Reply to Boghossian Jerrold J. Katz Philosophical Issues, Vol. 4, Naturalism and Normativity. (1993), pp. 142-152. Stable URL: http://links.jstor.org/sici?sici=1533-6077%281993%294%3C142%3ARTB%3E2.0.CO%3B2-U Philosophical Issues is currently published by Ridgeview Publishing Company.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/rpc.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact
[email protected].
http://www.jstor.org Fri May 18 08:52:47 2007
PHILOSOPHICAL ISSUES, 4 Naturalism and Normativity, 1993
Reply t o Boghossian* Jerrold J. Katz
1. Boghossian is skeptical about my claim to provide a new intensionalism. He concludes that the position seems "mystifying" and "implausible". I think this judgment is mistaken. The mistake arises in part from the pervasive influence of Frege's semantics and in part from my own failure to properly clarify the nature and significance of my new intensionalism's departure from Frege's semantics. Boghossian's thoughtful comments give me a welcome opportunity to try again to say what is new about the new intensionalism. Over the last decade, I have been arguing against the notion nearly universal among philosophers- that every intensionalist is a Fregean. Despite those efforts, the notion remains entrenched in contemporary philosophy. Boghossian thinks this is as it should be. He tries to articulate the reason that, in spite of my protestations, I, like other intensionalists, must be a Fregean. He points out that I think "[i] that expressions of natural language have sense as well as *I feel fortunate t o have had such distinguished philosophers as commentators, Paul Boghossian, Paul Horwich, Roger Gibson. I a m grateful t o them for stimulating comments which have helped clarify and extend my thinking on many topics in MM. I want especially t o thank David P i t t and Virginia Valian for many useful comments on a n earlier draft of this and the two following replies.
reference, [ii] that senses are abstract objects.. . and [iii] that they are the proper objects of study in a theory of meaning for natural language" (p. 135). He reminds us that [i]-[iii]is "the essence of the view that Frege placed at the center of the philosophy of language". It is true that Frege and I both endorse the statments [i]-[iii].But we do not understand them in the same way because we define the critical notion of sense differently. Frege defines sense as the determinant of reference, whereas, in using the definition (D), I define sense as the determinant of sense properties and relations. T h e result is two very different conceptions of what it is for an expression of natural language to have a sense. Intuitively, there is something peculiar about defining sense as the determiner of reference. The peculiarity is, I think, that defining sense in this way defines it on the basis of non-intensional concepts. It is like defining pronounciation on the basis of non-phonological concepts or defining number on the basis of non-arithmetic concepts. In Katz (1992, pp. 689-720), I argued that, in defining sense on the basis of the concept of reference and in developing his systematic semantics to serve his logicist purposes, Frege ends up with a conception of what it is for an expression to have a sense which is essentially extensionalist. To obtain a intensionalist conception, we have to start with a definition of sense based on purely intra-intensional concepts, i.e., on concepts belonging to the family of concepts in question, as with the development of phonology and arithmetic, and we have to develop our systematic semantics free of the influence of extra-linguistic purposes. My principal concern in trying to obtain such a conception of what it is for a n expression to have a sense was to get intensionalism out of the trouble that Frege's conception got it into. That trouble first surfaced early in the history of analytic philosophy. In P I , 79-88 and related sections, Wittgenstein argued -clearly with Frege's notion of sense as the determiner of reference in mind- that the possibilities for reference in natural language are not completely constrained by the senses of words. Wittgenstein pointed out, inter alia, that, if we say that the property of being the leader of the Israelites in their flight from Egypt is the Fregean sense of "Moses", then we get the absurd consequence that, in a counterfactual situation where the Israelites did not flee Egyptian captivity, Moses did not exist. T h e trouble, as Wittgenstein notes in P I , sections 80 and 87, is not confined to proper nouns, but arises also with common nouns. Wittgenstein's powerful insight that Frege's definition of sense too severely constrains the possibilities of reference is a t the heart of the criticisms of intensionalism by Putnam, Kripke, and their followers.
All of the responses to Wittgenstein's insight and the criticisms stemming from it were based on the assumption that the troublesome notion of sense is the intensionalist's notion of sense. Wittgenstein threw out the notion of senses as objects. (Quine took a similar position, refusing to countenance intensional objects for any aspect of the extra-logical vocabulary of natural languages.) Putnam, Kripke, and their followers elaborated on Wittgenstein's argument. All of them discarded Frege's notion of sense in connection with names and natural kind terms in favor of a Millian position. Kripke seems to preserve Fregean senses for other types of expressions, but Putnam ("The Meaning of 'Meaning"', Minnesota Studies zn the Philosophy of Science, Vol. VII, 1975, pp. 131-193) replaces Fregean senses with a motley of syntactic, semantic, and empirical factors designed to guarantee that intension determines extension no matter what sort of earth is the referential domain. In preserving a pure notion of sense but abandoning referential determination, (D) provides the basis for a new response to Wittgenstein's insight. With (D), we can concede Wittgenstein's point about the unconstrained possibilities of reference without either rejecting or restricting the notion of senses as objects. (D) thereby makes possible a new intensionalism which has the central explanatory advantages of Frege's semantics without its disadvantages. Hence, Boghossian is not completely correct in saying that "The key.. . to the new intensionalism is the rejection of the claim that sense determines reference". The key is really the new conception of sense in (D); rejecting the claim that sense determines reference is the option that the key unlocks. Recognizing this is the first step in answering Boghossian's question of ". . . how exactly is [the rejection of the claim that sense determines reference] to be understood". Boghossian is right that, on my view, "Senses are, by their nature, not the sorts of objects that have underived extensions -their intrinsic, non-relational, context invariant properties do not suffice to determine extensions for them" (p. 137). But he is wrong in saying t h a t there is a problem about how the intrinsic properties lead to rejection of the sense-determines-reference principle. Let us suppose with Boghossian that tokens of "cube" express CUBE -the sense of "cube". As Boghossian goes on to say, "it is a fact about CUBE that it is true of all and only cubes" and, further, that "our token word, too, is true of all and only cubes". Seeing the case as paradigmatic, Boghossian thinks I am forced to acknowledge the very Fregean principle that I claim to reject. "Where", Boghossian asks, "is the room for play here?"
T h e fact that there are occasions where a token must refer to all and only the cases that it would refer to if the sense-determinesreference principle were true does not show that sense determines reference. I t only shows that there are cases that make it look as if that's so. But it is exactly those apparently standard, apparently paradigmatic cases, that mislead us into thinking that sense determines reference. To see where "the room for play" is, we have to stop looking at what is basically a theoretical issue from the perspective of particular cases. Instead, we have to look at the issue of what the relation between sense and reference is from the perspective of the new conception of sense in (D). Once we have seen where the room for play is, we will return to particular cases to see how something can fall under the sense of a linguistic token, but not be a Fregean case of sense determining reference. Frege's definition of sense institutes an essential connection between sense and reference. In making senses grammatical structures, that is, part of the internal structure of sentences, (D) makes senses independent of any intrinsic connection with reference. As a consequence, the theory of sense no longer has the burden of explaining what relation sense has to reference. The burden shifts from the theory of sense to the theory of reference, where it takes the form of explaining how the referential principles which interpret a language for a domain are related to the sense system in the language. It is thus a mistake to claim that referential facts such as that "cube", meaning what it means, cannot "fail to apply to something just in case it is a regular solid of six equal square sides" are "at the heart of [my] way of looking at things" in the theory of meaning. On my view, the reason that "cube" applies to cubes -when and where it does apply to them- is one that is external to the theory of meaning and internal to the theory of reference. Facts about what this or that word is true of are not grammatical facts, but referential facts. The rooin for play, which Boghossian asks about, is provided by the separation between the theory of sense and the theory of reference. But it is easy to miss. It is natural to think that the separation of those theories comes to nothing more than demoting the sensedetermines-reference principle from the theory of sense to the theory of reference, where it is Fregean business as usual. This is quite wrong. The separation opens up the possibility of a new position on the relation between sense and reference. The new position answers Boghossian's query. T h e extensional positions remain the same. Extensionalists such as Wittgenstein and Quine say that sense plays no role in reference: sense is neither necessary nor sufficient for fixing reference.
But there is a new intensionalist position. Previously, we had two forms of Fregean intensionalism, pure and impure. Pure Fregean intensionalism says that sense determines reference: sense is necessary and sufficient for fixing the reference. Impure Fregean intensionalism such as direct reference theories like Kripke's say that reference is direct in the case of names and natural kind terms, but take it is a standard Fregean position or something very much like it on the rest of the extra-logical vocabulary. (D) provides us with the option of rejecting both forms of Fregean intensionalism. We now have a new position, the new intensionalism, which says that sense mediates reference: sense is necessary, but not sufficient for reference. Boghossian does not appreciate the potential of this feature of the new intensionalism, as I think is clear from his contrast of my position with that of narrow content theorists. He says that I do not mean what those theorists mean, namely, ". . . that senses aren't the sorts of things that can determine their own extension" (p. 138). In fact, I mean precisely that. My position is the same as the narrow content theorist's in this respect. Of course, the narrow content theorist thinks that sense determines certain mental causation, while I think it determines certain grammatical properties and relations. But we both think that it is in the nature of senses to determine something other than reference. Because Boghossian does not see that my notion of sense corresponds to the narrow content theorist's in this respect, he is led to mischaracterize my position as holding that (a) "although senses are the sorts of things that determine their own extensions," ( b ) "nevertheless, the extension of a given token expression need not be identical with the extension of the sense that it expresses" (p. 138). Boghossian comments that he doesn't find this position "terribly plausible". I don't either. The new intensionalism denies (a) and asserts (b). It construes "need not" to express an a priori possibility, not an factual claim. Whether or not the extension of an expression token is identical with the extension of the sense of the expression token depends on the referential conditions for the token. Given the new intensionalism's separation of sense and reference, those conditions are a matter of context. In particular, whether something's falling under the sense of a n expression token is sufficient for it to be the referent of the token is a matter of context. Whether it is depends on the relevant
extra-linguistic, contextual factors being in place. Given that such factors are in place in a use of language, the expression token applies to objects in the extension of its sense. T h e new intensionalism thus allows the extension of an expression token to be identical with the extension of the sense of the expression token. But, even in the hypothetical case in which all expression tokens apply to objects in the extension of their sense, the Fregean conclusion that sense determines reference could not be drawn. The fact that the referent falls under the sense of the token in all cases can mean only that the relevant contextual factors are in place in all of them. To see what such contextual factors are, I want to examine some examples where they are not in place. This examination will explain why a case of something falling under the sense of an expression token is not automatically a Fregean case of sense determining reference. It will also enable us to comply with Boghossian's demand for convincing examples to support (b). T h e contextual factors which cause an expression token to apply to something outside the extension of its sense often involve the falsehood of the belief on which the speaker has chosen a description to pick out his or her intended referent. Perhaps the most famous example of this sort is Donnellan's case of the woman who uses a token of "the man in the corner drinking champagne" to refer to a man drinking sparkling water. In this example, the expression token has the sense THE M A N IN THE CORNER DRINKING CHAMPAGNE, but refers to a man drinking sparkling water because, as Boghossian puts it, "it is so clearly the case that it is that man that the woman had in mind and of whom she falsely believed that he was drinking champagne". Donnellan's case is as straightforward an example of a referent that does not fall under the sense of the expression token as one might want. This case and many others parallel to it in themselves are enough to comply with Boghossian's request to provide convincing examples of (b). Even if "Donnellan's phenomenon" did not seem "to generalize to general terms and to predicates", it is still a powerful counter-example to the charge that ( b ) is implausible. In point of fact, however, Donnellan's phenomenon does generalize to general terms. Let us return to the case in MM (p. 145) that I took to be a typical instance of such generalization, namely, Cotton Mather's uses of "witch" to refer to a woman accused of witchcraft. One such case is Cotton Mather on Gallows Hill saying -with obvious reference to Martha Cory- "The witch is almost dead". From the standpoint of decompositional semantics, it would be extraor-
dinary if Donnellan's phenomenon did not generalize to such a case since there is only an irrelevant syntactic difference between the term "witch" -which means WOMAN WITH MAGIC P O W E R S OBTAINED FROM A N EVIL SPIRITand the term "man in the corner drinking campagne" -which means MAN I N THE CORNER DRINKING CHAMPAGNE. Given that we have a case in which a token of the latter fails t o refer to a man drinking champagne, there is no problem finding a corresponding case in which a token of the former term fails to refer t o a witch. Since the context contains the fact that there are no witches, it falsifies the belief on which Cotton Mather has chosen the description t o pick out his intended referent Martha Cory. All that is required for Cotton Mather's token to apply to something outside the extension of its sense is that, first, the context falsify that belief without revealing that the referent is a non-satisfier, and, second, t h a t it clearly be the case it is Martha Cory that Cotton Mather had in mind and of whom he falsely believed that she is a witch. Thus, suppose that everyone else on Gallows Hill is male and taken by the others t o be a God-fearing ~ h r i s t i a n . ' There is considerable variation among cases to which Donnellan's phenomenon generalizes. For instance, the false belief about the referent may be the addressee's rather than the speaker's or there may be no false belief a t all. The practical joker who delights in fooling visitors with his bowl of imitation fruit says to one, "Take the tasty apple", referring to the only apple-looking piece of imitation fruit in the bowl, which the addressee naively takes. Here, the expression token refers t o a wax apple, not to something in the extension of TASTY APPLE. In the variant of this case where the addressee is not taken in, there is no false belief about the referent. Boghossian says that "it is presumably not the case that when people use general terms.. . they similarly have some specific set of things in mind which are clearly the intended extension of the term as then used" (p. 139). But in the literal use of general terms, people do typically have some specific set of things in mind. Consider an example similar to but simpler than the one Boghossian gives. If l ~ o ~ h o s s i a nexample 's "There are some witches around these parts" is in fact not one I would accept, but I apparently said something in conversation which misleadingly suggested t h a t I would. We agree t h a t uses of t h a t sentence do not work t h e way Cotton Mather's use of "The witch is almost dead" does. As I see i t , t h a t is because those uses are tokens of a n existence sentence, and as such their assertion is something like 'Something is a P (e.g., a witch, around these parts)'. Such cases differ from t h e Cotton Mather case, where t h e term appears as an expression in referential position rather t h a n within a n expression in predicate position.
I say, "The jar contains 3 thousand beans", it is true that I do not mean that it contains kidney beans, fava beans, navy beans, or lima beans, but that it contains 3 thousands objects satisfying the condition of being beans. That is the specific set of things I have in mind as the intended extension of the term as then used. Even here we have Donnellan's phenomenon. If my statement is an entry in Boston's annual bean guessing contest, my entry cannot legitimately be disqualified on the grounds that, although the things in the jar number exactly 3 thousand, they are only wax replicas of beans. Finally, let us return to Boghossian's case of a use of "cube". The cube's falling under the sense of that token is in part due to its sense being REGULAR SOLID OF SIX EQUAL SQUARE SIDES and in part due to the relevant contextual factors being in place, e.g., the belief on which the speaker has chosen the description "cube" to pick out his or her intended referent is true. Thus, Boghossian's case is a case of a cube falling under the sense of a token of "cube", but it is not a Fregean case of sense determining reference. 2. Boghossian expresses doubts concerning whether my view about the relation of sense and reference "is going to help with" any "major problems in the philosophy of language". Over the last few years, I have tried to show that there are a number of problems where it does.2 In the context of the preceding discussion, the illustration that comes first to mind is the argument in MM (pp. 216-224) that the position on sense and reference in the new intensionalism shows that Putnam's influential science fiction cases are not counter-examples to the classical theory of meaning. Putnam's ("It Ain't Necessarily So", The Journal of Philosophy, 1962, pp. 658-671) case in which past applications of "cat" refer to Martian robots works the same way as the Cotton Mather case. In both cases, the context falsifies the belief of the speaker(s) on the basis of which the description was chosen -in the one case, that there are witches and that they are women who appear and behave in the way Martha Cory does and, in the other, that there are feline animals and that they are things which appear and behave in the way the robots do. In both cases, the intended referent is not revealed to be a non-satisfier. In both, it is clear that it is the refere n t ( ~ t)hat the speaker(s) had in mind and of whom he(they) have 2See, for example, my papers "Why Intensionalists Ought Not be Fregeans" in Truth and Interpretation, Basil Blackwell, 1986, pp. 59-91, "Has the Description Theory of Names been Refuted?" in Meaning and Method, Cambridge University Press, 1990, pp. 191-192), and "Names Without Bearers", T h e Philosophical Review,in press.
the false belief. Given that Putnam's case works the same way as the Cotton Mather ,case and that the latter works the same way as Donnellan's, Putnam's case is one in which robots have been referred to as "cat" under a false description FELINE ANIMAL. Hence, Putnam's case is no more of a counter-example to the claim that "cat" analytically entails "animal" than Donnellan's case is a counter-example to the claim that "man in the corner drinking champagne" analytically entails "man drinking an alcholic beverage". An even more straightforward case where the new intensionalism helps in the philosophy of language is Putnam's (1975, pp. 135-136) Twin Earth argument. He states explicitly that, in addition to holding (I) that meanings are in the head (or something equivalent in the case of Platonist theories of meaning), intensionalism holds (11) that intension determines extension. The counterfactual fact that "water" applies to XYZ on Twin Earth (rather than applying to H 2 0 as it does on Earth) falsifies the conjunction of (I) and (11). But, unless the option of denying (11) has been excluded, arguing, as Putnam does, that the Twin Earth example refutes (I) -and hence refutes intensionalism- fallaciously goes from the denial of a conjunction to the denial of one of its conjuncts. In showing that the option denying (11) has not been excluded, the new intensionalism blocks Putnam's argument against intensionalism and provides intensionalists with another reason -ironically the very case on which that argument is based- to eschew Fregean semantics. 3. Boghossian expresses doubts about MM's (pp. 163-174) solution to Kripke's paradox about rule following. I will review the solution and then discuss Boghossian's doubts. If my solution works, it is another illustration of where the new intensionalism helps with a major problem in philosophy. In Wzttgenstein on Rules and Private Language (Harvard University Press, 1982)' Kripke asks what fact about my finite mind grounds the claim that I mean 125 in literal uses of "sixty eight plus fifty seven". That fact must somehow justify my thinking that the referent of that expression is the third member of the triple (68'57,125) in the infinite set of triples of numbers which is the plus function and not the third member of a triple (68,57, N) ( N = 125) in a n infinite set of triples like the quus function. As Kripke puts it (1982, p. 54), "the idea in my mind is a finite object" so why "can it not be interpreted as the quus function, rather than the plus function?" Since a n idea in my mind can represent only a finite set of triples and since the finite set is not only a proper subset of the infinite plus function but also a proper subset of the infinite quus function (as well as other such functions), the idea cannot discrimi-
nate between either (or any) one of those functions. Kripke's thought is that the finitude of our ideas makes it necessary to project from the finitely many addition triples with which we are acquainted to the infinite plus function, and that, once projection enters the picture, we are stuck with the problem of how a finite sample can rule out deviant projections. Without a solution to the problem, my claim that my use of "sixty eight plus fifty seven" means 125 is no better justified than a claim that it means some other number. Kripke is right that my idea must be finite, but it does not have to be the idea of a set of triples of numbers. On the new intensionalism, my finite idea can be an idea of an intension of a numerical expression rather than an idea of an extension. As competent speakers of English, we know the sense of expressions of the form "x plus y", say, ( P ) , and we also know the sense of ( P ) The number that is reached from the number m by the process of taking its successor, then the successor of the successor, and so on, repeating this process n times in all. the numerical terms which can replace "x" and "y". Thus, the fact about my finite mind which grounds my meaning 125 in literal uses of "sixty eight plus fifty seven" is the fact that, in such uses, my communicative intention reflects my grammatical knowledge of the compositional sense formed from the sense (P) and the sense of those numerical terms. Such knowledge is finite, just as the sense of which it is knowledge is finite, since both the knowledge and the sense are combinations of more elementary units, mental entities in the former case, abstract senses in the latter. Neither my grammatical knowledge nor the sense of which it is knowledge functions as a basis for projection. In virtue of that finite knowledge, I grasp the finite sense of "sixty eight and fifty seven" which picks out the number 125. When I use "sixty eight plus fifty seven" to refer literally to the sum of sixty eight and fifty seven, I refer to 125 because, since it is the sum of sixty eight and fifty seven, it falls under the sense "sixty eight plus fifty seven". Even Kripke (1982, pp. 53-54) concedes that there is no special problem here: the plus function, not the quus function, falls under the sense of "plus" because "it simply is the nature of a sense to determine the referent it determines". Of course, on the new intensionalism, it is not in the nature of senses to do this, but it doesn't matter here whether 125 is the referent of "sixty eight plus fifty seven" because, as Fregeans would say, it is the nature of the sense of that expression to determine 125 as its referent or because, as I would say, the relevant contextual factors are in place and 125 falls under the sense of that expression.
O n the new intensionalism, the problem of how I can rule out deviant projections on an acquaintance with finitely many addition triples does not arise because projection never enters the picture. I use "sixty eight plus fifty seven" with the intention to refer to the referent of the sense of the expression. Both my grammatical knowledge of the sense and the sense are finite objects. Under the appropriate contextual conditions, I refer to the third member of the triple (68,57,125) in the plus function. This is not because I cleverly employ what I recall from my previous acquaintance with addition triples to rule out projections to the quus function and its ilk, but because of an abstract relation between two abstract objects, the sense of "sixty eight plus fifty seven" and the number 125. Boghossian says that he does not see how the new intensionalism solves Kripke's paradox: "the problem about grasping infinite extensions.. . simply resurfaces for senses" (p. 141). There is no such problem. On the new intensionalism, the problem is not "grasping" infinite extensions but fixing them, and that, I have argued, is achieved through grasping senses which are finite, even if their extensions are infinite. Boghossian believes "senses are in every respect themselves infinitary objects" -hence, "subject to the problem about rule following" (p. 141). But senses are not "infinitary objects" in the new intensionalism. On a quite natural construal of Fregean semantics, the only way an infinite set of triples of numbers can be the referent of a sense is for the sense to be a mode of presentation of the infinite sequence which presents it, as it were, triple for triple. If Boghossian were thinking of such a position, infinite extensions would be an unresolvable problem. But, in the new intensionalism, compositional senses are not "infinitary objects". They are finite objects because they are finite combinations of finitely many finite component senses. Our grasp of senses, based on structural representations of such finite objects, can be finite, too, because what they represent is finite. No problem about grasping infinite extensions "resurfaces for senses" because grasping a sense is not fixing an extension and senses are not infinite.