I’m interested to see that Poland has suddenly entered the rankings on this website at number 2, following the USA at number 1 and then the UK. I wonder whether any Polish visitors would care to tell me why–I’m curious.

Meaningless Names

If an expression has meaning, it should be possible to say what that meaning is. Meaning should not be something ineffable. Dictionaries say what meaning is—they specify the meaning of words. But they don’t contain names (or very few).[1] What would they look like if they did? They would certainly be extremely long: all the names of people past and present laid out alphabetically with accompanying definitions. It is not clear what form the definitions would take, however—how would the names be defined? There are two theories of the meaning of names and each recommends a different method of definition: the description theory and the direct reference theory. According to the first, a dictionary entry would take the form “N means the F”; according to the second, it would be “N means x”. For example: “Plato means the teacher of Aristotle”, and “Plato means Plato”, respectively. One specification attributes certain properties to the bearer of the name, while the other simply specifies the bearer. In the case of a general term like “cat” these would be equivalent to the following: “Cat means a small carnivorous mammal of such-and-such an appearance”, and “Cat means cats”. In short, one view says that names mean properties and the other says that names mean objects. So, imagine two dictionaries of names each using one or the other format: would these be good dictionaries? They would not. The first would suffer from the problem of selection—which property specification should be put into the entry for a given a name? People associate different properties with the same name, at a time or at different times, so the choice appears arbitrary; no definition in these terms provides a constant interpersonal meaning. No meaning in a common shared language gets specified, but dictionaries specialize in such general public meanings. All we have is idiosyncratic beliefs about the bearer—hardly the stuff of a shared language. Also, the underlying theory has been subjected to devastating criticism, and we don’t want a dictionary of names to presuppose a false theory. This is simply not what names mean, so we shouldn’t say that this is what they mean in our dictionary. According to the second theory, names mean objects, so a dictionary along these lines will say as much. The trouble here is that such a dictionary will be entirely uninformative: it will simply say things like “Plato (the name) means Plato (the object)”. This is like saying “Cat means cats”: it doesn’t articulate the meaning at all. But meanings need to be informatively specifiable, or else the dictionary is useless (assuming it is a dictionary of a language in that language). Also, the underlying theory is controversial at best, false at worst. We don’t want a statement of meaning to presuppose a questionable philosophical theory of meaning. Thus, we can’t do with names what dictionaries do with other words: that is, we can’t say what names mean in the standard fashion.

The reason for this, I suggest, is that names have no meaning, given that words have meaning only if we can say what that meaning is. They are, as the tradition maintains, mere tags or labels, devoid of meaning. Some names have meaning, such as “Shorty” or “Big John” or “Autumn” or “June” or “Tuesday”, but most names don’t. If they have no meaning, then we cannot state their meaning; that is why we cannot state it. For the same reason, we can’t state the meaning of horses or toadstools or accents or pauses or sighs. Names are used in communication, as are many things, but they are not thereby elements with meanings. It is true that as so used they have denotations, but the denotation isn’t part of their meaning—for they have none. Names are meaningless tags for objects denoted by speakers. It might be wondered how they can have denotation but no meaning to determine the denotation. The answer is that there are terms that fix their reference but don’t determine their meaning. Suppose we say that the reference of “Plato” is fixed by “the object people call Plato”: that explains how the name latches onto one thing rather than another, but it tells us nothing about the meaning of “Plato”—nothing that could be usefully reported in a dictionary. So, names can feature in sentences which have meaning and truth conditions without themselves having any meaning.[2] Names don’t bring meanings with them that combine with other meanings to produce meaningful sentences (they have a use but no meaning). As theories of meaning, then, the two traditional theories are barking up the wrong tree—there is no meaning up there to start with. There may be truth in the theories in other respects, but not as contributions to the theory of meaning. They are certainly not paradigms for other classes of expressions that do indisputably have meaning. There is no such thing as a “semantics of names” if that means a theory of meaning for names. If names had meaning, it would be possible to compile a dictionary specifying their meaning; but it’s not, so they don’t. That’s why there is no dictionary devoted to names (even of famous people, as it might be Dictionary of the Names of the Stars).

Descriptions and demonstratives have meaning, as do prepositions and conjunctions, but not proper names. We don’t think we are expanding our vocabulary by learning new names; nor do we complain that our names misdescribe us. We recognize that they are just convenient devices with no intrinsic significance. If a name’s meaning were identical to its reference, then it would have a meaning, as the word “red” has a meaning by denoting the color red; but then it would not be a mere empty label. That theory would imply that the meaning would perish with its bearer (as Wittgenstein once remarked), which is not what meanings are supposed to do. But associated descriptive beliefs are too variable and contingent to constitute meaning, so this theory is ruled out too. Names have no meaning at all, in the way other words do, so there is nothing there for theories of the meaning of names to be theories of. A lot of recent philosophy of language has been chasing a mare’s nest.[3]   

[1] The OED contains the name “Jesus” and the names of the planets (though not “Shakespeare” or “London” or “Plato”), but the entries are hardly adequate to specify a meaning. For “Jesus” we read “the central figure of the Christian religion”: is that what the name meant to his family or the apostles? At best this is a reference-fixer for us now. Ditto for the entries purporting to give the meaning of the names of the planets.

[2] A pure causal theory allows for this possibility too.

[3] Oddly enough, the entire history of philosophical discussion of names has lost sight of their most obvious characteristic, which is difficult by this time to recover, namely that names are empty labels—meaningless sounds or scribbles conscripted to serve acts of speaker reference. They have no cognitive content, no intrinsic significance, no propositional potency. They are semantically vacuous; dead on arrival. Yet they do a useful job in acts of speech. See my “Completely Empty Names” for more on this theme.

Dictionaries and Meaning

Let’s stipulate that a theory of meaning is a specification of the meanings of all words, phrases, and sentences of a language.[1] If we knew that, we would know what meaning is, presumably. What form should such a theory take? I will suggest that it should take the form of a dictionary.  A dictionary tells us what words mean, and words compose phrases and sentences, so it ought to do what we desire—state what expressions of a language mean (finitely, compositionally). We will need to add a grammar, i.e., rules of sentence construction, but the hard work will be done by the lexical component. But how does a dictionary set about specifying meanings—what form does it take? We must inspect a good dictionary and see how it is actually done. I will choose the Concise Oxford English Dictionary (Eleventh Edition) and select two words at random: “cat” and “close”. For “cat” we find that word in boldface followed by an “n” for noun. Then we read: “a small domesticated carnivorous mammal with soft fur, a short snout, and retractile claws”. For “close” we get the same boldface but followed by “v” for verb, then “move so as to cover an opening”. Mark well the form of these inscriptions; you will notice a glaring use-mention mix-up. The word defined initially appears as mentioned, though not within quotation marks, given that it is said to be either a noun or a verb (as in ‘“cat” is a noun’). But immediately the word slides into being used as we are told what a cat is or what it is to close something (as in “a cat is a small domesticated carnivorous mammal” etc.). Nowhere does the dictionary explicitly say anything of the form ‘“cat” means…’. First it tells us what grammatical category the word belongs to; then it says what kind of thing the denoted thing (object or act) is. However, we would be within our rights to read into the entry something of the form “w means such and such”, where w is a word and “such and such” stands for a description of a thing. Thus it is that the dictionary specifies the meaning of a word: it says what kind of thing it is that the word denotes (expresses, refers to, indicates). As we might say, it spells out the meaning of the word by using a longer form of words that ascribes various properties to the thing in question. It makes the meaning explicit, displays it, articulates it, analyzes it. It would be wrong to say that the definition provides a synonym for “cat” or “close” (a thesaurus does that): “cat” has no synonym in English (unless we include “pussycat”) and “shut” does not occur in the dictionary entry for “close” (or vice versa). Synonyms in the strict sense don’t tell us what a word means by spelling it out; they just provide another word that has the same meaning. Only dictionaries linking one language to another do anything like that—they translate. An English dictionary doesn’t translate; it paraphrases. It is informative precisely because it does more than translate (and also less); it analyzes, disassembles. If you are in doubt about the exact meaning of an English word, a good dictionary will remedy your ignorance, add to your stock of knowledge; it will teach you something. It specifiesthe meaning—gives you the specifics, the details. It doesn’t just give you a word which happens to have the same meaning. Thus, it tells you the meaning of sentences containing the word in question; you can transfer your knowledge of word meaning into knowledge of sentence meaning—you don’t need to start again with the sentence, novel though it may be. The dictionary is therefore compositional: the meaning of the whole is a function of the meaning of the parts. This is something a theory of meaning ought to be. A dictionary (plus a grammar) does what a theory of meaning ought to do—provide a finite compositional analytic specification of the meaning of every well-formed expression of the language. It is not clear what else remains to be done, given the task we set ourselves.

It might be protested that such a theory fails to do the main thing a theory of meaning ought to do, viz. offer a central concept that captures all of meaning. It provides nothing analogous to truth conditions theories or verification conditions theories or use theories or picture theories or intention-based theories. Two points may be made in reply. The first is that this is actually a desirable property of the dictionary theory: such alternative theories are inevitably procrustean in that they fail to apply to all types of sentences (imperative, interrogative, expressive), as well as suffering from defects often pointed out. We don’t want a single property of all types of sentences, since sentences are of irreducibly different types; we just want a theory of word meaning plus combinatorial rules. Second, the theory is not as pluralistic as you might suppose: it specifies the meaning of all words in a similar manner, i.e., by ascribing attributes to things. Cats have the attributes listed and acts of closing likewise. We might say that the dictionary theory is an “attributional theory” of meaning: it says that all meaning involves property attribution.[2] Meaning is essentially attributive. Different attributes get attributed, but the common thread is the act or process of attribution. Typically, the attribution is complex not simple, so that a certain “holism” applies: the meaning of “cat” involves a bunch of attributes; the attributes are bundled together, synthesized. In addition, there is room for a theory of the nature of attribution, and of what is attributed. What kind of operation is this? Do different kinds of attribute produce different kinds of meaning? What is an attribute? There is plenty of room for philosophical thinking about meaning once we have settled on the general form of a meaning theory. Are all attributes made of sense data, or physical things? What is it to “grasp” an attribute? Are attributes platonic universals or conceptual entities? The dictionary theory is not a cop-out but a starting-point; it defines the general territory to be investigated. The central point of it is that meaning specifications involve attributing properties to things; they go outside of language narrowly conceived.[3] The meaning of “cat” involves facts about cats.

And so, we come to encyclopedias. People used to contrast dictionaries with encyclopedias—the former about meaning, the latter about facts. But this is a false dichotomy. A dictionary is a partial encyclopedia and a truly encyclopedic encyclopedia includes a dictionary. The latter proposition is true because a real encyclopedia will cover languages inter alia, and the former is true for the reason already stated—dictionaries contain facts about things (cats, closings). These are not disjoint enterprises. In order to specify meaning we need to advert to properties of things in the world, those comprised in the meaning. These properties are apt to be constitutive, or at least commonplace; they are a subset of the totality of properties possessed by a thing (or kind of thing). Semantics is also physics, chemistry, biology, psychology, common sense—whatever makes a thing the thing that it is. To specify meanings is to specify facts. Dictionaries and encyclopedias are not about different things; they are just more or less inclusive. Have you ever wondered why there aren’t any “conceptopedias”—alphabetized books devoted to the nature, analysis, and operations of concepts? Concepts are not the same as words or extra-mental facts, so shouldn’t there be a type of book that does for them what dictionaries and encyclopedias do for words and the world? Is it because there are really no concepts, or that conceptual analysis is lamentably underdeveloped? No, it’s because such a book already exists—in the shape of dictionaries and encyclopedias. So, there is no market for an extra book about concepts. These publishing divisions are arbitrary; they don’t correspond to book natural kinds. It isn’t that there are semantic truths and conceptual truths and factual truths all sealed off from each other; they bleed into one another. In particular, dictionary truths are a species of encyclopedia truths.

This conception of meaning might work for nouns, verbs, and adjectives, but what about conjunctions, pronouns, demonstratives, prepositions? We don’t normally look these up in the dictionary (when was the last time you looked up “and”?), but they are in there; and they receive a different kind of treatment from that accorded to the other three categories of word. For “and” we have “used to connect words of the same part of speech, clauses, or sentences”; for “not” we have “used chiefly with an auxiliary verb or ‘be’ to form the negative”; for “that” we have “used to identify a specific person or thing observed or heard by the speaker”. In these cases, the dictionary evidently opts for a use theory of meaning: it tells us how the words are used to form sentences not what the nature of some type of thing (object, action, characteristic) involves. Thus, we might conclude that meaning divides into two types—attributive and functional. That would be fine as far as the attributional theory is concerned; it merely requires us to recognize that there is a secondary kind of meaning that calls for a different account. We opt for a dualistic theory of meaning. However, there is another option: get creative with the attributional theory. Thus, we could suppose that “and” and “not” denote truth functions with characteristic logical properties; and “that” denotes, on its different occasions of use, certain types of objects with determinate properties. We go the denotational route and apply the attributional theory across the board. There is no serious challenge here to the attributional conception of meaning, just some minor tinkering.

The most substantive line of questioning comes from metaphysics. Okay, we can conceptualize meaning as holistic attribution of properties and look to the dictionary to provide the details, thus achieving what we set out to achieve, viz. specifications of meaning for an unlimited class of expressions. But none of this tells us what kinds of things constitute meaning, metaphysically speaking; so, we don’t yet know what meaning is. I think this is entirely correct: our aims were modest, or should have been (we weren’t trying to save humanity from itself or anything like that). For all we have said, meanings might be sense data, or platonic universals, or atoms in the void, or mathematical models, or words in the language of thought, or mental images, or ideas in the mind of God. It is perfectly true that none of these theories has been precluded. But that is really as it should be: don’t expect the theory of meaning to do your metaphysics for you. It isn’t a shortcut to anything, a “turn” that will get you to your destination quicker. It will tell you what a sentence—any sentence—means, but it won’t tell you the ultimate nature of the reality meant. The sentence “snow is white” means that snow is white, but it is a further question what snow ultimately is and what being white amounts to in the great scheme of things. Let’s keep these two questions separate, so that we can answer the former without being held hostage by the latter. To me it is quite reassuring that the problem of meaning can be (partially) solved by the not-so-humble dictionary.[4]

[1] This is Davidson’s original formulation of the aims of a theory of meaning, which he claims can be satisfied by Tarski’s theory of truth, suitably augmented. I am suggesting an alternative approach to the same question.

[2] Notice that dictionaries don’t have entries for nonsense words like “borogrove”, simply because nonsense words don’t attribute any determinate properties to their purported reference. Nor do they contain the likes of “grue” and “quus”, since these words refer to no natural kind with a fixed humanly-natural nature, however useful they may be philosophically.

[3] It has not failed to cross my mind that this position is not a million miles away from classic description theories. It’s a relevant fact that dictionaries don’t contain proper names (no meaning, you see). Mere labels don’t attribute.

[4] I am intending that the position outlined should be, and be seen to be, massively deflationary, given the recent history of the philosophy of language. There is much less to the theory of meaning, strictly so called, than has been supposed. In this respect I am outDavidsoning Davidson. He thought we could do theory of meaning on exiguous truth-theoretic foundations by wheeling in Tarski’s theory of truth; but I am suggesting doing away with even this degree of novel theorizing. The dictionary can do all the work we need. Radical? You bet.