Pragmatism and the Philosophy of Language

Danielle Macbeth
Haverford College

This article was originally published in Philosophy and Phenomenological Research, Vol. LV, No. 3, September 1995: 501-523, and appears here with the permission of Prof. Macbeth and the editor of PPR, Prof. Ernest Sosa. This article is copyrighted by Philosophy and Phenomenological Research, and permission has been granted only for it to appear at the Sellars site, and not for it to be used at another site or to be copied multiply.

1. In "Empiricism and the Philosophy of Mind",{1} Sellars argues that the notion of "self-authenticating nonverbal episodes" that would provide a foundation for empirical knowledge is a myth; nothing merely causal, not already in conceptual shape, could possibly play the justificatory role required of such a foundation. Rorty takes Quine, in "Two Dogmas",{2} to make the complementary point that the notion of analytic claims true by virtue of meaning, of self-authenticating verbal episodes that might provide a foundation for another sort for knowledge, is again a myth.{3} A third moment in the dismantling of the myth of a foundation--this time for the contentfulness of our thoughts rather than for the truth of our beliefs -- is due to Rorty himself. As he argues in "Realism and Reference"{4} (and again in Chapter 6 of Philosophy and the Mirror of Nature), the notion of reference as a nonintentional or "external" word-world relation that would ground our thoughts' representational bearing on things, and so explain how thoughts can so much as purport to be true, again involves illicit appeal to the idea that independent of what we take it as an object can have cognitive significance. An external relation of reference cannot serve as the unmoved mover of the contentfulness or aboutness of thought. Nor, if Quine is right about the breakdown of the analytic/synthetic distinction, can meanings or word-word relations play this role.

My aim here is not, however, to defend these arguments -- though I will try to clarify the pragmatist's motivation for claiming that neither the concepts expressed by natural kind terms nor the objects referred to by proper names are the given building blocks of contentful thoughts. My aim is rather to give a brief overview of this important strand of contemporary philosophy of language and to situate within it a conception of meaning and reference inspired by some themes in Kripke's work on reference and by Sellars' understanding of concepts and causal laws. While respecting the pragmatist's rejection of relations of reference and meaning, the account to be sketched seeks to avoid the puzzling pragmatist conclusion that meaning is a philosopher's fiction and reference a purely technical notion without application outside of formal semantics. It is intended as a first step on the way to an adequate understanding of the role of language in our epistemic commerce with the world around us.

2. It is natural to think of proper names as something like labels for objects rather than as descriptions true of objects. Names, it seems, represent objects directly, not (or at least not essentially) by way of features that uniquely characterize those objects. But if names are understood this way, if their function in language is to locate or to pick out individuals independently of and antecedently to the properties we ascribe to or deny of such individuals, then in the case in which a name has no bearer and so locates no individual, no property can be ascribed or denied by an utterance involving the name. If a name is a label, a direct representation of an object, then a sentence containing a bearerless name can have no truth-value. According to a familiar argument due to Russell, this consequence of the naive view is intolerable. The argument, in brief, is this.{5} (1) If a name is a label rather than a predicative expression then a sentence containing a name that has no bearer does not express a proposition. (2) Such a sentence does express a proposition. Therefore, (3) a name must function predicatively, as a disguised description. Unfortunately, while the argument is clearly valid, its conclusion is not very plausible. For, as Kripke has reminded us,{6} names, unlike descriptive or predicative expressions, do not have (or at least need not have) any inferential consequences or antecedents. Nothing at all need entail or be entailed by the fact that an individual has a certain name; the fact that some names do have such inferential connections to other expressions in the language is nothing more than an accidental feature of our practices. (A boy can be named Sue, and M Street in Washington could have been named anything you like.) Indeed, as both Kripke (in Naming and Necessity) and Donnellan{7} have argued, not only do names not function predicatively, they are not even essentially associated with privileged reference-fixing descriptions. The descriptions (uniquely identifying or not) that we believe to be satisfied by the objects we refer to are neither necessary nor sufficient to account for our use of names.

But if Russell's conclusion is false, that must be because at least one of his premises is false. McDowell has argued that it is his second premise, the idea that a sentence containing a denotationless name expresses a proposition.{8} Instead, McDowell argues, when one purports to think a thought that would be expressed by a sentence containing a denotationless name, one has no thought "precisely there"; for the thought that would be expressed by the sentence is not, in the absence of the relevant object, available to be entertained. In such a case, McDowell claims, one is merely under the illusion of entertaining a thought. But this conclusion too is quite implausible. We appeal to the thought of a subject to explain the fact that sometimes things seem to one to be a certain way when they are not. In this case, it seems there is nothing at all to account for it so much as seeming to a thinker that she entertains a thought.

The only alternative, however, is to reject Russell's first premise, the idea that a sentence containing a bearerless name can express no proposition. Such a sentence cannot have a truth-value, at least it cannot if the contrast with predicative expressions is to be maintained, but (contrary to what is standardly assumed) perhaps such a sentence does have a determinate meaning, that is, a cognitive value or Fregean Sinn -- as Frege himself seems to have thought.{9} If this idea can be made to work, then it follows that while a name has or need have no descriptive or predicative content associated with it, it is nevertheless not merely a label.

3. Much as we find it natural to think of names as something like labels whose role in language is to "point" to objects, so it seems plausible to think of predicates as general representations true of (or satisfied by) objects, but which are nevertheless meaningful independent of the objects, if any, of which they are true. A sentence containing a (general) term like 'phlogiston' or 'witch', for example, expresses a fully meaningful proposition in spite of the fact that there are not and could not be any such entities. That is, if our science is correct, it is not merely a cosmic coincidence, an accidental feature of the world, that phlogiston and witches are nowhere and nowhen to be found -- though it is presumably an accidental feature of the world that there are (say) no white crows or green cats. Assuming that our science is correct, the sort of thing that a witch (or phlogiston) is could not possibly occur. Hence, a predicate like 'is a witch' is not merely false for all applications to actual individuals, it would seem to be without application at all. There would seem to be no facts at all involving the notion of a witch, no facts either of the form 'x is a witch' or of the form 'x is not a witch' because nothing could possibly be like that. What this suggests is that predicates acquire their meanings not by virtue of what they are true of -- since some perfectly meaningful predicates are and could be true of nothing at all, but rather by virtue of the theories, broadly speaking, in which they occur.

The idea that the meaning of a predicate is articulated by its place in a theory is not without its problems, however. For while the form of a predicate can be given by its role in a theory -- ideally in the complete axiomatization of that theory -- a predicate so defined does not thereby have any empirical significance. The correctness of the application of a predicate depends on whether the relevant object has the property in question and requires that conditions of satisfaction be associated with the predicate. But the theoretical characterization of a predicate cannot determine its conditions of satisfaction any more than the inferential relations between colour terms, for example, which articulate the "logic" of colour terms, can fix which colour is correctly described as, say, red. Inferential relations, however extensive, at best provide only formal constraints on the use of predicative expressions. Like the axioms of a theory, they seem to require interpretation, application to actually experienced items, if they are to be empirically significant or correctly ascribable to some but not all objects.

The problem of the relationship between the meaning of a predicate and its correct application is especially severe for natural kind terms. Natural kind terms contrast with nonnatural (or nominal) kind terms by being doubly answerable to how things are. Not only are particular applications of these predicates correct or incorrect depending on how things are, but the (apparently) constitutive relations that they bear to other predicates in a theory are also revisable in light of our on-going experience. For example, while it is an empirical question whether a particular person is a bachelor, it is not an empirical question what it is to be a bachelor -- at least not if 'bachelor' is a nominal kind term. We can decide (for whatever reasons) to stop using the word 'bachelor' to mean 'unmarried male', and our use of the expression can shift without our being explicitly aware that it is doing so, but we cannot discover through an empirical investigation of bachelors that contrary to what we had supposed not all bachelors are unmarried males. In the case of a natural kind term, on the other hand, both the question of which things are of that kind and the question of what it is to be that kind of thing (and so of the empirical adequacy of our theory) are questions of empirical fact. According to our best theory, water is H2O, where what it is to be H2O is determined by the theory codified (in outline) in the periodic table of the elements. But we could be wrong about that. It cannot be the case, then, that all that is relevant to the meanings of predicative expressions like 'is water' or 'is H2O' is given by their roles in the relevant theories.

These considerations suggest that theoretical relations among predicates are not sufficient to determine the empirical significance or applicability of natural kind terms. Familiar arguments due to Kripke and to Putnam{10} suggest that they are also not necessary -- that (as Kripke remarks in the preface to Naming and Necessity) once one recognizes that the description theory is inadequate as an account of how the reference of a name is fixed it is "a short step to realize that similar remarks appl[y] to terms for natural kinds" (p. 5). The idea, in brief, is that much as our beliefs about a particular individual can undergo radical revisions while yet being beliefs about that very individual, so our beliefs about the nature of a particular natural kind (or about its ordinary observable properties) can be radically revised while remaining beliefs about that kind of thing. If, for example, it is possible to have true beliefs of the form 'this stuff is water' while having false beliefs about what it is to be water (perhaps one believes with Aristotle that it is one of the four elements, namely, the cold and the wet, and that it has its natural place between earth and air) then, it seems, one's beliefs about what it is to be of that kind can be neither constitutive of the meaning of the term nor what fixes its extension. What we take to be constitutive of the stuff water is quite different from what Aristotle took to be constitutive of it, and yet we seem clearly to have the same stuff "in mind". Since both we and Aristotle may be wrong about what it is to be water (or about the observable properties of water), no description is available to fix the kind. The meanings, and so the extensions, of natural kind terms do not seem to be constituted (either by our everyday beliefs or) by our theories about them.

4. If the arguments sketched above are sound, neither proper names nor natural kind terms function as predicative expressions whose meanings are fixed independent of what (if anything) satisfies them. It is tempting to conclude that names and natural kind terms must instead be constitutively related to particular objects and kinds of things, respectively, independent of any beliefs we may have about them. Rorty (following Quine) argues that this temptation ought to be resisted. The moral of the cautionary tales told by Kripke, Donnellan, and Putnam is not, on his view, that the reference of names and the extensions of natural kind terms are fixed by some word-world relation rather than by word-word relations -- either directly by a description or indirectly by their roles within theories. It is rather that reference and meaning are not fixed. Period.

Both externalists like Kripke, Donnellan and Putnam, and internalists like Russell, Blackburn and Searle,{11} take the various by now familiar examples in the literature to provide (or to purport to provide) evidence for or against one or other of the two opposing positions. They are taken to support either the claim that reference is fixed by an external, historical or causal chain, or the claim that it is fixed by some sort of description that the speaker has in mind. Rorty argues that these examples do not constitute evidence for either theory since they can be treated simply as cases where one set of considerations outweighs (on pragmatic grounds) another set. In the ordinary case of one's use of a name, for example, one has a variety of beliefs about the individual named and has, as well, acquired the name in a way that brings in contextual factors (i.e., one does not privately dub an individual identified by a Russellian definite description but rather acquires the name in a public context). And, in the ordinary case, the object that would be identified by appeal to one's beliefs (i.e., "internally") is just the one that would be identified by appeal to "external" factors, to factors involved in the circumstances of one's acquisition of the name. But in a case in which a subject is seriously confused or mistaken, the two methods can yield different answers to the question of what is referred to by a name used by that speaker. The speaker may herself resolve the conflict by stipulating, say, that she means by the name whatever is meant by it by those from whom she learned it, or that she means the one who was the cause of a certain perceptual experience she had on first becoming acquainted with the referent (i.e., in general, that her reference-fixing description is an "externalist" one{12}); but there is no guarantee of this and hence no guarantee that the answers arrived at by the two strategies will converge.{13}

The two strategies are not, however, merely alternative and sometimes competing strategies for answering the question 'who/what is being referred to?'. Usually, if one wants to know what another is talking about, one consults them. Only if one has reason to believe that the speaker is confused or seriously mistaken about the identity or nature of what is talked about must an independent or "externalist" account be given of the reference or extension. If, to borrow an example from Kripke, I believe that you have mostly false beliefs about Jonah then there is no point in asking you whom you refer to by the name 'Jonah'--not because there is a philosophically interesting sense in which you yourself do not know what you are talking about, but because my biographical purposes are better served by pursuing an externalist inquiry. Moreover, if as a result of such an inquiry I judge that you are really talking about a Hebrew prophet who is referred to in II Kings, it does not follow that I have somehow managed to discern a nonintentional or external semantic relation of reference between the word 'Jonah' (in you use of it) and a person. Where the surface grammar of the locution 'what you are really talking about' suggests such a relation the pragmatist, in a satisfyingly deflationary way, finds only a heuristic notion: "'[r]eally' here is just a matter of 'placing' the relative ignorance of the person being discussed in the context of the relatively greater knowledge claimed by the speaker."{14} We have no need of the idea of a properly externalist semantic relation between a person's words and what they refer to or mean, and no place for a practically useful distinction between what according to our best accounts there is and what there really is. Our everyday discourse, both externalist and internalist, about what others (or ourselves, since one can pursue an externalist inquiry on one's own behalf as well as on another's) are talking about does not justify either an externalist or an internalist (or a mixed) theory of the reference of names and/or the meanings of natural kind terms.{15}

Rorty himself endorses a descriptivist (or internalist) view both of proper names and of natural kind terms as the more felicitous account of our everyday notion of 'what someone is talking about'. But he does not thereby endorse the idea that when correctly analyzed such terms will be shown to function as predicative expressions whose meanings are fixed by word-word relations. Such a claim belongs to what Rorty thinks of as "impure" philosophy of language -- a domain of inquiry which, he claims, hopelessly aims to provide an ahistorical foundation for the theory of knowledge.{16} When he writes that "[f]or the realist, there were determinate moments in the past at which language threw out a fresh hook and succeeded in grappling the world at yet another point,"{17} the remark would seem to apply equally to the "realist" who thinks that the reference of a name or the extension of a natural kind term is fixed by an external relation of reference, and to the "descriptivist" for whom reference and extensions are fixed by descriptions "in the head" which individuals and kinds satisfy. The "grappling hook" which connects word to world is differently conceived in the two cases -- according either to the model of referring expressions as items that function as labels for objects, or to the model of predicative expressions as items that by virtue of their meanings are satisfied by or are true of objects -- but it is a grappling hook nonetheless, one that is determinately to fix what one talks about or says.

The idea that reference must somehow be fixed, either by a relation of reference or by the satisfaction of a predicate, is perhaps most compelling in the case of an initial baptism, whether it be of an object or a kind. But Rorty's point can be applied even to this case. Kripke claims that the reference of a name is identified at an initial baptism either descriptively or ostensively. Pure ostension cannot, however, fix the reference since mere pointing cannot succeed in discriminating what is to be the bearer of the name.{18} But neither can a description, or ostension plus a sortal, do the job -- for exactly the reasons that lead Kripke to deny that names have any descriptive content. Any description or sortal that one uses at a baptism to fix the reference can turn out to be false of the individual named.

Suppose, to borrow another example from Kripke, we discovered one day that the individual we call by the name of 'Nixon' is in fact a cleverly bioengineered robot, a creature that develops and behaves just like a human being but is controlled by radio signals from the headquarters of a nefarious organization. Would we, in such a case, have discovered that Nixon is after all a robot? Or would we have discovered that Nixon never existed? Had we been tricked into thinking that Nixon was a person, or was the illusion rather that there was a person, Nixon, at all? If it were the case either that some particular sortal were essentially associated with the name (either as its descriptive content or as what helps to fix its reference) or that the name just did either attach or fail to attach to some particular, then there would be a fact of the matter as to what we ought to say. But there is no such fact. One could argue (with the internalist) that we ought to treat our discovery as the discovery that the name is empty on the grounds that the baptism (and more generally the use of the name) was intended to be of a person. But an equally compelling (externalist) case could be made for the claim that the creature is Nixon, whatever we discover about it. Sometimes discovering that the individual is not the kind of thing we took it to be will seem a compelling reason to give up the name as bearerless; and sometimes the same discovery can be treated as a discovery about the named individual. And in some cases we will not know what to say, not because we do not have enough evidence but because an equally plausible story could be told either way.{19}

5. According to the pragmatist's historicist understanding of the relationship between referring and describing, on the one hand, and matters of meaning and matters of fact, on the other, neither relations of reference between words and objects nor relations of satisfaction between descriptions and objects, neither theoretical relations between terms nor matters of empirical fact, are intrinsically and nondefeasibly authoritative. None are given as the firm foundation of contentful thought. As I want now to indicate, even if this is right it does not follow that no useful distinction can be drawn between referring and describing, or between matters of meaning and matters of fact.

In Naming and Necessity, Kripke not only offers counterexamples to the claim that the reference of names is fixed by descriptions but also presents a modal argument based on the fact that names function differently in counterfactual discourse than descriptions do. As he points out, philosophers have tended to think of possible worlds purely qualitatively, as if they were distant countries whose inhabitants can be described but not referred to. But, he reminds us, in everyday discourse about counterfactual situations we often stipulate that the possibility being considered is that of some particular individual. This fact suggests a quite different picture of possibility:

we do not begin with worlds (which are supposed somehow to be real and whose qualities, but not whose objects, are perceptible to us), and then ask about criteria of transworld identification; on the contrary, we begin with the objects, which we have, and can identify, in the actual world. We can then ask whether certain things might have been true of the objects.{20}

Kripke himself occasionally seems to suggest that the two pictures are mutually exclusive alternatives. I suggest that we take them to indicate differences between names and predicates. The "distant countries" conception is inadequate to model the way names function in counterfactual discourse, but that is not a reason to reject it as a model of how predicates function in such discourse. Some counterfactuals are constructed purely qualitatively. In such a case, one combines predicates (for instance using truth-functional operators) to form a complex predicate and then imagines, or tries to imagine, a situation in which exactly one, or only some, or all individuals satisfy the predicate one has constructed. When one restricts one's view to the predicates of a language, it is quite plausible to say that one begins with worlds, that is, with the total array of possible states of affairs that can be described using the n-ary predicates of the language plus the usual logical resources. From this perspective, to discover truths about the actual world is to discover which of these possible states of affairs are states of affairs that are actually instantiated by existing individuals. What Kripke reminds us is that this is not the only perspective we can take. Given that the language also includes names for various actually existing individuals, we can also begin with such individuals and discover facts and counterfactuals about them.

On this reading of Kripke's discussion of counterfactual discourse, the point is that the (logical) relation of what is actual to what is possible is fundamentally different for the two sorts of expressions. In the case of predicative expressions, the actual world stands to other possible worlds much as one country, the one we are in, stands to the many others we could have inhabited but do not. In this case, the array of possible worlds is conceptually prior to the actual world. Were it otherwise, the idea of a property that as it happens is nowhere (nowhen) instantiated in the actual world and yet is a property an individual could have would be unintelligible. The distinction between those complexes of properties that are not and could not be jointly realized, on the one hand, and those that are not but could be jointly realized, on the other, requires that possibility precedes actuality in the case of predicates. But in the case of referring expressions priority falls to the actual world. Names that have no bearers in the actual world cannot function, even in counterfactual discourse, as names; for where the relevant object does not exist, there are no states of affairs in which it figures, nothing true or false to be said of it. The relevant object can be described, that is, one can specify a constellation of properties that could be jointly satisfied, and in some possible worlds are jointly satisfied, by just one individual. But they do not need to be satisfied by the same individual in each such world and because they do not, the idea that one can refer to the imagined object is senseless.{21} Referring to such an individual would require being in a world in which it exists -- and you just can't get there from here.{22}

In Naming and Necessity, Kripke summarizes his point about the way names function in counterfactual discourse in terms of a distinction between "rigid" designators and "nonrigid" designators. As he points out, two sentences, one involving a name and the other involving instead a Russellian definite description, which have the same truth conditions in the actual world, can have different truth conditions in counterfactual situations. In the case of a sentence containing a name, "there is a single individual and a single property such that, with respect to every counterfactual situation, the truth conditions of the proposition are the possession of the property by that individual, in that situation" (p. 10). In the case of a sentence that instead contains a definite description, the truth conditions are rather the possession of the property predicated by whatever (in that situation) satisfies the description in the subject position. But, by distinguishing names as rigid designators Kripke misleadingly suggests that they are merely one kind of designator -- as it were those descriptions that are necessarily true of the object rather than contingently true of it.{23} But this is not Kripke's view. His point is that names fundamentally contrast with descriptions, not that they are a species of description.

I have suggested that the distant countries conception of possibility is inappropriate for the case of referring expressions because names function differently in language than do predicative expressions. But we also assume that the pragmatist is right to reject the idea that reference is fixed (either internally or externally). What is wanted is an account that respects the distinction between reference to something and assertion about it without appeal to a relation of reference. I suggest the following.

Reference, like judgment or assertion--i.e., the endorsement of the application of a predicate, is both fallible and defeasible. Because it is, it should be thought of as some kind of commitment or belief. But the commitment (or belief) expressed in one's use of a name cannot have just the same form as the commitment that is expressed in a judgment (what I will call a propositional commitment); for, as Kripke's discussion of counterfactual discourse indicates, using a name is not like endorsing the application of a predicate. We cannot understand our use of names on the model of Russell's Theory of Descriptions. Rather than being a commitment to the truth of a claim (as Russell's theory would have it), referring seems to involve a commitment to an object, to its existence certainly but even more to its being the locus of the truth or falsity of the judgments (i.e., propositional commitments) one makes about it.

When one refers using a name and then asserts something of the object named (ascribes a predicate to it), it is that object that determines the truth or falsity of what one asserts. Moreover, if one fails to refer, if one's referential commitment is not after all entitled, then (on this view) one cannot succeed in asserting either. If nothing is picked out as the locus of the truth or falsity of what one says, one's words get no purchase on the world. There is in such a case no fact of the matter whether or not what one says is true because nothing has been identified as that relative to which the truth of one's claim might be assessed. But it does not follow that no proposition (or perhaps better, no Fregean Thought) is expressed by the sentence one utters. Since one's use of the name conveys a contentful commitment (to there being the relevant object to serve as the locus of the truth of one's claims), and the predicate is independently meaningful, we can understand what would have to be the case for the utterance to succeed in being an assertion, either true or false. But since one's referential commitment is not entitled, one cannot succeed even in making a propositional commitment -- though to a deluded user of the name it will seem that one is made.

The difference between referential and propositional commitments can be modeled on the difference between opening a bank account (with an initial deposit) and subsequent deposits to and withdrawals from that account. That is, we can think of acquiring the use of a name as opening an account in that name in one's mental economy. To open an account involves a judgment (an initial deposit), but it does not itself have the form of a judgment -- any more than opening a bank account has the form of a deposit. Each subsequent entry into the account, then, involves two commitments, one to the (existence and) identity of the object one now has in mind and the object in whose name the account was opened, and another to that object's having some particular property. Such entries have the form of singular object-dependent propositions. They are, or at least purport to be, beliefs de re about some particular individual, and contrast with one's general beliefs which are entered into one's cognitive economy but not into any account. But one can fail to be entitled to the commitments one makes in entering a judgment: perhaps the object does not have the property ascribed, perhaps the one now judged about only seems to be identical to the one originally referred to, and perhaps there is no such individual at all. Similarly, one can open two accounts for what is, unbeknownst to one, a single individual, and so can unwittingly hold contradictory views about one and the same individual -- as happens to Kripke's Pierre.{24}

Rorty argues that the distinction between what one talks about (either an individual or a natural kind) and what one says about it has been mistaken for a distinction between two competing semantic theories of names and predicates. I have suggested that it does not follow that there is no philosophically significant distinction to be drawn between referring to an individual using a name and describing or making a claim about some individual to which one refers. Rorty argues that since there is no semantic relation of reference (save for the purely technical notion used in a "pure" philosophy of language), there is no notion of reference that would clarify our epistemic commerce with the things around us. I have suggested that we can accept that a nonintentional external semantic relation of reference is of no use in "impure" philosophy of language, but can nevertheless admit the notion of reference. On the view pursued here, reference is not a non-intentional relation between a word and an object. Rather it has the form of a commitment to the expression having a referent.

6. In "Two Dogmas", Quine does not argue that the distinction between matters of meaning and matters of fact has been misunderstood. Rather he argues that there is and can be no such distinction at all. He aims to show that while the truth of any statement depends both upon the meanings of the words used and also on extra-linguistic fact, meaning cannot make an isolable contribution to the truth of a statement. If he is right then there is nothing to be salvaged of the idea that the meaning of a predicative expression is independent of what -- if anything -- it is correctly applied to. The argument Quine gives, however, rests on the particular way he understands the claim that the unit of empirical significance is not the statement but rather a whole theory (and in the limit the whole language). Quine rejects the verificationist thesis that the meaning of a sentence "is the method of empirically confirming or infirming it" (p. 37). But the problem (he thinks) is not with the general idea of verificationism; it is only with the unit to which such significance is accorded. On Quine's view, it is not individual statements but rather whole theories that are meaningful by virtue of confirmatory and infirmatory relations to extra-theoretical facts. Since terms in the language of the theory have meaning relative to the theory as a whole by virtue of states of affairs that would infirm or confirm the correctness of their application, and since it is not determined in advance just what a given state of affairs confirms or infirms (since "[a]ny statement can be held true come what may" and none are "immune to revision", p. 43), no sense can be made of the idea that there is an isolable linguistic component to a sentence. Quine's argument (at least on this reading of it) depends, in other words, not on verificationism about the analytic/synthetic distinction as such, but rather on a holistic version of verificationism as a theory of meaning generally. He clearly does endorse such a theory of meaning and on that basis correctly argues that the analytic/synthetic distinction must collapse: if the verificationist theory is true theory by theory (though not statement by statement) then there can be no distinction between statements true by virtue of meaning and statements true by virtue of (meaning and) fact. Given this argument, the fact that there are differences in our linguistic behavior that suggest such a distinction (e.g., the fact that "analytic" claims are taken to support counterfactuals but "synthetic" ones are not) is powerless to save it. Such behavioral differences can be no more than the expression of our relative unwillingness to give up certain claims.

Quine's modified verificationism is, however, untenable. For, as Rorty points out,{25} it falls prey to the myth of the given that Sellars exposes in "Empiricism and the Philosophy of Mind". Nothing merely causal, that is not already in conceptual shape -- nothing "extra-theoretical" -- can bear rational relations (e.g., of confirmation or infirmation) to claims. Perhaps even more obviously nothing that is not already in conceptual shape can play a role in the constitution of meaning: if it is not already meaningful then it cannot offer anything useful, but if it is already meaningful then the meanings whose constitution we were to explain have been illicitly appealed to. But Rorty does not seem to see that once this last vestige of the empiricist conception of concepts is given up it is Quine's argument against the analytic/synthetic distinction that collapses. If the interpretation of Quine's argument given here is correct then Sellars and Quine cannot be taken to make complementary points about the lack of a given foundation for knowledge.{26} Rather we should understand Sellars as having helped us to see exactly what is of value in Quine's argument, namely, that even analytic claims, supposedly "true by virtue of meaning" can be revised in light of one's on-going experience. Independent of appeal to the given, the fact that no statement is immune to revision does not entail either that meaning makes no isolable contribution to a statement or that there are no statements that express only constitutive relations between concepts.{27}

7. Even if one is willing to concede, in a general way at least, that the analytic/synthetic distinction is unproblematic as a distinction between (defeasible) matters of meaning and (defeasible) matters of fact, a question still remains about the nature of the meaning of predicates and in particular about the meaning of natural kind terms. It has already been pointed out that the meaning of a predicative expression cannot be accounted for solely by appeal to its role in a theory since a theoretical role cannot determine that to which the predicate correctly applies. Moreover, as Kripke and Putnam argue, the theoretical account of what it is to be a kind of thing is also not necessary to fix the extension of a natural kind term. Rather, they suggest, natural kind terms function referentially; their meanings and/or extensions are determined by actual instances of the kinds in question. Since, as I have just argued, Quine's argument against the distinction of fact and meaning is based on a mistaken, verificationist view of meaning, and since Rorty can show only that questions about what we are referring to are never definitively settled, it would seem that Kripke and Putnam could argue that the meanings of natural kind terms involve the relevant kinds. While our beliefs about which stuffs are, say, water are defeasible just as our beliefs about the nature of water are, this by itself does not show that the meaning of the natural kind term 'water' does not somehow involve the relevant stuff (whatever its nature) that is to be found in the actual physical environment. All that must be conceded is that not only do our beliefs about the nature of the stuff or kind change, but so also do the meanings we associate with the terms -- as happens (on the Kripke-Putnam view) when we take a different class of stuff than we had previously to provide the paradigm cases. An independent argument is needed to show that the very idea of an property- or stuff- or species-involving predicate is incoherent.

I argued above that the reference of a name cannot be fixed at an initial baptism either by ostension or by ostension plus a sortal or by description. Ostension fails to distinguish a kind of individual, and any judgment concerning the kind of individual one has in mind can turn out to be false of the individual named. Yet the reference of a proper name is fixed, if only defeasibly, in actual cases. It might seem that however the trick is done, the same story can be told for predicative expressions. We know that the norms governing the correct use of names must involve the notion of the identity of the individual as the same K over time, so why not say (as Putnam explicitly does say{28}) that the norms governing the correct use of a natural kind term involve the idea of the identity of a kind as the same K* (a higher-order sortal) over cases of it? Why can we not give an "ostensive definition" of, say, water as whatever stuff bears the relation 'same-liquid-as' to the liquid one points at?

We cannot because the definition starts a vicious regress. Given an independent account of the meaningfulness of predicates, the idea that one might identify an individual contextually (and defeasibly) as 'that (pointing) K' is perfectly intelligible. But without an independent account of the meaningfulness of, say, 'is a liquid', we cannot suppose that the meaning of a predicate like 'is water' might be contextually determined, even defeasibly, by an ostensive definition like 'that liquid'. Each ostensive definition of a natural kind term would require appeal to another natural kind to fix the relevant respect of sameness across instances. So without an independent notion of the meanings of at least some natural kind terms, the process of ostensively defining kinds and stuffs cannot get started.

The idea that it is possible ostensively to define a natural kind term requires the myth of the given, the idea that an object can as a matter of fact, independent of what we conceive it as, fix a principle of inclusion for things of that kind. The "short step" that Kripke and many others following him have taken from the idea that names function referentially to the idea that predicates--or at least those that are natural kind terms -- do so as well, is a misstep. If an independent account of the meaningfulness of predicates can be given, then we have the basis for (defeasible) contextual identifications of individuals as particular K's and so also the basis for an account of how reference is (defeasibly) fixed. But if predicates too are taken to have a referential component then the account collapses. Names can be object-dependent only if predicates are not. We need an account of the meaning of a predicate according to which it is articulated solely by word-word relations.

8. The idea that predicative expressions have both logical form determined by rules and empirical content derived somehow from experience is due ultimately to Kant. On that view of concepts, the rules governing the use of an expression provide it with a meaning that is wholly formal. As Kant says, a concept considered in abstraction from the objects to which it is correctly applied represents no object.{29} So it is often thought that while the form of a predicate may be given by (say) its inferential relations to other predicates -- ideally in the complete axiomatization of a theory, still such a set of rules must be interpreted if the predicate is to have any empirical significance. The primitive n-ary predicates of the theory, on this view, must be correlated, either directly or indirectly, with actually experienced properties and relations. Absent such a correlation, an interpretation or application, the axioms or rules are merely well-formed strings of symbols whose various logical relations can be investigated through the derivation of theorems but which are neither true nor false.

It seems clearly correct to say that the rules governing admissible inferences from sentences containing particular predicates to sentences containing other predicates do not by themselves impart empirical significance to predicates. To take symbols arbitrarily assigned syntactic form as n-ary predicates and to combine them into a consistent set of well-formed formulae that define their inferential relations is not to confer on them empirical significance. But in light of Sellars' argument in "Empiricism and the Philosophy of Mind", the prospects of doing so by appeal to entities in the world independent of what we take them as are slight--to say the least. Entities in the world independent of what we take them as cannot provide principles of application for the predicates we use. Once again, I suggest, we need to appeal not to external world-word relationships but instead to the notion of a commitment, though (again) one that is different from that of a propositional commitment or belief about how things are. For, as will be argued, we can understand the empirical significance or applicability of a natural kind term to derive from the fact that a subject endorses the inferential relations that articulate the ("formal") meanings of the family of predicates to which it belongs as the inferential rules governing the correct use of those predicates.{30} This cannot be the whole story since a question still remains as to what in the world the predicates apply to (given that no amount of inferential articulation determines that a particular predicate applies to this object rather than that), but before we can see how the story ends, we need to be clearer about the difference between what I will call conceptual commitments -- that is, a commitment to inferential relations as inferential rules, to a theory as empirically adequate -- and propositional commitments, judgments or claims about how things actually are.

I remarked above that although Kripke sometimes puts his point about reference in terms of the notion of a rigid designator, his aim is to contrast reference with description. If we want to understand reference we should not think of it merely as a mechanism that somehow keeps designation fixed across all possible worlds as if it were just like description only rigid, that is, necessarily true of just the individual it is true of in the actual world. Sellars makes an analogous point about constitutive inferential relations in Section IV of "Counterfactuals, Dispositions, and the Causal Modalities".{31} He suggests that if we want to understand meaning we should not think of constitutive inferential relations as differing from empirical generalizations by virtue of being necessary, by virtue of being generalizations that are true both in the actual world and in all other possible worlds. Such lawlike statements, in other words, are no more a special case of universally quantified statements generally than "rigid designators" are a special case of designators generally. Empirical laws are not descriptive claims about all possible worlds formally on a par with descriptive claims about all crows. Rather, Sellars suggests, they have the form of inference licenses which permit or enjoin commitment to particular descriptive claims given that one is committed to others. They have the form of rules for playing the game of giving and asking for reasons rather than the form of moves in the game.

From this Sellarsian perspective, Kant's idea that concepts involve laws governing the form of all possible experience and so are prior to and independent of actual experience is a fundamental insight. The only problem with it is the idea that concepts only involve laws -- that they have the form of laws but derive their contents from experience. For, as Sellars sees,

[t]he familiar notion . . . that the form of a concept is determined by 'logical rules', while the content is 'derived from experience' embodies a radical misinterpretation of the manner in which the 'manifold of sense' contributes to the shaping of the conceptual apparatus 'applied' to the manifold in the process of cognition.{32}

Rather, as he goes on,

[t]here is nothing to a conceptual apparatus that isn't determined by its rules, and there is no such thing as choosing these rules to conform with antecedently apprehended universals and connexions, for the "apprehension of universals and connexions" is already the use of a conceptual frame, and as such presupposes the rules in question.{33}

Sellars himself includes language-entry rules and language-exit rules among the rules articulating concepts, but their status is peculiar given that such rules cannot be understood to involve the notion of conformity "with antecedently apprehended universals and connexions".{34} On the account pursued here such rules are not necessary (and, I would argue, are useless for the purpose they are to serve of fixing the empirical significance of concepts) -- but to see that they are not requires being clearer on the nature and role of reference than Sellars perhaps was.

Following Sellars, I have suggested that we take concepts to be articulated by the inferential relations given in the theories that are constitutive of the family of predicates that express them. To endorse such predicates as having empirical significance, as terms for properties and relations that are actually (or at least can be) instantiated by existing individuals, is to commit oneself to (and so to become responsible for acting in accordance with) these inferential relations as the rules governing one's reasoning. At first, of course, such commitments are not explicitly made. Rather they are acquired in the course of one's entry into the language and are expressed in one's inferential practices. So long as they remain implicit in one's inferential practices and are not made explicit in claims, they can change or evolve but do not do so through any process of rational reflective critique. But once they are made explicit in the form of claims (expressible as counterfactual-supporting laws) they can, like any claims, be defended and used in defense, and can be explicitly endorsed or rejected. As Sellars says,

the presence in the object language of the causal modalities . . . serves not only to express existing commitments, but also to provide the framework for the thinking by which we reason our way . . . into the making of new commitments and the abandonment of old.{35}

Once having acquired the expressive resources necessary for the explicitation of the inferential rules implicit in one's practice, one can modify these rules and formulate radically new ones, and so also new concepts, in accordance with the rational process of inquiry characteristic of empirical science.{36}

9. If reference and meaning are not simply given but rather essentially involve commitments regarding the use of particular referential and predicative expressions -- commitments to semantic rules rather than to claims about how things are -- then they are in this sense "a priori": only one with a conceptual repertoire in use and referential commitments to individuals as loci of truth can find things to be this way or that. On this view, reference and inference underlie our capacity to judge of things that they are thus and so. But that is not to say that they are intelligible independent of our finding things to be thus and so. Because they are at once semantic and commitments, they come to nothing in the absence of propositional commitments much as one's commitment to one set of rules for a game rather than another comes to nothing if one does not play. This fact, I want now to indicate, provides the basis for an account both of how reference is (defeasibly) fixed and also of the empirical applicability of natural kind terms.

According to the way of thinking pursued here, our most basic judgments have the form of singular propositions; they are judgments of individuals referred to that they are thus and so. (One can then generalize on the basis of such claims.) These judgments are judgments rather than mere noises we make or conditioned responses to stimuli by virtue of the fact that the speaker has conceptual and referential commitments, and so also the capacity to reason--to play by the inferential rules (whatever they are) -- and at least in the simplest case, to recognize or reidentify the object referred to as the same (K) again. But, again, one need not at first be explicitly aware of the rules governing one's reasoning. Nor need one know just how one recognizes the object or even what kind of thing one takes it to be. All this will be expressed in a speaker's linguistic behavior but the speaker need have no consciousness of it. What is more important, because one's (referential) recognition of the object as the same (K) again is distinct from one's (propositional) recognition of it as having a property F, one can find oneself judging that an object o previously seen as having the property F now is seen as having the property G -- even if the conceptual rules by which one plays do not allow anything that is F to be or to become G. Were it the case either that our conceptual commitments were not complemented by independent referential commitments (if, say, we took referential commitments to have the form of propositional commitments), or that referential commitments were not complemented by conceptual commitments (all laws being conceived as empirical generalizations), this could not occur. In the former case, since (by one's own lights) nothing can be F then G, one would not find the one that was F now to be G but would rather see them as different objects. Any qualitative similarities between them would be overridden by the "impossibility" of F-things becoming G-things. In the latter case, there would simply be no conflict in the idea that o was F and is G. All that would mean is that a generalization we took to hold in fact does not. But we do experience such conflicts; they are the basis for our second thoughts about what is and must be. In such cases, there are three quite different responses we can have: perhaps the concept that involves an inferential incompatibility between an object's being F and its being G is not a useful one in the project of making intelligible the structure of things; perhaps one or other of one's propositional commitments is mistaken (perhaps o only seems to be F and G); and perhaps one has confused two different individuals, one of which is F and the other of which is G. Which is the most rational route to take depends on one's other commitments, and no decision one makes on the matter is irrevocable. Subsequent inquiry may provide further support for one's judgment on the matter, but it can equally well give one second thoughts.

Suppose, to return to Putnam's water example one last time, we discover this stuff that looks like water but on analysis find it to have the molecular structure XYZ. Perhaps there was a flaw in the analytic procedure or in one's execution of it (i.e., the stuff really is H2O and not XYZ); perhaps the analysis has altered the sample so that the stuff one began with (H2O) is not identical to the stuff one ended up with (XYZ); and perhaps water is not invariably H2O. Of course, it is also possible that the stuff just isn't water. The first point is that all are possible responses to one's conflicting commitments, though any one will likely seem more plausible than the others depending on the overall state of the relevant science at the time. The second point is that it is because we not only have a conceptual commitment (say) to water being H2O but also a propositional commitment to the effect that this stuff here is water (which also involves a referential commitment to the existence of the relevant stuff), and a referential commitment to the portion of stuff here now being (self-)identical to a particular portion of stuff there then, that our conceptual commitments are answerable to how things actually are. On this view, judgments of the form 'this liquid here is water' are no part of the meanings of our concepts, but without such judgments the notion of a natural kind term makes no sense. Conceptual commitments are empty without propositional (and referential) commitments, not because they are meaningless independent of a word-world connection but because their status as commitments requires that something hang on them and nothing does until and unless one uses them in judgments.

In a parallel way, referential commitments, while not involving propositional (and so also conceptual) commitments, are nevertheless empty without such commitments. They are as it were formally commitments in the absence of judgments, but merely to be committed to the propriety of using a certain expression as a name without being further committed to some claim about the individual named (that it is the one Joe was talking about yesterday, a scientist, the one I just bumped into, or whatever) is to have a commitment that makes no material difference. Searle is right: one does need to have views about the object one refers to. But so is Kripke: reference is not a species of description.

If this picture of our use of referring and predicative expressions is even roughly correct, it follows that judgments (i.e., propositional commitments) that we make about the reference of a name or the meaning of a natural kind term are essentially retrospective and revisable. The meaning of a natural kind term or the reference of a name is affected as much by what we will do (the decisions we will make in future cases of conflict) as by what we have done. The question of what one means or refers to is, in this respect, rather like the question of what the market value of a good is. Much as the market value of a good at a particular point in time is a function both of where the market has been (i.e., on previous market activity) and on where it is going (subsequent activity), so the reference of a name and the meaning of a natural kind term depend not just on how we have committed ourselves in the past but also (essentially) on how we go on to commit outselves. For example, if, in the Nixon case discussed above, we staunchly proclaim that the robot merely made us think that a person Nixon existed, but over time we find ourselves continuing to interact with it -- after all, it acts like a person -- perhaps even calling it Nixon (either in spite of our decision or in self-conscious reconsideration of it), then (at some point) it would not be unreasonable to say that by 'Nixon' we had always meant the robot (though for a long time we thought it was human). But even then the case is not closed and yet further investigations could well lead to a rather different view about "what we had always meant". As Rorty rightly reminds us, we cannot escape from history.

10. According to Rorty, an "impure" philosophy of language that would aim to clarify the role of language in our epistemic commerce with the world around us "represents a confusion between the hopeless 'semantic' quest for a general theory of what people are 'really talking about,' and the equally hopeless 'epistemological' quest for a way of refuting the sceptic and underwriting our claim to be talking about nonfictions."{37} I have tried to indicate that this need not be so. The account sketched here is fundamentally historicist; it makes no attempt to provide an algorithm for discovering what any term refers to or means, and indeed entails that any such attempt is futile, and it promises no certainty for our claims to knowledge. Nevertheless, there is clearly a sense in which it aims to outline "nonhistorical conditions of any possible development."{38} The two ideas would seem, then, not to be (as Rorty takes them to be) identical. They would come to the same thing if the notions of meaning and reference that we inherit from the tradition were fixed and immutable. If thinking of language as a "picture" of the world required the idea that language is "a set of representations which philosophy needs to exhibit as standing in some sort of nonintentional relation to what they represent",{39} then Rorty would be right to claim that we should give up the "picture" picture of language and knowledge. But in philosophy as in the rest of the culture we cannot escape from history -- either from the inferentially articulated notions that we find ourselves with as the legacy of our predecessors or from the task of critically reflecting on these notions. Like any concepts, the notions of meaning and reference are articulated by their inferential relations within a whole family of notions. I have here focussed primarily on one, on the idea we inherit from the moderns that reference and meaning are external or nonintentional relations; and I have suggested that we might better understand the role of language if we instead adopt a conception of reference and meaning based on the idea that they are particular kinds of commitments. As sketched here the account does not amount to a properly worked out alternative theory. My aim was only to show that the philosophy of language and in particular the notions of meaning and reference may yet have a role to play in our attempt to understand our knowledge of the world around us. As Sellars might have said: Philosophers have hitherto sought to analyze meanings; the task is to change them.{40}


{1} Wilfrid Sellars, "Empiricism and the Philosophy of Mind" in Science, Perception and Reality (London: Routledge and Kegan Paul, 1963). [Back]

{2} W.V.O. Quine, "Two Dogmas of Empiricism" in From a Logical Point of View (Cambridge, Mass.: Harvard University Press, 1953). [Back]

{3} Richard Rorty, Philosophy and the Mirror of Nature (Princeton: Princeton University Press, 1979), Chapter 4. My own view of how the two are related will be developed below. [Back]

{4} Monist 59 (1976), pp. 321-340. [Back]

{5} See Bertrand Russell, "On Denoting" reprinted in Logic and Knowledge, Robert Marsh, ed. (London: George Allen and Unwin, 1956), pp. 45-47. The formulation given here owes much to John McDowell's presentation of the argument in "Truth-Value Gaps" in Logic, Methodology and the Philosophy of Science, vol. VI (Amsterdam: North-Holland, 1982), p. 305. [Back]

{6} In Saul Kripke, Naming and Necessity (Cambridge, Mass.: Harvard University Press, 1980). See especially Lecture I. [Back]

{7} In Keith S. Donnellan, "Proper Names and Definite Descriptions" in Semantics of Natural Language, Donald Davidson and Gilbert Harman, eds. (Dordrecht: D. Reidel, 1972). [Back]

{8} See "Truth-Value Gaps" and also "Singular Thought and the Extent of Inner Space" in Subject, Thought and Context, Philip Pettit and John McDowell, eds. (Oxford: Clarendon Press, 1986). [Back]

{9} Gottlob Frege, "On Sense and Reference" in Translations from the Philosophical Writings of Gottlob Frege, Max Black and Peter Geach, eds. (Oxford: Blackwell, 1952). [Back]

{10} Hilary Putnam, "The Meaning of 'Meaning'" in Mind, Language and Reality (Cambridge: Cambridge University Press, 1975). [Back]

{11} The labels are John Searle's. See Chapter 9 of Intentionality: An Essay in the Philosophy of Mind (Cambridge: Cambridge University Press, 1983). According to the internalist, reference (and/or extension) is fixed by internal conditions of satisfaction, by some "Intentional content" that is associated with the term. For the externalist, reference (and/or meaning) is fixed "by virtue of some facts about the world quite independently of how those facts are represented in the mind" (p. 246). [Back]

{12} This is why Searle thinks that Kripke and Donnellan are really internalists or descriptivists. See Chapter 9 of Intentionality. [Back]

{13} This difficulty can arise even in a case in which the relevant object is fictitious. To vary a familiar example of Donnellan's, suppose that the sleepy child who believes that Tom is nice in fact arrived at this belief while watching a TV show featuring a character named Tom, though the child now vaguely thinks that Tom came in from the party to talk to him while he was watching TV. In such a case, the internalist strategy of asking him whom he is talking about may yield no answer at all (since perhaps no one satisfies most or any privileged subset of the claims he makes about "Tom"), and the externalist strategy of looking for the causal antecedents of his belief will yield a fictitious character -- which is not to say that the causal antecedents are fictitious. It is worth noting as well that analogous examples could be generated for natural kind terms. Imagine the result of pursuing the two strategies in attempting to fix the extension of Kant's term that we translate as 'gold'. In the one case, the extension would be all and only yellow metals; in the other, all and only gold (as we would say). A case involving a fiction can be built around someone's beliefs about, say, phlogiston. [Back]

{14} Philosophy and the Mirror of Nature, p. 292; cf "Realism and Reference", p. 326. [Back]

{15} Robert Brandom's excellent account of the role of 'refers' and its cognates in everyday discourse in "Reference Explained Away," The Journal of Philosophy LXXXI (1984), pp. 469-492, reinforces this point. [Back]

{16} On the other hand, "pure" philosophy of language, whose aim is to "systematize our notions of meaning and reference in such a way as to take advantage of quantificational logic, preserve our intuitions about modality and generally produce a clear and intuitively satisfying picture of the way in which notions like 'truth,' 'meaning,' 'necessity,' and 'name' fit together", is perfectly acceptable according to Rorty (Philosophy and the Mirror of Nature, p. 257). [Back]

{17} "Realism and Reference", p. 327. [Back]

{18} A name (or natural kind term) is properly used only to refer to its bearer (to things of that kind), to one and the same entity (kind) across various uses of the expression. But, as Geach argues in Mental Acts (London: Routledge and Kegan Paul, 1957), p. 69, "'the same' is a fragmentary expression, and has no significance unless we say or mean 'the same X', where 'X' represents a general term." Geach concludes that unless a common noun is given as a part of the content of a name, there can be no norms governing its correct use. The point would seem to apply equally to the case of natural kind terms. [Back]

{19} A similar example could be constructed for "baptisms" involving natural kind terms. For while Putnam would have us believe that even before we had any idea that water has a molecular structure, let alone what it is, we somehow meant by 'water' stuff with that structure, this just seems wrong (for familiar Quinean reasons). Even after we had discovered that water is H2O, we might have taken the stuff we discover is XYZ to be water. That is, we could treat the discovery not as the discovery that XYZ is not water but rather as the discovery that water is not invariably H2O. We may have, in a given case, good reason to go one way rather than the other, but the reason is not that it is simply a fact about the meaning of the term that it is so used. A further, more fundamental, problem with the idea that the meanings of natural kind terms are fixed at "baptisms" is discussed in § 7. [Back]

{20} Naming and Necessity, p. 53. [Back]

{21} Kant was quite clear on this point. He writes in the Logic, Robert S. Hartman and Wolfgang Schwarz, trans. (New York: Dover, 1974), Chapter 1 §1.15: "Since only single things or individuals are of an all-sided determination, there can be cognitions of an all-sided determination only as intuitions, not, however, as concepts; in respect of the latter, logical determination can never be considered as complete." As Quine might put the point: Exactly which possible fat man in the doorway is one referring to in such a case? See "On What There Is" in From a Logical Point of View, pp. 1-4. [Back]

{22} It is often thought that an operator like David Kaplan's 'dthat' ensures that we can refer to merely possible individuals. (He introduces the operator in "Dthat" in Contemporary Perspectives in the Philosophy of Language, Theodore Uehling, Peter French Jr., and Howard Wettstein, eds. (Minneapolis: University of Minnesota Press, 1979).) But this is a mistake. One way to think about Kripke's point is in terms of the distinction between imagining the possibility of some specific individual and merely imagining a possible state of affairs (i.e., its satisfaction by some individual or other). This distinction cannot be drawn for merely possible individuals since to alter any aspect of the description of such an individual is just to imagine another possible individual. [Back]

{23} Searle reads the argument this way. So, he thinks, it can be met by saying, first, that some descriptions function naturally, as it were, as rigid designators, and second, that any description can be made to do so. (See Intentionality, p. 258.) I have already indicated that the latter claim is false. The former, while perhaps true, is beside the point since we almost invariably lack any such description of the individuals about which we entertain counterfactuals. Rorty too seems to misunderstand Kripke's point here insofar as he takes it merely as a point about the constraints on an adequate semantic theory rather than as a fundamental insight into the respective roles of names and predicates in ordinary discourse. [Back]

{24} See Saul Kripke, "A Puzzle about Belief," in Meaning and Use, A. Margalit, ed. (Dordrecht: D. Reidel, 1979). [Back]

{25} Philosophy and the Mirror of Nature, p. 171. [Back]

{26} As already remarked, this is how Rorty reads them. Indeed, he takes Sellars to task for not getting Quine's point, for seeing through the myth of the given and then mistakenly hanging onto the myth of meanings. If I am right, Sellars saw far better than Rorty has the significance of the argument of "Empiricism and the Philosophy of Mind". [Back]

{27} Some of Rorty's more tendentious pragmatist conclusions seem to rely on his not recognizing this. See, for example, Philosophy and the Mirror of Nature or "The World Well Lost" in Consequences of Pragmatism (Minneapolis: University of Minnesota Press, 1982). See also Donald Davidson's "A Coherence Theory of Truth and Knowledge" in Truth and Interpretation, Ernest LePore, ed. (New York: Basil Blackwell, 1986). Like Rorty, Davidson sees that meaning cannot be explicated by appeal to experience (which, for such purposes, must be taken to fall outside of the realm of the meaningful or the conceptual). Since, like Rorty, he takes Quine to have shown nonetheless that there is no analytic/synthetic distinction, he (like Rorty) "loses" the world as the proper stage on which our cognitive lives are played out. [Back]

{28} In "The Meaning of 'Meaning'", p. 225. [Back]

{29} See the first Critique A147/B187; also the Logic, Chapter 1, § 1. [Back]

{30} A clear and simple example of the distinction between the meaning of a predicate and the question whether one does or should endorse it is given by Michael Dummett in Frege's Philosophy of Language (London: Duckworth, 1973). He writes (p. 454): "The condition for applying the term ['Boche'] to someone is that he is of German nationality; the consequences if its application are that he is barbarous and more prone to cruelty than other Europeans. . . . Someone who rejects the word does so because he does not want to permit a transition from the grounds for applying the term to the consequences of doing so." See also Robert Brandom's very useful discussion of this point in "Inference, Expression, and Induction", Philosophical Studies 54 (1988), pp. 257-285. [Back]

{31} In Minnesota Studies in the Philosophy of Science, vol II, Herbert Feigl, Michael Scriven and Grover Maxwell, eds. (Minneapolis: University of Minnesota Press, 1958). [Back]

{32} Wilfrid Sellars, "Inference and Meaning," Mind LXII (1953), pp. 313-338; quotation from pp. 336-337. [Back]

{33} "Inference and Meaning", p. 337. [Back]

{34} He describes them in "Language as Thought and as Communication", Philosophy and Phenomenological Research 29 (1969), pp. 506-522, as ought-to-be's, and contrasts them with rules of inference that are ought-to-do's. [Back]

{35} "Counterfactuals, Dispositions, and the Causal Modalities", pp. 302-303. [Back]

{36} In "Varieties of Understanding", in Reason and Rationality in Natural Science: A Group of Essays, Nicholas Rescher, ed. (New York: University Press of America, 1985), Robert Brandom discusses the role of logical vocabulary in making explicit in claims commitments that are implicit in our inferential practices. [Back]

{37} Philosophy and the Mirror of Nature, p. 293. [Back]

{38} Ibid., p. 9. [Back]

{39} Ibid., p. 295. [Back]

{40} What Sellars actually did say is, "The motto of the age of science might well be: Natural philosophers have hitherto sought to understand 'meanings'; the task is to change them." ("Counterfactuals, Dispositions, and the Causal Modalities", p. 288). [Back]

Formatted in hypertext by Andrew Chrucky, May 10, 1997.

Return to Problems from Wilfrid Sellars