Descriptivism, pretense, and the Frege-Russell problems.
Contrary to frequent declarations that descriptivism as a theory of how names refer is dead and gone, such a descriptivism is, to all appearances, alive and well. Or rather, a descendent of that doctrine is alive and well. This new version--neo-descriptivism, for short--is supposedly immune from the usual arguments against descriptivism, in large part because it avoids classical descriptivism's emphasis on salient, first-come-to-mind properties and holds instead that a name's reference-fixing content is typically given by (a cluster of) egocentric properties specified in terms of broadly causal relationships between a speaker and his environment: properties like being the actual individual called 'Aristotle' referred to by my informants' use of the name, being the actual individual called 'George Bush' whom I have seen/heard described as the U.S. President who started the second Gulf War, and so on. (1) (The addition of the rigidifier 'actually' is to make name-occurrences rigid; the rigidity of (most) names is not in contention.) What these neo-descriptivists claim is that the usual modal, semantical, and epistemological arguments against classical descriptivism don't get much of a foothold against this new version, especially if we don't insist (as we shouldn't) that speakers be able to state these properties on demand. It is enough, that these are properties implicit in the semantic judgments of speakers. Recent attacks notwithstanding, such a neo-descriptivism has struck many philosophers as a credible and worthy successor to classical descriptivism. (2)
It almost goes without saying that part of the conventional descriptivist wisdom accepted by these neo-descripdvists is that their theory is able to solve the problems that motivated the rise of descriptivism at the start of the twentieth century: the problems of why true identity statements can be informative, how to analyze singular negative existentials, and why co-referring names can't always be interchanged salva veritate in propositional attitude reports (the Frege-Russell problems, for short). (3) Almost without saying: for what led these neo-descriptivists to their version of descriptivism was not, by and large, the desire to solve these familiar problems but to get the account of reference determination right in the face of the Kripke-Donnellan arguments against classical descriptivism. Still, once we have a credible descriptivist account of the reference-fixing content of name uses, it seems that little stands in the way of directly using that account to solve the other problems, in particular the informativeness and negative existential problems. (Puzzles about belief are much less straightforward, as Russell already pointed out.)
Such a (strong) neo-descriptivism about content can be contrasted with a weak form of neo-descriptivism according to which reference is fixed in the manner described by strong neo-descriptivism but where it is not assumed that reference-fixing descriptions yield semantic content. (4) Until further notice, it is the strong form of the theory I will have in mind. (Strong) neo-descriptivists think that their theory can do what Frege and Russell wanted descriptivism to do, without running into the problems that face descriptivism's supposed nemesis--the New Theory of Reference, often touted as the new orthodoxy in philosophical semantics following the "death" of descriptivism. For the New Theory's commitment to a Millian or direct reference theory of names means that it can't explain in semantic terms just how true identities can be informative, or how true negative existentials can have any content, let alone a true content. (5) If, as Millianism holds, the semantic content of a name is just its referent, there just aren't enough contents to explain content in these cases. The neo-descriptivist has no such "not enough contents to explain content" problem.
I think that there is much to be said for the neo-descriptivist program, but I also think defenders of neo-descriptivism need to do rather more than they have done so far where the Frege-Russell problems are concerned. The locutions that give rise to Frege-Russell problems--"Frege-Russell locutions"--have certain features that reflect a kind of remarkable underlying phenomenology and should make us suspicious of any simple-minded application of neo-descriptivist ideas to these problems. By contrast, a number of New Theorists have drawn attention to these features in an attempt to argue that the New Theory is in fact well placed to solve the problems, despite the apparent lack of promise afforded by its semantic framework. The tool they offer in support is the idea of semantic pretense.
In this paper I shall argue for another way of understanding these features. After showing why we need to take them seriously, I argue that the appeal to semantic pretense by New Theorists won't work. Among other things, it makes a mystery of the modal properties of Frege-Russell locutions. Focusing on the informativeness and negative existential problems, I then argue for a wholly different way of understanding the relevance of pretense to the Frege-Russell problems. It turns out that, on this alternative model, neo-descriptivism has much to gain from an appeal to pretense.
2. The Phenomenology of Frege-Rumell Locutions
Consider again the Frege-Russell problems: the negative existential problem, the informativeness problem, and the various problems about belief and other propositional attitude reports. As Mark Crimmins pointed out in an important paper some years ago (1998), one remarkable feature of all the locutions implicated in these problems is their Janus-faced character. On the one hand, most of us think that the contribution made by the names involved in these locutions to what the locutions are used to communicate is in many cases not exhausted by the objects denoted by the names. Thus, 'Hamlet doesn't exist' is (used to say something) true without 'Hamlet' denoting anyone; 'Hesperus is identical to Phosphorus' is true and informative even though there is only one object denoted by the two names 'Hesperus' and 'Phosphorus'; 'Hammurabi believed that only Hesperus, and not Phosphorus as well, was visible in the evening' is true despite there being only one object denoted by these names. On the other hand, we can also detect an intuitive sense in which the truth of the communicated content seems to depend on the disposition of various objects of reference alone. Intuitively, it is because of Hamlet's not existing that 'Hamlet doesn't exist' is true; it is because of Hesperus and Phosphorus bearing the relation of identity to each other (something they needn't have done) that 'Hesperus is in fact identical to Phosphorus' is true; it is because of the distinct beliefs Hammurabi had about these two celestial bodies--identical bodies, as it turns out--that 'Hammurabi believed that only Hesperus, and not Phosphorus as well, was visible in the evening' is true. Putting things this way is a natural way of registering our sense of what's going on.
This sense is, of course, most familiar in the case of negative existential, where it has often been used to support the doctrine that there are genuine fictional objects. But Russell realized it was rather more general. This is how he puts the point in the case of identity statements:
When you say 'Scott is the author of Waverley', you are half-tempted to think there are two people, one of whom is Scott and the other the author of Waverley, and that they happen to be the same. That is obviously absurd, but that is the sort of way one is always tempted to deal with identity. (1985, 115)
Russell is surely right on both counts: we are half tempted to think of identity this way, even though the intuitive thought that there are two things we are identifying is absurd. But, absurd or not, this intuitive underlying "phenomenology" is not easily disposed of. There is a remarkable robustness about the phenomenology: it erupts into our linguistic practices, not just in our sense of what is going on, and not in a way that we can easily suppress. And, as Crimmins emphasizes, a phenomenology of this absurdly profligate yet robust kind infects all the Frege-Russell locutions.
Consider negative existentials again. One might expect the unchallenged assertion of a negative existential 'N doesn't exist' to be a conversation-stopper, since the name N is somehow declared to be empty, to have no application. But this is simply not the case. What typically happens after an utterance of a negative existential is further, often even busier, conversation, seemingly about the very object declared to be non-existent. This is scarcely surprising, for negative existentials often leave open too many possibilities, and so cry out for explanation or clarification ('What do you mean, Miranda doesn't exist. Is she a fictional character, perhaps, or one of Anne's imaginary friends, or ...?'). Consider, for example, the way Time Magazine answered Canadian Prime Minister Mulroney's puzzled question "Who is Murphy Brown?" after Dan Quayle's notorious attack on Murphy Brown's morals:
(1) Murphy Brown does not exist. She is a much-loved TV character played by Candice Bergen. Murphy is a blond media anchor-goddess and wiseguy and now a defiantly unmarried madonna. (Time International, June 1, 1992)
Or consider informative identities again, say:
(2) Hesperus is in fact identical to Phosphorus.
Competent speakers of English treat such a sentence as trivially equivalent to the following (apparently) plural identity sentence:
(3) Hesperus and Phosphorus are in fact identical.
Philosophers and linguists commonly think of the plural in such a case as a bogus plural, a syntactic product of the fact that we have a conjunction of two referring expressions. But that view is not reflected in the behavior of ordinary speakers. Even knowledgeable speakers are just as likely to phrase a statement like (3) in terms of a more blatant plural, for example,
(4) Those two famous celestial bodies of Babylonian astronomy, Hesperus and Phosphorus, are in fact identical.
Competent users of English simply do not see the second one as deviant, even if philosophers do.
Or consider belief reports again. We don't simply say:
(5) Hammurabi believed that only Hesperus, and not Phosphorus as well, was visible in the evening.
Knowledgeable speakers are just as likely to phrase this report in terms of a plural de re construction, for example,
(6) Of those two famous celestial bodies of Babylonian astronomy, Hesperus and Phosphorus, the first was believed by Hammurabi to be visible in the evening but not the second.
Here the attribution of plurality is no longer confined within a de ditto report, but seems to be made to the very objects the belief is about. (6)
3. Semantic Pretense
The question, then, is how to account for the phenomenology, given its remarkable robustness. But if it is indeed implausible to suppose that the phenomenology involves some kind of mistake on the part of speakers, aren't we then committed to taking the phenomenology seriously? And doesn't that in turn entail that we should adopt a realistic construal of the phenomenology, accepting that there really are objects that don't exist (perhaps Meinongian objects of some kind, as in Zalta 1988), and even objects that are distinct, yet in the relevant sense identical (the "guises" of Casteneda 1989, say)? And shouldn't we then say that it is the disposition of these objects that makes the various statements (1)-(6) true?
Mark Crimmins (1998) thinks that we can take the phenomenology seriously,),et steer a middle road between such a realistic construal of the phenomenology and a standard error theory. As a supporter of the New Theory of Reference, he believes that the very profligacy of the phenomenology can be turned to its advantage. Roughly speaking, even if the New Theory doesn't admit "enough contents to explain content," in Frege-Russell locutions speakers seem to talk as if there are contents aplenty, and this doing-as-if might be enough to generate a reasonable replacement proposition. More precisely, the view is that such locutions involve a kind of semantic pretense, a doing-as-if the world is a certain rather profligate way, in order to make a serious true-or-false claim about the world, with facts about what makes something true in the pretense fixing the genuine semantic properties of the serious claim.
The idea, which owes much to Ken Walton's make-believe account of fiction (1990), is broadly this. Suppose we have in place the notion of a game of make-believe: roughly, a stretch of imaginative activity, involving one or more actors, subject to rules that determine what is to be imagined as part of the game. In Walton's influential development of this idea, children as well as adults play such games on the basis of props that mandate that they imagine certain things. Thus, a children's game may require its participants to imagine that a certain oddly shaped stump is a bear, that actions done to the stump are acts against the bear, and so on. Another sort of game, indulged in by children as well as adults, involves reading or listening to a story, a game that requires its participants to imagine that certain events really happened (that there really was a famous detective called 'Holmes' who lived in Baker Street, for example). In Walton's terminology, a sentence, proposition, or utterance is fictional or fictionally true (true in the game of make-believe) when the conventions pertaining to props of the relevant kind require participants in the game to imagine its being true. (7) Principles that tell us what kinds of circumstances make a certain kind of proposition or sentence (-utterance) fictionally true are called 'principles of generation'.
Walton emphasizes the role of the imagination, but that suggests a degree and quality of involvement that may not be necessary to our participation in forms of make-believe. To use a phrase used earlier, it is often enough to "do as if" the world is a certain way, where that may involve forms of linguistic and cognitive activity that fall well short of full imaginative involvement. As to why speakers would engage in such less committed forms of pretense, one kind of reason deserves special mention, The following example suggests that doing so sometimes provides a natural and efficient way in which a speaker can make a serious assertion about the world. Consider Johnny's mother as she watches Johnny put a bungy cord around the stump and pull on it. Johnny's action makes it fictional that Johnny has lassoed a bear, a fictional truth that is likely to engage Johnny's imagination to a significant degree. But it is not what interests his mother as she impatiently tells him that "that's the second lasso you have broken on the bear this afternoon." What would make that utterance fictionally true is the obtaining of a certain real world truth condition: Johnny's breaking a second bungy cord on the stump. Her utterance shows her transient involvement in Johnny's game, an involvement that shows that she is "doing as if" the bungy cord is a lasso but not necessarily a strongly imaginative involvement in that game. Her purpose in this case is not primarily to pretend to assert something through her involvement in the game, but to genuinely assert something about the real world, namely that Johnny has broken a second bungy cord on the stump. She thereby indulges in what Crimmins calls a piece of semantic pretense, saying something that is, as she thinks, fictionally true in order to assert that the world is a certain way--the way that makes her utterance fictionally true.
Following Mark Richard (2000), let's call this phenomenon 'piggy backing'. More precisely, to piggy back is to make an utterance u within a pretense in which u has a real world truth condition c, thereby asserting a proposition that is in fact true iff c obtains. (For reasons that will become clear later, Crimmins thinks that we can't in general assume that what is asserted is simply that c obtains.) In Crimmins's application of this idea to the case of propositional attitude reports, an utterance of a sentence of the form 'Hammurabi believed that Hesperus was F and Phosphorus was G' is embedded in the pretense that there are two objects, one named 'Hesperus', the other 'Phosphorus', and that, thinking of a thing using the 'Hesperus' mode of presentation constitutes thinking of one of them, and that thinking of a thing using the 'Phosphorus' mode of presentation constitutes thinking of the other. The pretense thus employs principles of generation along the lines of:
P1 If (and only) if someone has a thought, belief, etc., involving the Hesperus-mode (alternatively, the Phosphorus-mode), then it is fictionally true that they have a thought, belief, etc. about the object named 'Hesperus' (alternatively, the object named 'Phosphorus').
Now consider my utterance of 'Hammurabi believed that only Hesperus, and not Phosphorus as well, was visible in the evening'. Given my little transient game of make-believe, it is thus fictionally true that I have said that Hammurabi believed of the object named 'Hesperus' that it was visible in the evening, while disbelieving of the object named 'Phosphorus' that it was visible in the evening. But in this kind of case my interest is clearly not in the content of the pretense itself but in what it says about Hammurabi's real belief state. According to Crimmins, we engage in this kind of pretense in order to piggy back: what I am trying to assert with my utterance (really assert) is a claim in fact true iff the real world truth condition of my utterance holds, and hence (by P1) a claim in fact true iff Hammurabi believed under the Hesperus-mode, while disbelieving under the Phosphorus-mode, that the object presented by the mode was visible in the evening.
It is far from obvious, however, how this kind of account might be extended to deal with a report like 'Sally knows that Hesperus and Phosphorus are in fact identical', for in the pretense the sentence 'Hesperus and Phosphorus are in fact identical' seems to say something strictly impossible. Crimmins's answer is that when we speak of identity we speak as if it is a relation that can hold between distinct objects. He suggests that such talk therefore involves an additional bit of pretense, namely that "with certain of our linguistic devices that normally express identity, we can express a relation which can hold between distinct objects" (1998, 35), where it is fictionally true that this promiscuous relation holds between two objects when these objects result from pretending-apart a single object, using two different modes of presentation. In particular, we have:
P2 If (and only if) the Hesperus- and Phosphorus- modes are modes that present the same thing, then it is fictionally true that the (distinct) objects named 'Hesperus' and 'Phosphorus' are identical.
It follows that, through my piggy backing on the pretense and relying on P2 (and an obvious embedding principle for propositional attitude reports), my utterance of' (Sally knows that) Hesperus and Phosphorus are in fact identical' is used to assert a claim that is in fact true just in case (Sally knows that) the Hesperus- and Phosphorus-modes are modes of the same thing.
So that is how semantic pretense allows us to accommodate the profligate object-involving phenomenology of propositional attitude reports as well as of plural sounding identity claims. Crimmins himself thinks that where attitude reports are concerned such an account vindicates important New Theory intuitions, since at one level it vindicates the idea that such reports can be seen as semantically innocent, with neo-Russellian logical syntax, even if at another level it suggests that we can't escape something like Fregean truth conditions.
That leaves us with just one other case, the case of negative existentials. This case had already been given a pretense treatment by Walton (1990); here I'll simply present Crimmins's version of Walton's view. Again appealing to modes of presentation, Crimmins thinks that in uttering something like 'Hamlet doesn't exist', the speaker pretends that all modes of presentation, including the Hamlet-mode, present something, and that, in addition, 'exists' is a predicate that partitions the domain of discourse into two sets, the things that exist and those that don't. The principle of generation for 'exists' then yields:
P3 If (and only if) the Hamlet-mode presents an individual, then it is fictionally true that the object named 'Hamlet' exists.
By piggy backing, it follows that an utterance of '(Sally knows that) Hamlet doesn't exist' is used to assert a claim in fact true just in case (Sally knows that) the Hamlet-mode doesn't present anything.
4. Some Problems
So there we have it: a unified explanation of how Frege-Russell locutions can be used to assert something genuinely true, one that accepts the apparent logico-semantic properties of these locutions and to that extent takes the accompanying phenomenology seriously, without, however, accepting a realistic construal of the phenomenology.
But despite its considerable elegance and power, this explanation faces some thorny problems. First of all, note the relatively underdetermined nature of Crimmins's understanding of modes of presentation. He says little more than that the Hesperus-mode is whatever allows us to classify together certain thoughts about Venus as Hesperus-thoughts (likewise for the Phosphorus-mode), adding that any such classification "may be obscure, artificial, vague, and ... vary from context to context" (1998, 8-9). (8) That is about all we are told about modes of presentation, and this silence contributes to the following problem. What we might call the 'modal' problem concerns the troubling issue of the modal content of what speakers assert through using Frege-Russell locutions. In order to know the proposition asserted with an utterance of a Frege-Russell locution, we must know what possible situations are described by this utterance, not just what actual situation is described. Piggy backing, however, yields only the actual truth condition of a fictionally true Frege-Russell locution, in a way that seems to resist extrapolation to other possible worlds. Take an utterance of 'Hamlet doesn't exist' or 'Hesperus exists'. Given our rather robust modal intuitions about such cases, it seems that the propositions asserted by means of such utterances can't simply be propositions about modes of presentation. After all, there surely are worlds where the Hamlet-mode does present something, but of which it is false to say that Hamlet exists (according to Kripkean orthodoxy, Hamlet couldn't have existed). Similarly, Hesperus could surely have existed without presenting via the Hesperus-mode.
Preferring to leave the idea of a mode open enough to allow for this conception of modes, Crimmins deals with this modal issue in a surprisingly indirect way. He argues that what is asserted with such an utterance is true at a world w just when w is so as to make the utterance fictionally true at w, and then relies on intuitions about when this is so:
[W]e need to ask which real possibilities make it fictionally true that there are possibilities in which "[Hamlet] exists". I believe the following ideas are defensible: normally, (a) no possible world makes it fictionally true that there is a possible world in which "[Hamlet] exists" ..., and (b) all and only possible worlds in which Venus exists make it fictionally true that there are possible worlds in which "Hesperus exists". (1998, 34-35)
But as a strategy for discerning modal content, this seems almost completely ad hoc. Nothing is said about what might possibly ground such intuitions, and yet something is sorely needed, given that ordinary, unproblematic facts about the fiction are not enough to bear the explanatory burden, while facts about modes turn out not to be enough either.
The modal problem raises questions about the very basis of the semantic pretense approach. The second problem for the approach, which might be termed the 'meaning' problem, relates to the semantic pretense treatment of negative existentials and identity statements in particular. For the semantic pretense approach to work, 'exists' and 'is identical with' must be taken to be special fictional predicates, whose meanings differ from the meanings they have in other contexts. But it seems best to accept such ambiguities only as a last resort. If we agree, with Crimmins (1998, 33ff.), that there is a perfectly common meaning of 'exists' on which it is a predicate true of anything whatsoever, and a perfectly common meaning of 'is identical to' on which it is a predicate that applies to anything and itself (nothing else), why not use these meanings to solve the problems rather than invent meanings that are suggested only by the desire to solve these problems in semantic pretense terms? Gareth Evans (1982, chap. 10), endorsing the view in Kripke 1973 that 'exists' is universal, has even argued that we should regard the universality of 'exists' as a constraint on an adequate analysis of negative existentials. Even if this is too strong, we might well insist that we should treat 'exists' this way if at all possible.
Worse, in certain cases the usual and fictional meanings must both be present if we are to avoid trouble. Consider again:
(4) Those two famous celestial bodies of Babylonian astronomy, Hesperus and Phosphorus, are in fact identical.
Given the usual understanding of numerical quantifiers in terms of identity, (4) seems to require the predicate of identity to be used in two different senses, else we have a contradiction of the form 'There exists x and y such that x [not equal to] y & x = h & y = p & ... & x = y'. (9) But, contradiction or not, the idea that two different meanings are involved surely can't be right, since there are familiar ways of rephrasing the claim that suggest that, all along, only one meaning was intended:
(7) Remember those two famous celestial bodies of Babylonian astronomy, Hesperus and Phosphorus? Well, it turns out they aren't really two: they are identical.
So the doctrine of semantic pretense faces some severe problems. (10) But if it falls, what remains? How else are we to make sense of the object-involving phenomenology of Frege-Russell locutions? Not, I think, by rejecting the phenomenology or embracing its commitments in realist Fashion. In the next section, I argue for a quite different way of understanding the role of pretense in at least the cases of negative existentials and identity statements. (The case of propositional attitude reports involves other complexities, and I have nothing further to say about them in this paper.) The final section returns us to the debate between New Theorists and neo-descriptivists.
5. Pretense Revisited
To see how pretense can help us understand the role of the object-involving phenomenology of existential and identity statements without running into the problems that face Crimmins's account, consider the class of what we might loosely call 'quasi-contradictions' and 'quasi-truisms: sentences, often quite mundane, whose form is explicitly contradictory or truistic and that use this feature to convey something that may well be both true and nontrivial. (11) As an example, think of an apparent contradiction like
(8) That woman is not a woman, or an apparent truism like
(9) That woman is a woman,
and think of the assertive utterance of such a sentence in a situation in which a speaker wants to say that his demonstration does (not) pick out a woman. For example, imagine (8) or (9) uttered while the speaker's audience is watching a play in which a certain person plays the part of a woman, in order: to say that the person playing the part is (not) a woman (both claims are in effect present in a famous scene in the movie Shakespeare in Love). (12) Or imagine it uttered by a speaker in the face of his audience's suspicion that a certain person in their view is a woman, in order to say that this person is (not) a woman.
Assuming that the logical form of (8) or (9) is ['That.sub.dem] F is (not) F', it is clear that what a speaker conveys in this way is not what such a sentence literally expresses: a contradiction or truism of some kind. (13) But neither does he simply convey some kind of Gricean implicature explicable in terms of Gricean mechanisms but not amounting to something that the speaker can rightfully be said to assert with his utterance. Confronted by an assertive utterance of (8) or (9), it seems entirely appropriate for the audience to report, "The speaker asserted that that woman is (not) a woman," where what the speaker is said to have asserted is a noncontradictory, nontrivial proposition stating, roughly, that some demonstrated individual is (not) a woman. That is the kind of account I shall assume in what follows. (14)
But how, in more precise terms, should we understand this asserted content and the way in which the audience works out this content? It would, I think, be a mistake simply to invoke something like Donnellan's referential/attributive distinction. That distinction is simply too crude to allow us to understand the full communicative purpose of such statements. This criticism can already be made of some of Donnellans' own examples. Both Donnellan (1966) and Kripke (1973) (citing Donnellan) give the example of referential uses of descriptions that are believed by the speaker and even his audience to misdescribe the intended referent (say, use of the expression 'the king' to talk about some usurper). But to call these 'referential' uses underdescribes the communicative situation. There is a striking sense in which the speaker does as if the description correctly describes the intended referent, and that he achieves his communicative purpose partly through knowing that his audience knows that he is doing as if the description is apt. (15) Unlike standard cases of the referential/attributive distinction, correction of the speaker would simply be misplaced.
Correction of the speaker would obviously also be misplaced in the case of (8) or (9). Indeed, the statement itself is the corrective. The extent to which the distinction fails to yield an adequate view of the communicative situation in (8) and (9) is further underlined by the following consideration. Suppose you have been talking to me about "that woman." (the one you claim is standing at the bus stop across the road from us). If, after peering into the rain, I begin to realize that you are confused by the light and the drifting rain, and that there is no object for your demonstrative to pick out (no person, no waving branch, etc.), I might challenge your references to "that woman" by insisting:
(10) Wait, that woman is not a woman. She is really just a figment of your imagination--there is nothing there.
Here there is no object for me to have in mind. No referential interpretation is possible (not even one involving a queer entity--a "figment of the imagination"--that you have temporarily confused with a woman.) (16) It seems that this kind of case involves a pure "doing as if" I am correctly describing someone, without there being any object for me to describe. All I can be doing in such a case is simply doing as if,, or pretending that, your (and now my) demonstrative reference is apt in order to assert that such demonstrative reference is entirely inapt when understood from outside of this pretense.
With that in mind, here is the suggestion about how in general to interpret (8) (and similarly (9)). A speaker's utterance of such a sentence shows him to be opportunistically engaged in pretense. The speaker pretends that his demonstrative reference is apt, typically by pretending, about a certain salient individual, that this individual is a woman, and at other times by pretending that there is a demonstratively salient individual who is a woman (as in (10)). (The salience is a function of features that, in the context, make it appropriate to describe an individual as a woman.) Focusing on the typical case for now, note that the speaker may indulge in such pretense for any of a number of reasons. Perhaps, for example, the best way to identify a particular individual is by playing along with the author's mistaken view that the individual in question is a woman. Or perhaps the speaker and his audience are watching a certain individual playing a female role in a play, where pretending that they are thereby watching a woman is something they are supposed to do because of the game of make-believe they are playing.
Whatever the reason for the pretense, the interpretative tension generated by the speaker's utterance of 'that woman is not a woman' now tells his audience that he is only pretending that his means of securing reference with 'that woman' singles out a woman. Understood apart from this pretense, his means of securing reference is therefore best construed as the determiner of semantic reference for a weakened complex demonstrative, say, 'that person' or even 'that object'. (How much weaker will depend on context and/or the speaker's intentions. I'll use 'that individual', which can be understood in either of these two ways.) The speaker pretends that this reference determiner successfully picks out an individual who is a woman, thus justifying his use, inside the pretense, of the stronger complex demonstrative 'that woman'. What the speaker then asserts is that, construed from outside of the pretense, this determiner does not single out an individual who is a woman. Now in the typical situation of utterance, both the speaker and audience know that there really is a demonstratively salient individual present. In such a case, therefore, the speaker succeeds in asserting that, outside of the pretense, his reference determiner singles out an individual who is not a woman. (In addition, of course, he thereby asserts of the individual referred to on the basis of this determiner that this individual is not a woman.) (17)
In short, (part of) what the speaker asserts with a typical utterance of the quasi-contradictory (8) is something like the following internally negated claim:
([8.sub.I]) Outside of the pretense that the underlying reference determiner (for the tacitly understood demonstrative 'that individual') secures reference to an individual who is a woman, it secures reference to an individual who is not a woman.
Similarly, (part of) what the speaker asserts with a typical utterance of the quasi-truistic (9) is something like the following claim:
(9) Outside of the pretense that the underlying reference determiner (for 'that individual') secures reference to an individual who is a woman, it secures reference to an individual who is indeed a woman.
Now of course, the form in which ([8.sub.I]) and (9) articulate this asserted content is clumsy, but that is because each is really only a schema. For the moment, I want to give the content in a neutral way, one that offends neither a neo-descriptivist nor a New Theorist. For neo-descriptivists, what determines the reference, at any world, of the speaker's tacit use of 'that individual' in the actual world is presumably something like the property being the actual individual of whom I am demonstratively aware. In that case, ([8.sub.I]) and (9) can be given particularly simple formulations: 'Outside of the pretense that the actual unique individual of whom I am demonstratively aware is a woman, this individual is (not) a woman'. New Theorists, who presumably have a preference for something like a causal theory of demonstrative reference, will prefer a different formulation.
The gloss I put on (8) and (9) harbors a suggestion about how to interpret the profligate phenomenology of informative identity statements and negative existentials. Instead of Crimmins's idea that in uttering such statements a speaker pretends that different modes of presentation (for example, the Hesperus- and Phosphorus-modes) present different objects, or that a particular mode (for example, the Hamlet-mode) presents an object, suppose we take the speaker to pretend that the reference determiners underlying his use of distinct names like 'Hesperus' and 'Phosphorus' secure reference to distinct objects, or that the reference determiner underlying his use of a name like 'Hamlet' secures reference to a particular object. (And, of course, not just the reference determiners for the speaker's use of these names; the pretense involves the thought that other people, differently situated, are also able to secure reference to these objects.) As before, there is no need to suppose that the pretense in question is of a deeply imaginative kind. While this will no doubt be so in certain cases, in other cases, it may be rooted in little more than the desire to "do as if" a certain way of seeing the world is correct simply in order to raise questions about that way. Whatever the initial reasons for the pretense, what interests me, as it does Crimmins, is how entering into pretense gives speakers a way of talking about the real world.
The emphasis on reference determination gives a rather different picture from the semantic pretense picture of why it is fictionally true that there are two objects, Hesperus and Phosphorus, or why it is fictionally true that there is such a person as Hamlet. It also gives a very different picture of what speakers manage to assert with their utterance of sentences like (1)-(4). Take a plural-sounding identity like (4):
(4) Those two famous celestial bodies of Babylonian astronomy, Hesperus and Phosphorus, are in fact identical.
Assuming, contrary to Crimmins, that 'is identical to' stands for the usual strict relation of identity, we have in place the thought that this attribution involves a contradiction inside a pretense. (18) That scenario is reminiscent of the interpretative tension that led to the interpretative gloss we were able to place on (8). As before, it is the speaker's use of the device of a blatant contradiction (this time, that distinct things are identical) that now allows his audience to understand that he is claiming the world to be different, in relevant respects, from the way his pretense depicts it to be.
It is easy to see what these relevant respects are. When distinct things are declared identical, the predicate is the relational one of identity, which holds pairwise only between an entity and itself: a single object. By analogy with (8), therefore, (part of) what the speaker asserts with (4) is something like:
(4) Outside the scope of the pretense that the underlying reference determiners for my use of 'Hesperus' and 'Phosphorus' secure reference to two celestial bodies (thereby warranting talk, inside of the pretense, of 'those two famous, etc, celestial bodies Hesperus and 'Phosphorus'), these reference determiners secure reference to an object X and an object Y related by identity; hence secure reference to a single object only.
Like (8I) and (9), (4) is clumsy, but that is once again an artefact of a desire to remain neutral for now. (4) is a schema that ought to commend itself to both New Theorists and neo-descriptivists. Neo-descriptivists will interpret it as a claim to the effect that certain properties--say, being the actual celestial body of my acquaintance called 'Hesperus' that, on the basis of my acquaintance, I take to be the brightest such body in the evening sky and being the actual celestial body of my acquaintance called 'Phosphorus' that, on the basis of my acquaintance, I take to be the brightest such body in the morning sky--rigidly apply to a single object once these properties are construed from outside of the pretense that they identify two objects. Allusions to pretense aside, this is just the kind of reading we can expect neo-descriptivists to favor. New Theorists will presumably appeal to a causal-historical theory of some kind. Assuming they agree that speakers have sufficient cognitive grasp on the means of reference determination (a point I return to below), they will prefer to tell the story in terms of the way in which the speaker pretends that his tokenings of the names 'Hesperus' and 'Phosphorus' have their causal-historical origin in distinct objects. The speaker's utterance of (4) then allows him to assert that, outside of this pretense, these tokenings have their origin in a single object.
Turn finally to the case of singular negative existentials, in some ways the most interesting and difficult of the Frege-Russell locutions. I claim that in their case it is the systematic interaction of negation, pretense, and the meaning of 'exists', again on the model of (8), that enables speakers to assert appropriate propositions. Recall the way in which someone who utters (8) as part of (10) succeeds in asserting that there is not even a demonstratively salient individual present. It seems that in this case what is asserted with an utterance of (8) is not the internally negated (8I) but the weaker externally negated (8E):
(8E) Outside of the pretense that the underlying reference determiner (for the demonstrative 'that individual') secures reference to an individual who is a woman, it is not the case that it secures reference to an individual who is a woman,
where the speaker's intention to rule out interpretation (8I) as an account of what he is saying may become evident only when he adds: "She is really just a figment of the imagination--there is nothing there."
So context may have a complex hand in determining the intended reading of (8) in such a case: the audience may need to do some disentangling. But how would a speaker mount an effective, unambiguous challenge to the thought that there is a woman to be demonstrated in the first place, a challenge not in need of disentangling in this way? Well, suppose we agree with Kripke and Evans that 'exists' stands for a universal property, one possessed by everything there is even if most things don't possess it essentially. (19) Now consider the statement:
(11) That woman doesn't exist.
Given that existence is universal, this is trivially equivalent to 'That woman who exists doesn't exist'. (Since 'who exists' is redundant, it is naturally elided.) Earlier I suggested that an utterance of 'That woman is not a woman' concerns the fate of the reference determiner for an appropriately weakened version of 'that [individual who is al woman', namely 'that individual'. By analog), (11) concerns the fate of the reference determiner for an appropriately weakened version of 'that woman [who exists]', namely 'that woman', hypothetically stripped of its association with existence. By parity with (8), it might now be thought that an utterance of 'That woman who exists doesn't exist' could be used to assert either the internally negated claim
(11I) Outside of the pretense that the underlying reference determiner (for the demonstrative 'that woman') secures reference to a woman who exists, it secures reference to a woman who doesn't exist
or the weaker externally negated claim
(11E) Outside of the pretense that the underlying reference determiner (for the demonstrative 'that woman') secures reference to a woman who exists, it is not the case that it secures reference to a woman who exists.
But since everything exists, the first reading (11I) is simply not available. Hence we are left with (11E) as the only reading, or, more simply (eliding the redundant 'who exists'),
(11) Outside of the pretense that the underlying reference determiner (for the demonstrative 'that woman') secures reference to a woman, it is not the case that it secures reference to a woman. (20)
Once again, (11) is only a schema. Neo-descriptivists will presumably take (11) to say something like 'Outside of the pretense that there is a unique actual demonstratively salient individual who is a woman, there is no such individual'--apart from the appeal to pretense, surely the kind of reading we would have expected from neo-descriptivists (and the kind of reading most of us would, pre-analytically, have assigned to (11)). Such a descriptivist reading of the pretense mechanism also secures a natural way of saying that there is not even a demonstratively salient person or object, let alone a woman: instead of saying, "That woman doesn't exist," for example, a speaker might say, "That person doesn't exist" in order to assert that, outside of the pretense that there is an actual person of which he is demonstratively aware, there is no such person.
What, finally, about negative existentials involving names rather than demonstratives, say 'Hamlet doesn't exist' or 'Vulcan doesn't exist'? They can be accorded the same kind of treatment. Like (11), negative existentials featuring names have the logical form of ordinary negated subject-predicate statements that have the existence of their bearers as a presupposition. In this sense they are no different from ordinary subject-predicate statements like 'George Bush doesn't lie'. Thus consider:
(12) Hamlet doesn't exist.
Take this to have the force of 'the individual who is Hamlet, and who exists, doesn't exist', an explicit case of a quasi-contradiction. (Similarly, 'Hamlet does exist' can be taken as a quasi-truism.) In uttering this sentence a speaker adopts the pretense that the reference determiner underlying his use of the name 'Hamlet' secures reference to some individual, and hence an individual who exists, and uses the resulting interpretative tension to assert that
(12) Outside of the pretense that the underlying reference determiner (for my use of 'Hamlet') secures reference to an individual, it fails to secure reference to any individual.
Neo-descriptivists will once again offer their own interpretation of such a schema. They will presumably take (12) to say something like 'Outside of the pretense that there is a unique actual individual called 'Hamlet' whom I haste seen or heard described as a Danish prince intent on avenging the death of his father, etc., there is no such individual'. An utterance of 'Vulcan doesn't really exist' will be deemed to assert something like 'Outside of the pretense that there is a unique actual planet responsible in thus-and-so a way for such-and-such perturbations in the orbit of Mercury, there is no such object'. And so on. Pretense aside, these are just the kinds of readings we would in any case have expected from neo-descriptivists, at least assuming normal knowledge on the part of speakers. (21) Causal-historical theorists would no doubt want to tell the story in a different way, say in terms of causal-historical chains that are pretended to be whole but are in fact broken. Pretense aside, this is in fact close to the story in Donnellan 1974.
6. Back to Descriptivism
Because of the role played by pragmatic considerations in this account of identity statements and negative existentials (specifically, the emphasis placed on the way speakers exploit the literal, semantic content of predicates like 'exist' in order to assert what is not semantically expressed), I'll call it the 'pragmatic pretense' account of such statements, to contrast with Crimmins's semantic pretense account. The preceding section assumed a kind of neutrality between the ways neo-descriptivists and New Theorists might interpret pragmatic pretense. In this final section, I argue that while each interpretation is capable of solving the modal problem facing Crimmins's appeal to semantic pretense (the problem of how to account for the necessity of statements like (4) and (12)), there is nonetheless reason to prefer a neo-descriptivist version.
To see why each interpretation is capable of solving the modal problem, take the neo-descriptivist interpretation first, and consider the question whether the neo-descriptivist version of (4)--say, '[Outside of the pretense] the actual brightest celestial body in the evening sky = the actual brightest celestial body in the morning sky'--is true or false when evaluated at worlds other than the actual world @. Since the role of a rigidifier like 'actually' is to shift truth evaluation and reference at a world w back to the actual world (Davies and Humberstone 1980), the answer depends solely on what the terms refer to at @. That way we get the welcome result that (4) is used to say something that is not just true but necessarily true, since every world containing Venus is a world where the actual brightest celestial body in the morning sky (Venus) is the actual brightest celestial body in the evening sky (Venus again).
Similarly, (12) is necessarily true: because there is no actual individual called 'Hamlet' whom we are acquainted with on the basis of reading Hamlet, there is no possible world containing this individual. The Neo-descriptivist's interpretation of the pragmatic pretense account thus yields a clear sense in which 'Hamlet doesn't exist' is used to assert something that is necessary rather than contingent.
But of course there is also a sense in which (4) and even (12) say something that is contingent rather than necessary: Hesperus mightn't have been Phosphorus (distinct celestial bodies might have been the brightest bodies in the evening and morning sky), and Hamlet might have existed (the Hamlet story, might have been recorded history). To its credit, Crimmins's appeal to modes of presentation is able to capture such a sense, but so, as it turns out, can the neo-descriptivist interpretation of pragmatic pretense. It is true that '[Outside of the pretense] the actual brightest celestial body in the morning sky = the actual brightest celestial body in the evening sky' holds at the actual world @ and at all worlds considered as counterfactual alternatives to @. But suppose we change our perspective, and instead consider worlds under the supposition that they are actual, rather than worlds considered as counterfactual alternatives to the actual world @. This is an utterly natural perspective for any pretense theorist, for the pretense treatment of sentences like (4) and (12) highlights the way speakers and listeners contemplating such sentences routinely do as if some other world is the actual world. It is now easy to see why (4) appears contingent. There deafly are worlds w, now considered as actual, such that in w the unique actual object that appears as the brightest celestial body in the evening sky is not the same as the unique actual object that appears in w as the brightest celestial body in the morning sky ("actual" from the point of view of w). Indeed, in uttering (4) the speaker pretends that the actual world is just such a world w. And the same can be said about (12). There are worlds w such that w contains an actual person called 'Hamlet' whom we are acquainted with on the basis of reading Hamlet ("actual" from the point of view of w). In uttering (12) the speaker pretends that the actual world is just such a world w.
This is a familiar result, of course, taken from the annals of two-dimensional modal semantics. It suggests that how (4) and (12) divide the class of possibilities is subject to two construals, yielding a sense in which (4) and (12) count as necessary and another sense in which they count as contingent. (22) But it is important to note that such an appeal to two-dimensionalism doesn't help only neo-descriptivism. It helps any account that classifies names as rigid designators, no matter how reference is seen as determined and no matter how semantic content is understood. Thus, consider a weak neo-descriptivist view according to which appropriate descriptions fix the reference of the uses of names (thereby ensuring their rigidity) but don't provide their content. Even if such a theory doesn't depict these descriptions as themselves rigidified, occurrences of such descriptions in (4) and (12) should he understood as rigidified if (4) and (12) are to capture their modal behavior. As before, this yields a solution to the modal problem (more generally, it means we can apply the two-dimensional modal framework to show how (4) and (12) are in one sense necessary, in another sense contingent).
Exactly the same type of appeal also shows how the New Theorists' interpretation of pragmatic pretense can solve the modal problem. New Theorists think that names are (de jure) rigid designators, and so must think that, if schemas (4) and (12) are to tell the full modal story, descriptions of the broad form 'the object standing at the origin of the relevant causal chains underlying the current use of N' should again be read as rigidified. In that case, New Theorists too can count (4) and (12) as necessary (and, in another sense, contingent)--an especially welcome result in the case of (12), since the New Theory's commitment to Millianism, combined with the view that fictional names are empty, makes it hard for New Theorists to identify any other suitable sense in which (12)is used to say something true, let alone something that is necessarily true. (23)
So the pragmatic pretense account provides New Theorists as well as neo-descriptivists (both strong and weak) with a means of solving Crimmins's modal problem. Despite this, I think the account favors a version of neo-descriptivism. My argument is a simple variant of a certain familiar argument for neo-descriptivism. Earlier I said that on the pragmatic pretense, account even New Theorists should agree that speakers are in implicit cognitive possession of the means of reference determination for the names they use, where these reference determiners are now taken as causal-historical chains. The reason is this: Apart from the fact that attributing such a grasp results in some welcome unification (it allows us to interpret the descriptive informativeness of what a speaker asserts with an identity or negative existential--a feature granted even by New Theorists--in terms of the informativeness of a story about the fate of certain causal-historical chains), (24) the role that the pragmatic pretense account assigns to pretense in such statements requires us to attribute such a grasp. After all, the account claims that speakers assertively uttering such a statement focus on a contrast they see as significant: they pretend that their reference determiners secure reference to two things, or to something, before asserting that in reality--that is, apart from the pretense--they secure reference to one thing, or to nothing. But that requires speakers to have a robust 'conception of that contrast: robust enough to explain how they are able to view the pretend-scenario as coherent, while knowing what it takes for the real world--the world as it is apart from the pretense--to be different in relevant respects from the world of the pretense. If the facts that determine reference are genuinely beyond a speaker's ken, no such robust conception seems possible.
The rest of the argument will be familiar. (25) A causal-historical theory that assigns a central role to speakers' implicit knowledge of relevant causal-historical chains looks indistinguishable from a version of descriptivism on which reference determination is a function of the fact that speakers implicitly conceive of the reference of the names they use in terms of such underlying chains. Why not, in that case, opt for some such version of descriptivism? That would unify cases susceptible to a causal-historical treatment of reference with cases ("descriptive" names) not susceptible to such treatment. The resulting "causal theory made serf-conscious" (as Brian Loar once put it) is precisely the kind of descriptivist theory of reference determination for ordinary names preferred by contemporary neo-descriptivists.
Let me emphasize the limits of this argument. There is nothing in what I have said that argues directly for interpreting pragmatic pretense in terms of a neo-descriptivist theory of semantic content for names. Both the preceding argument and the solution to the modal problem are already available to a weak neo-descriptivism about reference determination that resists understanding the notion of semantic content in neo-descriptivist terms (perhaps even preferring a Millian account). So we are left with a relatively modest conclusion: whatever the merits of neo-descriptivism as a theory of semantic content, from the point of view of interpreting the idea of pragmatic pretense there is much to be said for at least a neo-descriptivist theory of reference determination. Still, this conclusion is far from inconsequential, given that this paper began with the problem of neo-descriptivism's apparent inability to deal with the phenomenology of Frege-Russell locutions. If I am right in thinking that the phenomenology of identity statements and negative existentials is plausibly understood in terms of pragmatic pretense rather than semantic pretense, it thus turns out that neo-descriptivism has a great deal to gain from acknowledging the role that pretense plays in our understanding of such locutions.
This paper was written while I was a Fellow in the Philosophy Programme of the Research School of Social Sciences at the Australian National University. I am grateful to the School and the Programme for their support. Thanks also to audiences at the University of New England, Monash University, the University of Sydney, the Australian National University, and Victoria University of Wellington for numerous useful comments. I owe special thanks to an anonymous referee for this journal, and to David Braddon-Mitchell, Martin Davies, Richard L. Epstein, Frank Jackson, Jonathan McKeown-Green, and Daniel Stoljar. An early version of this paper appeared in a University of Auckland festschrift in honor of Krister Segerberg.
(1) See, for example, Searle 1983, Lewis 1984, 1994, Kroon 1987, and Jackson 1998a, 1998b. (Although I am assuming that these theorists take the reference-fixing properties associated with the use of a name as specifying something akin to the semantic content of that use, some would prefer to say that their theory does away with the notion of semantic content as ordinarily understood; cf. Jackson 1098b.)
(2) See Soames 2004 for an extended attack on such a causal-descriptivist account of reference. There are, of course, other descriptivist successors to classical descriptivism, for example, the pure metalinguistic description theory of Jerry Katz (1994, 2001), according to which a name N has a meaning (the bearer of N) that doesn't determine reference. Such a theory is very different in kind from the kind of neo-descriptivism considered here, even though it proposes solutions to some of the same problems.
(3) See Frege 1952 and Russell 1905. 'Frege-Russell problems' is a slight misnomer, since, unlike Russell, Frege did not discuss the problem of singular existential statements. (Frege 1979, 60, suggests that he would have construed them as misleadingly formulated metalinguistic statements.)
(4) Weak neo-descriptivism is thus what some call a descriptive theory of "intentionality" rather than "content." Cf. Stanley 1997, 568. As Stanley points out, Kripke's modal argument is ineffective against such a descriptivism even if the descriptions invoked are nonrigid.
(5) On the theory of direct reference, see, for example, Almog, Perry, and Wettstein 1989, Salmon 1986, Recanati 1993, and Soames 2002. Salmon prefers the label 'theory of direct reference' for the negative theory that proper names, demonstratives and other noncompound indexicals do not have descriptive semantic content, using 'Millianism' for the positive view that the semantic content of such a term is its referent. I shall use the more common positive reading for both terms.
(6) The robustness of such attributions of plurality is reinforced by examples that are quantified in form. Thus, forgetting names and even positions of the planets, I may say, "There are two celestial bodies marked in that early map of the heavens we saw that have turned out to be identical--but I have quite forgotten which celestial bodies." I may then add: "And Hammurabi believed one of them, but not the other, to be visible in the evening." Both the theory in Crimmins 1998 and the theory developed in this paper seem to have the resources to handle such quantified locutions.
(7) Like many New Theory, sympathizers, Walton and Crimmins think that sentences containing fictional names do not express propositions, so that it is strictly impossible to imagine the truth of the proposition expressed by a sentence like 'Hamlet is a Danish prince'. What can be imagined is that there is a proposition expressed by our utterances of 'Hamlet is a Danish prince'. Kripke's influential John Locke Lectures (1973) contain an earl), defense of such a view.
(8) In Crimmins 1992 such modes are said to be instanced by what Crimmins calls 'notions'--particular representations of objects that form part of thoughts. Grimmins 1998 is far less forthcoming, at one point even suggesting that "our access to certain modes of presentation comes from our capacities to participate in pretenses of just this kind" (26).
(9) Cf. Richard 2000, 218ff., which clearly takes the idea that an identity statement might involve an implicit or explicit contradiction to be (only slightly) less tolerable than the idea that we might be using '=' with two meanings.
(10) See also Stanley 2001. Stanley takes much of what he says as an attack on "hermeneutic fictionalism" in general (roughly speaking, any pretense account of some seemingly serious bit of discourse), although I doubt that his specific criticisms affect the sort of pretense view developed below.
(11) Earlier versions of part of the argument in this section occur in Kroon 2000 and 2001.
(12) The movie contains an exchange about the real sex of the person playing the part of Juliet in the first performance of Romeo and Juliet--as it turns out, Lady Olivia, Shakespeare's lover. With Queen Elizabeth I secretly in attendance, the Queen's Master of the Revels enters the theatre and loudly charges: "That woman [pointing to Lady Olivia] is a woman," whereupon the Queen rises from her seat and calmly retorts that he is mistaken. Imagine she used the actual words: 'That woman is not a woman'.
(13) The contradiction or truism need not be at the level of semantic content. If, for example, the semantic contribution off in 'That F is G' is at the level of character rather than content (Borg 2000), it follows that there is no possible context of utterance at which a sentence like (8) expresses a true content: contradiction enough for my purposes. Note that a speaker doesn't avoid the contradiction by including the qualifier 'really', as in 'That woman is not really a woman'. All the evidence suggests that 'really' in such cases is simply a pragmatic device used to highlight the fact that how matters stand may conflict with expectations or appearances. Since its chosen way of specifying how matters stand is in this case contradictory, its use doesn't by itself harbor a way of explaining away the contradiction.
(14) Such an assertion-attributing account also applies to broadly metaphorical uses of language (indeed, quasi-constradictions and quasi-truisms could be construed as a kind of metaphor). Thus, Mary might say, "The President is a pain," thereby asserting, and not just implicating, something that is not semantically expressed. What she asserts is that the President is behaving in an annoying manner, an assertion-attributing construal confirmed by the way we say, "Mary said/asserted that the President is a pain." (Cf. Soames 2002, 210ff., which cites such cases to support a certain account of the propositions asserted through attitude ascriptions.)
(15) What makes the citation of such cases in Kripke 1973 somewhat surprising is that Kripke's lectures are, in part, a sophisticated attempt to show the relevance of pretense to talk of fiction. But we often use fictional descriptions in order to say something serious about real people who feature in our fictions, for example, 'The famous bear-hunter has finally returned for dinner'. Such cases, which rely on a form of pretense, seem continuous with the cases of deliberate misdescription given by Donnellan and Kripke.
(16) Since (10) is supposed to be true in the absence of any demonstrated object, this provides further reason for taking the logical form of (8) at face value, and not, for example, construing it as a sentence whose underlying logical form contains contextually supplied fillers (say, 'That person over there whose appearance made us think it was a woman is not a woman', or even 'That so-called woman is not a woman'; cf. Borg 2000, n. 22). As for the thought that (10) might be made true by the literal presence of a "figment of the imagination" (a demonstratively salient figment, no less), this strikes me as barely credible. A "figment of the imagination" doesn't seem to be a kind of thing at all. Talk of "figments of the imagination" is best seen as metaphorical talk, allowing a speaker to signal what he sees as the role played by a person's imagination when the person actually believes that the world is as the speaker merely pretends it to be. In the spirit of Walton 1990, and contrary, to Kripke 1973 and Salmon 1998, it seems to me that predicates like 'is a failed posit', 'is an imaginary friend', and even 'is a fictional character' should be understood in a similar pretense fashion.
(17) We derive this (second) content simply by adopting the non-pretense perspective on the speaker's words in (8). Accepting a direct reference theory of content, New Theorists will understand this second content in an object-involving way. Not so those who subscribe to a neo-descriptivist account of content, since they don't acknowledge a second layer of content above the reference-fixing content.
(18) The contradiction is blatant in the case of (4), and especially (7), but in my view is also implicitly present in the simple forms (2) and (3). Speakers and their audience realize that beneath the surface of such locutions is the ever-present possibility of trivially equivalent plural formulations.
(19) Kripke 1973; Evans 1982, chap. 10. See also "Chakrabarti 1997. Nakhnikian and Salmon (1957) provide an early defense of the view that 'exists' has a use as a universal predicate, and there are signs of such a view in Russell's and even Frege's work (Frege 1979 holds that 'exists' so understood is merely a grammatical predicate). For present purposes, we can take 'exists' to stand for existence at some time or other, although it is not hard to incorporate the tensed use of 'exists' (as in 'Homer did exist, but Holmes didn't') into the account I defend. For a very different account of the use of 'exists' as a tensed predicate, see Salmon 1998.
(20) Note that the qualifier 'really', present in the alternative formulation "That woman does not really exist', plays no role in this derivation of (11), just as it played no role in the derivation of (8I) and (8E). As before (note 13 above), I take 'really' to have a purely pragmatic, contrast-emphasizing role, a contrast that often barely needs emphasis (which is why the word is readily omitted). Compare this to Evans 1982, chap. 10, which assigns a pivotal but problematic semantic role to 'really'. Because Ewans's account entails that 'Really A(t)' is true only if t genuinely refers, it rules out the evident equivalence of "That woman doesn't really exist' and 'Really [in reality], that woman doesn't exist', and fails to explain the use of 'really' in closely related locutions like (10).
(21) Because of the role assigned to pretense, the fact that an audience may have radically different knowledge from the speaker is only a limited barrier to communication on the present account. Thus, consider a knowledgeable speaker's utterance of 'Hamlet doesn't exist'. Given that the audience knows that it is part of the speaker's pretense that people can be acquainted with the person called 'Hamlet' in radically different ways, the audience knows that the speaker's utterance indirectly discredits a whole range of ways of trying to secure reference to such a person (cf. Walton 1990, 425ff.), including the method of deferring to other users of the name. Consequently, someone who only knows that "Hamlet is the person called 'Hamlet' I have heard others talk about under that name," and who then hears the speaker's words but is told nothing else (not even that 'Hamlet' is a fictional name), thereby still learns that--in his words--"Hamlet doesn't exist." Such a case is difficult for many other (partly) descriptive accounts of negative existentials, including the theory in Salmon 1998.
(22) In the current jargon, it all depends on whether the focus is on secondary/C-intensions or on primary/A-intensions. See Jackson 1998a, 2004, and Chalmers 1996, 2004 (the latter describes Chalmers's distinctive epistemic interpretation of two-dimensional semantics). Stalnaker (2001, 2004) urges a contrasting "metasemantic" interpretation of the two-dimensional framework, one based on a causal-historical view of reference.
(23) New Theorists like Salmon (1998) and Soames (2002, 89-95) maintain the necessity of (12) in the context of an account that lets fictional names like 'Hamlet' stand for abstract fictional objects. But such an account is of no use to the many Millians who think that true singular negative existentials feature genuinely empty names, not names for special objects. Perhaps the most promising alternative is Braun's (1993) view that negative existentials containing empty names semantically express true "unfilled" propositions-propositional structures that lack constituents corresponding to empty terms, and hence are true, and indeed necessarily true, to the extent that they involve the negation of certain unfilled and hence non-true propositions. But such a view also faces serious problems. Among other things, it is forced to count positive statements like 'Hamlet is a fictional character' and 'Vulcan is a failed posit' as non-true, even though these are surely no less true than negative statements like 'Hamlet/Vulcan doesn't exist', and in a sense even serve to explain the truth of the latter (cf. also Reimer 2001). Braun's way of making sense of the (necessary) truth of a negative existential like (19) therefore fails to make sense of the truth of certain statements to which the negative existential is explanatorily linked.
(24) Note that a speaker's grasp of such a story, leaves room for the informational role of distinctive, salient beliefs like 'Aristotle was a famous Greek philosopher'. For although the Donnellan-Kripke arguments against classical descriptivism show that this belief plays no direct reference determining role for the name 'Aristotle', casual-historical theorists and neo-descriptivists alike think that the casual history of such a belief (as the firmly held, but fallible, product of processes of information acquisition) is relevant to reference determination--a belief's causal history may be necessary for disambiguating the reference of a speaker's use of a name, say.
(25) Cf. especially Lewis 1984, 1994, and Jackson 1998a, 1998b. Critics include Devitt and Sterelny (1999, 61).
Almog, Joseph, John Perry, and Howard Wettstein, eds. 1989. Themes from Kaplan. Oxford: Oxford University Press.
Borg, Emma. 2000. Complex Demonstratives. Philosophical Studies 97:229-49.
Braun, David. 1993. Empty Names. Nous 27:449-69.
Casteneda, Hector-Neri. 1989. Thinking, Language, and Experience. Minneapolis: University, of Minnesota Press.
Chakrabarti, Arindam. 1997. Denying Existence. Boston: Kluwer.
Chalmers, David. 1996. The Conscious Mind. Oxford: Oxford University Press.
--. 2004. Epistemic Two-Dimensional Semantics. Philosophical Studies 118:153-226.
Crimmins, Mark. 1992. Talk about Beliefs. Cambridge: MIT Press.
--. 1998. Hesperus and Phosphorus: Sense, Pretense, and Reference. Philosophical Review 107:1-47.
Davies, Martin, and Lloyd Humberstone. 1980. Two Notions of Necessity. Philosophical Studies 38:1-30.
Devitt, Michael, and Kim Sterelny. 1999. Language and Reality. 2nd ed. Cambridge: MIT Press.
Donnellan, Keith. 1966. Reference and Description. Philosophical Review 75:281-304.
--. 1974. Speaking of Nothing. Philosophical Review 83:3-31.
Evans, Gareth. 1982. The Varieties of Reference. Oxford: Oxford University Press.
Frege, Gottlob. 1952. On Sense and Reference. In Translations from the Philosophical Writings of Gottlob Frege, trans. Peter Geach and Max Black, 56-78. Oxford: Basil Blackwell.
--. 1979. Dialogue with Punjer on Existence. In Posthumous Writings, ed. Hans Hermes, Friedrich Kambartel, and Friedrich Kaulbach, 53-67. Oxford: Basil Blackwell.
Jackson, Frank. 1998a. From Metaphysics to Ethics. Oxford: Oxford University Press.
--. 1998b. Reference and Description Revisited. In Philosophical Perspectives, vol.12, ed. James Tomberlin, 201-18. Atascadero, Calif.: Ridgeview.
--. 2004. Why We Need A-Intensions. Philosophical Studies 118:257-77.
Katz, Jerrold. 1994. Names without Bearers. Philosophical Review 103:1-39.
--. 2001. The End of Millianism: Multiple Bearers, Improper Names, and Compositional Meaning. Journal of Philosophy 98:137-66.
Kripke, Saul. 1973. Reference and Existence. John Locke Lectures (unpublished). Oxford: Oxford University Press.
--. 1980. Naming and Necessity. Cambridge: Harvard University Press.
Kroon, Frederick. 1987. Causal Descriptivism. Australasian Journal of Philosophy 65:1-17.
--. 2000. 'Disavowal through Commitment' Theories of Negative Existentials. In Empty Names, Fiction and the Puzzles of Existence, ed. Anthony Everett and Thomas Hofweber, 95-116. Stanford: CSLI Publications.
--. 2001. Fictionalism and the Informativeness of Identity. Philosophical Studies. 106:197-225.
Lewis, David. 1984. Putnam's Paradox. Australasian Journal of Philosophy 62:221-37.
--. 1994. David Lewis: Reduction of Mind. In A Companion to the Philosophy of Mind, ed. Samuel Guttenplan, 412-31. Oxford: Blackwell.
Nakhnikian, George, and Wesley Salmon. 1957. 'Exists' as a Predicate. Philosophical Review. 66:535-42.
Recanati, Francois. 1993. Direct Reference: From Language to Thought. Oxford: Blackwell.
Reimer, Marga. 2001. The Problem of Empty Names. Australasian Journal of Philosophy 79:491-506.
Richard, Mark. 2000. Semantic Pretense. In Empty Names, Fiction and the Puzzles of Existence, ed. Anthony Everett and Thomas Hofweber, 205-32. Stanford: CSLI Publications.
Russell, Bertrand. 1905. On Denoting. Mind 14:479-93.
--. 1956. The Philosophy of Logical Atomism. In Logic and Knowledge, ed. Robert C. Marsh, 241-54. London: Macmillan.
Salmon, Nathan. 1986. Frege's Puzzle. Cambridge: MIT Press.
--. 1998. Nonexistence. Nous 32:277-319.
Searle, John. 1983. Intentionality. Cambridge: Cambridge University Press.
Soames, Scott. 2002. Beyond Rigidity: The Unfinished Semantic Agenda of Naming and Necessity. New York: Oxford University Press.
--. 2004. Reference and Description. In The Oxford Handbook of Contemporary Analytic Philosophy, ed. Frank Jackson and Michael Smith. Oxford: Oxford University Press.
Stalnaker, Robert. 2001. On Considering a Possible World as Actual. Proceedings of the Aristotelian Society, supp. vol. 75:141-56.
--. 2004. Assertion Revisited: On the Interpretation of Two-Dimensional Modal Semantics. Philosophical Studies 118:299-322.
Stanley, Jason. 1997. Names and Rigid Designation. In A Companion to the Philosophy of Language, ed. Bob Hale and Crispin Wright, 555-85. Oxford: Blackwell.
--. 2001. Hermeneutic Fictionalism. In Midwest Studies, vol. 25: Figurative Language, ed. Peter French and Howard Wettstein, 36-71. Oxford: Blackwell.
Walton, Kendall. 1990. Mimesis as Make-Believe. Cambridge: Harvard University Press.
Zalta, Edward. 1988. Intensional Logic and the Metaphysics of Intentionality Cambridge: MIT Press.
University of Auckland