Truth and Meaning

Truth and Meaning

What have truth and meaning got to do with each other? A dominant view has it that the two are deeply connected—specifically, meanings are truth conditions.[1] The view comes in several varieties, but the central thought is that the meaning of a sentence consists of the state of affairs that would make it true. Alternatively, meaning is given by the conditions of the world that would render the corresponding sentence true. Thus the meaning of “snow is white” can be identified with the condition of snow being white, i.e. the state of affairs consisting of snow being white, i.e. snow being white. Truth conditions can hold or fail to hold (as with “snow is black”), so that false sentences can still be meaningful; but when they do hold they coincide with facts (otherwise they are merely possible states of affairs). Meanings are accordingly to be identified (generally) with non-psychological states of the world—states that could exist without the existence of minds. There were states of affairs (truth conditions) before there were minds and language. Language hooks onto these antecedent states of affairs and thereby becomes meaningful. Meanings are thus an extra-mental matter—combinations of objects and properties, according to a standard position (sets of possible worlds in one version).

There is a fundamental difficulty with this doctrine (as well as several non-fundamental difficulties): meanings are always for someone but truth conditions are not necessarily for anyone. No sentence is meaningful except in relation to a speaker or hearer; there is no such thing as a meaningful sentence that no one understands or could understand. Some sentences are difficult to understand, but no sentence has a meaning that transcends human understanding, i.e. contains concepts that no one possesses. A sentence to be meaningful must be meaningful to someone. Meanings are inherently graspable things; they are essentially human. They are objects of apprehension. But the same is not true of truth conditions: being objective aspects of the world, there is no guarantee that they are grasped by humans. They may not be humanly graspable at all. Certainly many states of affairs eluded human knowledge for a long time, and no doubt there are many that still do: they existed but were not apprehended by us. None of these constituted meanings. They can only enter the realm of meaning if they are grasped—and they may not be. Some speakers may not even grasp the state of affairs of snow being white, so the sentence “snow is white” will not have a graspable truth condition for them; the sentence is not meaningful tothem. But then meanings can’t be truth conditions: for meanings are essentially graspable while truth conditions are not. Truth conditions are the wrong kind of thing to constitute meanings—too divorced from a speaker’s understanding. Meanings are speaker-relative, but truth conditions are speaker-independent. Intuitively put, meanings are necessarily things that speakers know, but truth conditions are not necessarily things that speakers know. This is why it is possible to be a skeptic about truth conditions but not about meanings: maybe we don’t know what states of affairs there are in the world, but we surely know what our words mean. Then how could the latter be the former? In short, meanings are psychological, but truth conditions are not.

Here we see a sharp contrast between truth conditions and every other factor that meanings have been identified with. Verification conditions are not speaker-independent in the way truth conditions are, since they depend on attributes of the speaker, viz. his powers of knowledge acquisition. A sentence is always verifiable (or not) for a speaker. Similarly for the use of a sentence: this too is an attribute of the speaker—his ability to perform acts of speech with the sentence. Equally in the case of image theories: images are items existing in people’s minds not realities that can exist independently of human consciousness. Ditto for intentions and beliefs. It is generally assumed that whatever meaning is it had better relate to the speaker’s mind or behavior or brain, but truth conditions theories locate meaning in the extra-mental world, where it can possibly transcend human knowledge. But even when the states of affairs are known, they are logically of the wrong nature to constitute meaning, since meanings are meanings for someone but truth conditions are not truth conditions for anyone. To be a state of affairs is not to be a state of affairs to someone, but meanings are always meanings to someone. Meanings are always meaningful to someone, but states of affairs are not always meaningful to anyone—they are not intrinsicallyaccessible to the mind. Meanings are things that are communicable or usable in thought, but states of affairs are not subject to this constraint; they exist independently of the human subject. They can’t be meanings because they are too objective.

It might be replied that meanings should not be identified with truth conditions themselves but with mental representations of truth conditions. Truth conditions are admittedly too mind-independent to constitute meaning, but the same cannot be said of mental representations of them. Two questions arise about this reply. First, what is the nature of these mental representations? Perhaps they are senses construed as modes of presentation: this is obscure for many meanings, but in any case the suggestion threatens to make truth conditions redundant in the theory of meaning—why not just make do with the mental representations themselves? Second, the question must arise as to what constitutes the meaning of these mental representations: won’t this lead to the same problem as before? If their meaning is constituted by truth conditions, then we have chosen the wrong kind of thing for meaning to be; but if not, we have moved to another theory of meaning altogether. This is very clear if the mental representation is a sentence in the language of thought: it either has a truth conditions type of meaning or it does not—the former is untenable, while the latter amounts to abandonment. No, the theory must be that meanings aretruth conditions; but then we are faced with the objectivity problem. We are locating meaning too far “outside the head”, making it hostage to ignorance and skepticism. To repeat: states of affairs are (generally) extra-mental in nature and bear no necessary relation to a speaker’s understanding, while meanings are necessarily objects of understanding. Identifying the two cannot then be correct. What a speaker means by an utterance is not what obtains in the world when that utterance is true. The two are subject to quite different epistemic constraints.

What then is meaning? I have no pat answer: none of the alternative theories strikes me as adequate for one reason or another. But that’s okay: maybe we just don’t know what meaning is (lord knows it has been a problem). Wittgenstein repudiated the truth conditions theory of the Tractatus in the Investigations, but he put no new theory in its place; and he was Wittgenstein. Certainly the considerations adduced in the Investigations are far more psychological and speaker-centered than the abstract truth conditions theory of the Tractatus, but they don’t add up to a nice neat theory of what meaning really us. Wittgenstein saw that human meaning cannot emerge from objective states of affairs, but he also rejected both image theories and simple behaviorist theories. His solution, in effect, was to question the hunt for a theory. I prefer to say simply that I don’t know. What I would accept is that truth conditions are connected to meaning in some way: we do often grasp the truth conditions of our sentences. We know that “snow is white” is true if and only if snow is white, and this fact is somehow connected to the meaning we grasp. So we can say that meaning somehow involves or leads to or points to truth conditions, but we can’t say how. Meaning serves to bring truth conditions into view, rendering them visible to us: we see that the sentence is true under certain conditions. Meaning makes truth conditions apparent to us, but not because it is identical to truth conditions. If meaning were use (whatever exactly that means) we could say that the use of a sentence renders its truth conditions manifest to us: we see the relevant state of affairs in the use. But this is horribly obscure and leaves us none the wiser. Still, one can appreciate how the truth conditions theory gained its popularity: truth conditions are connected to meaning somehow and in certain cases (but not when they transcend human knowledge); we just can’t see how. Indeed, it is problematic how this can be so, given that truth conditions are external to the human subject: how can we apprehend objective truth conditions in virtue of any fact about our inner nature? It’s not as if we literally perceive them with our senses, or snub our toe against them. So we not only don’t know what meaning is; we also don’t know how meaning brings truth conditions into the picture, as it apparently does. But if I am right, we at least know that it isn’t truth conditions—we have some negative knowledge. And of course our not knowing what meaning is explains a lot about the history of the subject (all that floundering and fighting).

I began by asking what truth and meaning have to do with each other. I just argued that truth conditions cannot constitute meaning as a matter of deep principle, putting aside questions of extensionality and so on. But there is another line of attack that is far more obvious and well trodden, namely that many meaningful sentences are not true at all, such as imperatives. The limitation of this style of argument is that it is open to ingenious response: it might be contended that imperatives are really disguised indicatives, or that the spirit of the truth conditions theory can be preserved by switching to obedience conditions. However, in the light of what I argued above these maneuvers look suspiciously ad hoc: for the thrust of the present objection is simply that the concept of truth plays no obvious role in the theory of the meaning of non-indicative sentences. If a language consisted solely of imperatives, the truth conditions theory would look like a non-starter. What does the meaning of thesesentences have to do with whether states of affairs obtain? They command actions; they don’t describe states of affairs. Or suppose a language was totally expressive of emotions with no world-directed assertions in it at all: wouldn’t it contain meaningful sentences with nary a hint of truth conditions? There is just nothing in the concept of meaning as such to entail that truth conditions are necessarily involved. So long as there is communication there is meaning, but communication can take many forms. If truth conditions constitute meaning for indicative sentences, that is a special case by no means generalizable to other sentence types. But even in that case there is a principled problem about identifying meaning with truth conditions—the objectivity problem. It is noteworthy that such a theory only gained traction relatively recently with the work of Frege, early Wittgenstein, Carnap, Tarski,[2] and Davidson: before that it had no defenders. Earlier theorists took a far more psychological approach, preferring images, ideas, feelings, thoughts, or behavior. Did they tacitly sense that truth conditions place meaning too far beyond the human subject? Did they realize that states of the world are not intrinsically meaningful or meaning conferring? Meaning is always meaning to someone, but the world doesn’t point to the human subject in this way. Meaning refers itself to us, as it were, but the world makes no such reference—so how could it be meaning? How can facts be meanings given that facts bear no essential relation to the beings that grasp meanings? The fact that snow is white is not a fact for us, but the meaning of “snow is white” is a meaning for us. Identifying the two looks like a category mistake. Alternately put, possible worlds are not worlds for subjects, but meanings are intrinsically subject-directed. As Wittgenstein would say, meanings are part of our form of life, but objective states of affairs are external to that (they constitute the form of the world). Meaning can certainly be about the objective world, but it isn’t the same thing as that world.[3]

Here is another way to put the same basic point. Meaning essentially attaches to symbols: no symbols, no meaning. If I say, “’Snow is white’ means that snow is white”, this is by no means equivalent to any statement that omits reference to the symbols “snow is white”. But notoriously, if I say, “’Snow is white’ is true”, this is equivalent to a statement that omits reference to those symbols, viz. “snow is white”. Statements about truth are equivalent to statements about the world, but statements about meaning are not thus equivalent. So statements about truth point directly to the world whereas statements about meaning do not. Meaning has to do with symbols, but truth has to do with reality. How then could truth conditions constitute meaning? How, in particular, could the statement “’snow is white’ means that snow is white” be analyzable by the statement “’snow is white’ is true if and only if snow is white”? The word “true” cancels quotation, but the word “means” does not. Truth takes you to the world, but meaning keeps you within the domain of symbols. Thus meaning and truth are logically different kinds of concepts. Accordingly, we cannot hope to explain meaning by invoking truth: truth belongs out in the non-symbolic world while meaning belongs with symbols grasped by the mind.

[1] See, for example, Davidson’s “Truth and Meaning”.

[2] It is very much a moot point whether Tarski’s theory of truth qualifies as a truth conditions theory in the classic sense. He does not speak of truth conditions or states of affairs in the development of his definition, and it is not unreasonable to take it that the notion amounts to nothing more than a sentence of the meta-language. He is not referring to states of affairs or anything of the kind but simply using meta-language sentences on the right-hand-side of biconditionals referring to object language sentences. It is quite unclear that he is advancing a truth conditional semantics in the classical sense (in fact, I think he is not).

[3] Is there some kind of use-mention confusion at work here?