Descriptions and Non-Existence

Descriptions and Non-Existence

Semantics would be easier if there was no such thing as non-existence (if non-existence didn’t exist). Then we could simply assign an existing reference to any referential-looking term. We wouldn’t have the problem of empty terms: all meaning would be explicable by means of existing entities. We would have a fully denotational semantics, perhaps supplemented by Fregean sense. In particular, we could assign an existing reference to any definite description—we would have no problem of empty descriptions. This problem is what led Russell to propose his theory of descriptions, which reshapes the logical form of description-containing sentences (they are not singular terms at all). That theory would not be needed if non-existence were not a thing. Of course, some philosophers (e.g., Meinong) have denied that so-called empty descriptions are really empty: they propose a kind of shadowy existence (“being”) for things like the golden mountain or the king of France. These things exist in another realm, or in a different way, or to a lesser degree (hence they are said to “subsist”). Russell wasn’t happy with this kind of ontological largesse, so he was thrilled with his shining new theory; but its entire motivation depends on an ontological assumption, namely that some things don’t exist (or subsist either). That is, the attraction of the theory depends on an ontological position or presupposition; it makes semantics depend on ontology (metaphysics). If Meinong were right, we wouldn’t need Russell’s theory—we could just stick to the practice of assigning denotations to descriptions. We wouldn’t need an alternative revisionary anti-referential semantics: quantificational, conceptualist, formalist, fictionalist–whatever avoids an unpalatable Meinongian ontology. All this arises because of the problem of non-existence—a problem in metaphysics. It doesn’t arise by direct consideration of the sentences in question; it arises from elsewhere. It isn’t as if Russell took a long hard look at definite descriptions and saw that they conform to his paraphrase; he inferred that his theory must be true or else we land in ontological hot water (the fetid Meinongian swamp). He deduced semantics from ontology.

But what if he was wrong about the ontology? What if Meinong was right all along? I am not saying he is right; I am just saying that he might be. It is a substantive metaphysical question—and we don’t want our semantics to be hostages to the fortunes of our metaphysics. Suppose for a moment that Meinong is right: then Russell’s theory is unnecessary, a revision without a reason. For there is nothing intrinsic to descriptions that warrants its adoption. This is why no one proposed it before—it looks wrong. It has counterintuitive consequences. It is complicated and contrived. Students find it hard to understand. No one would talk like that. Only the desire to escape the clutches of Meinong could make it seem attractive. To a Meinongian, it seems like a desperate attempt to avoid the obvious truth. Think of it this way: in a possible world in which Meinong is right Russell’s semantics is unnecessary, pointless, and unattractive. If we are actually living in that world, it is bad philosophical linguistics. But surely, we don’t want to take that risk—we don’t want our semantics to hang by an ontological thread. Semantic theories shouldn’t be justified by ontological considerations alone. It would be different if Russell’s theory had other cogent justifications, but in fact its sole motivation derives from (questionable) ontological assumptions (and did Russell ever really refute Meinong?). It is therefore methodologically misguided. It needs to be established on quite different grounds. It’s like trying to justify a non-referential semantics of predicates by insisting that Platonic universals don’t exist; maybe they don’t, but you don’t want to adopt a revisionary semantics of predicates on that basis alone. This is a debatable metaphysical question, not a datum that can be wheeled in to rule out a perfectly natural semantic theory (viz., predicates denote universals). You may have an ontological beef with physical objects, but do you want to keep them out of semantics when they seem like the perfect tool for the job? And you might be wrong about the non-existence of physical objects, in which case you have ejected them falsely from your semantics.[1]Don’t mix up semantics and metaphysics! Don’t let your metaphysical views shape your linguistic views! In particular, you shouldn’t let logical form depend on ontology, which is exactly what Russell does (this is why his theory is so exciting). Grammatical form doesn’t depend on ontology, so why should logical form? The meaning of “the” should not be made to depend on whether the golden mountain exists.

Let me try to make the situation vivid by constructing a thought experiment of a familiar form. Suppose that in the actual world there are no Meinongian objects, and suppose that speakers know this. Russell proposes his theory and everyone is happy with it. Now consider a possible world in which there are Meinongian objects and everyone believes in them. Wouldn’t it be correct to say that Russell’s theory fits the actual world but not the stipulated possible world? We could even suppose that the speakers in both worlds have no opinion on the truth of Meinongian ontology and are precise physical duplicates with the same internal mental states. Then an externalist will want to say that definite descriptions have different meanings in the two worlds in virtue of the external ontological facts, despite the internal identity of the speakers. Logical form will accordingly be different in the two worlds. If we make semantics depend on ontology, this is the kind of result we get. But doesn’t it seem wrong to make logical form depend on such facts? The ontological difference shouldn’t generate a semantic difference in respect of logical form. Russell thinks the lack of existence forces a revisionary logical form, but in the Meinongian possible world there is no such pressure for discerning that kind of logical form. Better to deny that the facts of ontology can determine semantics to this degree. Individual meanings may not be (completely) in the head, but surely logical form is (grammatical form certainly is). Something is wrong with Russell’s methodology.

Suppose you are agnostic about Meinong’s ontology: should you then be agnostic about the semantics of descriptions? If you are an agnostic about God, should you be an agnostic about the meaning of “God”? No, you want a semantic theory that is neutral with respect to such ontological questions. It is true that Russell’s theory is neutral about the existence or otherwise of the objects that the description purports to describe, but it is not neutral in its motivation—it presupposes that Meinong is wrong. That puts the theory in a needlessly precarious position, since the presupposed metaphysics might be mistaken. One wants to be able to defend the theory on less vulnerable grounds, by appeal to the very nature of the description. But its revisionary character makes this difficult (as Strawson in effect pointed out): the description seems very much like a singular term.  We get a rough equivalence in Russell’s paraphrase, but not a precise synonymy. We can’t even apply the word “refers” or “denotes” to the description if Russell is right. So, the linguistic data don’t prima facie support the theory; its support comes from a presupposed anti-Meinongian ontology. And isn’t it true that the historical enthusiasm for the theory came from its anti-Meinongian credentials—its rejection of Meinongian extravagance? But that is a frail basis for a semantic theory—the wrong kind of basis. It’s rather like saying that proper names don’t name people because you don’t believe in selves (perhaps on Humean grounds), proposing instead that they are not singular terms at all. In a way Russell’s theory is not psychological enough (psycholinguistic enough); it relies too heavily on theses concerning things outside the mind. It needs to be more internalist (in a Chomskyan sense)—more about the brain or the cognitive-linguistic system. What empirical evidence is there that definite descriptions are really quantified conjunctions? Where is the cognitive science that demonstrates that linguistic proposal? From this point of view, the rightness or wrongness of Meinong’s ontology looks irrelevant.[2]

[1] What if the speakers of the language are explicit Meinongians, have been for centuries, have it in their genes? Are we to say that their language is Russellian? They will openly disagree with his theory, perhaps regarding it as preposterous, so how can it be the true theory of their definite descriptions? Russell’s ontology is not theirs, and neither is his semantics.

[2] The issue can be compared with possible worlds semantics: how can the existence or otherwise of possible worlds determine the viability of a semantic theory of natural language modal expressions? That is a metaphysical question extrinsic to the syntax and semantics of such expressions. What people think about possible worlds might be relevant, because it concerns psychology, but the actual existence of them seems beside the point. Certainly, it is hard to justify a semantic theory just by asserting or denying a particular ontology. Ontology and semantics are separate domains.