[1] This doesn’t mean there is nothing “erotic” about the unconscious: there is room for the erotic in all sorts of personal relations, and indeed in the joys of discovery (“Eureka!”). In psychoanalysis the erotic is understood as the “life force” (anima in Plato) and contrasted with Thanatos (the “death instinct”). We can preserve this aspect of Freudian theory without supposing that the male infant actually desires sex with his mother (or the female infant with her father—the so-called “Electra complex”). There is indeed a lot of “positive affect” in the unconscious, but it need not take the form of thoughts of actual incest! Freud’s specific theories do a disservice to his general conception.

 

 

Completely Empty Names

 

It is time we faced up to some uncomfortable truths about proper names. There have been two theories about them, neither very intuitive, commonly known as the description theory and the direct reference theory. The two theories are radically opposed to each other and each faces formidable difficulties. The simplest way to state the difficulties is that (a) functioning names can lack individuating descriptive content and (b) they can lack an existing bearer: thus they can be empty of sense and empty of reference. On occasion they can be empty of both: a name might be used by someone who associates no uniquely identifying description with it and which also lacks a bearer. For example, you might use the name of a Greek god(“Hera”) and not be able to distinguish this god from others, and neither does the god in question exist. This is what I am calling a completely empty name—empty of both sense and reference. Yet the name has a use. Theorists will sometimes resort to reference when sense is lacking, and resort to sense when reference is lacking, but this tactic won’t work when both are lacking. Notice that both theories face blindingly obvious problems: for names don’t look like descriptions (where is the “the”), and meanings don’t get buried in the ground when bearers of names do. But it is felt that such counterintuitive results are required by theory, these being the only theories available. Odd that the only possible theories look so unpromising on their face, but hey that’s philosophy for you! Thus philosophers pick their poison, and gag on it too (all good fodder for doctoral dissertations). No matter that speakers can hardly ever produce a suitable description, or that it might turn out that all our names lack bearers (we are name-users in a vat): a bit of vigorous bullet-biting can steel us to such embarrassments, and ingenuity is always able to come up with something. Still, some faint hearts might feel that this is not a happy situation and cast about for a less rebarbative theory. And it is not as if there is no logical space for an alternative theory: not all usable expressions of natural language fit the sense-reference paradigm. What about logical connectives or stress patterns or words like “boo” and “ha”? They have a use but it would be stretching a point to declare them susceptible to the sense-reference scheme: they have neither, but they have a job to do. Can’t proper names fall into this general category? Their meaning isn’t constituted by a description or by their bearer but by whatever it is that confers a use on them. Thus they may be empty of both sense and reference and yet possessed of a usable meaning—significance, “semantic value”, communicative function. We just have to find out what that is; then our troubles with names will be over.

            That sounds like a nice resolution, pleasantly irenic and thrillingly eliminative (vaguely Wittgensteinian), but it runs into an obstacle: names are nouns that occupy the subject position in sentences, like descriptions and demonstratives, so they are not like sentence connectives or stress patterns or exclamatory noises. We use them to denote things and they enjoy liaisons with other denoting expressions, so how can they not be components of sentences with that kind of function? Just look and listen and you will see and hear them occurring as parts of sentences playing a specific semantic role—how then can their meaning be constituted by anything other than sense or reference? Don’t they carry modes of presentation of some sort, and aren’t they intended to refer to things? To say they mean neither sense nor reference sounds like saying they possess the null semantics, so it has to be one or the other (or both as in “dual component” semantics). For them to be “completely empty” would be for them to have no linguistic function, to be nonsense, like rocks and table salt. And yet the name “Hera” would appear to be perfectly usable, entirely normal, nothing to worry about, despite its semantic impoverishment. We therefore have a puzzle about names: they can have a use and hence a meaning, but the only candidates for that meaning can be absent from the scene—no definite description or existing bearer in sight. They ought to mean nothing, but in fact they are as meaningful as any name, no matter descriptively well endowed or firmly linked to a real object. Is there any way out of this puzzle?

            I will develop what I call the “parasite theory of names”. Names are not self-sufficient sources of semantic nourishment; they derive their nourishment from a host that is rich in semantic nutrients. First, we must note that what appears to be part of a sentence is not necessarily really a part of it. The specific sensorimotor systems that are employed in speech and writing provide outputs that contain names, but the underlying abstract language faculty may be differently constituted—it might not contain any names at all.[1] Names are imported into natural language use from an outside source, chiefly for purposes of communication; they are not part of the original language faculty designed for aiding thought. We don’t think in names, though we do use them to talk to other people—they are useful for that purpose. In the language of thought we have descriptions and demonstratives, but we don’t have names. Names are introduced as tags or labels for things that people often can’t identify by means of descriptions (and may not even exist): they are parasitic on more basic (and arguably innate) categories of expression. These basic expressions have descriptive content—they are not mere labels—and they anchor names in discourse; but names themselves lack such semantic features. Names can function without themselves possessing sense and reference because they are linked to other expressions parasitically: they work a bit like winks used in communication—they don’t mean anything in themselves but they can be “translated” into genuine words. Indeed, a system of winks could in principle replace a stock of names. A semantic parasite like this will have not have sense and reference considered in isolation (like a description), but it will borrow from words that do, thus achieving a utility in communication. Thus neither a description theory nor a direct reference theory will capture their mode of semantic functioning (their “meaning” roughly). Maybe some words in the language of thought have a purely descriptive content, and some may well be subject to a directly referential semantics (e.g. words for sensible qualities); but the names employed in outer speech (auditory and visual) don’t participate in such a semantics. They are non-descriptive labels designed to overcome a knowledge problem (people often don’t know enough to pin their intended reference down). Thus they have a use but they don’t have the kind of meaning that other expressions have—they have a kind of zero semantics. They have a pragmatics but they don’t have their own proprietary semantics: they are linguistic parasites. In a sense they defer to the “experts” located in LOT. If they were basic they would need a semantics to ground them—such as a descriptive or referential semantics—but they are anything but basic. The mistake of the tradition was to assume that names are basic and then fret about their inherent semantics, thus oscillating between descriptions and objects as sources of name meaning; but once we recognize that names are parasitic we can accept that neither type of theory applies to them—which is exactly how things appear. In particular, they don’t have the kind of meaning postulated by description theories—hence the familiar counterexamples. But neither do they have a meaning constituted by objects themselves—hence the problem posed by empty names. Completely empty names are thus not at all anomalous or defective: they do their job perfectly well by being anchored to semantic machinery in the head. They simply don’t have any originalsemantics, any self-sufficient content; they are just convenient ways to get things across to people. We could speculate that they were appended to the language faculty late in the game, long after LOT evolved to service thought; and they have nothing in them analogous to descriptions. Nor do they possess any mechanism that could deliver a reference (unlike descriptions): we merely stipulate a reference for them and then use them to refer—they couldn’t do it without us. Semantically, they are blank tablets, mere counters in a game, so it is pointless to try to develop a “semantics of names”. Descriptions and demonstratives, yes (and other types of expression), but not names; they are, as we naturally say, meaningless labels. So theories of names have been chasing a mare’s nest: there simply is nothing that a mare’s nest could consist of, since horses don’t build nests. Likewise, there is no such thing as a name’s meaning, unless we mean by that “whatever it is that names do in communication” (mares have nests too if “nest” just means “a place they raise their young”). In the same sense “um” and “ah” have no meaning, though they play a role in discourse. Names look like they must have a meaning because they occur in the sensorimotor externalizations of internal sentences of LOT (the abstract language faculty), but it turns out that this appearance is deceptive because names are not parts of the innate lexicon (as accents are also not). The mistake is thinking that names are a semantic category analogous to descriptions or demonstratives, equally basic, equally non-parasitic. But names by themselves could never exist; they need support from elsewhere. The correct conclusion, then, is that names are the wrong kind of thing to subject to semantic analysis of the traditional kinds. We can sensibly have a description theory of descriptions (trivially) and a direct reference theory of basic sensory terms (as well as a two-level theory of demonstratives), but it is a category mistake to seek such a theory of names. This then is the solution to the puzzle of names outlined earlier: the language game of names is not the same as other language games used to refer to things, so it is wrong to seek a theory of names based on theories appropriate for other types of expression. And that is why the theories that have been proposed are so wide of the mark.[2]

 

Colin McGinn             

 

[1] For this kind of perspective, see Chomsky, Why Only Us: Language and Evolution (2016) and What Kind of Creatures Are We? (2016).

[2] We might compare the situation with a comparable mistake in the case of demonstratives: supposing that they must follow the paradigms of descriptions and names. Either demonstratives are really descriptions or they are really proper names, but they don’t appear to be either, so we have a puzzle. But the puzzle is resolved once we recognize that demonstratives function according to their own principles, in which context plays a crucial role. By analogy, names are rightly seen as neither descriptive nor directly referential but rather as parasitic and extrinsic to the internal language faculty proper. They are, just as they appear, meaningless labels arbitrarily assigned to objects about which we wish to communicate. They don’t mean associated descriptions and they don’t mean external objects: they don’t mean anything—though we may be said to mean something by using them. They are precisely not like descriptions or primitive terms for sensory qualities (e.g. “red”). Names are sui generis and should not be modeled on descriptions or demonstratives or primitive symbols for simple qualities.     

 

 

Naming and Contingency

 

Let’s accept that names have no meaning, as a distinguished tradition contends.[1] They may have a reference or denotation but they have no sense or connotation. Names lack “descriptive content” and are not synonymous with definite descriptions. Predicates are alien to their semantic functioning. They belong to a different semantic category from descriptive phrases (or demonstratives). Then an elementary consequence follows: names cannot give rise to analytic truths. For analytic truth is truth in virtue of meaning, but names have no meaning: they lack the semantic dimension that generates analytic truth. Descriptions have this dimension, so they readily produce analytic necessities (“The bachelor standing in the corner is unmarried”), but names are devoid of the kind of content that could underlie such necessities. This seems confirmed by simple inspection: sentences of the form “ais F” are never analytically true (the point is often cited as an argument against description theories).  You can never generate an analytic truth from “Aristotle” by combining it with a predicate true of him, though this is easily done by combining descriptions of Aristotle with such a predicate.[2] A generalization thus appears indicated: names cannot give rise to analytic necessities—simply because they lack the property responsible for such necessities, viz. meaning (sense, connotation). Sentences containing names are always synthetic; analytic necessity never follows from the semantics of names. So names belong naturally with contingency not necessity. Descriptions, on the other hand, readily lend themselves to necessities in virtue of their intrinsic semantics, because they possess a content that generates necessities (they have definitions and constituent semantic structure).[3] A monograph entitled “Describing and Necessity” would be aptly so named, while “Naming and Necessity” would raise eyebrows if intended to suggest that names and necessity have anything much to do with each other. One might even prefer “Naming and Not Necessity”, because names are not capable of producing necessities in virtue of meaning, having none.

            However, plausible as all this may sound, there appears to be an obvious counterexample to the generalization stated, namely sentences of the form “a is a”. Aren’t such sentences analytic tautologies, as with “Hesperus is Hesperus”? Such a sentence is necessarily true, known a priori, and inferable from mere mastery of the name “Hesperus”. It has all the marks of analytic truth. But on reflection this verdict should strike us as paradoxical, since we have just seen that names lack the property necessary for conferring analytic necessity, viz. meaning. How can our sentence be true in virtue of the meaning of “Hesperus” if that word lacks meaning (the reference of the name can’t produce such necessary truths)? It therefore looks as if the alleged counterexample cannot be genuine, so something must be said to explain it away. That is not so hard to do. First, notice that “a is a” sentences are highly unusual and not part of ordinary language—no one ever says things like this (except in a philosophy class). Second, are they even about things like the planet Venus—if so, what are they saying about it? Third, it seems obvious that their purport is something like the following: “The thing called “Hesperus” is identical to the thing called “Hesperus”—which is a metalinguistic tautology containing two occurrences of a single definite description. That is not generally true of sentences containing “Hesperus” (unless we decide to make it our general theory of names), but in this case it is what the peculiar sentence is really saying. But then the analytic status of “Hesperus is Hesperus” traces to a hidden description, which of course has descriptive content: the sentence has the same logical form as, “The inventor of bifocals is the inventor of bifocals”. The name “Hesperus” is not here occurring as a used proper name but occurs quoted inside a definite description, so we don’t have a counterexample to our plausible generalization. This is simply a case of a logically misleading sentence that can be generated from a natural language (though it is not generally used as part of natural language). When names occur in propria persona they never give rise to analytic necessities (how could they given their lack of meaning?); but they can occur in a nonstandard way in certain types of sentence (as quoted expressions in a metalinguistic definite description). The case is not unlike the sentence “I’m John Smith” said by way of personal introduction: it clearly means something like “I’m called ‘John Smith’”, in which the name is mentioned not used. Strictly speaking, “Hesperus is Hesperus” is a non-sentence, a linguistic monster, but we easily hear it as a quick way to express the descriptive metalinguistic equivalent (compare “There are a lot of John Smiths in the world”).

            This point can be reinforced by asking whether names ever have synonyms. Analytic truths arise from synonyms, but where are the synonyms for ordinary proper names? They don’t even occur in dictionaries, let alone thesauruses: there is no list of synonyms for “John Smith”. Don’t say “John” is synonymous with “Johnny”: this is a matter of alternative versions of the same name not real synonymy of different names. Names are just tags or labels (as we are assuming) so they don’t have synonyms—words with the same meaning. Different words can express the same concept, but names don’t express any concept—so we don’t have the phenomenon of synonymy, i.e. different words for the same concept. Different versions of the same name are really pseudo synonyms, rather like abbreviations of common words in casual speech (e.g. pronouncing “living” without the “g” at the end). Nor would it be plausible to suggest that names are synonymous with themselves alone, so that “Hesperus is Hesperus” contains a synonymy: no word is ever synonymous with only itself, since synonymy requires identity of meaning and hence a meaning that can be otherwise expressed. Names don’t function like regular words (and arguably are not words at all), so they can’t do what genuinely meaningful words do, i.e. produce analytic truths. They look like words, but so do such non-words as “Ha” or “Boo” or “Um”. Names can be used by speakers to perform acts of reference, but that doesn’t qualify them as real semantic units—almost anything can be used to perform acts of reference (e.g. blinking).[4] In any case it would be wrong to try to assimilate names to other types of expression that participate in genuine synonymies.

            It might be said that names admittedly never generate analytic necessities, but don’t they occur in sentences expressing synthetic necessities so that it would be wrong to dissociate them altogether from the concept of necessity? What about “Aristotle is a man” and “Paris is a city” and “Hesperus is Phosphorous”? I would not wish to deny that such sentences express necessities (non-analytic ones), but it doesn’t follow from their existence that the names in them have anything to do with the necessities stated. On the contrary, the necessities in question (“de renecessities”) can be reported using sentences of different semantic types, including descriptive sentences. The necessity doesn’t arise from the name but from the fact: facts of natural kind and facts of identity. The necessity reported has nothing intrinsically to do with the names used to report it but simply arises from the nature of the fact reported. So a monograph on de re necessity might be misunderstood if entitled “Naming and Necessity”, suggesting perhaps that the two topics are connected. Names have nothing to do intrinsically with either analytic de dicto necessity or synthetic de re necessity. Descriptions, on the other hand, have to do with both, since they are the ground of analytic necessity and they express de re essential properties of objects. For someone wishing to point up a connection between types of reference and types of modality, the titles “Naming and Contingency” and “Describing and Necessity” would be the best choices. Of course, the word “and” by itself entails no claim of connection (though it may conversationally imply it), being usable simply to make a list; but if it is desired to suggest a deep connection between language and modality, then the titles just mentioned would be appropriate. In particular, names are indissolubly connected to contingency in that they are incapable of generating necessary truth, whether analytic or synthetic.[5] Necessity either comes from concepts or from the world, but names express no concepts and are not part of extra-linguistic reality. They are merely convenient labels devoid of descriptive content and not integral to de re necessities: they cannot function as necessity generators (unlike concepts and properties). It is as if names have never even heard of necessity and want nothing to do with it. When God created names he still had more work to do to bring necessity into the world.

[1] I defend this position in “Naming and Knowledge”.

[2] There are such things as descriptive names (oxymoronic as it sounds), since we can simply stipulate that a name is to abbreviate a description, and these will give rise to analytic truths. But they are far from being the general case (Mill could have accepted their logical possibility).

[3] Demonstratives also feature in analytic truths in virtue of an appended predicate, as in “That bachelor is unmarried”. 

[4] It can even be argued that names lack reference as a matter of their intrinsic character: it is speakers who refer by using name not names themselves. But in the case of descriptions the words themselves contribute to determining reference; it isn’t just the speaker’s act that confers reference. This is why the content of a description can conflict with the speaker’s intention, which is not the case with names. The reference of a name is a matter of speaker stipulation alone, but a description carries its own reference-fixing content. Name reference is essentially pragmatic while description reference works via semantic content, i.e. meaning. (These are thorny issues, but I take it the basic contrast is plain enough.)  

[5] Someone may object that rigid designation can lie behind necessary truth, so names are logically connected to necessity, as in the necessary truth of “Hesperus is Phosphorus”. Both names designate the same object in every possible world; therefore the identity statement is a necessary truth. The same is not true for definite descriptions, which are non-rigid designators. But this is confused: descriptions can be rigid designators and thus generate necessary identity statements; and not all statements containing names yield necessary truths, e.g. “Plato taught Aristotle”. It is the necessity of the identity relation that lies behind the necessity of “Hesperus is Phosphorus” not the occurrence of names in this statement.  And we would get the same result by substituting rigid descriptions into the sentence (“the planet composed of such a such a chunk of matter”). Names are not distinctive of necessarily true identity statements; still less do they produce such necessities (identity itself does). 

 

 

Universals in Thought

 

We introduce the concept of the particular because we observe distinctness in the world. We introduce the concept of the universal because we observe similarity in the world. I see the cat as distinct from the computer and I conclude that particulars exist; if I didn’t I wouldn’t have much use for the concept of a particular. But I also see that several particulars are similar to each other and I conclude that there are respects of similarity, which I call universals; if I didn’t I wouldn’t have much use for the concept of a universal (suppose I live in a world without similarity). Thus the particular-universal distinction finds a foothold in our description of the perceptible world, from which it might fruitfully migrate: we use it in our general conception of the nature of objects.[1] We might say that it explains what we observe: we see distinctness because particulars exist, and we see similarities becauseuniversals exist. We describe the world as consisting of particular objects with shared properties (attributes, features, qualities, characteristics, forms). Evidently this is a sensible way to proceed, even if we subsequently disagree about what particulars and universals are (are particulars sets of universals or are universals sets of particulars?). But it is also true that we describe thoughts using the same distinction: they have a role in psychology as well as in ontology. For thoughts are about particulars and also about universals: the same duality pervades the mental world as the non-mental world. Why do we talk and think this way about thoughts? Why, in particular, do we think that thoughts have something in common analogous to universals—with the very same entity cropping up in different thoughts? It isn’t that thoughts instantiate universals as objects do (they are not red, square, etc.), so the rationale cannot be the same in both cases; we certainly don’t see that distinct thoughts share the same universal. So what is it that explains our readiness to postulate identity of universals across distinct thoughts? Hugely many different individual thoughts can concern the same universal—or so we suppose—so why do we insist that it is the same entity that crops up in each case? And why must this entity approximate to Plato’s conception of a universal? Why are universals considered an indispensable part of psychology?

            First, let us note the troubles of theories that try to do without them. Suppose we limit ourselves to particulars of various sorts—mental, linguistic, or even immaterial. Then we run up against the problem that we can’t capture the identity of the thought from one instance to another: I can’t think the same thing of two objects when I think that both are red, say. For the particulars in question will vary from instance to instance, according to time or person. If we try to circumvent this problem by appealing to types of particular, then we are back with universals. The general concept can’t be equated with a series of particular happenings; we need something in common between all the cases. We need, that is, the universal RED. Sameness of ascribed property means sameness of property ascribed.  We need something abstract and general to capture the commonality across thoughts. Nor will it help to bring in dispositions, since these vary over time depending on external circumstances and the thinker’s internal psychology. My disposition to use “red” in a certain way at time t need not be preserved to a later time at which I use the same word, despite the fact that I ascribe the same property at the later time. Dispositions to behave in certain ways don’t track meaning or attribute denoted, so they can’t play the role of universals as traditionally understood.[2] We need the robust identity provided by a traditional universal to explain how the same thing can be thought about different things, or about the same thing at different times. But why exactly do we need the same universal at any occurrence of the thought, so long as some universal is present? It can’t be a dated particular or a changing disposition, but why do we insist on invoking the very same universal whenever language appears to suggest it? Why must “red”, say, always be assigned the same universal? Couldn’t this be some sort of oversight or result of laziness? Why don’t we follow the example of “bank” and assign different universals in different contexts? True, that would be contrary to common sense, but it would be nice to know if the usual practice has anything further to recommend it. Why are we so wedded to the idea of universal invariance as between different thoughts and sentences? Why not a more anarchic approach?

            The answer is that logical reasoning requires it. Consider the classic example: All men are mortal; Socrates is a man; therefore Socrates is mortal. We have all been taught that for an argument to be valid there must be no equivocation in the premises and the conclusion. Thus the name “Socrates” must have the same meaning in the conclusion as it has in the second premise, i.e. it must designate the same particular. But the same is true of the predicates, and this requires that “mortal” must have the same meaning in the conclusion that it has in the first premise. That in turn requires that the same universal be assigned to it not some distinct universal (e.g. being vegetarian). So validity depends on constancy of universals through premises and conclusion. We can bring this out be rephrasing the argument using nominative forms: Anything that has humanity has mortality; Socrates has humanity; therefore Socrates has mortality. What we call a universal is just whatever is needed to render this argument valid, which it clearly is: it must not vary between premises and conclusion, as associated particulars and dispositions will. Thoughts occur in trains of logical reasoning and we need entities that will respect the prohibition on equivocation. Thus universals come into the psychological picture; they don’t just belong to the non-psychological world. They have a kind of double life, being related both to similarity among particulars and to validity of argument: we need them in both domains. Again, what they are exactly remains to be determined, but we can’t do without them in some form if we want to make sense of object similarity and argument validity. Postulating them is not some sort of gratuitous mysticism or ontological indulgence; it is warranted by facts that need explaining—the existence of similarity and logical validity. There cannot be logic without something like Plato’s universals, i.e. something constant across particular sentences or thoughts. This means that general concepts cannot be characterized without reference to universals. When we employ concepts we re-apply the same universal over time, which is what happens in a sequence of thoughts that form a logical argument. Token thoughts therefore share universals as particulars share universals—the former by means of intentionality, the latter by means of instantiation.

[1] The word “universal” is not ideal, entrenched as it is, since so-called universals are not universal—not all things have them. Not all things are red or square or even or good, only some are; so what are called universals are really partials, parochials. But there is no natural term that conveys the idea accurately: the closest I can find would be “shareables”, but that is not a felicitous expression. So we may as well stick to “universals” noting that the things in question are precisely not universally distributed. Notice too that the term “particular” is also not particularly descriptive and must be used in a technical sense: how “particular” is a grain of sand? These are both technical terms (not that there is anything wrong with that).

[2] I am here adverting to the kind of argument against dispositions voiced by Kripke in Wittgenstein on Rules and Private Language.