Names Without Bearers*

by Jerrold J. Katz



Millians claim that names have no sense (linguistic meaning). If that were so, then names that have no bearer would have no meaning at all. But, in that case, sentences like (la) and (1b) would be meaningless.

(1a) There is no Santa Claus.
(1b) Santa Claus does not exist.

They cannot obtain a meaning compositionally, since one of their major constituents--the predicate in (la) and the subject in (1b)--is meaningless, and they cannot obtain a meaning non compositionally, since they are not idioms. But such sentences seem perfectly meaningful. They receive neither a raised eyebrow from the ordinary speaker nor an asterisk from the linguist. Moreover, people unhesitatingly use them to make true statements about reality. Thus, they are a problem for Millians.

In the present paper, I will argue that the problem cannot be solved within the Millian position, but can be solved outside it. I will argue that this problem is part of a complex of problems, involving Kripke's famous puzzle about belief and a somewhat less well known difficulty about names with multiple bearers, all of which, for the same reason, are insoluble within the Millian position. I will try to show that what has stood in the way of a solution to this complex of problems is Frege's conception of sense.



It is customary to distinguish between the study of language and the study of language use. The object of study in the former is English, French, Chinese, etc., and the object of study in the latter is the linguistic behavior of their speakers. The study of languages concerns the grammatical structure of sentence types, and the study of language use concerns the pragmatic structure of tokens of sentence types. Given this distinction, the problem for the Millainn position divides into two problems, one raised by the meaningfulness of sentences types like (1a) and (1b) and the other raised by the illocutionary straightforwardness of uses of their tokens. With respect to the former, neo-Millians like Donnellan and Kripke have, as far as I can tell, bitten the bullet, denying that such sentences have a meaning, presumably because they recognize that their account of proper nouns does not permit such sentences to receive a compositional meaning. Neo-Millians concentrate their attention on the latter problem. Their strategy is, I believe, to vindicate their overall position by complementing their criticisms of the description theory with a satisfying explanation of how speakers make true statements in their use of meaningless sentences like (la) or (1b).

The focus of philosophers concerned with names has largely reflected the neo-Millians' focus on the area of language use. This is because their criticism of the description theory has been widely accepted as showing that no description theory of proper names is tenable. As a consequence of this general focus on the area of language use, the problem that neo-Millians face in the area of language has gone virtually unnoticed. This is unfortunate. The cause of philosophical understanding as well as the cause of descriptivism would have been better served had that problem received the attention it deserves.

There is a variety of facts that spell trouble for the Millian position in the area of language. As already indicated, there is the central fact that sentences like (la) and (1b) are perfectly meaningful in English. If (la) and (Ib) are meaningful, we can easily argue that it is false to claim that proper nouns have no meaning. Since (1a) and (1b) are not idioms, their meaning is a function of the meaning of their constituents and the syntactic relations among their constituents. If the meaning of (1a) and (1b) is compositional, then those sentences could not be meaningful unless their constituent `Santa Claus' had a meaning to contribute to their compositional meaning. Hence, it is false to claim that proper nouns have no meaning.[1]

Another troublesome fact is that pairs of sentences like (la) and (1b) are synonymous. Their synonymy is evidence for the meaningfulness of such sentences, since they could not have the same meaning if neither had a meaning. Yet another is that the conjunction of (la) and (1b)--that is, (2a)--is redundant, and still another is that (2b) is contradictory.

(2a) There is no Santa Claus and Santa Claus does not exist.
(2b) Santa Claus does not exist but there is a Santa Claus.

These last two facts also argue for the meaningfulness of (la) and (1b), since a conjunction could not be either redundant or contradictory if its conjuncts had no meaning. What makes the problem that Millianism faces in the area of language formidable is the clarity and robustness of the linguistic intuitions supporting such facts.

Therefore, in the area of language, a strong case can be made against Millianism in connection with the phenomena of names without bearers. In section 5, I will show how the case can be strengthened in connection with the phenomena of names with multiple bearers. But, however strong a case is built up, neo-Millians can respond that no case restricted to the area of language can be decisive against them because the issues in that area are only part of the general issue between Millianism and descriptivism. The general issue, being broader and more complex, cannot be settled on the basis of evidence from one area. No matter how clear and compelling that evidence might be, the general issue must ultimately be decided on the basis of which position provides the best overall account of names both in the language and in use.

Some neo-Millians might want to make a stronger response. They might say that the intuitions on which the "contrary evidence" rests must be illusory because there is nothing for those "intuitions" to be about. Mill and the neo-Millians have demonstrated that names are not synonymous with descriptions, and therefore there is only one remaining problem, which is a problem for everyone, namely, to explain, without appealing to linguistic meaning, how sentences like (la) and (ib) are used to express truths about reality.

This stronger response fails for two reasons. First, our semantic intuitions about (la) and (1b) are too clear and robust to go away so easily. Accepting the stronger response would leave us not with a comfortable feeling of having satisfactorily resolved all the issues outside of those concerning use, but rather with the queasy feeling that our semantic intuitions in the area of language have been repressed. It is just as unsatisfying to insist that Millian arguments show that the intuitions must be illusory as it is to insist that the intuitions show that the arguments must be fallacious. In both cases, we are left with an embarrassment. Without scrutinizing the arguments or the intuitions, the only judicious thing to say is that there is a puzzle about how to reconcile an apparently irresistible set of arguments with an apparently immovable set of intuitions.

Second, although it is true that Millian arguments are effective against nearly all versions of the description theory, they are not effective against all. There is considerable confusion on this point. The basic problem is that the term `description theory' has been taken to denote Fregean description theories in which the sense of a name is required to determine its referent. But the term can also denote a non-Fregean description theory in which sense is not required to determine reference. This kind of theory is still a description theory because it also takes a name to have a sense that-- assuming sufficient resources on the part of the language--can be expressed by a description. A non-Fregean description theory agrees with Frege that proper nouns and their tokens have sense, but disagrees with him on the nature of senses and on the role that the sense of a linguistic type plays in fixing both the sense and reference of its tokens. The range of description theories is thus wider than it is commonly taken to be. But because of the enormous influence of Fregean semantics, theories in the non-Fregean portion of this range have gone unnoticed. As we shall see, some of them escape the Millian arguments that have turned the tide against descriptivism.

In earlier publications, I have proposed a non-Fregean description theory.[2] The theory, which I refer to as "PMT," is a pure metalinguistic description theory in that, on its account of the senses of a names, (i) the sense of a proper noun mentions (or represents) the noun of which it is the sense, and (ii) the sense of a proper noun contains no condition on being a bearer of the name beyond being a bearer of the name. As a first approximation to PMT's formulation of the sense of names we can say that the sense of a proper noun `N' has the form
(A) The thing which is a bearer of `N' .[3]

(I postpone an explanation of (A) until section 5 in order not to interrupt the argument here and also in order to locate it where it is used.)

From the perspective of PMT, the significant semantic difference between proper and common nouns is that the former have a sense that is metalinguistic. From the perspective of PMT, the Fregean and Millian positions are both importantly right and importantly wrong about the semantics of proper nouns. Fregeanism is right in holding that proper nouns as well as common nouns have sense, but wrong in holding that the senses of proper nouns are just like those of typical common nouns.[4] Millianism is right in holding that there is a significant semantic difference between common and proper nouns, but wrong in characterizing the difference as that between nouns with sense and nouns without sense. PMT steers a middle course between Fregeanism and Millianism.

Frege seems to have formed his notion of the senses of proper nouns by generalizing directly from the senses of common nouns. In the case of a common noun, there is a basis for taking a sense to be a representation of one or another aspect of the referent, to be a certain way in which it presents itself to us. Recall Frege's famous analogy with observing the moon through a telescope, in which the referent of a word is compared to the moon and its sense to "the real image [of the moon] projected by the object on the glass in the interior of the telescope.'[5] Such a conception of sense precludes pure metalinguistic senses for names, since a pure metalinguistic sense of a name cannot be regarded as being "projected by" a hearer of the name. Quine need not have spoken exclusively for extensionalists when, in "Two Dogmas of Empiricism," he famously decried the notion of meaning as "what essence becomes when it is divorced from the object of reference and wedded to the word."

All metalinguistic description theories explicate the sense of proper nouns--but not common nouns--in terms of a relation between the noun and the objects that bear it as their name. But although all metalinguistic description theories get the semantic difference between proper and common nouns right, not all of them get the senses of proper nouns right. Such theories differ from one another in what they take the bearer relation to be. PMT is a pure metalinguistic description theory because it takes the bearer relation to involve no real property or relation in addition to the purely nominal bearer relation that makes a sense metalinguistic.

That difference is critical. Whether a metalinguistic description theory avoids or repeats the mistake of Fregean description theories that makes them vulnerable to Millian arguments depends on whether the theory takes the bearer relation to be pure or impure. One metalinguistic description theory that repeats the Fregean mistake is the theory held by Russell and Kneale, in which the relation is formulated as "the object called `N'."[6] As a consequence of adulterating the bearer condition in the senses of names with real properties and relations, impure metalinguistic description theories forfeit the advantage of having gone metalinguistic in the first place. Including such a property or relation in the bearer condition makes the theory vulnerable to the very counterfactual arguments that refute straightforward Fregean versions of the description theory. This was clearly demonstrated by Kripke when he pointed out that the name `Socrates' would apply to Socrates even if his contemporaries never called him that.[7]

The moral is clear. If the advantages of having metalinguistic senses are to be realized, a metalinguistic description theory must be pure. But this raises the question of why a philosopher, having taken the significant step of constructing a metalinguistic description theory, then adulterates pure metalinguistic conditions with real properties of bearers. No doubt, some intensionalists have had their own idiosyncratic reasons for including real properties of bearers, but I think there is a general reason, one that exerts almost irresistible pressure on intensionalists to do so. That is the recognition that a pure metalinguisitic condition will not satisfy the Fregean requirement that sense determine reference. The fear is that unless the metalinguistic condition associated with a name contains sufficient real properties of its bearer to determine the referent in applications of the name, the condition will not even count as a sense.

Here we clearly see the pernicious influence of Fregeanism at work. The equation of Fregean intensionalism with intensionalism entails the equation of descriptivism with Fregeanism descriptivism. As a consequence, Fregean descriptivism and Millianism are taken to exhaust the range of positions on the semantics of names. Accordingly, descriptivists are saddled with the Fregean requirement that senses determine reference, and those who are attracted to a metalinguistic formulation of the sense of names are forced to adulterate their metalinguistic formulation. But once the formulation contains real properties of bearers, metalinguistic description theories are vulnerable to Millian counterexamples. It is thus not at all surprising that the best efforts of descriptivists have proven futile in the face of Kripke's innovative uses of Millian modal counterexamples.

To see why equating Fregean intensionalism with intensionalism is wrong, it is necessary to step outside the issue between descriptivists and Millians and consider the broader issue between intensionalism and extensionalism. Elsewhere, I have argued at length that the criticisms of Donnellan, Kripke, Putnam, etc. are good criticisms of Fregean intensionalism, but not of intensionalism per se. The problem that their criticisms reveal can be traced to Frege's definition of sense as the determiner of reference. Once this etiology is appreciated, we will want to ask whether a non-Fregean definition of sense can provide an intensionalist semantics free of such problems.

The first philosopher to spot the problem with having sense determine reference was Wittgenstein. He argued, clearly with Frege in mind, that the possibilities for reference are not completely constrained by the senses of words (see, for example, Philosophical Investigations, sections 79--88). Wittgenstein's solution was to throw out the notion of senses as objects, replacing it with his own notion of sense as given in use. Quine's treatment of the extra-logical vocabulary of natural languages was to entirely eliminate intensional objects. Similarly, Donnellan's, Putnam's, and Kripke's solution to the problem in connection with proper nouns was to adopt Mill's position. But these anti-intensionalist approaches are not the only solutions to the problem with Frege's definition of sense. We can concede Wittgenstein's point that referential possibilities are not completely constrained by sense without either abandoning or restricting the notion of senses as objects. Instead of throwing out the notion of senses as objects, we can throw out the notion of senses as determiners of reference.

When we stop to think about it, we may wonder what a definition of sense is doing defining that notion in terms of the determination of reference. A definition of sense that is not formulated in terms of intensional features is surely peculiar, in the manner of a definition of pronunciation that is not formulated in terms of phonological features or a definition of number that is not formulated in terms of arithmetic features. From this perspective, what we want in a definition of sense is not a condition on which sense is a determiner of referential properties and relations, but one on which it is a determiner of sense properties and relations.[9] Thus, we want a definition like (B).

(B) The sense of an expression is that aspect of its structure that is responsible for its sense properties and relations, that is, having a sense (meaningfulness), sameness of sense (synonymy), multiplicity of sense (ambiguity), repetition of sense (redundancy), opposition of sense (antonymy), and so on.

Unlike Frege's definition, (B) does not compromise the autonomy of the domain of sense by basing the definition of sense on the concept of reference. (B) secures the autonomy of the domain by making senses grammatical objects, that is, objects of study in the linguistic investigation of the structure of sentences in natural language. Since sense structure is independent of any intrinsic connection with reference, the question of the role that senses play in reference is shifted from the theory of sense to the theory of reference--where it belongs.

Thus, (B) gives rise to a new intensionalism that provides an alternative to Fregean intensionalism in the controversy between descriptivists and Millians. Since, on this new intensionalism, sense is required to be just informationally rich enough to explain the sense properties and relations of expressions, descriptivists are no longer forced to stuff senses full of reference-fixing information. There is now no obstacle to a pure metalinguistic theory, such as PMT, on which the senses of proper nouns are as informationally poor as instances of (A). PMT thus steers a middle course between Fregean and Millian positions on the issue of the sense of proper nouns.

For the same reason, PMT also can steer a middle course between those positions on the issue of the reference of proper nouns. With the new intensionalism, there are now three positions on the reference of names, as well as three positions on their sense. The Fregean position is that the sense of a name determines its reference; that is, sense is necessary and sufficient for fixing the reference of names. The Millian position is that reference is direct; that is, sense is neither necessary nor sufficient for fixing the reference of names. The new intensionalist position is that sense mediates reference; that is, sense is necessary but not sufficient for reference.[10]

The basic reason that sense is necessary for reference is that referential uses start with knowledge of sense. In the case of literal use, knowledge of the sense of a linguistic type provides the speaker with the condition for literal use of its tokens. Roughly speaking, the sense of a linguistic token is literal just in case it is the sense of its linguistic type, and the referent of a linguistic token is literal just in case the sense of the token is literal and the referent is in the extension of that sense. Thus, the literal application of `gorilla' to a gorilla and the literal application of `Hitler' to the Nazi dictator depend on knowledge of the sense of `gorilla' and knowledge of the sense of `Hitler'. In the case of nonliteral use, too, use starts with knowledge of sense. The nonliteral application of `gorilla' to an NFL tackle or the nonliteral application of `Hitler' to a college dean depend, respectively, on an analogy between the muscular power of the ape and that of the football player and on an analogy between the tyrannical behavior of the Nazi dictator and the tyrannical behavior of the dean. Such analogies presuppose knowledge of literal applications of `gorilla' and `Hitler', which, as we have seen, involves knowledge of the sense of a name.

The reason that the sense of proper names is not sufficient to determine the reference of their tokens is that it is too meager.[11] Speakers who know the sense of `London' know that literal tokens of that proper noun type denote the contextually definite bearer of the name, but that knowledge does not tell them who or what the bearers are, or even whether there are any tokens of the type. Extra-linguistic information of various sorts is required to know who or what the bearers of the proper noun `London' are and whether an utterance or inscription is a literal occurrence of a token of this type. Given this information, the speaker still requires further extra-linguistic information to know which of the bearers is the referent in the context.

PMT combines the virtues of the Fregean and Millian approaches. Frege was right to say that names have sense, but Mill, too, was right to say that names are marks with the function of tags for distinguishing objects from each other on the basis of the orthography of the tnarks.'[12] Mill was wrong to say they are meaningless marks. I think Mill's mistake was to think that names can function as tags only if they are meaningless. As (A) shows, this is not so.

Names can function as tags in spite of being meaningful, provided that their meaning represents them as marks whose orthography, rather than any real properties of their bearer, is what is critical to their functioning as tags. Now, the new intensionalism can represent the meaning of names in this way because it drives a wedge between the semantics of names and their tag function. Its definition of sense locates information about the orthographic character of proper nouns and about their tag function in their sense, and, at the same time, relocates the determination of reference, identification of actual bearers, in the pragmatics of language use.



Fregean descriptivism and Millianism each have a fatal flaw. The former's is very well known. It cannot account for the referential stability of names under counterfactual transformations that falsify contingently true descriptions of their bearers. The latter's is not as well-known. It cannot satisfactorily account for the meaningfulness of sentences like (la) and (1b) or, as we shall see, for their use in making true statements about reality. The flaws, despite the great differences between the positions, have a common source. Both are consequences of accepting the Fregean definition of sense. Fregean descriptivists generalize from their view that common nouns have Fregean senses to the conclusion that proper nouns have Fregean senses, failing to see that descriptions expressing contingent aspects of the referents of proper nouns cannot determine their referents in counterfactual cases. Millians, seeing the modal lay of the land quite clearly, properly conclude that names do not have Fregean senses, but, equating sense with Fregean sense, they go on to conclude improperly that names do not have senses at all. Accordingly, descriptivism runs into trouble in connection with names with bearers, while Millianism runs into trouble in connection with names without bearers.

In this and the next section, I want to explain how, by rejecting the equation of sense with Fregean sense, PMT escapes both flaws. In this section, I will briefly review and expand upon my earlier explanations of how PMT escapes the neo-Millian criticisms of Fregean description theories. In the next section, I will describe PMT's account of the semantics of sentences like (la) and (1b), thereby showing how it escapes the flaw in Millianism.

Kripke's Jonah case is the most powerful of his direct counter-examples to the traditional description theory because it is a case in which all the descriptive information we have about the bearer of a name is false.'[13] Since the story of Jonah in the Bible contains the only information we have about him, it does no good to invoke Wittgenstein's and Searle's move of taking reference to depend on satisfaction of a sufficiently large number of (perhaps weighted) descriptions.'[14] In the case of this example, traditional descriptivists are forced to say--quite counterintuitively--that Jonah did not exist. But, on PMT, the sense "Jonah" contains no information drawn from the Biblical story, and hence it does not have to say that Jonah did not exist.

Kripke bases a number of criticisms on his noncircularity condition that the use of a name must not enter a loop that prevents fixing its referent.[15] This and other informativeness conditions express the Fregean requirement that senses have to do their job of determining reference. Since senses of names in PMT have no such job, instances of (A) cannot run afoul of Kripke's condition. Given (B), the only way senses of names can fail to be sufficiently informative is if they cannot determine sense properties and relations.

Kripke also argues that the sense of `Nixon' cannot be something like "named `Nixon'" because (3a) is false but (3b) is true.[16]

(3a) Nixon might not have been Nixon.
(3b) Nixon might not have been named "Nixon."

PMT agrees. PMT denies that simply replacing the second occurrence of `Nixon' in (3a) with `named "Nixon"' or some other such expression preserves the semantic structure of the false statement that Kripke is assuming one makes in a use of (3a). No doubt, such a replacement would turn (3a) into something like the predicational sentence (3b), which makes a true statement in a world where Nixon exists but does not hear the name `Nixon'. But, on PMT, the semantic structure of (3a) is something like the identity sentence (3c)

(3c) The contextually definite thing that is a bearer of `Nixon' might not have been the self-same contextually definite thing that is a hearer of `Nixon which expresses the false statement that Kripke is assuming one makes in the use of (3a) in such a world.'[17]

PMT, as a theory of sense, is neutral on the choice between Kripkean and other theories of the referential structure of the language. PMT allows, but does not require, the use of names to be governed by a pragmatically introduced rigidifying stipulation. Rigidity does not have a grammatical source the way that definiteness does. It is not names themselves--the proper nouns of a language--that have the property of rigidity, but their tokens, as Kripke's account recognizes. On his account of rigidity, the rigidity of an utterance (or inscription) of `Nixon' derives from the user's "stipulation" that he or she is using that token of the proper noun `Nixon' to speak of the same contextually specified individual in every possible world--that is, that he or she is speaking, to put it in Kripke's words, "of what might have happened to him."[18]

Thus, independently of any assumption about rigidity, we want to formulate an account of (3a) on which the senses of its occurrences of `Nixon' are instances of (A), and (3a) is represented as making a false statement about Nixon even in worlds where he does not bear the name `Nixon'. That formulation is (3c). (3c) expresses such a statement because, due to the anaphoric character of the second term in (3c), its referent is not determined independently, but is whatever the referent of the first term is. Extensionally speaking, the second term is just the first one again. Thus, if the first term is about Nixon, then the statement that uses of (3c) make is ipso facto about Nixon.

The first term can be about Nixon even on the counterfactual supposition that he does not bear the name `Nixon' (or even that the world in question does not contain the name `Nixon') because, on PMT, there is no claim that Nixon bears the name `Nixon' in the world in question (or that the name exists in that world). All that PMT claims is that Nixon bears the name `Nixon' in our linguistic community--of which there can be little doubt insofar as (3a) is a sentence of our language. Thus, (3c) represents the (necessarily) false statement that uses of (3a) make, namely, that it is possible for Nixon to have been other than himself
Uses of (3a) make a (necessarily) false statement because the second occurrence of `Nixon' is interpreted as anaphorically related to the first. Since that is so, everyone's account of the statement that uses of (3a) make has to represent the anaphoric relation, including Kripke's. The rigidity of the two occurrences of `Nixon in uses of (3a) does not deliver the coreference of those occurrences. Rigid designation tells us only that an occurrence of a proper noun (both inside as well as outside a sentence) denotes the same thing in every possible world. Therefore, the two occurrences of `Nixon' in (3a) can each rigidly designate without codesignating. There can be tokens of (3a) in which the (rigid) designatum of the first occurrence of `Nixon' is the Richard M. Nixon, while the (rigid) designatum of the second is Thelma Catherine ("Pat")

Nixon--the speaker thinks Nixon is a master female impersonator--or perhaps some other Nixon. Hence, some prior coreferential interpretation of the occurrences of `Nixon' in (3a) is required to preclude disjoint reference of the two occurrences of `Nixon'. No doubt, uses of (3a) typically involve such a prior coreferential interpretation, so that PMT's, Kripke's, or anyone else's coreferential interpretation reflects such typical usage. Still, on everyone's account, the (necessary) falsehood of (3a) depends on interpreting its occurrences of `Nixon' as coreferential to the Richard M. Nixon.[19]



In this section, I will first explain the schema (A). In the course of this explanation, I will show how the new intensionalism handles the Pierre and Paderewski problems and the problem presented by names with multiple bearers. Then I will explain how the new intensionalism handles names without bearers in existence sentences.

The predicate in (A) indicates that the sense of a name contains a representation of the name.[20] As discussed elsewhere, this automatically provides a solution for Kripke's puzzle about belief.[21] On PMT, there is no problem about the consistency of Pierre's beliefs because the subject term in the belief acquired in Paris is `the thing which is a bearer of "Londres"' while the subject term of the belief acquired in London is `the thing which is a bearer of "London"'.

The predicate in (A) also indicates that the sense of a name expresses the condition for being a bearer of a name as part of the sense of the name. I will not try to present an account of what it is to be the bearer of a name. Such an account will have to be provided at some point in the development of PMT as part of a descriptive semantics for natural languages. But we do not need to provide it here because the nature of bearerhood is not part of the issue between the Millian position and PMT. PMT can, in principle, make use of a variety of accounts of bearerhood, including Kripke's and Donnellan's.[22]

The indefinite determiner `a' associated with the bearer relation in (A) represents the linguistic fact that a name is not limited to a single bearer. One name can have many bearers-- for example, all the John Smiths bear the name `John Smith'. This aspect of PMT vindicates Kripke's claim that the fact that a name can have many bearers is completely compatible with rigid designation, hut it delivers compatibility without resorting to what is for Kripke a quite uncharacteristically counterintuitive proposal to regiment our idolects so that "uses of phonetically the same sounds to name distinct objects count as distinct names," a proposal that, he observes, associates him with classical description theorists (``this is not an issue between us'', that is, himself and the classical description theorist) [23]

Kripke and the classical description theorists are not entirely of one mind. The latter say that descriptions give the meaning of a name, and hence they have to say that proper nouns with multiple bearers are ambiguous. But this is clearly wrong. It is not that the proper name `John Smith' has a large number of senses; it is rather that the name has a large number of bearers. The proper name `John Smith' is, as I shall say, referentially equivocal. If, in speaking to philosophers, I say, "Nagel taught in New York City," their problem will not be choosing between senses, as it is in the case of "Let's meet at the bank," but choosing between Ernest and Thomas. Moreover, if each name that had k bearers had k distinct senses, birth records would be a major part of linguistics.

Kripke avoids such problems because he is not a description theorist. But, as a Millian, he has related problems. Instead of proliferating senses for names like `John Smith', he has to proliferate names. On Kripke's proposal, there are thousands of names `John Smith', each of which designates a different bearer, even though, intuitively, there is only the one name `John Smith' with thousands of bearers. Also, on his proposal, the thousands of names `John Smith' are each different in spite of the fact that all of them are the same in phonology, syntax, and sense, and, hence, linguistically the same. Further, the proposal does not provide an intuitively plausible treatment of idiolects in the hardly rare cases where the speaker's competence includes the knowledge that a name has two or more bearers. To put the point another way, it is intuitively quite clear that when two John Smiths meet, they can (truly) remark, "Oh, we have the same name." On Kripke's proposal, they cannot.

The main source of this counterintuitiveness is a failure to separate features of language (the possibility of a proper noun having multiple bearers) from features of language use (the necessity of tokens of a proper noun to refer to a single contextually definite bearer). On PMT's separation, there is only one proper noun `John Smith' in English, but there are a large (but indefinite) number of tokens of this type in the use of English. These tokens fall into many distinct equivalence classes, such that each member of one class designates (perhaps rigidly) a different John Smith from the John Smith designated by the members of the other classes.

The separation of features of language from features of use enables us to pinpoint the exact place where Searle goes wrong in arguing that "the sense of `This is Aristotle"' cannot be anything like "This object is spatio-temporally continuons with an object named `Aristotle'" because "not just any object `Aristotle' will do. `Aristotle' here refers to a particular object named `Aristotle', not to any."[24] The argument goes wrong right here where Searle uses the fact that the sentence token has definite reference to try to show that the type of which it is a token cannot have a sense that allows the possibility of multiple bearers. As we shall see, the definite element in the sense of a proper noun--which Searle overlooks in this argument--requires the referent of a literal token to be restricted to one of the bearers.

The representation of the possibility of multiple bearers within the sense of a proper noun also handles Kripke's Paderewski case.[25] Given that names have a sense of the form (A), a competent speaker knows that it is possible for two occurrences of a name to refer to different bearers of the name. Marie can thus consistently believe the sentences "Paderewski had musical talent" and "Paderewski had no musical talent," even though the musician and politician are one and the same person. To put the point another way: On PMT, full sensitivity to the sense structure of proper nouns would prevent Marie from applying conjunction introduction to the sentences expressing her beliefs to obtain "Paderewski had musical talent and had no musical talent."

The internal structure that (A) ascribes to the senses of names makes it possible to associate different temporal indices with the entire sense of a name and also with the bearer relation part. These indices express, respectively, the temporal designation for applications of the name and the interval over which the bearer relation in question obtains. Normally, the time designated by the former falls within that designated by the latter, but it need not, as is shown by (4a) and (4b).

(4a) Jacqueline Kennedy was the only one close to the Kennedy family not to appear at William K. Smith's trial.
(4b) Cary Grant will quickly adjust to this name and have no regrets about losing his name `Archibald Leach'.

A name change does not preclude the subsequent use of the obsolete name-bearer correlation, and a name change's not having yet occurred does not preclude the use of the (future) name-bearer correlation. Middle-aged Americans can use (4a) among themselves to refer to Jacqueline Onassis because her previous name remains salient in their collective consciousness. Studio executives can use (4b) among themselves to refer to Archibald Leach because they know that actors do what the studios tell them to do. The studio executives refer in a way that is similar to the way the then Senator Gore might have referred to the then Governer Clinton during the 1992 presidential campaign in saying, `President Clinton will not send troops to the Persian Gulf."[26] (See footnote 27 below.)

The determiner `the' associated with the term in referential position represents the linguistic fact that proper names are definite--for example, the proper names `God' and `Satan' are as definite as the noun phrases `the lord' and `the devil'. This feature of a proper name limits the reference of its literal uses to one bearer per use. The definite determiner is thus the source of the pragmatic task, typically imposed on the speaker, of distinguishing the intended bearer from other bearers known to be such in the context. When there is just one such correlation known in the context, or when there is more than one but the context already distinguishes them, the referent of the token can be unequivocal without special effort on the speaker's (or anyone else's) part. When special effort is required, typically the speaker supplies descriptive information to distinguish the bearer to which he or she intends to refer. For example, if the speaker is required to distinguish the Jonah of the Bible from other Jonahs, he or she might describe the intended referent as the bearer of whom the author(s) of the Bible told the Jonah story. Thus, on PMT, Kripke's circularity condition, like other informativeness conditions (for example, Grice's maxim of quantity), enters as a constraint on this pragmatic task.

There is, as it were, a division of labor. The objects in the domain of the language are first filtered by the bearer condition, then by contextual knowledge of name-bearer correlations, and finally by descriptive information introduced to make the referent contextually definite. Descriptive information is required that makes it unequivocal in the context to which of the known bearers of the name the token refers. It is required to distinguish one of the known bearers from the others, but not, beyond that, to distinguish that bearer from everything else in the domain of the language. If descriptive information were required to be absolutely identifying with respect to all name-bearer correlations, rather than only relative to the contextually known correlations, communication would fail far more often than it does.

Thus, PMT differs sharply from traditional description theories in connection with the level at which descriptive information operates in communication, the source of such information, and the role it plays. Instead of coming in at the semantic level, as Fregean semantics requires of traditional description theories, descriptive information comes in at the pragmatic level. The source of such information is not the language, but the context. Further, the role of descriptive information is not to provide the sense content of names, as it is on traditional description theories, but to situationally distinguish the bearer of the name to whom the speaker's utterance refers from the other bearers whom the speaker and audience also know to bear the name. When descriptive information fails to perform its function, as in the "Nagel" case above, the result will be referential equivocality (perhaps in certain cases referential vacuity).

Descriptive information plays essentially the same role in the use of tokens of bearerless names to make statements about reality. Someone who wants to use `Saint Nicholas' to make the true statement that there never was a Santa Claus, but who is worried about being taken to make the false statement that there never was a Saint Nicholas who was bishop of Myra, might distinguish the intended bearer by saying, "There never was a Saint Nicholas living at the North Pole." It is not clear how to formulate the constraint on descriptive information in cases where bearerless names are involved. One possibility is to say that the information must distinguish the intended bearer from the other known bearers in other possible worlds where all of the bearers, factual and fictional, exist (with respect to contexts in those possible worlds that are relevantly similar to the one in question). Another possibility is to say that the descriptive information must distinguish the bearer in the relevant story. (See the second part of footnote 28.) For our purposes, the precise formulation is not important.

Definiteness can be obtained without a Kripkean stipulation that rigidifies the reference of a class of tokens of a name, but, as indicated above, it can also go hand and hand with such a stipulation. PMT is compatible with Kripke's account of counterfactual uses of names, just as it is compatible with his account of the bearer relation. Fregean intensionalism is not. Frege's definition of sense, on which a sense structure like that in common nouns is the source of the uniformity in the reference of proper nouns, precludes Kripke's pragmatic account of the source of that uniformity, but (B), on which sense structure cannot be the source of the uniformity in reference, does not preclude such an account. Kripke correctly saw both the need to distinguish between the tasks of giving a meaning and fixing reference and the hopelessness of relying on a conventional description theory for either task. But, because he understood the notion of sense in Fregean terms, be had to deny that names have sense in order to capture the uniformity in their reference. With (B) instead of Frege's definition, we obtain a grammatical explanation of Kripke's distinction between giving the meaning and fixing reference that preserves his important suggestions about bearerhood and rigid designation.

We now come to sentences like (la) and (1b). The fact that proper nouns have a sense on PMT means that such sentences do not pose the problems in the area of language for PMT that they do for Millianism. Further, since such sentence types are meaningful, the account of how their tokens come to have a meaning in literal uses of language follows directly from the standard intensionalist account of how tokens of sentence types come to have a meaning in literal uses of language: the sense of literal tokens derives from the sense of their type. Thus, no special account is required. We shall see that a special account is required on the Millian approach.

Let us suppose that the sense of sentences of the form "There is an (no) N" and "N does (not) exist" is something like (C):

(C) Something (Nothing) is the thing with the property P that is a bearer of `N', [27]

where `P' marks the place at which descriptive information may be introduced into instances of (C) to make the reference of tokens of `N' contextually definite by distinguishing their intended bearer from the other known bearers. On (C), the assertion that is made in the literal use of a token of (la) or (1b) (in application to reality) is an assertion about the things in the world, to the effect that none of them is the such-and-such thing that is a bearer of `Santa Claus'. In contexts where there is just one name-bearer correlation for the name `Santa Claus', those tokens can be used to make the assertion that nothing in the world is the unique thing that is a bearer of `Santa Claus'. In contexts where there are two or more known correlations, descriptive information will be required to flesh out the statement in order for it to express the speaker's intention. Supposing it is known that some clever people name boats and dogs `Santa Claus', a speaker might use a token of (la) or (1b) to assert that nothing in the world is the thing who bears the name `Santa Claus' and brings Christmas presents to good little boys and girls.



From the standpoint of solving the problems that (la) and (1b) raise for the Millian position, it is essential not to throw out the intensionalist baby with the Fregean bath water, because when the baby goes, so does all chance of adequately handling sentences like (la) and (1b). This is clearest in connection with the problem that such sentences raise in the area of language. No Millian, to my knowledge, has tried to solve that problem.

The situation is less clear in connection with the problem those sentences raise in the area of language use. Both Donnellan and Kripke have tried to solve it. Each, as might be expected, has come up with an interesting proposal. Donnellan's is an account of how uses of putatively meaningless existence sentences with bearerless names might in themselves express truths. Kripke's proposal, in contrast, is an account of how speakers can express true propositions in the use of such sentences even though their utterances, in themselves, do not express a proposition. In this section, I will argue that Donnellan's proposal is not a good solution to the problem that sentences like (la) and (1b) raise in the area of language use. In the next section, I will argue that Kripke's proposal is not a good solution either.

Donnellan claims that a child who says (5)

(5) Santa Claus will visit tonight

"has not . . . expressed a proposition," asking "Given that [(5)] is a statement about reality and that proper names have no descriptive content, then how are we to represent the proposition expressed?"[28] This is, of course, a rhetorical question, since there is no way to represent the proposition on the Millian position. Millians deny that such sentence types are meaningful, and as a consequence also deny that a speaker's tokens of the type have a sense. Moreover, there is no de re proposition expressing what a speaker states--truly or falsely--in the use of such a sentence. On the de re notion of proposition, the same proposition can be stated "with the aid of other and different singular expressions, so long as they are being used to pick out the same individual" and make the same attribution.[29] Since there is nothing for the speaker to pick out in the case of (5), it cannot express a de re proposition.

On the Millian position, then, whether a sentence like (la), (1b), or (5) expresses a proposition in the de re sense depends on the empirical facts. Accordingly, astronomers on both sides of the controversy about the existence of Vulcan expressed no propositions in their uses of sentences like `There is (is not) a planet Vulcan in the vicinity of Mercury' and `Vulcan is (is not) the cause of Mercury's perturbations'. Had Vulcan existed, they would have expressed true (or false) propositions, but, as things are, they failed to express any proposition. What is troubling about this is that the Millian position seems to provide us with no account of the Vulcan controversy independently of the facts about its actual resolution, that is, no account of the controversy on which it would have been the same even if the other side had won. Controversies are disputes; disputes involve conflicting propositions (in some sense of the term) that are communicated and impart information, but, as the Millian position seems to preclude such propositions, it seems we cannot say, what we intuitively want to say, that a particular belief of the astronomers on one or the other side would have been the same even if the empirical facts had been different--or that if there had turned out to be a planet Vulcan, the very propositions that are false would be true.

Consider another example. Suppose that you are a publisher and that you have just read a fascinating manuscript purporting to be a biography of Tex Tannenbaum, the legendary Jewish cowboy of the Old West. Initially, you believe the manuscript to be authentic, but a routine check produces evidence that it is a hoax. Presumably, the story itself does not change. The sentences you thought were true turn out (as you now think) to be false, but that's it. You say, "Too bad the story turned out to be fiction." But suppose further checking reveals evidence that the story really is true. You now decide you simply don't know whether the story is fact or fiction. The Millian position does not account for the story, the object that remains invariant through your changes in propositional attitude from belief that the story is fact, to belief that it is fiction, to agnosticism.

In the same vein, the child who truthfully asserts (6)

(6) 1 believe Santa Claus will visit tonight

expresses a proposition containing a propositional object of belief. The empirical fact that there is no Santa Claus is not involved here. Given the truth of a use of (6), its complement sentence expresses the object of belief, that is, the propositional object to which the speaker of the sentence is related in virtue of the speaker's having the belief. Since the complement sentence in (6) is (5), the fact that (5) contributes a propositional object to (6), together with standard assumptions about compositionality, entails that (5) expresses a proposition. Thus, in the case of (5), the existence of a proposition, in some quite ordinary sense of this notion, cannot depend on the empirical facts about the residents of the North Pole. It looks, therefore, as if, Millianism notwithstanding, existence sentences with bearerless names have to express propositions in the familiar, purely linguistic sense of the term.

Donnellan himself seems to accept that sense of proposition, and hence fails to maintain a consistent position. In rejecting a solution based on Russell's treatment of names "in the strict logical sense", Donnellan says,

There simply are no meaningful existential statements involving these "genuine" names But, of course, we cannot countenance this about ordinary proper names, for it does make sense to say, "Homer existed" or `Santa Claus does not exist.

Thus, it looks as if Donnellan is now saying that sentences like (la) and (1b) make sense and are "meaningful existential statements." That would be inconsistent with what had been said about negative existential sentences with bearerless names. If such sentences "make sense," are "meaningful," then there is a notion of proposition--where propositions are senses or meanings--on which sentences like (1a) and (1b) express a proposition. Furthermore, in the same discussion, Donnellan seems to endorse that notion of proposition when he equates `having a meaning' with `expressing a proposition' in remarking "this will come to saying that, for example, `Socrates does not exist' means the same thing as (expresses the same proposition as) some other sentence."[31]

Let us now consider another sort of difficulty. Donnellan's semantics contains three categories of semantic objects: the meanings of sentences, the propositions they express in use, and the truth conditions for the statements made in the use of sentences. Donnellan distinguishes the first category from the second in the following way:

[I]f, at the same moment of time, one person were to say, `Smith is happy," a second "You are happy," a third "My son is happy," and a fourth "I am happy," and if in each case the singular expression refers to the same individual, all four have expressed the same proposition.[32]

The meanings of these sentences vary with the different senses of their grammatical subjects, while the propositions that the speakers express are the same when the referent of their subject is the same.

Donnellan distinguishes propositions from truth conditions in the following way:

If you say `Henry is bald" and I say "George is bald," we express the same proposition if the person you referred to by using the name `Henry" and the person I referred to using the name `George" are the same person. But what you say is true if and only if the person you referred to... when you used the name "Henry" has the property of being bald; whereas what I say is true if and only if what I referred to by using the name "George" has the property of being bald. The truth conditions are different because they must be stated in terms of what is referred to by different expressions, in the one case my use of the name "George" and in the other your use of the name "Henry." Yet we may express the same proposition.[33]

The distinction between meanings and propositions is clear, as is the distinction between propositions and truth conditions. What is less clear is the distinction between meanings and truth conditions. Donnellan says that the truth conditions for negative existential statements that a speaker might make with (la) and (1b) are satisfied when the history of the relevant uses of the name ends in "a block," that is, when those uses end with "events that preclude any referent being identified."[34] He says that the rule for the truth conditions of negative existential statements is not a meaning analysis. The rule, Donnellan writes,

[does] not tell us what such statements mean or what proposition they express. This means that in this case we are divorcing truth conditions from meaning[35]

While it is true for Donnellan that the general rule for the truth conditions of sentences like (la) and (1b) is not itself a meaning analysis, a meaning analysis of such sentences might nonetheless show their meaning to provide the particular truth conditions generalized in the rule. Donnellan says that the truth conditions for the statements in the use of sentences like `Homer is a great poet' are that there is an individual with the appropriate historical relation to the relevant uses of `Homer' and that that individual has the property of being a great poet. Since the individual with the appropriate historical relation to the use of `Homer' is the appropriate bearer of `Homer', talk about these statements and their truth conditions does not essentially differ from PMT's talk about metalinguistic meanings/truth conditions. The truth conditions in both cases are simply that the appropriate bearer of `Homer' has the property of being a great poet. Moreover, since a block is nothing but the absence of an individual with an appropriate historical relation to the relevant uses of the name, the truth of (la) and (1b) in both the case of Donnellan's theory and in the case- of PMT comes down to there being no appropriate bearer of the name `Santa Claus'. There is no essential difference between Donnellan's non-meaning based, metalinguistic truth conditions and PMT's meaning-based, metalinguistic truth conditions.

Both Donnellan and I reject Fregean description theories. Neither of us takes an approach like Kripke's, which denies that uses of sentences like (la) and (1b) in themselves express truths about reality. Under these conditions, only a pure metalinguistic condition can account for the truths expressed by sentences like (la) and (1b). Where Donnellan and I differ is on how tokens of such sentences obtain the metalinguistic clause(s) of their truth conditions. I say that clauses associated with a proper noun in the sentence come in the standard way from its linguistic meaning via the relation of that meaning to the literal meaning and reference of its tokens. Donnellan says that such clauses come in a nonstandard way and are, as it were, an emergent property at the level of use.

I see two (related) difficulties with saying this. One, indicated already, is that divorcing truth conditions from meaning in the case of sentences like (la) and (1b) but not in the case of other sentences makes the former an ad hoc deviation from an otherwise general pattern in the way literal token meaning reflects type meaning. That the deviation is made under the pressure of Millan doctrine ought to some extent to count against that doctrine.

The other difficulty is a loss in economy. Since Donnellan's semantics has a category of meanings as well as categories of propositions and truth conditions, it already contains a potential locus for metalinguistic truth conditions in the case of names in the category of meaning. It is only Millian
doctrine that prevents Donnellan from locating the source of such conditions in that category. Thus, Millian doctrine dictates "divorc[ing] truth conditions from meaning" and creating a third category of truth conditions endogenously associated with the use of sentences like (la) and (1b). But the creation of such an ideologically safe haven for the indispensable metalinguistic conditions comes at a cost in economy.

Donnellan has to create an independent category of semantic objects with essentially the same role as meanings, whereas PMT locates the indispensable metalinguistic conditions in the category of meanings and has them filter down. It thus does not have to posit such conditions independently; because Donnellan's theory cannot do this, it does have to posit them independently. Hence, there is a straightforward Occam's razor argument against Donnellan's theory, namely, that positing a category of metalinguistic truth conditions that emerge in the use of sentences like (la) and (1b) unnecessarily proliferates semantic categories.

Do Donnellan's metalinguistic truth conditions help the neoMillian position deal with Frege's familiar problems about identity and oblique contexts from "On Sense and Reference"?[37] Consider the problem that on the Russellian notion of proposition, `Hesperus is Phosphorus' expresses the same proposition as `Hesperus is Hesperus', so that either both state an empirical discovery or both state an a priori triviality. Donnellan's metalinguistic truth conditions seem to help with this Fregean problem. But as Donnellan applies them in the case of bearerless names, they do not.

Donnellan wants to say that the French-speaking child who asserts (7a) and the English-speaking child who asserts (7b)

(7a) Santa Claus does not exist
(7b) Pere Noel n'existe pas

"have learned the same fact and, on that account, ... have expressed the same proposition.'[38] But he notes that his own treatment of the truth conditions of statements with proper names, like Frege's Begriffsschrift treatment of identity statements, turns statements expressing facts about the world into statements about the language.[39] To avoid making truth conditions depend upon facts about particular names in the case of sentences like (7a) and (7b), Donnellan modifies his treatment to make both the truth conditions of such sentences and the propositions they express the same whenever the names in the sentences exhibit a "historical connection between the blocks involved."[40] Since, as Donnellan observes, the blocks in the case of the names `Santa Claus' and `Pare Noel' are connected, (7a) and (7b) have, on his modification, the same truth conditions, state the same fact, and express the same true statement.[41] But, of course, the same is then true for (8a) and (8b).

(8a) Santa Claus is Santa Claus.
(8b) Santa Claus is Pere Noel.

Hence, Donnellan faces the problem about identity that Frege was faced with and that he solved in replacing his Begriffsschrzft treatment of identity with one involving senses.[42] Since Millianism prevents Donnellan from using the senses of names in treating (8a) and (8b), he has to deny the intuitively obvious fact that (8a) is trivial while (8b) is informative.

As we saw above, Frege's problem about oblique contexts arises for Millians in connection with sentences like (6). Since a conventional Millian position has nothing to serve as a suitable object of the propositional attitude in cases like (9a)

(9a) Many little children believe there is a Santa Claus,

the question here is whether Donnellan's metalinguistic truth conditions fill the gap. But they do not provide a suitable object. When the Donnellan-type truth conditions associated with `there is a Santa Claus' are taken as the object of belief in (9a), the sentence is represented as asserting that an appropriate number of children believe that the relevant uses of the name `Santa Claus' do not end in a block. This, however, is clearly quite different from the assertion that an appropriate number of children believe that Santa Claus exists. The same point emerges in connection with a counterfactual like (9b)

(9b) Had Socrates not existed, Plato would not have taken up philosophy,

which does not assert a causal relation between a feature of the use of the name `Socrates' and Plato's not taking up philosophy.



Kripke's proposal differs from Donnellan's.[43] Kripke agrees with Donnellan that in and of themselves, utterances of (la) and (1b) have no meaning and express no proposition. But, contrary to Donnellan, he thinks that in and of themselves, they have no truth conditions and make no statement, either. Kripke thinks that proper names are as senseless as nonsense words like `Bandersnatch', and models his account of (la) and (1b) on what he takes to be the natural account of (10).

(10) Bandersnatches do not exist.

The account runs like this. Since the only source of `Bandersnatch' is Lewis Carroll's nonsense poem "Jabberwocky," it is a nonsense word. We cannot even say that it is a contingent truth that Bandersnatches do not exist. We have no notion whatever of the circumstances under which there might be such things. Hence, (10) expresses no proposition and its utterances make no assertion about what is the case. Since `Santa Claus' is as meaningless as `Bandersnatch', the same things are to be said about (la) and (1b). That does not mean we do not make true statements when we use (la) and (1b) to deny that the fictional character Santa Claus actually exists. Kripke thinks that those who use such sentences in that way speak "carelessly," but that this "careless" speech somehow conveys the same proposition that a "careful" person asserts in the use of a sentence like (11).

(11) There is no true proposition that Santa Claus exists.

Before examining Kripke's proposal, I want to mention a number of differences between proper nouns and nonsense words that, I think, cast doubt on the use of nonsense words as a model for the semantics of proper nouns. First, proper nouns are not etymologically exceptional in the manner of nonsense words like `Bandersnatch'. Proper nouns like `Clark Kent', `Sherlock Holmes', `Peter Parker', etc., are ordinary English nouns whose source is not a nonsense context in a work of fiction and whose use is not constrained by such a context. They appear in a variety of meaningful speech and prose, frequently referring to actual people. Second, proper nouns are not prima facie senseless in the manner of `Bandersnatch'. Rightly or wrongly, many philosophers from Frege to the present have unhesitatingly taken names both with and without bearers to have a sense. No one, I should think, would be tempted to offer a Fregean description as the sense of `Bandersnatch'.

Kripke is surely correct that we have no notion of the circumstances under which there might be a Bandersnatch. If we had, Lewis Carroll would have had to provide it. Not only did he not, but it is doubtful that he could have provided it and still had the nonsense word `Bandersnatch' he needed. The contrast with the proper noun `Sherlock Holmes' is striking. Unlike Lewis Carroll's "Jabberwocky," which is nonsense, Conan Doyle's A Study in Scarlet is not nonsense. As a consequence, in the case of `Sherlock Holmes', we do have a notion of the circumstances under which there might be a Sherlock Holmes--namely, A Study in Scarlet turns out to be biography instead of fiction, to be a chronicle of the exploits of Dr. Joseph Bell of the Edinburgh Infirmary.[44] The existence of such circumstances undermines the claim that the name `Sherlock Holmes' appearing in the book is on all fours with `Bandersnatch'.

Moreover, if someone questions our use of (10) by asking, "Who or what does not exist?" then, true enough, there is no answer. It doesn't help to say "Something frumious," or "Something named `Bandersnatch'," or even `Something to which `Bandersnatch' applies," since on Kripke's assumptions, `Bandersnatch' could not apply to anything. In contrast, if someone responds to our use of `Sherlock Holmes does not exist' by asking, "Who or what does not exist?" there is an answer-- namely, that the detective hero of A Study in Scarlet does not exist.[45]

Kripke's proposal contains both a material and a logical difficulty, each of which by itself is enough to show that the proposal does not explain the use of tokens of (la) and (1b) to assert the nonexistence of Santa Claus. The material difficulty is that what a speaker wants to say in a "careless" use of sentences like (la) and (1b) cannot, in fact, be said with (11). The speaker wants to assert the nonexistence of Santa Claus, but (11), instead of asserting this, asserts the nonexistence of certain propositions. Mightn't there be no such propositions, yet be a Santa Claus? Philosophers who do not believe in propositions cannot in good conscience use (11) to break the news about the true source of the Christmas presents their children receive. (ii) is not a "careful" way of saying something that is "carelessly" said in uses of (1a) and (1b), but a way of saying something different.

The logical difficulty with the proposal is that (11) contains the clause (12)

(12) Santa Clans exists,

which Millians have to count as being as senseless as (la) or (1b). Hence, if (la) and (1b) have to he replaced with (11), the clause (12) in (11) has to be replaced by (13).

(13) There is a true proposition that Santa Claus exists.

But the replacement does not help, because the result of the replacement is (14)

(14) There is no true proposition that there is a true proposition that Santa Claus exists,

and (14) still contains the senseless clause (12). Clearly, further such replacements only continue the regress. Moreover, the senseless clause, or some replacement containing it, cannot just be omitted. It has to appear because the noun phrase `the proposition' in (11) requires a complement to specify the proposition in question. Since the required complement must use rather than mention the name `Santa Claus', the "careful" sentence itself will be counted as senseless on Millian grounds. Hence, as a consequence of the uneliminable occurrence of (12) in any "careful" way of talking, those ways of talking are as meaningless as the "careless"



If the argument thus far is correct, and if Donnellan's and Kripke's approaches are the only ones open to Millians, there is only one way left for neo-Millians to try to save their theory. They have to provide new criticisms of pure metalinguistic description theories that show those theories to have even worse problems than neoMillian theories. In his Shearman and John Locke lectures, Kripke makes some new criticisms of pure metalinguistic description theories. In this section, I want to show that these criticisms do not raise problems for PMT.[47]

The basic reason that the criticisms do not raise problems for PMT is that they apply only to metalinguistic description theories that represent sentences like (la) and (1b) on the basis of a paradigm like (15a) or (15b).

(15a) `N' denotes nothing.
(15b) `N' has no referent.
(15c) Moses does not exist.
(15d) Moses led the Israelites out of captivity.

So represented, a sentence like (15c) is about the name `Moses'. Kripke is certainly right to say that theories that analyze (15c) as asserting that `Moses' has no referent (denotes nothing) are implausible because that assertion is not at all the same as the assertion that Moses does not exist. But PMT does not base its representation of a sentence like (15c) on a paradigm like (15a) or (1 5b). Rather, PMT bases its representation on the paradigms (A) and (C). Thus, neither (15c) nor (15d) is about the name `Moses' because neither has a semantic form containing a term referring to the name. (15c) is about the world, and (15d) about Moses.

In a related argument, Kripke points out that it is possible forMoses to have lived, led the Israelites out of bondage, and so on, but not have had the name `Moses'. But, as can be seen from the discussion of (3a), PMT does not deny this possibility. In addition to not analyzing sentences like (15c) as having the form (15a) or (15b), PMT is not committed to claiming that the name `Moses' in (15c) is a name that Moses had during his own life, in ancient Israel, or anything of the sort. PMT is committed to nothing more than that Moses bears `Moses' in our linguistic community, which is guaranteed by the fact that (15c) and (15d) are sentences of our language. On PMT, (15c) would state a historical falsehood in the circumstances Kripke is entertaining.

Kripke points out that the name in sentences like (15c) is used, whereas metalinguistic description theories represent it as mentioned. But mentioning per se does not raise a problem for such theories; a problem arises only when a theory represents a mentioned name as the object of predication in sentences like (15c), since only then does the theory represent the sentence as making an assertion about the mentioned name. Theories that employ the paradigm (15a) or (15b) do represent the quoted name as the object of predication, and hence they incorrectly analyze (15c) as being about the name `Moses'. However, PMT is different. Since PMT represents the sense of (15c) as an instance of (C), it does not represent the quoted occurrence of `Moses' as the object of predication in the semantic structure of (15c), but represents it rather as a component predicate of the predicate. PMT represents (15c) not as being about `Moses' but as being about the world.

Similarly. PMT also represents (15d) as about Moses rather than `Moses'. (15d) is of subject-predicate syntactic form, with `Moses' as the subject and `led the Israelites out of captivity' as the predicate. On PMT's representation of the sense of (15d), the sense of its subject is an instance of (A). Thus, the sense of the subject is not something like "the name `Moses'" but "the definite thing that is a bearer of `Moses'." Hence, uses of (15d) and sentences like it are also not represented as about the name in them; they are represented as about the contextually definite bearer of the name Moses'.

Kripke also argues that metalinguistic theories cannot explain the fact that the statement we make in a use of (16)

(16) `Sherlock Holmes' has no referent

is comprehensible to people who know nothing about Conan Doyle's fictional detective, whereas the statement we make in a use of (17)

(17) Sherlock Holmes does not exist

is not. Again, the argument does not apply to PMT. There is no problem about the comprehensibility of (16), and on PMT, people who know nothing of Conan Doyle's stories will find the statement that we make in uses of (17) to be incomprehensible because, having no idea that a bearer of `Sherlock Holmes' is the fictional detective in those stories, they do not know to whom the statement purports to refer.[48]



Names can have any number of bearers--none, one, or many. In the case where a name has one and only one bearer, Millianism makes its best showing, and it is not hard to see why. Millians are extensionalists with respect to names, and when names have a unique bearer the extensionalist strategy of using reference as a surrogate for sense seems to work because, in this case, referential structure tracks sense structure tolerably well.

But the extensionalist strategy makes a strikingly poor showing in both the case of bearerlessness and that of multiple bearers. Just as the full-fledged extensionalist's use of reference as a surrogate for sense breaks down in cases where there is sense but no reference, like Frege's "the least convergent series," so the neo-Millian's use of Russellian propositions as a surrogate for senses of sentences breaks down in cases like (la) and (1b). In those cases, as well as other cases of bearerless names (such as Frege-type identity sentences and belief sentences), and in cases of names with multiple hearers, referential structure does not track sense structure.

Successful theories of language and language use require not only senses for names, btit senses with the proper decompositional structure. Such theories must have something like (B) as their account of sense and (A) as their account of the sense of names in order to have access to the internal components of the senses of names. Without such access, the problems that bearerlessness, unique bearerhood, and multiple bearerhood present will remain with us. With it, it seems possible to free ourselves from them.

PMT is free of the problems that Millians have in explaining the meaningfulness of sentences types like (1a) and (1b), and of the problems they have in explaining how the tokens of (la) and (1b) make true statements about reality. It is free of the Pierre and Paderewski puzzles, which are only another example of the failure of reference to track sense. Further, PMT is free of the problem that traditional description theories have with the relation of names to their hearers in counterfactual situations, as well as the mention/use problems that Kripke pointed out in connection with some pure metalinguistic description theories. Finally, PMT is free of the pressure to idealize natural language that both Fregean description theories and neo-Millian theories are under in connection with names having multiple bearers.

PMT thus provides an account of names with none of the major problems that have haunted attempts to understand names since those problems first emerged in the clash between Mill's and Frege's semantics.



[1] See Frege's related argument that the meaningfulness of sentences containing bearerless names implies that proper names have sense, in Gottlob Frege, Introduction to Logic,' in Posthumous Writings, ed. P. Long, R. White, and R. Hargreaves (Oxford: Basil Blackwell, 1979), 191--92.

[2] Jerrold J Katz `The Neoclassical Theory of Reference, in Contemporary Perspectives in the Philosophy of Language, ed. P. A. French, T. F. Uehling,Jr., and H. K. Wettstein (Minneapolis: University of Minnesota Press, 1979),103--24, and Jerrold J. Katz, "Has the Description Theory of Names beenUniversity Press, 1990), 31--61.

[3] ` Has the Description Theory of Names been Refuted?" 39--45.

[4] For examples illustrating the kind of senses that Frege takes proper nouns to have, see Gottlob Frege, "On Sense and Reference," in Translations from the Philosophical Writings of Gottlob Frege, ed. M. Black and P. Geach (Oxford: Basil Blackwell, 1952), 58. For a critical discussion of Searle's suggestion that the senses of common nouns might be metalinguistic like those of proper nouns, see "Has the Description Theory of Names been Refuted?" 53. It is worth adding here that Searle could not make use of a schema for the senses of common nouns like "` N" means `object to which `N' applies"' because this schema is no more adequate than the one he offered in John R. Searle, "Proper Names," Mind 67 (1958): 166--73. ``Bachelor" means "object to which `bachelor' applies"' will not explain the redundancy of `unmarried bachelor', the antonymy of `bachelor' and `spinster', and the analyticity of `Bachelors are male'. Moreover, if we expand the sense of `bachelor' beyond the instance of the schema to include the concepts required to explain such sense facts, then the instance of the schema is no longer necessary to explain sense and has to he removed on grounds of simplicity.

[5] "On Sense and Reference." 60.

[6] Bertrand Russell, `The Philosophy of Logical Atomism," in Logic and Knowledge, ed. R. C. Marsh (London: Macmillan, 1956), 241--54; William Kneale, "Modality, De Dicto and Dc Re," in Logic, Methodology, and Philosophy of Science, ed. E. Nagel, P. Suppes, and A. Tarski (Stanford: Stanford University Press, 1962), 622--33.

[7] Saul Kripke, Naming and Nece6sity (Cambridge: Harvard University Press,1980), 68--70.

[8] Jerrold J. Katz, `The New Intensionalism," Mind 101 (1992): 689--719, and `Has the Description Theory of Names been Refuted?" 46--53. For discussions of the questions in somewhat different contexts, see Jerrold J. Katz, `Why Intensionalists Ought Not Be Fregeans," in Truth and Interpretation, ed. E. LePore (Oxford: Basil Blackwcll, 1986), 59--91, and Jerrold J. Katz, The Metaphysics of Meaning (Cambridge: MIT Press, 1990), 21--162.

[9] Quine is popularly thought to have shown that it is circular to use sense properties and relations to define the notion of sense, but see, inter alia, Jerrold J.. Katz, "The Refutation of Indeterminacy," Journal of Philosophy 85 (1988): 227--252. Quine appears to have dropped the criticism; see Willard Van Orman Quine, "Comments on Katz," in Perspectives on Quine, ed. R. Barrett and R. Gibson (Oxford: Basil Blackwell, 1990), 198--99, and The Metaphysics of Meaning, 199--202.

[10] "Has the Description Theory of Names been Refined?" 46--50.

[11] Generally, on the new intensionalisin, sense not only does not determine what uses of expressions refer to, it does not even determine what they mean. That is to say, the sense of an expression in the language, a linguistic type, is not sufficient to determine the senses of its tokens. Extralinguistic information of various sorts plays a significant role. Moreover, insofar as information necessary for fixing the reference of a token conies from its sense, the sense of linguistic types is, as it were, doubly insufficient to fix the reference of their tokens. See Jerrold J. Katz, "Reply to Boghossian," in `Symposium on The Metaphysics of Meaning," Philosophical and Phenomenological Research 54(1994): 161--70.

In "Reference and Definite Descriptions," Philosophical Review 75 (1966):

281--304, fn. 9, Donnellan argues that it is not clear that linguistic tokens always have a determinate sense. The new intensionalism is not committed to there always being a determinate sense for linguistic tokens, any more than it is committed to there always being a referent, or an unequivocal one.

[12] John S. Mill, A System of Logic, vol. 1 (London: Parker, Son, and Bourn, 1862), 36--37.

[13] Naming and Necessity, 67 and 87. It is worth noting that cases of the Jonah type refute neo-Fregean views like that in Graeme Forbes, "The Indispensability of Sinn," Philosophical Review 99 (1990): 535--63. Forbes claims that "the cognitive significance of the sense [of a proper name is `the person/thing this body of information is about'," where the information in question is a specified, but revisable, dossier of descriptive conditions (538--39). But in the Jonah case, the informarion in the dossier-- which is everything we have--is all false of Jonah, and, presumably, of everyone else. Hence, Forbes's theory mistakenly claims that in the counterfactual case, Jonah didn't exist. Note, further, that this difficulty is not avoided by introducing the notion that the subject of the dossier is "the object at the start of the channel along which the de re information in the dossier has flowed to the thinker" (510). The information in the Jonah dossier might have flowed from the improvisations of an ancient actor whose performance was then written down by the author(s) of The Book of Jonah. In this case, the actor is the source of the information in our dossier, not Jonah.

[14] L. Wittgenstein, Philosophical Investigations (Oxford: Basil Blackwell, 1953), sec. 79, and "Proper Names," 166--173.

[15] N aming and Necessity, 71 and 8 1--82.

[16] Ibid 48--49.

[17] 'See "Has the Description Theory of Names been Refuted?" 40--45 for the original statement of the response presented below. I think the view that proper names in natural language are predicates, in Tyler Burge, "Reference and Proper Names," Journal of Philosophy 70 (1973): 425--39, is not correct. I cannot examine Burge's view here, but the point needs to be made that the examples cited (see 429--43]) are not clearly evidence that proper nouns are predicates. The examples need not be so understood, and even if they are, they show only that proper nouns, like common nouns, occur in predicate phrases. Nothing is said to show that they occur as predicates in those phrases. Another point is that Burge's understanding of the idea that a name enters into its own application conditions construes those conditions pragmatically, so that his conception of the role of names in communication fits nicely with a Millian position like Donnellan's.

[18] Nonuing and Necessity, 49.

[19] The point is, if anything, more evident in cases where the terms are different names having a common bearer, such as in `George Orwell might not have been Eric Blair'. (I put the example this way to avoid the salient but, in this context, misleading reading of `Eric Blair might not have been George Orwell'--for example, if Blair did not become a writer.) If we want to use this sentence in place of (3a) to unequivocally express the type of false statement that (3a) expresses, then, since it is all the more clear here that rigidity has to be supplemented with coreference, on everyone's story there will have to be some indication of the anaphoric relation.

[20] (A) entails that different proper names have different senses. This consequence can be independently supported in various ways. It is confirmed by the fact that `Is the White House the White House?' is trifling (a self-answered question), but `Is the White House la Maison Blanche?' is not; it is also confirmed by the fact that `John believes Germany is in Europe' can be true while `John believes Deutschland is in Europe' is false. As I point out in the text, this consequence of (A) provides a simple solution to Kripke's puzzle about belief. It also explains why lexical proper nouns like `Louis' and `Ludwig' are not synonymous and why the English name of the philosopher is `Ludwig Wittgenstein', not `Louis Wirtgenstein'.

But the fact that proper nouns have different senses does not entail that they are not in some sense translatable. Phrasal proper nouns like `the White House' and `la Maison Blanche' are translations. Exactly what such translation consists in, however, is not clear. Those proper nouns are trans lated by translating their common noun constituents term for term, but their reference is not fixed by what would be their compositional meaning were they common nouns. The name `the White House' could remain the name of the building despite the white paint turning grey from air pollution and the building being converted from the residence of the first family to a museum. As I see it, the proper conclusion to draw about translation is that it can be based on a weaker relation than synonymy or perhaps on a family of weaker relations. Independent support for this view is found in Tyler Burge, "Self-Reference and Translations," in Translation and Meaning: Philosophical and Linguistic Approaches, ed. F. Guenthner and M. Guenthner-Reuter (London: Duckworth, 1978), 137--56.

[21] Saul Kripke, "A Puzzle about Belief," in Propositions and Attitudes, ed. N.Salmon and S. Soames (Oxford: Oxford University Press, 1988), 102--48. The point in the text is made at greater length in "Why Intensionalists Ought Not Be Fregeans," 59--91, and in "Has the Description Theory of Names been Refuted?" 32.

[22] The Cambridge University philosophical logician W. F. Johnson anticipated the theory that the bearer of a name is the individual who was given it in an appropriate "baptismal" ceremony. See W. E. Johnson, Logic, vol. I (New York: Dover, 1964), 84--85. He writes that `we should justify our applying the name `Roger Tichhorne' to the man presented in court" on the grounds of "the presumed identity of the man before us with the man to whom his godparents had given the name." Johnson also has a similar notion of transmission: "the application of a proper name amongst those who continue to use it with mutual understanding" (84).

[23] Na;ning arid Necessity, 7--S. Note that the one example Kripke presents to mitigate the counterintuitiveness of the proposal--that is, the question `How many names are there in this telephone book?' (8 n. 10)--is not persuasive, because the total number of names in phone books listing many people with the same name is obtained by the dubious device of counting each inscription of a name as a distinct name rather than as a distinct token of the same name.

[24] 'Proper Names," 169.

[25] "A Puzzle about Belief," 102--48.

[26] The fact that present knowledge of a future name-bearer correlation can ground the reference of a use of a name when there is no bearer at the time raises a problem for the causal (historical) theory. Insofar as the "baptismal ceremony" in which Archibald Leach acquires the name `Cary Grant' has not occurred at the time the studio executive uses (4b), the theory has to invoke the concept of a contemplated baptismal ceremony to explain the speaker's use of the future name to make a statement about Archibald I.each, in order to avoid backwards causation. This seems to he a serious departure from the central idea of a causal (historical) chain.

[27] No paradox of nonbeing arises here. Being an instance of (C), the sense of a sentence like (la) and (1b) predicates something about the world, not something about a nonexistent being. Essentially the same accountt of existence sentences appears in Jerrold J. Katz, Cogitations (New York: Oxford University Press, 1988), 98--105. Note that any account also requires the temporal indices discussed in connection with (A). Suppose Gordon Glockenspiel, the sole person so cursed, changes his name to `Gordon Smith'. Once the change is official, (i) makes a true statement but (ii) makes a false one.

(i) Nothing is a bearer of `Gordon Glockenspiel'.
(ii) Gordon Glockenspiel does not exist.

Now, recall our earlier point that a name change does not eliminate the use of the name bearer relation. So Gordon's friends might subsequently gossip, `Have you heard, Gordon Glockenspiel changed his name last month? Now he's Gordon Smith." Moreover, if name changes did completely expunge name-bearer relations, the sentence (ii) would not be false but true, in the manner of sentences like (1b). The gossip about Gordon and the falsehood of (ii) depend on the relation in the past between the name `Gordon Glockenspiel' and the present Gordon Smith. Given appropriate temporal designations in instances of (C), the proposition expressed by the false token of (ii) is not the truth expressed by (i) hut something like the false proposition `Nothing in the world now is the unique thing which once was a bearer of "Gordon Glockenspiel"'. Thus, there will have to be a temporal designation expressing the time at which the assertion is made about the world and another expressing the time at which the bearer relation obtains.

[28] Keith S. Donnellan, "Speaking of Nothing," Philosophical Review 83 (1974): 20--21, and11, for the notion of a proposition.

[29] Ibid., 2 1--22.

[30] Ihid., 22.

[31] Ihid., 21.

[32] Ibid., 11.

[33] lbid., 28.

[34] Ibid., 22--25.

[35] Ibid., 25.

[36] lbid., 22

[37] Translations from the Philosophical Writings of Gottlob Frege, 42--78.

[38] "Speaking of Nothing," 27]

[39] Ibid., 27.

[40] Ibid., 29--30.

[41] Ibid., 27--30.

[42] Not that Frege succeeded in explaining this difference; see "Has the Description Theory of Names been Refuted?' 32--33.

[43] The proposal is made in Kripke's widely circulated and highly influential Shearman Lectures, "Empty Reference," and 1973 John Locke Lectures. I discuss his proposal even though neither set of lectures has yet been published because it has played so important a role in the continuing philosophical controversy about names that no attempt to settle the controversy that ignores the proposal could reach a definitive conclusion. Furthermore, for my argument for PMT to fail to respond to such a wellknown and influential proposal would seem like evasion and would perhaps leave the impression that no response can be made. Worse yet, failing to consider the Shearman and John Locke lectures would leave the impression that PMT would not stand up to the criticisms of pure metalinguistic description theories in them (see section 8). Kripke may, of course, no longer hold some of the views in "Empty Reference" and the John Locke Lectures.

[44] Conan Doyle is reported to have based Sherlock Holmes on Dr. Joseph Bell, a former teacher famous for his exceptional diagnostic powers.

[45] This point does not depend on the fact that there is a story about Santa Claus. The sentence `Jack Robinson does not exist', where the name is the name in the expression `before you can say jack Robinson', contrasts with (it)), even though the etymology of `Jack Robinson' does not trace this name back to any actual or fictional bearer.

[46] Kripke's formulation uses rather than mentions the name in the complement of the noun phrase `the proposition', hut we might try a metalinguistic formulation in the hope of avoiding the difficulties presented in the text. We might replace (ii) with, say, (11') or (II").

(11') There is no true proposition: `Santa Claus' has a referent.
(11") `Santa Claus exists is not true.
(11"') `Santa Claus exists' expresses no true proposition.

But neither of these formulations provides a satisfactory account of (1a) and 1b). (11') has the difficulty, discussed in the text, that it is about propositions whereas (1a) and (1b) are not. (11") has the parallel difficulty that it is about a sentence. Both (11') and (11 ") are vulnerable to a version of the very arguments that, as we shall see in section 8, Kripke himself deploys against analyzing sentences like (1a) and (1b) as `"Santa Claus" has no referent'. (11"') has the further difficulty that it asserts a relation between a particular sentence and the collection of propositions, an assertion that, depending on the grammar of the language, might he compatible with there being a trite proposition that Santa Claus exists.

[47] Note that on a quite straightforward interpretation, Donnellan's truth conditions for statements (`Speaking of Nothing," 25) are, in effect, (15a) or (15b). If so, Kripke's criticisms raise problems for Donnellan's account of the statements that uses of (la) and (1b) make. If my responses to those criticisms are right, a pure metalinguistic account of such statements cannot avoid Kripke's criticisms without adopting the essential features of PMT.

[48] Kripke argues that, supposing `Sherlock Holmes' is also the name of someone's dog, a metalinguistic description theory would make it impossible to assert the truth that Sherlock Holmes does not exist, because on such a theory, uses of (17) would assert the fallsehood that `Sherlock Holmes' has no referent. Kripke's case is like the Saint Nicholas case in section 5. On PMT, speakers wishing to refer to Conan Doyle's detective in contexts where it is known that there are other hearers, such as a canine named `Sherlock Holmes,' are required to make their use of the name contextually definite, presumably, on the basis of suitable information from Conan Doyle's stories. Of course, a speaker might fail to make his or her reference contextually definite, but in that case the utterance is simply referentially equivocal. But in neither the case where the speaker's utterance unequivocally states that Conan Doyle's detective does not exist, nor the case where it equivocally expresses that statement and the statement that Sherlock Holmes the canine doesn't exist, does PMT represent the speaker as asserting that the name `Sherlock Holmes' has no referent.