Correlational Semantics

Correlational Semantics

I will describe some possible uses of correlational semantics. I don’t say I subscribe to these uses; I offer them as a gift to those with anti-realist or fictionalist yearnings in certain areas.  It may help ease some discomfort caused by such yearnings. Let’s begin with a relatively simple case: feature-placing sentences like “It’s raining”. The utterance of such a sentence can be true yet contain no reference to the place at which it is raining—for example, London. The sentence is not semantically comparable to “London is rainy”. The word “it” is not a referential singular term denoting London, even though the truth of the utterance depends on the fact that rain is falling in London. There is a correlation between uttering “It’s raining” while in London and the statement “Rain is falling in London”, but nothing in the former denotes London. The former is true in virtue of the latter but it isn’t about what the latter is about in the sense that it contains a term that refers to what “London” refers to: the two sentences don’t mean the same thing even at the level of reference. The phrase “It’s” is a kind of dummy subject expression, not a genuine referential term. Correlation is not same as denotation, even though truth may depend on correlation. We may say that our sample sentence alludes to its correlate (it presupposes a particular place, typically known to the speaker), but it contains no term that denotes that place. We know that it rains at places (where else might it rain?) and we assume that that is what is going on in the present instance, but epistemics is not semantics. We might even say that our sentence connotes a place where it is raining while not denoting any such place.[1] Places belong in its semantic periphery, so to speak. Now consider fictional names like “Hamlet” as it may occur in a sentence like “Hamlet is a prince”. That sentence seems true (Hamlet isn’t a pauper or a porcupine): we can insert it into the formula “s is true if and only if p”. Yet Hamlet does not exist, so (we might say) the sentence ought not to be true. What is it that makes it true? The obvious answer is Shakespeare’s intentions: he decided to create a character named “Hamlet” who was a prince. The name “Hamlet” doesn’t denote Shakespeare or his intentions, but anything true of Hamlet is due to Shakespeare’s intentions. There is semantic correlation but not semantic denotation. And anyone familiar with fictional names understands this kind of dependence-without-denotation; it is part of our linguistic competence. Thus, statements containing fictional names are true in virtue of the author’s intentions, which are correlated with the name; but they don’t refer to such intentions—quite possibly they refer to nothing. The correlation explains the truth of the statement, but there is no denotation relation connecting the two. Truth and denotation come apart; the former doesn’t require the latter. It is therefore possible to maintain both that the statement is true and that it has no reference, because reference comes from elsewhere, in the shape of a correlated statement that does refer. The connoted (not denoted) statement does the truth-conferring work, leaving the original statement to luxuriate in its non-referential indolence. We can combine fiction with truth by availing ourselves of correlational semantics. Otherwise, we are saddled with truth without reference, or no truth at all. Hamlet doesn’t exist, but statements about him can still be true, because “Hamlet” is correlated with Shakespeare’s intentions, which do exist. Thus, anti-realism does not imply lack of truth (or weird kinds of truth). The anti-realist does not have to deny the obvious.

Now that the form and point of correlational semantics is made clear, we can extend it to other areas. Some philosophers (named “positivists”) have maintained that theoretical entities are mere fictions: but then how can statements about them be true? Correlational semantics supplies an answer: these statements are correlated with other statements about existent non-theoretical entities such as experimental results, sense-data, retinal stimulations, or what have you. These entities confer truth even when the statement correlated with them refers to nothing real. Correlation steps in where denotation fails. Statements about electrons, say, can be true even if there are no electrons, because they are made true by non-electrons—non-existence is no bar to truth. We simply detach truth from reference and existence; and, indeed, there is nothing in the concept of truth itself to compel a rigid connection, because truth simply requires correspondence (or the possibility of disquotation) not reference and existence. No sentence is true but reality makes it so, but the sentence need not denote this reality. We can have a correspondence or redundancy theory of truth without building in reference in the sentence declared true; for we can appeal to correlated sentences or states of affairs. The concept of truth itself doesn’t even require a referential structure (grammar). Truth is more general than reference, more capacious. Similarly, it can be maintained that folk psychology refers to nothing real and yet its propositions are true, since they are made true by propositions about the brain.[2] It is true that I believe that snow is white even though there are no such things as beliefs, because that statement is made true by the condition of my brain as it controls my behavior. Thus, one can consistently be a mental eliminativist and also ascribe truth to mental propositions—rather like the anti-realist about fictional characters. For there is something other than mental reality that can confer truth on such propositions; and surely it is true that I believe that snow is white (even though that phrase refers to nothing, according to the eliminativist). Correlational semantics allows you to have it both ways (if both ways appeal to you). Correlation not reference; connotation not denotation.

I suspect the philosophers most well disposed towards correlational semantics will be ethical expressivists. Their official view is that ethical sentences do not report facts, are not true, and cannot be said to be known; there are no values “in the world”. Or better put: since there are no values in the world, ethical utterances cannot be true, even though they appear to be true. Wouldn’t it be better to accept that they are true but that they don’t denote values-in-the-world? That is what correlational semantics allows: expressivism combined with truth. Suppose an experience of pleasure occurs and someone remarks “That’s good”: the expressivist says that no evaluative property is thereby ascribed to the pleasure, since there are none such; instead, the utterance is like an outburst of emotion or an order to act in a certain way. But there is an alternative story: the utterance is true in virtue of the existence of the pleasure but no property is ascribed by the word “good”. The word isn’t even a predicate (in some versions of the doctrine). The sentence is not true in virtue of a denoted evaluative property but in virtue of the correlated pleasure property—correlation not denotation. The world does contain pleasure and it makes such propositions true, but evaluative discourse does not refer to this property. The evaluative force comes from the attitude expressed not from the state of affairs that makes the sentence true. Thus, truth is compatible with expressivism concerning value; we are not forced to reject ethical truth just because ethical propositions don’t refer to ethical properties. Put differently, the truth-makers belong to the supervenience base not the supervening values (they are add-ons derived from human psychology). Again, I am not saying I accept this position, only that it exists in logical space and has attractions for a convinced expressivist, particularly one who refrains from withholding truth-values from ethical statements. Something objective is involved in making ethical statements true, but it isn’t the denotation of ethical words; rather, it is what those words connote in the way of correlation. We know these correlations, so we accept the truth of what is said; but this doesn’t entail that the correlated facts are denoted by moral terms—they are not. Accordingly, we get to be moral anti-realists and accept objective ethical truth-makers. Everyone is happy (well, not everyone). Some words don’t denote anything real, but that doesn’t stop them appearing in true statements, thanks to suitable correlated realities. Only referentially empty sentences with no correlates will turn out not to be true or be incapable of truth, such as ungrammatical or meaningless types of sentences. The sentence jumble “It thing number gone” is not true and has no true correlate; nor does “All mimsy were the borogroves”; nor “Colorless green ideas sleep furiously”. But it is possible for a sentence to contain empty terms (no denotation) and the sentence be perfectly true; it has, we might say, parasitic truth. Correlational semantics is designed to accommodate these cases. We must give up the dogma of denotational truth.[3]

[1] See my “On Denoting and Connoting”.

[2] See my “Semantical Considerations on Mental Language” and “Ontology of Mind”.

[3] There are other reasons we might want to give up the dogma of denotational truth: a devotion to hard redundancy theories, adherence to coherence or pragmatic theories, skepticism about the notion of reference (itself having several sources). The point I would emphasize is that nothing in the concept of truth analytically implies that true sentences or propositions must have a referential semantics: truth does not require referential relations between sentence parts and whatever in the world makes the sentence true. In principle, a sentence could have no such structure and still be true. Some linguists and others have toyed with the idea of a pre-referential level of meaning in the child’s understanding of language (“RED!”, “COW!”, “MAMA!”); such a level would not preclude application of the concept of truth.