Some things are difficult, like climbing Mt. Everest or K2. Others are impossible, like climbing an overhanging sand dune. Metaphysics, all will agree, is at least difficult, indeed very difficult. But is it really impossible? Are metaphysicians barking up the wrong sand dune? That depends on how successful are the arguments that have been brought against metaphysics. In our century, anti-metaphysical arguments tend to begin by considering language, meaning and the relations between language and the world. Because our knowledge of the world is made possible only by language, we are told, the limits on our ability to use language meaningfully to talk about the world amount to limits on our knowledge, and indeed on the world itself insofar as we can meaningfully speak of it. "The limits of my language mean the limits of my world," according to Wittgenstein (1889-1951). We have access to the world, if at all, only through the veil of language.

This viewpoint resembles a story told by many philosophers ever since the 17th century. Because our knowledge of the world is made possible only through our immediate sense experience -- our sensory impressions or ideas -- our knowledge of the world is limited by what we can infer from the properties of our ideas. We have direct access to our ideas or experience, but not to the world itself. Unless we can somehow pierce the veil of ideas, the metaphysician's desire to talk about how things really are is doomed to failure. A number of twentieth-century philosophers replace ideas in this scheme with language as the interface between us and the world. Let us see some of the ways they have done so.

The most famous 20th-century anti-metaphysical argument -- or the most infamous -- may well be one advanced by the logical positivists. They held a theory of meaning according to which a sentence or statement is cognitively meaningful only if it is in one of two categories. The statement must be either analytically true (true, that is, "by definition," as supposedly are `Bachelors are unmarried males' and `2 + 2 = 4'), or else empirically verifiable or testable. The positivists arrived at this meaning criterion partly to give an account of meaning and method in science, partly to bring precision and progress into philosophy, if not to eliminate philosophy entirely. They deemed much philosophy prone to obscurantism if not outright nonsense. Metaphysics especially offended in this regard, given what struck them as the bombastic pseudo-profundity of its claims, or at least the claims of transcendental metaphysics about some supersensible reality. Positivists were fond of ridiculing statements like "The Absolute is beyond time" and "The Nothing itself nothings."

Such statements clearly are not analytic. Hence they must be synthetic, synthetic statements being defined as those that are not analytic. But the suspect metaphysical statements are not necessary truths either. The reason is that according to the positivists all necessary truths are analytic; there are no synthetic necessary truths. Nor are the suspect statements testable by anything we could possibly experience. It follows that they are neither analytic nor testable; they are cognitively meaningless, hence neither true nor false. In the words of Hume (1711-1776), a philosophical ancestor of the positivists, metaphysics can therefore "contain nothing but sophistry and illusion," and we should "commit it then to the flames."

Metaphysics was not the only victim of the positivist meaning criterion. Value judgments were also ruled cognitively meaningless and thus neither true nor false. Consider even so plain a judgment as that torturing infants is monstrously wrong. This judgment is not true simply by virtue of the meanings of the terms involved, hence not true by definition and not analytic. Nor is it empirically testable. Therefore, it must be counted not true and not false. It merely expresses an emotion -- the speaker's disapproval of such torture -- together with an imperative or exhortation to others to disapprove of this as well. There are no moral facts, no objective values, only our emotions projected onto the value-neutral real world.

Logical positivism is now almost universally thought to have failed, and not only as an argument against objective values and the possibility of metaphysics. Further study of meaning and method in science, and of the actual history of science, has shown that the meaning criterion does not yield a good account even on its home ground. Scientific statements often are not accepted or rejected on the basis of some narrow test-procedure, but in light of more global considerations (such as coherence with other science, comparative simplicity, comprehensive explanatory power, a progressive history of problem-solving, and so on). Furthermore, it now appears that truth in logic and mathematics cannot be a matter simply of "definition." One reason is that the mathematical truths have been shown to outrun what can be derived from any axioms the positivists would count as analytic. Finally, there is considerable doubt as to whether the positivists' theory of meaning satisfies its own standard. Theories that fail to satisfy the standards they lay down for all theories are said to be self-referentially inconsistent.

All in all, positivism asserts a number of sharp contrasts, between verifiable and unverifiable, mathematical and factual, analytic and synthetic, necessary and contingent, observation and theory, the given and the conjectural, fact and value, science and metaphysics. These contrasts have all been rejected in much recent philosophy. They are rejected even by some who wish to make philosophy more scientific, or who remain sympathetic to the spirit of positivism.

One historian of 20th-century philosophy concludes that "logical positivism ... is dead, or as dead as any philosophical movement ever becomes." As to positivism's legacy, some of it may be unfortunate -- perhaps the persistent belief among some that philosophy is not about the world but about language, or that philosophers should always use the technical tools of mathematical logic, or that like scientists they should make no value judgments. But in the rush to bury positivism (or to heap abuse on the grave), too few reflect on the extent to which it helped render obscurantist varieties of metaphysics suspect if not meaningless, and fewer still imitate the positivists' generally high standards of argument, their intellectual honesty, or their rejection of pretentious generalizations in favor of careful attention to crucial detail.

Ironically, a number of arguments against positivism, some believe, may be far more devastating to metaphysics than the meaning criterion ever was. These are arguments against the sharp contrasts entailed by positivist doctrine. One such contrast is that between the contingent and the necessary. If metaphysics must consist of principles that are necessary, and if the necessary/contingent distinction must be rejected and with it all talk of necessary truth, metaphysics is impossible. So too is it impossible if and to the extent that metaphysics is committed to some notion of the given, or to a sharp observation/theory distinction, or to the analytic/synthetic distinction.

How might a metaphysician be committed to the analytic/synthetic distinction? One way is by construing method in metaphysics as largely a matter of conceptual analysis. For example, in dealing with the relations between persons (or minds) and bodies, the metaphysician might start by analyzing various mental concepts. Just as on analysis the concept "bachelor" is found to mean "unmarried male," a mental concept like "knows French" might be found to mean "utters French sentences on appropriate occasions." Other analyses, like this one, would be constructed out of some relatively unproblematic elements, perhaps out of terms familiar to everyone (common-sense concepts). Or an analysis might be based on terms-of-art invented on the spot to express some other fundamental matter that only the metaphysician has had the wit to see. In either case, the analysis would be presented not as empirically true but as true by virtue of the meanings of the very terms involved, hence as analytic. Thus the analysis would yield a priori knowledge of some necessary truth about minds, about experience, about understanding and its structures, even about the world so far as it is experienceable or intelligible.

But what is supposed to be wrong with the analytic/synthetic distinction? After all, there surely is a big difference between `Bachelors are unmarried males' and, say, `Phosphorus melts at 44C'. We can easily imagine experiments that would force us to give up the latter, but what could ever make us change our minds about bachelors' being unmarried males? That's just the way we use the term `bachelor', a convention.

True enough. But the classical analytic/synthetic distinction does not merely recognize this sort of difference between such sentences. It goes on to claim that `Bachelors are unmarried males' is made true by our convention, made true not by the world but simply by our use of words, true by virtue of their meaning alone. According to the classical analytic/synthetic distinction, the difference between the two kinds of sentence is a matter of one's being made true by our meanings, the other by the world. It is this doctrine of truth by convention that is the target of arguments against the analytic/synthetic distinction, not the obvious fact that there is a considerable difference between the two kinds of sentence.

The most powerful attack on analyticity is that of Quine. Quine argued that every attempt to provide an adequate explanation of truth-by-meaning had failed. Thus suppose I try to explain the truth of `Bachelors are unmarried males' by pointing out that the key terms mean the same thing. Obviously I then owe you an explanation of "means-the-same," or synonymy, and of how it can generate truth. Suppose I adopt an explanation of synonymy, as many philosophers have, according to which two terms are synonymous just in case they are necessarily true of the same things. The trouble, Quine argues, is that `necessarily true' is as much in need of explanation as is `analytic'.

Worse, what we normally call necessarily true are simply those matters in whose truth we have the greatest confidence and whose role in our world-view is central; so too for what we call conventions. Notoriously, such matters can turn out to be false after all. Examples have included Euclidean geometry, set theory, and the laws of elementary logic. The first, declared necessarily true by Kant, proved false of our world in light of Einstein's theory of gravitation. The second, some have speculated, may likewise need to be revised in future physics, by replacing one of the axioms of what is called Cantorian set theory with an alternative. And the third, according to some physicists and philosophers, must give way to a "quantum logic" in order to resolve certain paradoxes about the quantum world.

In each case -- and in many others -- what once seemed true of any possible world, hence true independently of how this one in particular is, now seems no different from any other of our "high-level" theories about this world. What makes them true (or false) is the world, Quine thinks, not our conventions or meanings or some other allegedly a priori affair. This is the point of his slogan, "Nothing is immune to revision."

Furthermore, many things we may take to be necessarily true actually started out as generalizations about our experience of the world. `Nothing can look reddish green' evidently is an example. Things can look reddish blue (purple) or bluish green (turquoise), but never reddish green. A number of philosophers do take it to be a necessary truth that nothing can look reddish green. But research in color vision strongly suggests that our acceptance of this truism is based on quite contingent features of our visual processing. Had we evolved differently, things could have sometimes looked reddish green, just as they now sometimes look reddish blue or bluish green. Indeed, in a striking recent experiment subjects do report seeing reddish green. What seemed indubitably a necessary truth to a number of philosophers now appears not even to be true.

None of this proves conclusively that the analytic/synthetic and necessary/contingent distinctions are mistakes. Conclusive proof of anything is beyond the reach of philosophy, and also of science and much if not all else. But more than a few philosophers have been persuaded by these and related considerations to reject any such contrasts.

A further consideration, also advanced by Quine, is that the very notion of meaning involved in the doctrine of truth-by-meaning is itself doomed. According to Quine, meaning in the traditional sense, and sameness-of-meaning, are not empirically testable. The reason is that there can be a difference in the meaning of a couple of utterances without any appropriate difference in either the behavior or the brain or the external circumstances of the speaker or listener. Thus the meaning is not determined by such empirically testable matters as behavior, brain states, or external conditions. This "indeterminacy of meaning," when taken together with Quine's argument against truth-by-convention, would spell the end of all talk of truth-by-meaning, analyticity, necessity, and a priori knowledge, together with everything they imply. It would also spell the end of metaphysics if and to the extent that metaphysics is committed to any of these.

Some philosophers believe that Quine's thesis of the indeterminacy of meaning would spell the end of metaphysics in another way: it would undercut any alleged correspondence between words and the world that would entitle metaphysicians (or anyone else) to regard their utterances as determinately true of the world. Let us see how this might happen.

In a famous example, Quine invites us to consider the utterance of `Gavagai!' by the speaker of an alien language at the very moment a rabbit bounds across the trail in front of us. We are naturally inclined to translate the utterance as `Rabbit!' or perhaps `There goes a rabbit!'. But the alien might be referring instead to only a part of the rabbit (an "undetached rabbit part"), or to a condition or stage of the rabbit (a "rabbit stage"), or even to a part of the spatially and temporally scattered object composed of all rabbits everywhere everywhen (a part of the "rabbit fusion"). All these, and many more, are present equally. Pointing at the rabbit is simultaneously to point at them all. If we point and ask the alien, "Gavagai?" the response in each case will be "Yes," or what we translate as "Yes." There will be no difference in the alien's behavior that could determine what the referent is (rabbit, undetached rabbit part, rabbit stage, or whatever). Nor will there be any difference in the alien's brain states that could be correlated with what is being referred to: even if there were a difference on different occasions of seeing a rabbit, the difference could as well be correlated with, say, an undetached rabbit part or a rabbit stage, as with the rabbit.

Since there is nothing else that could determine the meaning, Quine contends, meaning is indeterminate and reference inscrutable. And since all truth is supposedly determined by matters that are empirically accessible -- indeed for Quine all truth is determined more narrowly still, by truth at the level of physics -- it follows that there is no truth of the matter as to meaning and reference, hence none as to what the alien is talking about.

Nor are you and I in any better position than the alien. Precisely the same considerations apply to us. Hence there is no fact of the matter as to what we are talking about. If so, our words are cut off from the world. We can hardly regard them as corresponding determinately to anything in the world. Worse, once we deny determinate reference and aboutness, and with them any objective link between our language and independent reality, reality itself can begin to look problematic. If it is only relative to some interpretation of my words that it is true that there goes a rabbit, then whether there goes a rabbit seems itself to be relative to some way of interpreting or understanding. The idea of a mind-independent world begins to flicker and fade.

One widespread response to all this is to insist, "I just know what I mean by `rabbit'! It's absurd to think otherwise, and there must be something wrong with any argument that suggests I don't!" This response seems to presuppose that I have a priori knowledge of meaning, knowledge that is guaranteed from within consciousness. I cannot be mistaken as to what my words mean, because the meaning is self-evidently given or present to consciousness; what my words are about is a matter of what I intend them to be about, and I cannot be wrong as to what my intentions are. The trouble with this meaning-rationalism, as it has been called, is partly that it presupposes a notion rejected by Quine and many others, that of a priori knowledge. Likewise, Quine rejects any notion of a self-evident given; again, nothing is immune to revision.

There is another problem with insisting that I just know my meaning. It seems to block any account of meaning that can be smoothly integrated with what we know of human beings from the sciences. If we are naturally evolved creatures, as biology assures us, it would appear that language and meaning too have naturalistic explanations. Now suppose we insisted that meaning is determined solely by private, self-validating acts of intending that are beyond the reach of public test. This would imply that there can be a difference in meaning even when there is no difference accessible to empirical science. It would follow that there can be no naturalistic account of meaning. Indeed, according to some philosophers -- Brentano (1838-1917), for example -- that is the very point of meaning-rationalism: human beings, via language and meaning, transcend any account that would reduce intentionality to natural or causal relations of things. As Quine himself remarks, Brentano's thesis of the irreducibility of intentionality is of a piece with the thesis of the indeterminacy of meaning. It's just that they draw opposite conclusions from the irreducibility: Quine, that there is no fact of the matter as to meaning; Brentano, that there can be no naturalistic science of human beings.

Perhaps neither conclusion quite follows. Perhaps there are naturalistic theories of meaning according to which meaning is irreducible to behavior, intentions, brain states, the speaker's present circumstances, or the like, and yet there is a fact of the matter as regards meaning. It turns out there are such theories, and we consider a couple in the next chapter. They represent important alternatives to the bleak choice between Quine and Brentano, between no fact of the matter as to what we are talking about and no notion of meaning that can be integrated with what we know of human beings from science. If some such theory proves correct, our words are not cut off from the world in the way imagined, and metaphysics is safe at least from this sort of threat.

Meanwhile, notice that even if we conclude that there is a fact of the matter as to meaning, this does not by itself restore analyticity, necessity, or the a priori. For the theory of meaning in light of which we conclude there is a fact of the matter could turn out to be inhospitable to truth-by-meaning, necessity, a priori knowledge, and the like.

There are further sources of skepticism about reference and about other relations between words and the world (word-world relations). One is structuralism, according to which (among other things)meaning is entirely a matter of word-word relations internal to a language. In particular, the meaning of a word or phrase is identical with the set of its "syntagmatic" and "paradigmatic" relations, as they are called. Two words are syntagmatically related when they are arranged in sequence to make a well-formed string (a "syntagm" or sentence), as can `cats' and `purr'. Two words are paradigmatically related when they are associated (in memory) on the basis of some similarity or connection between the concepts they signify. For example, `meow' is associated with `purr', hence paradigmatically related to it. In our language, the two words belong to an associative family evoked by `cat'. The meaning of a word is exhausted by its syntagmatic and paradigmatic relations.

According to Saussure (1857-1913), who is widely regarded as the founder of structuralism, it follows that "Language is a system of interdependent terms in which the value of each term results from the simultaneous presence of the others." Change a word's position in this system or structure of relationships with other words, and you change its meaning. The distinctive semantic contribution of each word is determined by how it differs from other words as regards the relationships they enter into in the language. "Signs function ... not through their intrinsic value but through their relative position.... In language there are only differences."

Some are attracted to structuralism by what they believe it implies about the world. They believe it implies that like language, the world is to be construed as "made up of relationships rather than things." An item in the world, so far as we can talk about it at all, is defined not by what it is intrinsically or independently of other items, and not by its "essence" (if any), but by its relations to all the other items -- that is, by its relationship in a structure. The significance or sense of an item in the world is determined by the meaning of our word for it, hence by the linguistic relationships that constitute the meaning. On this view, a language creates its own reality: "Language ... allows no single, unitary appeal to a `reality' beyond itself. In the end, it constitutes its own reality." Each society, via its language, covertly projects some such system or structure of relationships onto the world. There is no language-independent reality that constrains such projection (a view we encounter further in Chapter 4, when we discuss what a thing is). "Reality is carved up in various ways according to the ... patterns of sameness and difference which various languages provide." Metaphysics, to the extent it presupposes otherwise, is fundamentally misguided.

These conclusions form a version of conceptual relativism, according to which a thing does not exist in itself but only relative to a conceptual scheme. Such relativism may go beyond what Saussure himself was prepared to assert. Rather than inquire into the relation between language and reality, he seems to defer the question, or to treat it as not part of the autonomous discipline or theory of language he has in mind. Within that discipline, questions about the external conditions and causes of the use of linguistic signs are to be put aside. Thus Saussure can say that his "definition of language presupposes the exclusion of everything that is outside its ... system."

Nevertheless, this official silence or agnosticism about word-world relations is nearly as odd in a theory of meaning as their outright rejection. Not only common sense but several theories of meaning hold that words do on occasion determinately refer to language-independent items in the world, and that a word's referential function is an important dimension of its meaning. Thus the most striking feature of structuralist accounts of meaning is their omission of reference. Their notion of a linguistic sign is not one that relates a word and a thing, but rather a sound and a concept. The sign is made up of a signifier (a sound) and a signified. What is signified is not some language-independent item (a rabbit, say) but a concept or idea (someone's rabbit-concept). The very identity of a concept or idea is determined by relations internal to the language: "Without language, thought is a vague uncharted nebula. There are no pre-existing ideas, and nothing is distinct before the appearance of language." Whether the whole sign -- the signifier-plus-signified-concept -- determinately refers to some language-independent item in the world is left open by some structuralists and denied by others.

Either way, a language is treated as an entirely autonomous system, to be explained without reference to anything beyond itself. In this regard it is like chess, as Saussure is fond of saying: "Just as the game of chess is entirely in the combination of the different chess pieces, language is characterized as a system based entirely on the opposition of its concrete units." The notion of check is entirely defined within the rules of chess; it refers to nothing beyond. We could change the rules so that the king may be taken (as in "blitz chess"); nothing external to the game constrains us. So too for language. One structuralist goes so far as to say, "The word `dog' exists, and functions within the structure of the English Language, without reference to any four-legged barking creature's real existence." If so, our words are cut off from the world; there can be no hope of a notion of their being true of independently existing things. At most they are true relative to some society's way of projecting a system of relationships onto the world.

What lies behind this rejection of reference? Structuralists, even those who do not reject but are merely agnostic about reference, emphasize the "arbitrariness" of signifiers and signs. There is first of all arbitrariness in the association between signifier and signified. We could have used the sound of `dog' to signify our concept of cats, the sound of `cats' to signify our concept of clowns, and so on. Second, there is arbitrariness in the association (if any) between the whole sign and the real thing. "There exists no necessary `fitness' in the link between ... `tree' ... and the actual physical tree growing in the earth. The word `tree', in short, has no `natural' or `tree-like' qualities." Hence there is no intrinsic connection between the sign and what it means. It follows, we are told, that "the sign ... must be defined as a relational entity, in its relation to other signs." For some structuralists, at least, the arbitrariness means there can be no reference to any four-legged barking creature's real existence.

But how does it follow from the arbitrariness that the sign is related only to other signs? How does it follow that there can be no reference? One way it would follow is on the assumption that a word can refer only by somehow resembling or picturing its referent. The idea is that the word `tree' can refer to trees only if it has tree-like qualities, and `dog' to dogs, one presumes, only if it has four legs and barks, or at least pictures or mirrors these. But this amounts to a naive resemblance-theory or picture-theory of reference, one which we have no reason to accept and plenty of reason to reject. There are alternative theories of reference according to which `dog' refers to dogs without barking, and indeed without resembling or picturing or mirroring dogs in any way. We encounter two such theories in the next chapter, but we needn't accept them in order to recognize that we can use a word to refer to a thing even if we could have used it to refer to something else, and even if the word neither resembles nor pictures nor mirrors the thing. Indeed, some structuralists themselves would reject any such naive account of reference. Rather than argue there can be no determinate word-world relation of reference, they remain officially silent on the issue.

What happens to the issue in deconstruction, as it is called, particularly in that of Jacques Derrida? Derridean deconstructionists take over much of structuralism while questioning the rest. In particular, they question what they see as structuralism's lingering commitment to "independent meanings," in its talk of pre-existing concepts as what signifiers signify. According to structuralists -- certainly according to Saussure -- the structure or system of concepts has a kind of transcendent ideality that exists apart from the signifiers that happen to signify the concepts. According to Derrida, Saussure "accedes to the traditional demand for ... a `transcendental signified'." But there can be no such signified, no independent meaning or system of concepts. Rather, the signified concepts, so far from having a "commanding force," are themselves created, temporarily stabilized, and changed by the ways we happen to use words. "Language in its `creative' uses outruns what might be accounted for in terms of purely `structured' or pre-existent meaning." Derrida excels in detailing how a seemingly necessary way of conceptualizing things corresponds to no fixed position in some independently given system of concepts. Instead, it is a quite contingent result of our habitually talking in a certain way. We could as well talk in another way, and often we should. In effect, Derrida agrees with Saussure in assuming linguistic meaning is "differential," in the sense of being determined by the relations that hold between the words in the language. But he thinks Saussure did not quite see the full implications of this move, which Derrida happily exploits to the hilt by treating the concepts themselves as differential.

Again reference drops out. Again there is silence on or outright rejection of determinate reference to some language-independent world. But now meaning is completely indeterminate, perhaps even more so than for Quine. There is nothing beyond our words -- nothing beyond the text, no "foundation" -- that could determine their meaning: not some "dictating referent that stands independent of the referring agencies of discourse," not the author's or speaker's intentions, not behavior, not brain states, and above all not some intrinsic fit or transparency between the signifier and the signified. This is the point of Derrida's famous -- or notorious -- slogan: "There is nothing beyond the text." There is no foundation in light of which to determine the meaning of a word. The very effort to resuscitate determinate meaning is condemned as nostalgia for the lost foundation.

But what if determinate meaning is not a matter of foundations? What if determinate meaning need not be transparent or present or given to the speaking subject or to consciousness? It is true that the tradition against which Derrida is reacting presupposes some such "metaphysics of presence," according to which the meaning of an utterance is something present to the consciousness of the speaker, so that the speaker cannot be mistaken about the meaning. But it hardly follows from the rejection of such foundationalism that there can be no determinate meanings and no determinate reference to language-independent items in the world. What determines such matters might be something other than presence or any foundational affair, as we see in the next chapter.

Occasionally, deconstructionists seem to reason that because any sentence can function in many different contexts, in which it will have different meanings, it can therefore have no determinate meaning. But this presupposes that a sentence would have a determinate meaning only if it expressed that same meaning in all contexts, or perhaps only if there were some one favored context in which it had that meaning. This is like saying that the word `red' would have a determinate meaning only if it expressed that very meaning in all contexts. Yet obviously in some contexts it denotes the color, in others a member of the Communist Party, in still others a wine, and so on. Plainly, it is far more plausible to say that in each context of use `red' has a determinate meaning, even though there is no one meaning it has in all contexts. If you and I are discussing wines and you say, "Try the '73 red," then in this context `red' certainly seems to denote a wine, and determinately so.

There is a similarity between what motivates the foregoing argument from multiple contexts against determinate meanings and what motivates the structuralist's uneasiness about reference. The idea in both cases is that there would have to be some non-arbitrary, necessary connection between a word and what it determinately means. In the present case, the alleged necessary connection is held to be destroyed by there being no one favored context in which the word has a meaning. A word has no "essence," hence none that ties the word to the extra-linguistic item or items it might denote, nor do the extra-linguistic items themselves have an essence that could bind certain words onto them. The resulting arbitrariness is supposed to imply that there can be no determinate meanings. But why should we suppose that the only way words can correlate with or map onto meanings or onto the world is via some such necessary or essential tie? Again, there are plausible theories of meaning that require no such tie, as we see in the next chapter.

Actually there may be a reason why structuralists and deconstructionists alike suppose there must be some such necessary connection between a word and what it means. They may have been influenced by meaning-rationalism, as we called it earlier, a view that recurs in much continental philosophy (but not only there). They may share the meaning-rationalist's presupposition that if there are determinate meanings, the subject must have a priori knowledge of them. They may believe further that such knowledge could only be gained by way of some appropriate necessary connection between the word and what it means; otherwise the word's meaning could not be suitably present to consciousness, as it must be if the subject is to have a priori access to the meaning. Since there is no such necessary connection, hence no such presence, there can be no a priori knowledge of what a word means, hence no determinate meaning at all.

This line of thought is no stronger than the meaning-rationalism that underlies it. Perhaps there is an alternative account, as we see in the next chapter, according to which meaning is determinate but not known a priori, and the determinacy is not at all a matter of presence. Instead, we can sometimes be mistaken as to what we mean, hence mistaken as to what we are thinking about.