The Space of Issues and Options
The Space of Issues and Options
Abstract and Keywords
This chapter presents an overview of central issues that must be confronted in developing a decent account of meaning, with various positions that might be taken with respect to them, and with some of the arguments that can be given for and against these positions. Topics addressed include forms of scepticism about meaning, the pros and cons of reductionism, the relationship between language and thought, the compositionality of sentence-meanings from word-meanings, the normative import of meaning, individualism, externalism, and the promise of a deflationary theory.
Each expression of a language surely means something—there is some fact as to what it means; but the nature of such facts is notoriously obscure and controversial. Consider the term “dog”. It possesses a distinctive literal meaning in English, and this feature is closely associated with various others—for example, that we use the word to help articulate certain thoughts; that it is appropriately translated into the Italian “cane” and the German “Hund”; and that we should try to apply it to dogs and only to dogs. But such characteristics range from the puzzling to the downright mysterious. Does thought itself take place in language? How might ‘little’ meanings (like that of “dog”) combine into ‘bigger’ ones (like that of “dogs bark”)? What is it about that word's meaning that enables it to reach out through space and time, and latch on to a particular hairy animal in ancient China? And there is a ramified profusion of further questions, as we shall see. So it isn't surprising that philosophy abounds with theories that aim to demystify these matters, to say what it is for a word or a sentence to have a meaning.
This introductory chapter aims to map the terrain of alternative suggestions. To that end I will mention the central issues that must be confronted in developing a decent account of meaning, together with the various positions that might be taken with respect to them, and some of the arguments that can be given for and against these positions. Be warned, however, (p.2) that the immediately following discussions are cryptic and sketchy—something of a mad dash through the literature. They are intended merely to provide an orienting background to the line of thought that will be elaborated at a more reasonable speed in the rest of this work.
2. MEANING SCEPTICISM
It is sometimes maintained that the expressions of a language really do not, as we might naively think, possess meanings—but accounts of this sceptical kind may be more or less radical. At the most extreme there is a theory that, as far as I know, has never been seriously proposed, namely, that there are no semantic phenomena at all, that no word stands for anything, and that no sentence is true or false. Such a view is hardly credible: for no one who understands the word “dog” could doubt that it picks out dogs (if there are any dogs); and no one who understands the sentence “dogs bark” could doubt that it expresses a truth if and only if dogs bark; and so on. However, there are less radical forms of meaning-scepticism that do have adherents.
For example, one might deny (with Quine1) that there are any facts concerning the meanings or referents of foreign expressions (including the expressions of compatriots, who seem to be speaking the same language as oneself). This is not as chauvinistic as it may initially sound; for it amounts to a general and unbiased scepticism about the objectivity of translation. Quine's position is based on his ‘indeterminacy thesis’: namely, that linguistic behaviour at home and abroad—which he takes to provide the only facts with the potential to establish the correctness of any proposed translation (p.3) manual—will in fact be consistent with many such proposals; so we can rarely fix what a foreigner (or any other person) means by his words. But a number of counters to this argument have appeared in the literature. One response (pioneered by Chomsky2) is that the failure of the phenomena of word-usage to settle how an expression should be translated would not result in there being no fact of the matter, but merely in a familiar underdetermination of theory by data (i.e. in a difficulty of discovering what the facts of translation are). Another common strategy of reply (e.g. Horwich3) is to argue that Quine has adopted too narrow a view—too behaviouristic—of what the non-semantic meaning-constituting features of word-use may be; that they actually include, not merely assent–dissent dispositions, but also (for example) causal relations amongst such dispositions; and that once such further evidence is taken into account, the alleged indeterminacy disappears. To illustrate using Quine's famous case: although we may be prepared to assent and dissent, in the same environmental circumstances, to “There's a rabbit” and “There's an undetached rabbit-part”, we tend to assent to the second as a consequence of having assented to the first, not vice versa; and that causal fact can be a ground for deciding which of two co-assertible foreign sentences should be translated into one and which into the other.
A different and relatively mild form of semantic scepticism would countenance facts about what refers to what and about the truth conditions of sentences, but would renounce any finer-grained notion of meaning, such as Fregean ‘sense’. Thus there would be no respect in which co-referential terms (such as “Hesperus” and “Phosphorus”) would differ in meaning. One (p.4) source of this scepticism might be a Millian/Russellian rejection (Salmon,4 Donnellan,5 Crimmins & Perry,6 Lycan,7 Soames8) of the argument typically offered in support of fine-grained meanings: namely, Frege's argument that they are needed in order to accommodate our intuition that (for example) ‘believing Hesperus is Phosphorus’ is not the same thing as ‘believing Hesperus is Hesperus’. But it remains hard to see much wrong with that reasoning.9
Another widespread motivation for embracing the mild form of scepticism is the Davidsonian view that compositionality (the dependence of our understanding of sentences on our understanding of their component words) requires that fine-grained meanings be abandoned in favour of mere truth conditions and their coarse-grained determinants.10 But again one might well prefer a Fregean point of view: one might suppose that the state of understanding a complex expression is identical to the state of understanding its various parts and appreciating how they are combined with one another. In that case compositionality will have a trivial explanation, and there will be no pressure to adopt Davidson's truth conditional account of it.11
(p.5) Finally there is a so-called ‘non-factualist’ form of meaning-scepticism, which Kripke12 takes Wittgenstein13 to be urging. The idea is that although we may properly and usefully attribute meanings to someone's words, we should not think of these attributions as reporting genuine (‘robust’) facts about that person, but rather as implementing some quite different speech act—something along the lines of ‘expressing our recommendation that his words be taken at face value’. Of course, there is a perfectly legitimate deflationary sense of “fact” in which “p” is trivially equivalent to “It is a fact that p”; and when we attribute a meaning we obviously suppose there to be a ‘fact’, in that sense, as to what is meant. Thus non-factualism faces the problem of specifying what makes certain facts ‘genuine’ or ‘robust’ ones; and this has not so far been satisfactorily resolved. For example, it might be tempting to identify them as those facts that enter into causal/explanatory relations. But then—since it is pretty clear that a word's meaning helps to explain the circumstances in which sentences containing it are accepted—the Kripkensteinian position would be pretty clearly false. Alternatively, it might be said that the ‘genuine’/‘robust’ facts are those that are constituted by physical facts. But in that case non-factualism would boil down to a familiar form of anti-reductionism, and one would be hard-pressed to see anything sceptical about it.
Amongst non-sceptical accounts of meaning, some are reductionist, others are not: some aim to identify underlying non-semantic facts in virtue of which an expression possesses its meaning; others take this to be impossible and aim for no more than an epistemological story—a specification of which (p.6) non-semantic data would tend to justify the tentative ascription of a given meaning.
Reductionist theories are typically motivated by a general sentiment to the effect that, since we humans are fundamentally physical beings, i.e. made of atoms, all our characteristics—including our understanding of languages—must somehow be constituted out of physical facts about us. However, many philosophers are unconvinced by this line of thought—arguing that the majority of familiar properties (e.g. ‘red’, ‘chair’, ‘democracy’, etc.) resist strict analysis in physical terms, and therefore that the way in which empirical facts are admittedly somehow grounded in the physical need not meet the severe constraints of a reductive account. In response to this point, it may be observed that although some weak form of physical grounding might suffice for certain empirical properties, others—those with a rich and regular array of physical effects—call for strict reduction. Otherwise, given the causal autonomy of the physical, those effects would be mysteriously overdetermined. In particular, the fact that the meaning of each word is the core-cause of its overall use (i.e. of all the non-semantic facts concerning the acceptance of sentences containing it) would be explanatorily anomalous unless meaning-facts were themselves reducible to non-semantic phenomena. However, as plausible as these considerations might be, the only solid argument for semantic reductionism would be an articulation and defence of some specific theory of that form. Conversely, the best anti-reductionist argument is that no such account has been found despite strenuous attempts to construct one.
Reductionist approaches of various stripes will be the focus in what follows; so I won't dwell on them now. As for anti-reductionist proposals, amongst the most prominent in contemporary analytic philosophy are those due to McGinn, McDowell, Davidson, and Kripke. McGinn14 argues that our not having (p.7) managed to devise a plausible reductive account of ‘understanding’ should be no more surprising or embarrassing than our inability to give such an account of other psychological features, like bravery or kindness. McDowell15 gives this perspective a Wittgensteinian gloss: since our puzzlement about meaning is merely an artefact of self-inflicted mystification, the illumination we need will have to come from a rooting out of confusions rather than from the development of a reductive theory, and so there is not the slightest reason to expect there to be such a thing. Davidson16 combines that anti-reductionist metaphysics with a neo-Quinean epistemology of interpretation: the most plausible translation manual for a foreign speaker's language is the one that optimizes overlap between the circumstances in which her sentences are held true and the circumstances in which we hold true the sentences into which hers are to be translated. And Kripke17 sketches a superficially similar idea (on behalf of Wittgenstein): it is reasonable to tentatively suppose that someone means plus by a symbol of hers when she deploys it more or less as we deploy the word “plus”. But note that in Kripke's view, unlike Davidson's, such norms are not to be regarded as specifying the evidence for a species of ‘genuine’ fact.
4. LANGUAGE AND THOUGHT
A further bone of contention is the relationship between overt, public languages, such as English and Chinese, and the psychological states of belief, desire, intention, and other forms of thought, which these languages are used to articulate and (p.8) communicate. The central issue here is whether or not thinking itself invariably takes place within a language (or language-like symbol-system). Is it the case, for example, that the state of ‘believing that dogs bark’ consists in accepting (perhaps unconsciously) some mental sentence whose meaning is dogs bark? The overall shape of any account of meaning will depend on how this question is answered.18
Consider, to begin with, the philosophers who would deny that thinking is inevitably linguistic. Within that group there are those (such as Grice19) who maintain that the meanings of public-language sentences derive (in virtue of our intentions and conventions) from the propositional contents of the beliefs, etc. that they are typically used to express. Thus “dogs bark” means what it does because of our practice of uttering it in order to convey the belief that dogs bark. But this approach fails to address the problem of how certain configurations of the mind/ brain come to instantiate the intentions and beliefs they do. Then we find those—arguably Wittgenstein20 and Quine21—who would solve this problem by supposing that public language meanings are ‘prior’ (in a certain sense) to the contents of thoughts, i.e. that one can see how a given state of the mind/ brain comes to possess the conceptual content it does by reference to the meaning (independently explained) of the public expression with which it is correlated.
Alternatively, there are theorists who maintain that all human thinking takes place within a mental language—either a universal ‘Mentalese’ or else a mental form of English, Italian, etc. (depending on the speaker). Of these theorists, many (e.g. (p.9) Fodor,22 Schiffer,23 Loar,24 Sperber and Wilson,25 Neale26) advocate a two-stage theory: first, an account of how the terms of a mental language come to mean what they do; and, second, a neo-Gricean account of how the meanings of someone's overt public language derive from those contents.
However, as we shall see in the appendix to Chapter 2, it might be argued that the agreements and explicit intentions invoked by Grice rely on public language meaning, and so cannot constitute it; that the link between a sound and its mental associate is fixed at an early age; and that their common meaning derives from the joint possession of the same meaning-constituting property, e.g. the same basic use, or the same causal correlations with external properties. Therefore, it is best to suppose that there is a single way in which meaning is constituted, applying equally well to both mental and overt languages. Such an approach would obviously have to be non-Gricean. And it would be especially compelling if each of us thinks largely in our own public language. From this point of view—suggested by Gilbert Harman,27 and argued in Chapter 7—it seems especially clear that there can be no substantial difference between an account of the contents of thoughts and an account of the literal semantic meanings of the sentences that express them.
It is uncontroversial that, apart from idioms, the meaning of any complex expression-type (such as a sentence) depends on (p.10) the meanings of its component words and on how those words have been combined with one another. But there is little consensus on how this obvious fact should be incorporated within a full story about meaning.
A common assumption is that compositionality puts a severe constraint on an adequate account of how an expression's meaning is engendered. For it requires that the facts in virtue of which a given sentence means what it does be implied by the structure of the sentence together with the facts in virtue of which the words mean what they do. And, given certain further commitments that one could well have, this condition may be difficult to satisfy.
For example, verificationists (e.g. Schlick28) maintain that the meaning of each sentence consists in the way in which we would go about establishing whether or not it is true (—from which it follows that no untestable hypothesis could be meaningful). And they go on to say (in light of compositionality) that the meaning of each word must consist in the constant ‘contribution’ it makes to the various ‘methods of verification’ of the various sentences in which it appears. But this point of view suffers from the fact that no one has ever been able to spell out what these contributing characteristics are. In addition, it is hard to see why one should not be able to construct sentences that, despite being neither verifiable nor falsifiable, nonetheless possess meanings in virtue of their familiar structures and the familiar meanings of their parts. Thus compositionality and verificationism do not sit well together.
Davidson's influential thesis (mentioned in section 2) is that compositionality may be accommodated only by identifying the meanings of sentences with truth conditions and the meanings of words with reference conditions; for one will then be in a position to derive the former meanings from the latter by exploiting the methods deployed in Tarski's definitions of truth. (p.11) And this idea sparked energetic research programmes aimed towards extending the types of linguistic construction (e.g. to those involving adverbs, indexicals, modalities, etc.) for which this treatment may be given, and towards finding a notion of ‘truth condition’ that is strong enough to determine (or replace) meaning. Doubts about whether such problems can be solved tended to be dismissed with the response that since natural languages are evidently compositional, and since there is no alternative to the truth-conditional way of accommodating that characteristic, there must be solutions, and so our failing to find them can only be due to a lack of ingenuity.29
In a similar vein, Fodor and Lepore30 also brandish a ‘substantive compositionality constraint’. In their case, the aim is to knock out various accounts of word meaning. For example, they argue that the meaning of a term cannot be an associated stereotype, since the stereotypes associated with words (e.g. with “pet” and “fish”) do not determine the stereotypes associated with the complexes (e.g. “pet fish”) in which those words appear. Clearly this argument presupposes that there is a certain uniformity in how the meanings of expressions are constituted, i.e. that whatever sort of thing (e.g. an associated stereotype, or a reference/ truth condition) provides the meanings of words must also provide the meanings of the complexes formed from them.
An alternative picture—one that will be developed in Chapter 8—would oppose this uniformity assumption (including the Davidsonian implementation of it). Indeed, it would oppose giving any general account—covering the meanings of complexes as well as words—of the sort that could leave open the question of whether the former could be determined by the latter. Instead, its account of complexes would presuppose (p.12) compositionality; for it would say that the meaning of a complex expression is constituted by the facts concerning its structure and the meanings of its words. For example, the property, ‘x means THEAETETUS FLIES’, would be constituted by the property, ‘x is an expression that results from applying a function-term that means flies to an argument-term that means THEAETETUS’. In that case, any reductive account of word-meanings—no matter how poor it is—will induce a reductive account of complex-meanings that trivially complies with the principle of compositionality. Thus that principle cannot help us to decide how the meanings of words are constituted.
Focusing now on what does engender the meaning of a word, we find a much debated division between theories that favour analyses in evaluative terms and those that do not. There is an intimate relation (emphasized by Kripke31) between what a word means and how it should be used: for example, if a word means dog then one ought to aim to apply it only to dogs; therefore one should not apply it to something observed swinging from tree to tree. And many philosophers (e.g. Gibbard,32 Brandom,33 Lance and Hawthorne34) have drawn the conclusion that meaning must somehow be explicated in terms of what one ought and ought not to say—hence, that meaning is constitutionally evaluative. Thus it could be, for example, that the meaning of “not” is partially engendered by the fact that one ought not to accept instances of “p and not p”.
(p.13) In opposition to this conclusion it can be argued that the ‘factual’ effects of a word's meaning (namely, someone's disposition to accept certain sentences containing it) would be difficult to explain if meaning were evaluative rather than ‘factual’. And in opposition to the reasoning behind that conclusion, it can be argued that the evaluative import of a meaning-property isn't enough to make that property constitutionally evaluative. Killing, for example, has evaluative import; one ought not to do it. And this could well be a basic evaluative fact—not explicable on the basis of more fundamental ones. But we may nevertheless give an account of killing in wholly non-evaluative language. So why not take the same view of meaning?
The answer, perhaps, is that, unlike killing, meaning is a matter of implicitly following rules (Wittgenstein,35 Brandom); for the patterns of word-use that a speaker displays are the result of corrective molding by his community. But even if one concedes that meaning is constitutionally regulative—i.e. a matter of rule following—this is not to say that attributions of meaning are evaluative. No doubt, the notion of ‘its being right to follow a certain rule’ is evaluative. But the notion of ‘a person's actually following that rule’ surely lies on the other side of the ‘fact’/value divide.
Moreover, it would remain to be seen whether meaning is fundamentally regulative—for one might aspire to analyse rule-following in entirely non-normative, naturalistic terms. Some philosophers (e.g. Kripke and Brandom, in the works just cited) contend that this is impossible. They argue that any analysis of ‘implicitly following rule R’ would have to depend on an a priori specification of the naturalistic conditions in which an action would qualify as mistaken, and that such an account cannot be supplied. But there are others (e.g. Blackburn36) who maintain that the required account can be supplied. And yet others, (e.g. myself, in Chapter 5 of this book) who reject the (p.14) requirement, claiming that the relevant notion of ‘mistake’ is defined in terms of ‘following rule R’, rather than vice versa, and proposing analyses that do not satisfy it. Thus one might suppose that S implicitly follows R when, as a result of corrective reinforcement, it is an ‘ideal law’ that S conforms with that rule—where the notion of ‘ideal’ is the non-normative, naturalistic one that is often deployed in scientific models, e.g. the ideal gas laws.
According to some philosophers (again following Kripke) a consequence of these normativity considerations is that meaning is an essentially social phenomenon; so a ‘private language’ is impossible. For the implicit rule-following which must be involved in a person's meaning something allegedly depends on activities of correction displayed within his linguistic community. And this conclusion is independently supported by the observation (Kripke,37 Evans,38 Putnam,39 Burge40) that we in fact do interpret people, not merely on the basis of their own idiosyncratic usage of words, but also on the basis of what their community means. Thus if a girl, reporting what she has learned at school, says “Kripke discovered other worlds”, we take her to be referring not to whichever individual satisfies some definite description that she happens to associate with the name—there may be no such description, or it may pick out the wrong guy—but rather to Kripke, i.e. the person her teacher was referring to, who was in turn referring to the same person as (p.15) his source of the name was referring to, and so on. And when Putnam—a self-confessed incompetent with tree-names—says, pointing to a big shrub, “Is that an elm?”, we take him to have asked whether it's an elm, i.e. whether it's what the experts would call “an elm”. His own defective practice with the word does not fix what he means by it. (See section (h) of Chapter 2 for further discussion.)
Opposed to this conception, however, there are a number of philosophers (e.g. Chomsky,41 Crane,42 Segal43) who maintain that there is a kind of meaning, better suited to psychological explanation, whereby what each person means is constituted by facts about that person alone and is conceptually (though not causally) independent of what other people do. These theorists could either deny that this individualistic brand of meaning is constitutionally regulative; or they could accept that it is, but regard the rules as sustained by self-correction. They may allow that we also have a notion of communal meaning, and that this is the notion that is typically deployed in ordinary language when we speak of what someone means. But, if so, they will contend that it is derived (e.g. by a sort of averaging) from the more fundamental notion of idiolectal meaning—so that communal meaning is not appropriate for explaining a particular person's thoughts and actions.
Alongside the distinction between ‘communal’ and ‘individualistic’ accounts, there is a distinction between those theories according to which what we mean by our words (at least certain words) depends on the physical environment of their deployment, and those according to which meanings are wholly “in the head”.
(p.16) The former (‘externalist’) perspective came to prominence with Putnam's44 famous thought experiment. Since Oscar's physical duplicate on Twin Earth is surrounded by a liquid that, despite its superficial appearance, isn't really water, we are reluctant to say that the doppelganger's word “water” refers to the same thing as our word does—even though, since he and Oscar are intrinsically identical, their internal uses of it are exactly the same. Thus it would seem that the facts that provide certain terms with their meanings must include aspects of the outside world.
On the other hand, it has been argued (Fodor45, White,46 Jackson and Pettit,47 Chalmers48) that words like “water” have a certain indexical character—that their reference depends (as in the case of “I”, “our”, and “here”), not merely on their fixed meanings in English, but also on the context of their use. One method of implementing this idea would be to suppose that the meaning of “water” is constituted by an acceptance of
x is water ↔ x has the underlying nature, if any, of the stuff in our seas, rivers, lakes and rain.
In this way (as Putnam himself appreciated) the usual twin-earth intuitions may be somewhat reconciled with internalism. Twin-Oscar would mean the same as Oscar, but would refer to something different.
An especially prominent form of externalist view is one that explains the meaning of a word in terms of its reference, which (p.17) is then explained in terms of one or another naturalistic relation between the word and some aspect of the world (Devitt49). More specifically, Stampe50 and Fodor51 have developed (each in their own way) the idea that
w means F ≡ w is causally correlated with fs,
where the lower-case “f” is to be replaced by a predicate (e.g. “dog”) and the capital “F” is to be replaced by a name of the concept that the predicate expresses (e.g. “DOG”). Alternatively, Millikan,52 Dretske,53 Papineau,54 Neander,55 and Jacob56 have offered versions of the idea that
w means F ≡ the (evolutionary) function of w is to indicate the presence of fs.
However, a good case can be made that the relational form exemplified by all such accounts, viz.
w means F ≡ R(w, f)
is incorrect, and that the motivation for implicitly insisting on it is defective. For the reason one might be drawn to such an account is that meaning has truth-theoretic import; if a word means dog, then it is true of dogs; so sentences containing it are about dogs. And, in general,
w means F → (x)(w is true of x ↔ fx).
But this implies—assuming some reductive analysis of ‘w is true of x’ as ‘wCx’—that whatever constitutes the meaning-fact (p.18) must entail ‘(x)(wCx ↔ fx)’, and so must indeed be an instance of the form, ‘R(w, f)’. However (as we shall see in detail in Chapter 3)57 this line of thought is undermined by the plausibility of deflationism with respect to truth and reference: namely, the idea that these are non-naturalistic, logical notions—mere devices of generalization. For, if that is correct, then the presumption that ‘w is true of x’ has some reductive analysis would be mistaken.
Thus the import of deflationism is that we should not require a reductive theory of meaning to have the relational form
w means F ≡ R(w, f)
Nor—which comes to the same thing—should we expect, given some proposed reductive analysis of a specific meaning-property, to be able to explain why it holds (e.g. why a word with this particular use must mean dog, and must be true of that set of objects). And nor—again equivalently—should we require an account that will enable us to read off what each word means from information about its use. Consequently, our inability to devise a theory that does satisfy such constraints—an inability which has been convincingly demonstrated by Kripke,58 Boghossian,59 and Loewer60—should not tempt us to doubt (as they do) the prospects for a reductionist account. It should rather confirm what we might well have already recognized—that these constraints should never have been imposed in the first place.
The legitimate basic requirement on an adequate analysis of a meaning-property is exactly what one would expect from consideration of reductions elsewhere—i.e. in biology, physics, (p.19) etc.—namely, that the alleged underlying property must contribute to explanations of the symptoms of the superficial property. Thus ‘being magnetic’ reduces to having a certain microstructure in virtue of the fact that something's possession of that microstructure explains why it exhibits the attraction– repulsion behaviour that is symptomatic of being magnetic. Similarly, ‘U(w)’ provides a good analysis of ‘w means F’ if and only if ‘U(w)’ contributes to explanations of the symptoms of meaning F. But the symptoms of a word's meaning F are its having a certain overall use (that of the word “f”). Therefore, ‘U(w)’ constitutes the meaning of “f” just in case it explains (in conjunction with extraneous factors) the differing circumstances in which all the various sentences containing “f” are accepted. And there is no reason why the satisfaction of this adequacy condition should dictate analyses that take the relational form.
10. PROMISING DIRECTIONS
The preceding survey of alternative views of meaning suggests that there are reasonable prospects for an account that is (a) non-sceptical, (b) reductive, (c) applicable to both overt and mental languages, (d) focused in the first instance on word-meaning and trivially extendable to sentence-meaning, (e) not evaluative or fundamentally regulative, (f) applicable to both communal languages and idiolects, (g) internalist, and (h) deflationist, in the sense of not having to take the form of a relational account, ‘w means F ≡ R(w, f)’, which would incorporate a naturalistic analysis of truth.
These features are characteristic of so-called use theories of meaning, deriving from the work of Wittgenstein61 and Sellars,62 and also known as “conceptual (or functional) role semantics” (p.20) (see Field,63 Block,64 Harman,65 Peacocke,66 and Wright67). According to the version of it that I favour, and which will be developed in subsequent chapters, the meaning of each word, w, is engendered by its ‘basic acceptance property’—that is, by the fact that w's overall use stems from the acceptance (in certain circumstances) of specified sentences containing it. A singular virtue of this proposal is that we have a plausible model—namely inference—of how such a property might, in conjunction with other factors, explain a word's overall use (i.e. the acceptance-facts regarding every sentence containing the word). Consequently, we can see how the just mentioned condition on an adequate account of meaning constitution might be met.
Given the enormous variety of things that are done with language, we should not expect there to be much similarity between the basic acceptance properties of different predicates. Perhaps those of colour words resemble each other to a fair degree; and similarly there could well be resemblances within species names, numerical predicates, evaluations, mental terms, etc.; but as we move from one such type to another there is likely to be a considerable divergence of structure. In particular, there is no reason to anticipate that the basic acceptance property of predicate “f” will generally have the form ‘R(w, f)’. Indeed, one might question whether it ever will.
Nonetheless it will not be hard to account for a word's referential and normative character. We have the pair of fundamental schemata:
w means F → (x)(w is true of x ↔ fx),
(x)(w is true of x ↔ fx) → one's goal should be that of accepting the application of w to x only if fx.
Therefore, once we have established (on the basis of the above-mentioned adequacy condition) that a word's meaning F is constituted by its having a certain basic acceptance property, then its principal referential and normative characteristics are trivially accommodated.
Two further features of this proposal are worth emphasizing (and will be treated in greater depth in the next chapter). First it is ‘non-holistic’ in the sense that it incorporates an objective separation between those sentences that are held true as a matter of meaning and those sentences whose acceptance is not required by meaning alone. This anti-Quinean68 distinction is drawn on the basis of explanatory priority: the meaning-constituting uses are those that are responsible for the others. Thus one may rebut the claim that ‘use theories’ inevitably lead, for better (Block,69 Harman70) or for worse (Fodor and Lepore71), to holism. Second the theory is ‘non-atomistic’ in the following sense: it implies that the existence of words with certain meanings requires the existence of further words with certain different meanings. After all, the meaning of a word can be engendered by the acceptance of some particular sentence containing it only if the other words in that sentence are understood appropriately. This is not the extreme and implausible view (condemned by (p.22) Dummett72) that the meaning of every word depends on the meanings of every other word in the language. What is required, rather, is that there be a limited stock of interrelated basic meanings on which all others asymmetrically depend.
11. FURTHER PROBLEMS
This introductory survey provides no more than the briefest of discussions of some of the many important issues and options confronting a theorist of the nature of meaning. And those dimensions of controversy that I have mentioned are merely the most central ones; there are others that have not yet been considered, but which any satisfactory account must come to grips with. Let me end by listing four of them that will be taken up in the next chapter.
(I) It is not unnatural to think that whenever a word is used the speaker invests it with a certain meaning, and that if he uses the same word (i.e. sound-type) on another occasion, then he may or may not decide to invest it with the same meaning. It may seem, therefore, that the meaning of an unambiguous word-type should be explained in terms of the uniform meaning given to its various tokens; similarly the meanings of ambiguous word–types should be explained in terms of the several meanings distributed amongst its tokens. But this tempting picture is at odds with the various accounts we have been considering. For example, according to Fodor's theory, the meaning of a type is engendered by a causal correlation between its tokens and exemplifications of a certain property. And the other accounts also attribute meaning, in the first instance, to word-types. Thus we must address the following couple of questions. Can there be a reductive account (perhaps a modification of one of those discussed above) that applies initially to word-tokens? And if, on (p.23) the contrary, type-meaning is indeed primary, then how—given the phenomenon of ambiguity—are we to account for the meanings of specific tokens?
(II) We have been concentrating on our notion of ‘the meaning of a word in a given language’. But there are other meanings of “meaning” that also stand in need of explication, especially:
(a) What the speaker means on a given occasion by some word—where this is some temporary modification of its meaning in the language as a whole. The notion of meaning in which “The President” may be used, in virtue of the speaker's local intentions, to mean “The current President of France”.
(b) What is said, in a given context, by the utterance of some sentence, the proposition expressed by a sentence-token. The notion of meaning in which “I am hungry” means different things depending, not on the speaker's intentions, but on who is speaking, and on when the utterance is performed.
(c) The conventional pragmatic content of a term, its illocutionary force (going beyond the de dicto propositional constituent that is expressed by it). The respect of meaning in which “but” differs from “and”, and in which “I promise to go” engenders a specific obligation.
(d) The full information conveyed by the making of a given utterance, i.e. its ‘conversational implicature’, that which the hearer may infer from the speaker's deciding, in the circumstances, to say what he does. The respect of meaning in which “There's no milk left” can mean “Would you buy some?”.
(e) The non-literal meanings of an expression, including metaphorical and ironic meanings.
It is not implausible that the kind of meaning on which I have been focusing here (and on which I will continue to focus in (p.24) subsequent chapters) is fundamental, i.e. that the other kinds are best explained in terms of it. But this assumption may be justified only on the basis of defensible concrete proposals (Grice,73 Sperber and Wilson,74 Neale,75 Recanati76).
(III) On the face of it, an expression's having a certain meaning consists in its standing in the relation, ‘x means y’, to an entity of a special kind—a meaning-entity. Consequently, one would expect a reductive theory of any particular meaning-fact to be the product of two more basic theories: first, an analysis of the general meaning-relation; and second, an analysis of the particular meaning-entity involved. But it is not obvious how to square this expectation with any of the reductive proposals discussed above, since they do not appear to be divisible into components of this sort. In light of this tension, it would seem that at least one of the following theses must be defended: (1) that meaning-facts do not in fact have the just-mentioned apparent structure; or (2) that their reduction does not in fact require analyses of their constituents; or (3) that some form of non-semantic ‘grounding’ of them, weaker than reduction, is the most that can be expected; or (4) that certain analysantia of the sort considered above (e.g. that such-and-such sentences containing w are accepted underived) can in fact be factored into one part that analyses the meaning-relation and another that analyses a particular meaning-entity. In the next chapter I will be defending a version of the fourth strategy: meaning-entities are identified with basic acceptance properties, and the meaning-relation is reduced to the relation of exemplification between words and those properties.
(IV) According to Quine's thesis of radical indeterminacy there are few foreign expressions whose correct translations into English are grounded in objective facts. But even if Quine is (p.25) 99 per cent wrong (for the reasons mentioned in section 2), it may be that the correct translations of some expressions are nonetheless indeterminate. For example, Brandom77 and Field78 have argued that a language's words for the two square-roots of minus one may be used so similarly that there will be no properties that might constitute the distinctive meaning of one of them (and thereby constitute its translation into “i” rather than “2i”) that are not also possessed by the other. But any such prospect is a threat to semantic reductionism. For it is not easy to see how that doctrine, in any of its specific forms, can be reconciled with the concession that there is even a single term whose meaning is not constituted by non-semantic facts.
What is plain from the above review is that research into the nature of meaning must confront a formidable cluster of interlocking problems. I would suggest, however, that if we adopt the neo-Wittgensteinian use-theoretic perspective outlined in section 10, a coherent network of plausible solutions to them may be found. My hope for the following chapters is to vindicate this conjecture.79
(2) Chomsky, N. (1975), “Quine's Empirical Assumptions” in Davidson and Hintikka (eds.), Words and Objections; idem (1987), “Reply to Review Discussion of Knowledge of Language”, Mind and Language 2: 178–97.
(3) Horwich, P. G. (1998), Meaning, chap. 9.
(5) Donnellan, K. (1989), “Belief and the Identity of Reference” in French, Uehling, and Wettstein (eds.), Midwest Studies in Philosophy, vol. 13, Contemporary Perspectives in the Philosophy of Language II 275–88.
(8) Soames, S. (2002), Beyond Rigidity: The Unfinished Agenda of Naming and Necessity.
(9) Frege, G. (1952), “On Sense and Reference”, Translations from the Philosophical Writings of Gottlob Frege, P. Geach and M. Black (eds.). For a defence of Frege's argument, see S. Schiffer (2004), The Things We Mean.
(10) Davidson, D. (1984), Truth and Interpretation. There is some controversy as to whether Davidson himself advocates an elimination of meaning in favour of truth conditions, or an analysis of meaning in terms of truth conditions. For reasons given in chap. 8 fn. 5, I myself favour the second of these interpretations.
(18) I will use expressions in capital letters to name meanings. Thus, “DOG” names the meaning of the English word “dog”; “I AM HUNGRY” names the literal English meaning of “I am hungry”, etc.
(21) See his Word and Object.
(24) Loar, B. (1981), Mind and Meaning.
(27) Harman, G. (1982), “Conceptual Role Semantics”, Notre Dame Journal of Formal Logic 28: 252–6; idem (1987), “(Non-solipsistic) Conceptual Role Semantics”, in E. Lepore (ed.), New Directions in Semantics.
(29) See chap. 8 for elaboration of the difficulties confronting Davidson's account of compositionality.
(30) Fodor, J., and Lepore, E. (1991), “Why Meaning (Probably) Isn't Conceptual Role”, Mind and Language 6: 328–43; idem (1996), “The Pet Fish and the Red Herring: Why Concepts Arn't Prototypes”, Cognition 58(2): 243–76.
(31) See his Wittgenstein: On Rules and Private Language.
(45) See his Psychosemantics.
(51) See his Psychosemantics.
(58) See his Wittgenstein: On Rules and Private Language.
(61) See his Philosophical Investigations.
(64) Block, N. (1986), “Advertisment for a Semantics for Psychology”, Midwest Studies in Philosophy 10, (eds.) P. French, T. Uehling, and H. Wettstein.
(65) Harman, G. (1982), “Conceptual Role Semantics”, Notre Dame Journal of Formal Logic 28: 252–6; idem (1987), “(Non-solipsistic) Conceptual Role Semantics”, in E. Lepore (ed.), New Directions in Semantics.
(67) Wright, C. (2001), Rails to Infinity.
(69) Block, N. (1994–5), “An Argument for Holism”, Proceedings of the Aristotelian Society 95: 151–69.
(74) Sperber, D., and Wilson, D., Relevance.
(78) Field, H. (1998), ‘Some Thoughts on Radical Indeterminacy’, The Monist 81: 253–73.
(79) I am grateful to Ned Block, Tim Crane, Michael Devitt, Stephen Schiffer, and Barry C. Smith for their comments on a draft of this chapter.