at this stage, we may leave the notion of a "way of thinking" at an pre-theoretical level. But he also indicates that he has a more sophisticated notion in mind.

If the intuitive notion needs to be supplemented,¹ we can appeal to the general idea of an account of what makes it the case that a thought is about the object which it is about; two people will then be thinking of an object in the same way if and only if the account of what makes the one person’s thought about that object is the same as the account of what makes the other person’s thought about that object. (p. 294)

Evans does not say much more about this idea yet, but it will become important later.

One of Evans's key innovations here is the thought that, on this conception, sense is, in a certain respect, not independent of reference: "If sense is a way of thinking of reference, we should not expect to be given the sense of an expression save in the course of being given the reference of that expression" (p. 294). This leads to the idea that it is perfectly consistent with Frege's ideas to recognize what Evans calls "Russellian singular terms": terms that could not have the sense they do without also having a referent (p. 296). And that idea will play a crucial role in what follows.

Evans tends to talk of Russellian terms as ones that could not have the sense they do without having some referent. But most interpreters have concluded that what he really means is: without having the very referent they do. So the senses of Russellian terms are specific to a particular referent.

On pp. 301-3, Evans argues in detail that there is no argument from the claim that terms with the same referent can have different senses to the conclusion that those terms are non-Russellian, i.e., that they could have the sense they do without having a referent at all. The issue is whether senses, as Evans understands them, can do the work that Frege needs them to do: explain differences of cognitive value. Evans says that they will, so long as his principle (P) is satisfied:

If the account of what makes a subject’s thought T₁ (about x to the effect that it is F) about x is different from the account of what makes his thought T₂ (about x to the effect that it is F) about x, it is possible for the subject coherently to take, at one and the same time, different epistemic attitudes towards the thoughts he entertains in T₁ and in T₂. (p. 301)

Evans claims that (P) is "very plausible" and, more importantly, that there is no reason to believe that the only accounts that could satisfy (P) would involve the subject having some identifying description of x.

The evaluation of principle (P) obviously depends, to a large extent, upon how the notion of an "account of what makes a thought about the object it is about" is spelled out, and we'll get more help with that notion later in the paper. But Evans does seem to think that, even absent such details, (P) is plausible. Is it? Why or why not?

In section III, Evans begins his positive account of what the sense associated with a given utterance of "today" might be. Evans's first point is that of course there is a particular way one must think of a given day in order to be thinking of it as 'today': not any way of thinking of that day will do. This does not by itself help us to understand what way that is, or what 'ways of thinking' are, in general. But Evans goes on to link this way of thinking of a day with certain cognitive dispositions that a thinker might have. The idea, I think, is clearer with "here": To think of a place as here is, in part, to be disposed to judge thoughts about 'here' as true or false on the basis of one's perceptions of one's immediate environment. What makes it possible to have this disposition with respect to a given place is simply being there (and making use of certain of one's perceptual and cognitive capacities). In particular, one does not need to be able to identify that place by description.

Evans writes: "To give an account of how a thought concerns an object is to explain how the subject knows which object is in question" (p. 303). What does Evans mean by that? How is it different from being able to pick the object out by description? (These are not easy questions. They go to the very foundations of Evans's view.)

Evans later (p. 315) gives an abstract summary of how this is supposed to work. He imagines these 'accounts' taking the form:

The subject S is thinking of object X at time T iff R(S,X,T)

I.e., some particular relation must hold between the subject and the object of thought at a given time for the subject to be thinking of the object at that time; which relation this is is supposed to fix the sense of the thought in question. What makes Evans's view externalist is his denial that the subject need have any thoughts about this relation: It's enough to stand in this relation to the object; one does not have to think of oneself as standing in that relation (see p. 320).

Evans then claims that, to understand a particular utterance of "Today is F", one must think of the day in question as today, meaning: One must think of it in such a way that what makes one's thought be about that day is one's disposition to judge thoughts about it as true or false in the way described. So there are two parts to Evans's view here: A claim about thought, specifically, about what the senses of certain 'indexical' thoughts are like; and a claim about language, which is meant to explain what kinds of thoughts one must have to understand indexical utterances. (Note that these views are, to some extent, independent.)

We'll not discuss section IV. It is on 'cognitive dynamics'. The central question is e.g. whether it is possible to have the same thought tomorrow that one expresses today by saying "It's cold today". Specifically, how does one have to think of today tomorrow in order to be having the same thought? Obvious answer: One has to think of the day as yesterday. But why? It's much less clear what the answer should be to corresponding questions about 'here'-thoughts, but the key idea is to try to investigate the circumstances under which one's belief persists over time, or as one moves about, and Evans grounds that idea in our capacity to keep track of objects over time. The section is extremely suggestive, and has been influential, but there is too much in this paper for us to discuss everything.

In section V, Evans turns his attention briefly to "I"-thoughts. There are two main topics of discussion. The first (pp. 312-4) is whether the idea of "unshareable" thoughts is so unFregean as to be completely unavailable to Frege. Evans denies that it is, but the question how Frege should be read is not the critical one here.

Evans's main argument here consists in: (i) distinguishing shareability from objectivity; (ii) claiming that it is the latter that really matters to Frege, since objectivity is what distinguishes senses from ideas; and (iii) insisting that unshareable thoughts can still be objective. How successful is this argument? What does the objectivity of an unshareable thought consist in?

The second topic concerns the extension of this idea to "now" and "here". As Perry had noted, Frege will need a "primitive" way in which one might think of each time and each place, which Perry derides as "very implausible". Evans responds that there is nothing at all implausible about the idea that certain ways of thinking of objects should only be available if one stands in certain spatial, temporal, or causal relations to those objects. That is, his view is that, in virtue of being in a certain place, one is able to have a certain kind of thought about that place (a 'here'-thought about it) that one otherwise could not have. (Note that the view is not, absurdly, that one can only think of that place if one is there. It is that one can only think of the place in a certain way if one is there.)

Evans does not actually talk much about demonstratives (as opposed to indexicals) in this paper—he develops views about this topic in his book Varieties of Reference—but in a way they are the most natural case for his view. The idea would be that, at least for so-called "perceptual" demonstratives, the way of thinking that is associated with them involves connecting thoughts that involve that way of thinking with certain information one is receiving perceptually. Can you see how this might allow Evans to deal with the Enterprise case?

In section VI, Evans argues that Perry's view is, at best, a "notational variant" of Frege's. If you understand lambda-notation, then you might want to read through this section. The argument is not difficult. Evans also argues that, so understood, Perry fails to say anything illuminating about the senses of indexicals. (Evans's was infamous for his cutting sarcasm, and that is very much in evidence here. His example is not to be imitated.)

---

¹I think this must be a kind of joke: Of course we cannot rest content with the 'intuitive' notion. It's not even clear there is any intuitive notion of 'a way of thinking of a thing'.

At the end of "Dthat", Kaplan briefly discusses the question how the referent of an uttered demonstrative is determined. He mentions three options:

1. The demonstrative refers to the object to which the speaker intends to refer.
2. The demonstrative refers to the object that various 'contextual cues' would lead a "linguistically competent public observer" to pick out. These 'cues' include particularly the demonstration (e.g,. pointing) that typically accompanies an utterance of a demonstrative.
3. The referent is primarily determined by 'contextual cues', but the speaker's intentions help "disambiguate" vague demonstrations.

In "Dthat", Kaplan endorses (3). In later writing, however—in particular, in "Afterthoughts" (one of the related readings)—he abandons that view in favor of (1).

Reimer here argues that Kaplan ought not to have changed his mind. More precisely, she wants to argue that the demonstrations that typically accompany utterances of demonstratives play at least some role in determining to what the uttered demonstrative refers. The discussion here is limited, as Reimer notes, to so-called "perceptual demonstratives", where the uttered demonstrative is supposed to pick out some object that both the speaker and her audience can perceive.

Reimer claims that there are three sorts of cases that pose a problem for (1):

1. Cases in which the speaker clearly demonstrates some object other than the one to which they intend to refer.
2. Cases in which there is a demonstration but there is no "directing intention".
3. Cases in which there is a "directing intention" but there is no associated demonstration.

As Reimer notes, such cases are ones in which the predictions made by Kaplan's new and old views come apart.

As sensible as this last suggestion may seem, one might for other reasons think it odd to build one's account of demonstrative reference around cases that are "unusual" or "atypical". This sort of issue arises several times throughout the paper. I'll not ask a question about this, as the question would be very vague and difficult. But if anyone wants to think about this....

The first sort of case Reimer considers involves her grabbing a set of keys and saying "These are mine", but just as she says this she mistakenly grabs a set of keys next to hers. Surely, Reimer says, in this case, her intention to refer to her own keys does not make it the case that she does so.

Reimer's discussion of this case is very brief. She seems to regard it as a straightforward refutation of view (1). Is it? What responses might be open to a defender of that view? (Is it true, in the example Reimer gives, that she does not intend to refer to the keys she picks up? Is it possible to use the distinction between speaker's reference and semantic reference here?)

The second sort of case Reimer considers is the Carnap–Agnew case that Kaplan introduced in "Dthat". Reimer claims that such cases pose a problem for view (1) for two reasons. First, there is no "directing intention", since Kaplan requires such an intention to be directed at a perceived object. So, second, since Kaplan's version of view (1) was only supposed to apply when there is a directing intention, we need some other account of what fixes the reference in this case. Presumably, it is the demonstration. But, as Reimer says, if the demonstration plays such a role when there isn't a directing intention, why shouldn't it also play a role when there is one?

As Reimer notes, Kaplan could respond to the first worry by broadening the notion of a "directing intention" so that it might be directed at unperceived objects. How plausible might that be? What are her worries about this move? (And is it true, in the example Kaplan gives, that he does not intend to refer to the picture behind him?)

If the response to the first worry just mentioned works, then the second worry evaporates. Why?

In the third sort of case Reimer considers, one makes a demonstrative utterance, such as "That dog is Fido", but makes no demonstration at all. Reimer argues that, in this case, one does not manage to refer to any dog at all: her claim is that "that", as it functions in "that dog", is an adjective, playing the same sort of role as "spotted"; that adjective is meat to mean something like "demonstrated"; so "that dog" is co-referential with "the demonstrated dog" or "the dog I am demonstrating".

Reimer's account seems to entail that "that F" can never appropriately be uttered without an accompanying demonstration (at least in cases of perceptual demonstratives). Is that true? Can you think of any counterexamples to this claim? Can you modify the case Reimer mentions—her dog in the park—in such a way that the utterance of "That dog is Fido" seems both appropriate and true despite the absence of any demonstration?

Note that answering the questions just asked seems to presuppose our getting somewhat clearer about what a "demonstration" is supposed to be. So what are demonstrations exactly?

Siegel's goal here is to defend something akin to Kaplan's view from Reimer's criticisms. Siegel holds that, in the case of perceptual demonstratives, "the reference of [a] use of a demonstrative is fixed by a perceptually anchored referential intention" (p. 1). More precisely, Siegel's view is limited to what she calls "Basic Cases", which have three features: The speaker perceives a certain object; they intend the demonstrative they utter to refer to that object; and that intention is "anchored" by the perception, i.e., the intention and the perception are related in some intimate way. (Siegel elaborates this idea a bit further on p. 3, but ultimately leaves it relatively unexplained.) Siegel's claim is that, if these three conditions are satisfied, then the demonstrative refers to the intended object and does so because of the speaker's intention.

In section 2, Siegel motivates her view, which she calls Limited Intentionism, by highlighting the role that perception of an object plays in typical uses of demonstratives. In section 3, she presents Reimer's example about her keys, which is a prima facie counterexample to Limited Intentionism. Siegel agrees with Reimer that the speaker refers not to her own keys (as she intends) but to her office-mate's keys. But Siegel insists that the speaker also has another intention: an intention to refer to her office-mate's keys that is grounded in (anchored by) her tactile (rather than visual) perception of those keys.

As Siegel notes, in Reimer's original version of the example, the speaker utters "These are my keys" while she is grabbing her officemate's keys. She thus has, at that moment, a perceptual (tactile) experience of those keys, and so can have a perceptually anchored intention to refer to them. Are there ways of re-casting Reimer's example that would undermine this response? (Generally: Whenever someone responds to an example by emphasizing a certain feature of it, you should immediately ask yourself whether that feature is really essential.)

In this example, then, the speaker has multiple intentions that conflict with one another. The crucial question for Intentionism (of any variety) then becomes which of these intentions is supposed to fix reference. What's needed is thus "a criterion that says which one of a speaker's various intentions fixes demonstrative reference" (p. 7).

It seems worth emphasizing that, even if Reimer had grabbed the correct set of keys, she would still have multiple intentions, according to Siegel. It is just that, in this case, they would not conflict.

Siegel offers such a criterion in section 4: the intention that fixes reference must be one that is perceptually anchored. That takes care of the case described by Wettstein. But Reimer's case might seem to be one in which the speaker has two perceptually anchored intentions: After all, they feel their office-mate's keys, to which they intend to refer, and also see their own keys, to which they intend to refer. Siegel argues that Reimer's example isn't of quite this structure: The speaker isn't looking at their keys while also feeling them. But there are other examples that do have that structure. Siegel argues, however, that, in such cases, it's not clear that the case is a counterexample any longer: Rather, in such circumstances, the demonstrative is (more or less) ambiguous.

Siegel writes: "As Reimer describes the case, the speaker does not lack eye-hand coordination. She is not looking at her own keys while grabbing her officemate's; rather, she first sees her keys and then, without continuing to look at them, grabs the wrong set" (p. 6, my emphasis). In fact, though, Reimer does not seem to specify this aspect of the case one way or the other. But the crucial question is whether this aspect of the case makes a difference. What difference exactly does Siegel need it to make? Does it actually make that difference?

In section 6, Siegel addresses a possible counterexample to her view. In this case, a speaker intends to tell someone to pick up a certain ball they can see, but they manage to point (we can imagine, quite directly) at a different ball (to which they may have no perceptual connection at all). Initially, Siegel concedes, one might suppose that they have referred to the ball at which they pointed, not the one to which they intended to refer. But, she argues, this judgement is unstable: There are very similar examples in which the contrary judgement seems just as reasonable. Most importantly, she argues that we cannot just defer to 'the' judgement of a "linguistically competent public observer" (as Kaplan put it). Such people's judgements will vary depending upon their background knowledge, and so forth.

Siegel discusses the Carnap-Agnew example in this connection. Can you construct a version of that example in which it seems 'intuitive' that Kaplan actually manages to refer to the picture of Carnap, despite the fact that it has been replaced with a picture of Agnew?

The central idea in this paper is that neither the associated demonstration nor the speaker's intentions is what fixes the reference of an uttered demonstrative. Rather, reference is supposed to be determined by what Mount calls "mutually recognized salience". Mount does not explain in detail what 'salience' is supposed to be, beyond saying that it involves 'perceptual or cognitive focus'. But the significance of the notion can be illustrated by a version of an example that seems to have been due to David Lewis.

Imagine that you are watching a dance team when suddenly one of the participants falls over. "She must be tired", your friend says to you. In the context, it may be completely clear to whom your friend has referred, and that is because, as it is usually put, the dancer in question has "made themselves salient" to both of you by falling over. Mount's idea, then, is that this "mutual salience" is what makes the demonstrative refer to the object to which it does. Mutual salience can be achieved in a variety of ways, including by pointing, and through the audience discerning the speaker's intentions. But what matters, according to Mount, is whether mutual salience is achieved, not how it is achieved.

In "Three Views of Demonstrative Reference", listed as optional for last session, Reimer argues that a speaker's intentions can sometimes determine reference, but only if there is no associated demonstation (as in the example just given). On the other hand, Reimer had claimed, if there is an associated demonstation, then it always over-rides the speaker's intentions. Her argument largely involves examples of the sort we saw in her other paper.

In response, Kent Bach distinguished two kinds of intentions, which Mount calls "thought intentions" and "referential intentions". A 'thought intention' is just any intention to refer to an object. But a referential intention is, specifically, "the intention that one's audience identify, and take themselves to be intended to identify, a certain item as the referent by means of thinking of it a certain identifiable way" (Bach, p. 143). For example, one might intend that one's audience should focus upon a specific dog, and do so because it is the dog to which one has just pointed. So this sort of intention is specifically tied to how one intends to communicate with one's audience. Moreover, Bach argues that, since these intentions are specifically communicative, they are the ones that matter for reference.

How, in terms of this distinction, might we account for Reimer's example of the keys?

On p. 151, Mount argues that the specific nature of referential intentions, in Bach's sense, all but guarantee that his view will answer the question to what a particular uttered demonstrative referred in the same way that Reimer's does, at least whenever there is a accompanying demonstration (and they agree which demonstration that was). What's her argument?

In section 2, Mount begins the argument for her own view. She develops a version of the Carnap-Agnew case in which the picture that is most salient to the conversational participants is that of Carnap, despite the fact that the speaker has pointed to the picture of Agnew. In this example, neither of the participants actually looks toward the wall. Rather, they 'know' (that is, they think they still know) what picture hangs there, and they both take the speaker to have been referring to the picture of Carnap. So, in this case, Mount claims, salience overrides the demonstration.

In the second paragraph on p. 153, Mount mentions, very briefly and in passing, several other variants of the Carnap-Agnew case one might construct. Construct some. Does Mount's view seem to give a plausible account of those cases?

Moreover, Mount claims, "It is the salience itself, not the speaker's intention per se, which is more important here" (p. 153). That does not make the speaker's intention irrelevant, however, since it "can influence what is salient to the hearer, if that intention is apparent for any reason" (p. 153). So both demonstrations and the speaker's intention have a role to play, but the role is one of making an object salient. As Mount emphasizes, this makes salience not an objective notion, but one that is relative to particular speakers, what they know about one another, and so forth.

Can you give an example that illustrates this latter point? That is, an example in which what is salient to the speaker and her audience very much depends upon what they know about each other, so that it would not be salient to an arbitrary third party who overheard them?

Mount's view, then, is that an uttered demonstative refers to the object that is mutually recognized as maximally salient, and that it fails to refer, if there is no such object.

One might argue that, if mutually recognized salience is what determines reference, then the referent must also match the speakers' referential intentions. How might such an argument go? What lesson should we draw from it? (There are some relevant remarks on pp. 159-60.)

In section 3, then, she considers a series of objections to this view. The general worry here is that Mount's view seems to make the speaker's ability to refer to an object too dependent upon the audience's receptivity.

Mount argues that we need to be careful about who we consider to be a participant in a conversation, and that doing so should alleviate some of the worry about the audience-dependence of reference on her account. And she claims that "If attentive and cooperative interlocutors cannot figure out what the speaker intends for the demonstrative to refer to, then the speaker has failed to fulfill her obligation. Reference has failed because the audience misunderstood" (p. 156). Does that seem reasonable?

A more principled worry is that Mount's view conflates the metaphysical sense of "determine" with the epistemological one. She argues in response that, although there certainly is a distinction here, these two notions are, in the case of demonstratives, closely related. How satisfying is this response?

In section 4, Mount reinforces her view by arguing that it explains why, in many cases, no demonstration is required for demonstrative reference to succeed: because the object in question may already be sufficiently salient. Moreover, Mount claims, there is no reason to think that salience should be important when there is no demonstation, but be irrelevant the moment there is one. Demonstrations are required only in so far as they are needed to bring an object to salience.

Can you give other, natural examples of uses of demonstratives without a demonstration? And here's a question for those with some experience with linguistics: Might some uses of demonstratives that might otherwise be classed as anaphoric instead be treated as involving mutual salience?

In section 5, Mount further argues that her view also explains another role that speakers' intentions have often been thought to play: helping to disambiguate an otherwise ambiguous demonstration.

Mount notes that "there are no standard requirements for what a speaker must do to make something maximally salient to her interlocutors, although speakers do have a repertoire of linguistic and non-linguistic strategies from which to choose" (p. 160). Can you think of any examples of such strategies besides the ones Mount mentions?

Finally, in section 6, Mount argues that it is a virtue of her view that it encompasses also what Quine called 'deferred ostension': uses of demonstratives to refer to objects other than the one at which one directly points. A simple example would be pointing at a car covered with parking tickets and saying, "That person is going to be unhappy". Mount argues that there is really no alternative to her view in this case: There is no general way of explaining what the relation is supposed to be between the object directly ostended and the one indirectly ostended. Mount's point, ultimately, is that, in all such cases, what matters is which object is mutually salient to the conversational participants. It does not matter how it came to be so.

Early in the paper, Mount mentions that she "will proceed as if it makes sense to talk about demonstratives themselves referring", contrasting that view with one according to which "speakers just use demonstrative expressions to refer the audience to an object, and reference succeeds when the audience identifies the object intended in the way intended" (p. 146). She argues in the last section that her arguments would be just as compelling under the latter, more Strawsonian construal. In fact, however, one might suggest that her arguments would be much stronger under that construal, indeed, that many of the objections she considers in section 3 could be much more easily answered from a Strawsonian point of view. How so?

This paper is focused on issues concerning belief attribution: the meaning of sentences of the form "N believes that P". As we know, there are reasons to suppose that substitution of co-referential expressions should fail in such contexts: It would seem, prima facie, as if

(H) Hammurabi believes that Hesperus is visible in the evening.

could be true even though

(P) Hammurabi believes that Phosphorous is visible in the evening.

was false. But some people have been willing to bite this bullet and claim that substutition of co-referential expressions is permitted everywhere (excepting quotation).

Such a view is known as a Russellian view of attitude ascriptions. Its main advantages are (i) that it gives a relatively simple semantics for such sentences and (ii) that it saves us from having to answer difficult questions about what role, exactly, senses are supposed to be playing.

Of course, someone who thinks that (H) and (P) must either both be true or both be false needs to explain why we have an "intuition" to the contrary. And one obvious way to do this is to insist that the contrast between (H) and (P) is not semantic but pragmatic: Even if (P) is true, perhaps uttering it would in some way be misleading. The details of how that account might go will not matter for our purposes.

Saul gives a number of examples, at the beginning of the paper, to illustrate the problem in which she is interested. With one exception, most of these involve superheroes. And one might wonder, as indeed some have, if these are not special cases. Can you think of other sorts of examples? A particularly important question here is whether we can always generate such examples for pairs of names that give rise to Frege cases.

Note that the claim is that there is a felt contrast between these two sentences

(10) He hit Clark Kent once, but he never hit Superman.
(10*) He hit Clark Kent once, but he never hit Clark Kent.

and that it is the same sort of felt contrast as between (H) and (P). And it's important to note that, as in that case, the contrast between (10) and (10*) does not depend upon our being ignorant of the identity of Clark and Superman.

The question is how we should respond to these examples. Saul first considers whether we might simply say that, in fact, (10) can be true even though (10*) cannot be, since "Clark Kent" and "Superman" do not refer to the same thing. She quickly discards a couple very unpromising attempts (pp. 103-4) and then turns to the suggestion that we should think of the names as referring to "temporal phases". The natural application of this idea is to the "Leningrad" vs "St Petersburg" case: The former refers to the "time slice" of a certain city in Russia lasting from 1924-1991; "St Petersburg" refers to the "time slice" lasting from 1703-1924 and then from 1991 to the present. As Saul notes, it's harder to know how to allocate temporal parts to Clark and Superman, but she sets that worry aside, having more serious objections to press.

The first problem is that this view threatens to make

(11) Superman is Clark Kent.

false, since the whole point is that the time slice that constitutes Superman has to be different from the time slice that constitutes Clark. If so, however, then cases like

Lois believes that Clark can fly.
Lois believes that Superman can fly.

aren't actually counterexamples to the substitution of identicals. And we probably do not want to have to say that Superman isn't Clark Kent (though there have been some who have explored that idea). How many times does he get to vote?

There are two ways to avoid that consequence: One is to postulate a more complex semantics for proper names, so that "Superman" sometimes refers to a person and somtimes refers to a time slice of a person. An alternative might be to locate the ambiguity in "is", so that, in (11), "is" means something like: are temporal phases of the same temporally unified object. The problem with this view is that it is insufficiently general, as examples like:

(12) Clark Kent can fly, though he conceals this fact.

show. So it looks as if only the ambiguity view will do.

Saul does not say very much about this "ambiguity" view, other than that it "will be undesirable to many". Does it seem plausible? What pros or cons might it have?

If proper names are ambiguous in this sort of way, then one might wonder whether that shows that (H) and (P) aren't, in fact, counterexamples to substitution. What else would one need to show to make that claim?

So, it looks as if it is going to be difficult, at best, to defend the view that (10) can be true. But we do not have to explain the contrast between (10) and (10*) semantically: We can instead explain it pragmatically.

Although Saul does not mention this point, one might think that her examples (1*) and (1**) pose serious problems for the ambiguity view. Why?

Saul remarks that the pragmatic strategy works most smoothly when what we need to explain is why we are reluctant to say something like "Clark went into the phone booth, and Clark came out": The utterance would implicate something false. What falsehood might it implicate? She also suggests that it isn't as obvious how to explain why we do say things like (10). Presumably, it ought to be because utterances of (10) can implicate something true, even if they are themselves false. What true might they implicate?

At this point, however, Saul reveals the trap she has been laying for the Fregean:

The proponent of this [pragmatic] response...now has a choice. She must decide whether or not to accept a perfectly parallel account of our intuitions about attitude reports.... According to [it], substitution of co-referential names in attitude reports preserves truth conditions but may result in the generation of new, and misleading, pragmatic implicatures. It is these implicatures which result in our (mistaken) tendency to say that [(H)] may betrue while [(P)] is false.... The main argument against [this] theory has been that it requires the violation of our intuitions about substitution. But the current approach to substitution in simple sentences requires what is apparently a perfectly parallel violation of intuitions, accompanied by a perfectly parallel appeal to pragmatics. The advocate of this approach owes us a reason for supposing that one set of intuitions deserves to be taken so much more seriously than the other. (pp. 106-7)

This does not seem a particularly comfortable place for the Fregean to be. And that might make one want to explore again the possibility that (10) can be true. As we'll see next time, some have gone that route.

Saul's argument here depends crucially upon the claim that the cases of (H) and (P), on the one hand, and of (10) and (10*), on the other, are "perfectly parallel". Might it be possible to resist that claim?

Saul's paper "Substitution and Simple Sentences" inspired quite a number of different attempts to solve the problem she'd introduced. We look at the first of them here, by Graeme Forbes, and Saul's response to him.

In earlier work (cited in the paper), Forbes had developed a broadly Fregean account of attitude attribution. The details of the account do not matter very much here, and he explains it well enough for our purposes in the paper. The basic idea is that, in a sentence like

(i) Lois believes that Superman can fly.

there is an implicit reference to a "way of thinking" of Superman that is picked out by the words used in reporting the belief. So (i) is supposed to have the meaning:

(i') Lois believes, so-labeled, that Superman can fly.

where that means (roughly) that the belief in question is one Lois would herself report using those words.

Here's an updated version of the example of Quine's that Forbes mentions. Some years ago, there was a football player named "William Perry" who was so large that he was known as "The Fridge". So The Fridge was so-called because of his size. But William Perry was not so-called because of his size. So substitution fails. Why? Because 'so-called' introduces an implicit reference to the version term used, and substitution changes which term is so referenced.

As Forbes notes, many of Saul's examples plausibly involve psychological attitudes of some sort. And, in that case, whatever machinery a Fregean might have available to account for failures of substitution in attitude contexts can equally well be deployed to account for those examples. As Forbes also notes, however, not all of Saul's examples are amenable to that same treatment.

Can you spell out in more detail than Forbes does how his analysis of (8) is supposed to work?

On the other hand, however, Forbes suggests that a somewhat similar move is still available in other cases. Thus:

(9) Clark went into the phone booth and Superman came out.

might be read as:

(11) Clark, so-attired, went into the phone booth and Superman, so-attired, came out.

A somewhat similar suggestion, based upon Forbes's discussion on pp. 111-2, might be that (9) should be interpreted as:

(11') Clark, so-presenting, went into the phone booth and Superman, so-presenting, came out.

One might expect this sort of proposal to generalize more easily to other cases.

Forbes really does not say very much about how this kind of proposal is supposed to generalize to other cases. So consider some of the other cases Saul mentions: How would Forbes want to handle them? Are there cases that are particularly difficult for this kind of view?

In Saul's reply, she raises several problems of detail for Forbes's proposal. It should be obvious that, on a case-by-case basis, there might well be some sensible account, broadly along Forbes's lines, to be given. But the underlying worry here is that these various moves are ad hoc. What one would like is some general and principled method for determining what the "literal" content of a sentence like (9) is supposed to be that could generate a difference of truth-value between it and

(10) Clark went into the phone booth and Clark came out.

But Forbes says almost nothing about this.

Saul also raises a more principled worry, however, about exactly what "modes of self-presenation" are supposed to be. What is it supposed to be for Clark to "present as Clark"? Saul considers several possiblities and argues that none of them work for all cases.

Are there any plausible proposals about what it might mean for Clark to "present as Clark" that Saul has neglected? Are there cases for which those proposals seem not to work?

Ultimately, the lesson of Saul's discussion seems to be that such "modes of self-presenation" need to be non-psychological if they are going to do the work Forbes needs them to do, and it just isn't clear how they could be. But if they are psychological, then that means that all of Saul's cases have somehow to be unmasked as involving attitudes, even cases like (9) that do not seem to involve attitudes.

What's your own view about what we should say about these sorts of cases?

There are no reading notes for this paper.