Can There Be a Theory of Meaning?

I will do what is never done: list all the important properties of meaning, in no particular order. First, meaning is combinatorial: meanings combine to form phrase-like and sentence-like structures. Second, meanings are world-correlated: they refer, link to reality, represent how things are. Third, they are use-determining: what a word means fixes its use. Fourth, meaning is truth-related: meanings fix the truth conditions of sentences. Fifth, meanings vary in type: they are different for different types of word (names, predicates, connectives, etc.). Sixth, meanings are precise, specific, fine-grained, discriminating: they track distinctions in reality (and beyond). Seventh, they are graspable by minds: it is possible for some types of minds (not all) to apprehend and understand them. Eighth, they have a history: they have a past, a present, and a future; they can come and go, and change over time. Ninth, they are tied to grammar (syntax): there are grammatical rules governing their normal expression (word order, embedding, etc.). Tenth, they are normative: there are right and wrong ways of combining them (or their linguistic expression), and of applying them. Eleventh, they are capable of forming nonsense: there is no guarantee that combining them will make sense. Twelfth, they are communicable: you can convey to another person what you mean; they are not generally private. Thirteenth, they are invisible and impalpable: they can’t be sensed by the human senses. Fourteenth, meanings are cognitive (not affective): they belong to the ratiocinative part of the mind (though they can link with the affective part). Fifteenth, they are thought-related: they enter into the contents of thought and are used in thought processes. Sixteenth, they are connected to knowledge: they are both known in themselves and vehicles of knowledge. Seventeenth, they are potentially infinite: they can be combined to produce an infinite totality of meanings (the idea of infinity is built into them). Nineteenth, they can interact with context to generate further meanings: this is particularly evident with indexical expressions. Twentieth, they are connected to logic: whether an inference is valid depends on the meanings of the elements composing the inference. Twenty-first, meanings are embedded in sensorimotor systems: they can be the objects of both hearing (seeing, touching) and speaking (with voice or signs). Twenty-second, meanings are (or can be) musical: they have rhythm and melody (it isn’t just words considered apart from their meaning). Thus, meaning is multifaceted, protean, and promiscuous; it has its fingers in many pies, and its fingers are highly adaptable. It isn’t simple. It can’t be quickly and easily summed up. What have philosophers, psychologists, and linguists had to say about it? Frege stressed the meaning-world connection (sense and reference, truth). Wittgenstein stressed the meaning-behavior connection (use, language games, communication). Russell stressed the meaning-mind connection (sense-data, acquaintance). Chomsky stressed the meaning-grammar connection (combination according to rules). Grice stressed the meaning-intention connection (speaker meaning and sentence meaning). Psycholinguists stress the sensorimotor links of meaning (reception and production of speech). Austin stressed the meaning-speech act connection (performatives, perlocutionary effects). Quine stressed the meaning-ontology connection (quantifiers and existential commitment). Davidson stressed the meaning-logic connection (logical form, recursive structure, truth conditions). Lewis Carroll stressed the meaning-nonsense connection. All these connections exist and are well worth studying. Some excellent results have been obtained. But there is nothing we could call a unified semantic theory: there is no theory that captures all the aspects of meaning itemized. There is no one concept that can be used to tie it all together (truth, verification, use, etc.). Indeed, a survey of the multiple aspects of meaning suggests that nothing could constitute such a theory; the most we can hope for is sub-theories dealing with meaning’s several aspects (dimensions, properties). Compare matter: is there a single unified theory of matter? Notoriously not: gravitational theory will not mesh with electromagnetic theory. Physics thus deals with matter in two separate branches (and it doesn’t deal with every aspect of matter, e.g., its aesthetic properties). So why should the “theory of meaning” admit of unified treatment? Similarly, must there be a unified “theory of mind”? There certainly isn’t one, despite some valiant (and less than valiant) attempts: what we call “the mind” is just too various for that to be possible. So, we should really accustom ourselves to something less all-encompassing in meaning studies—theories dealing with specific aspects of meaning.[1] For meaning (to repeat) isn’t simple; it isn’t some kind of ontological primitive. It isn’t conceptually one-dimensional. Whoever thought it was?[2]

[1] As an example of different sub-theories of meaning, consider Frege’s theories of the meaning-world connection and of the combinatorial nature of meaning. The first employs the apparatus of sense and reference; the second employs the apparatus of function and argument—each addressed to different properties of meaning. They are logically independent. He also had separate theories of what he called tone and force. He didn’t try to do it all with one concept.

[2] The musical and nonsensical aspects of meaning have been relatively neglected. The linguistic and musical parts of the human brain are evidently connected, but the same is not true of other animals. Why this should be is something of a mystery. In song (and poetry to some degree) they come magically together: words become lyrics associated with melody (pitch patterns). In the case of nonsense, we have a bit of a paradox: the things equipped to form meaningful wholes can readily fall into sheer nonsense, and we like the result (hence the popularity of Lewis Carroll). They are, so to speak, perpetually on the brink of nonsense, only too ready to tip into it. We can imagine beings that speak only nonsense and are happy with that. Meanings have meaninglessness built into them, contrary to what we might expect given their original purpose. The meaningless is just the flipside of the meaningful, its shadow or sibling.

A parody of the Beatles’ Here Comes the Sun.

Here Comes the Sun Again

Here comes the sun again

Smiling its salesman’s smile

Sun, you are so seductive

You want to hold me for a while

Oh sun, you’re irresistible

How can I turn you away

When I feel your warmth on my face

I know it will be a good day

Sun, sun, sun, here you are now

Sun, sun, sun, yet so far now

But now I know you better

I know your wily ways

You tricked me into loving you

While you sent down your deadly rays

We’ve taken down the barrier

So you can bring your heat

We’re like a dog trapped in a car

Hoping to be released

Sun, sun, sun, there you go now

Sun, sun, sun, keep it low now

You feel so close to home today

Your breath can easily roast us away

Every day feels like another scorcher

Life has become a sunny torture

Sun, sun, sun, please go away now

Sun, sun, sun, leave us to pray now

Sun, sun, sun, it’s enough now

Sun, sun, sun, we’ve had enough now

Sun, sun, sun, we’ve had enough now…

This paper follows on from “The Cruel Gene” and “Pain and Unintelligent Design” on this blog.

Evolution of Pain

Pain has evolved over many millions of years. Presumably it had primitive forms that were subjected to natural selection. It was honed and whittled, modified and amplified. There are now several species of pain, each no doubt fine-tuned to achieve a certain end. It is very widespread among animals—if not universal, then close to it. If we think of it as a mental organ, we can model its evolution on the evolution of organs in general: there is a kind of structure-function matching, with the organ designed to have certain functional effects (ultimately survival and reproduction). It probably evolved very early given its survival utility, and it has persisted robustly over the millennia; it is unlikely to become obsolete or extinct. We can think of it as the eyes of the body: it “sees” danger and reacts to it. It is analogous to a perceptual faculty. Perhaps it evolved in prey animals (which includes nearly all animals, since nearly all are prey to some animal at some point), because they need to be sensitive to the teeth and claws of predators. Perhaps some animals perceive it more exquisitely than others, depending on their vulnerabilities. A lack of body armor would favor greater sensitivity to pain—there is no need for pain perception if your armor is impregnable to predators (and rivals). What we can be sure of is that pain was rigorously selected for according to its costs and benefits, like any other biological trait. It feels and acts the way it does for good reasons. There are two puzzles about this. First, the costs are considerable: yes, you avoid the painful stimulus, which is all to the good, but you also incur serious liabilities because of general debilitation. An animal in pain is often hobbled by the pain, unable to function, like an athlete in pain. True, this can be helpful if rest is needed, but nature is not always so obliging as to allow for such rest. Second, the pain gives rise to other effects than those of avoiding the painful stimulus, such as overt expressions of pain (grimaces, cries). What is the point of these? It is hard to avoid the conclusion that they are communicative: they let other animals know you are in pain. But this is puzzling, because it isn’t as if other animals will automatically come to your aid—instead of sensing weakness and attacking you. You might want to keep it a secret that your right leg is hurting, or that you have a headache and are not up for a session of head butting. So, pain might not be the simple adaptation we assume: too much downside and too many collateral effects. We can take it that this is not the result of insufficient fine-tuning, which will be rectified in due course; on the contrary, pain mechanisms must be highly adapted by now, not still replete with glitches. Pain responses have been well-nigh perfected over evolutionary time; natural selection has made them as wondrous and efficient as eyes. This will no doubt involve making pain as painful as is compatible with proper functioning, which is obviously pretty damn painful. Good pain is bad pain, so far as evolution is concerned (those selfish genes!). Pain has evolved to be bad, not mild and tolerable. And it has produced a biologically marvelous trait, truly spectacular: pain is remarkably bad, devilishly so. Natural selection has done its job and done it well: it has produced organisms that really hurt—as it has produced eyes that see really well. It has brought pain to the pitch of perfection, survival-wise; it’s hard to believe it could get any worse, subjectively speaking. It has labored long and hard to make us suffer, to ruin our days, even to make us wish for death. Quite an achievement! The genes must be proud of themselves, but the animal must carry the burden. That is the logic of evolution by natural selection playing itself out: animals have whatever traits enable the genes to survive, pleasant (orgasm) or unpleasant (pain). But there is a residual puzzle: why not use reflexes instead? The patellar reflex allows the organism to move rapidly and effectively, but no pain is involved; same for the blink reflex. So, why must a stubbed or squashed toe be accompanied by intense pain—why not arrange the nervous system so that the foot is quickly and reflexively withdrawn but without the intervening agony? The pain sensation doesn’t seem necessary to the function; it seems like a gratuitous (indeed sadistic) add-on. The only thing I can think of is that the pain is somehow necessary for ongoing flexible voluntary behavior in the presence of harmful stimuli, as in managing a broken bone or a burn. But though that seems true as a matter of empirical fact, it is difficult to see why it has to be true. Presumably it has just turned out over the course of evolutionary time that pain is a more efficient way of handling injury than a purely reflexive and pain-free method; but why this should be remains obscure. So, the existence of pain is something of an evolutionary puzzle, especially given its functional downside (its phenomenological downside counts for nothing in the evolutionary game). It clearly evolved over millions of years and is close to universal, but it’s puzzling why it exists at all as an adaptation. It must have evolved in multiple species many times (convergent evolution), but its rationale is far from obvious (unlike the eye). Pain is puzzling, biologically and philosophically, despite its undeniable reality. Some animals do quite well without it—jellyfish, insects, worms—but many live by it (literally), despite its manifest aversiveness. It is the only adaptation bequeathed to us that we would rather be without—that is intrinsically nasty. Suffering may be adaptive to life, but it is also the bane of existence. It is really the only thing that can make life not worth living, yet it exists to serve life. Pain is an enigma that we could live without.[1]

[1] Imagine the amount of pain that has existed over the course of evolutionary time in all animal species. It doesn’t bear thinking about. Earth is the planet of pain.

Are There Synthetic Truths?

There is a tradition, stemming from Quine (but not pre-dating him), claiming that the concept of analytic truth is undefined, or ill-defined, so that the analytic-synthetic distinction cannot be made sense of; accordingly, there are only synthetic truths. I think this is the opposite of the truth. Nothing is true but reality makes it so, Quine tells us, contrasting reality with meaning. And it is routine to hear that a synthetic truth is one that is true “in virtue of the world” (not language)—or perhaps in virtue of “the facts”, or “extra-linguistic reality”. But these formulations are by no means clear and run into immediate objections. In the first place, language and meaning are clearly part of reality or the world, so reality or the world cannot be construed as excluding them: yes, nothing is true but reality makes it so, but the reality can be either meaning or non-meaning. It might be thought that there is an easy fix: just say that a synthetic truth holds in virtue of non-linguistic reality. Now the problem is that some synthetic propositions are about linguistic reality—take any statement about the languages of the world or what a particular word means. Clearly, we need to say that a synthetic truth is non-analytic, but (a) that would not suit Quine (though it’s fine by me) and (b) it is purely negative. This throws us back to talk of reality and the world and facts: we need to say that synthetic truths concern the part of reality that doesn’t include facts constituted by meaning relations (or something along these lines). Now a new problem confronts us: what is this entity and what composes it? Is it a totality of facts, a totality of objects, of objects and properties, of facts and values, etc.? This is beginning to sound ominously metaphysical, inherently contentious, and probably meaningless. And there is this problem: presumably we can’t mean that a given proposition is true in virtue of the whole of this world, or else each proposition is made true by the same thing; so, we need to say that parts of the world correspond to each synthetic proposition (unlike analytic propositions). This puts us on a familiar and dreary path: logical atomism, negative facts, complexes of objects, and the difficulty of supplying identity conditions for the putative truth-making entities. We didn’t think we had signed up for this rigmarole when we declared that there are synthetic truths! But we started talking about “the world” so what did we expect? We are thus driven back to saying that synthetic truths are non-analytic truths, with no positive characterization of them given (as well as uncritical acceptance of analytic truth). And here the nub of the matter emerges from the mist: the concept of the synthetic is parasitic on the concept of the analytic. It is the concept of truth in virtue of meaning that wears the trousers (as Austin would say): we need that concept if we are to carve out a territory labeled “synthetic truth”. The latter concept has no content without the former concept. It would be different if we could define “synthetic” as (say) “perceived by the senses”, because then we wouldn’t have to make do with the amorphous concept the world; but that is obviously too narrow, too restrictive. The problem is that the class of truths that are not analytic is a ragbag possessing no inner unity; it isn’t a natural metaphysical kind. Its sole content is given by the concept of the analytic, which trades in the restricted (and legitimate) concept of meaning. Note that “synthetic” is not a very descriptive term; it is really just a label for what is left over when we have listed all the analytic truths. The concept lives and dies by its contrast class. But that means that we have no unitary positive concept of the synthetic suitable for bringing together all that is supposed to fall under it. Moreover, we have a decent criterion of identity for meanings, viz. substitutivity inside belief contexts, but we don’t have such a criterion for whatever is supposed to make synthetic propositions true. Hence the vague and woolly talk of the “world” and “reality”. There is the general concept of truth and the specific concept of analytic truth, but there is no concept of synthetic truth save a hopelessly disjunctive one. It is the “craving for generality” that leads us to think that we have identified a metaphysical natural kind. All we really have is a philosopher’s invention—a contrivance, a fabrication. It is certainly not a concept of science, or even of linguistics. So, we cannot hope to base a philosophy on the concept of synthetic truth, saying such things as that all truths are synthetic; all that can mean is that no truths are analytic, relying on an antecedent grasp of that concept. This is obviously no help to a disciple of Quine, but even a firm believer in analyticity won’t be encouraged by a concept so negatively defined. I think myself that analytic truth is adequately defined as truth in virtue of meaning alone, but that synthetic truth has no adequate definition—certainly not one in terms of truth in virtue of “the world” or “reality”. You might hope to define it as “what is known by empirical investigation”, but that is objectionably epistemic (what if the truth can’t be known?), and also fails to count knowledge of one’s own mental states as knowledge of synthetic truths.[1] There is, in fact, little effort to define the category of the synthetic, so that the concept is left at a vague and intuitive level, explained more by example than by general definition. To say that all meaningful statements are (or must be) synthetic is therefore meaningless metaphysics, which is ironic in the circumstances. We literally don’t know what it means (Kant has a lot to answer for). In its ordinary use the word means “not genuine; unnatural” (OED) and that does indeed correspond to something real: for the idea of a synthetic truth is a synthetic idea, i.e., a fake classification. There are no synthetic truths (but plenty of analytic truths); there are just truths that fail to be analytic—as there are many types of truths that fail to be ethical or psychological or economic or about me. Not belonging to the class of analytic truths is not a way of forming a unified class of other truths. We can distinguish classes of truths within the non-analytic class, but no positive trait unifies them, whereas analytic truths form a well-defined class. It is the concept of analytic truth that is in good theoretical shape not the concept of synthetic truth. And isn’t this what any card-carrying rationalist would say?[2]

[1] The OED is uncharacteristically unhelpful here: as its second definition of “synthetic” it gives us “having truth or falsity determinable by recourse to experience” (this is relegated to Logic). This is a frankly epistemic definition of what should be a logical or metaphysical concept, and is therefore a kind of category mistake. It fits the concept of the a posteriori better. Also, how is it supposed to apply to the synthetic a priori, as in arithmetic? Such truths are not verifiable by “recourse to experience” yet are supposed synthetic. Nor do we know that we have beliefs by sense experience (and what other kind of experience is intended?). It looks to me as if the dictionary makers took the philosophers on trust and did the best they could with a dubious (spurious) notion.

[2] The concept of empirical truth is similar in view of the enormous variety of things that fall under the concept of experience. How unified is this class really? Is it as unified as the class of a priori truths? Empiricism is just not a monolithic unified theory, and nor is “synthetism” defined as the doctrine that all truths are synthetic. The metaphysics of these theories is not well defined, i.e., what it is to be an experience or a synthetic truth. More dogmas of empiricism (rationalism fares much better in this respect).