CHOMSKY John Collins
5. Problems with externalism
5.1 Meaning and truth
In general terms, to specify the meaning of an expression is (at least) to specify its contribution to the truth conditions of the constructions of which it is a possible constituent; but truth and reference are external relations; so, meaning essentially involves externalia. Chomsky (2000, p. 174) does not so much dispute the intimate connection, but rather suggests that semantic theorizing is really a part of syntax, in a broad sense. The basic point here is that the kind of semantic theorizing that assigns unique external ‘semantic values’ to linguistic material often does not reflect, and is not constrained by, our independent concep- tions of the relevant objects. So, of course, truth and reference are external relations, but to specify linguistic meaning is not to specify independent externalia to which lexical items refer or are true of. To see the point here, consider an adaptation of one of Chomsky’s examples:
(12) Bill had the book in mind for four years, and was very pleased when it sold 1,000 copies in the first week, especially since it weighs two pounds.
If we take book to refer to books, (12) would appear to commit us to some thing that can be abstract/content (be in one’s mind), be a type
that can have copies, and be concrete (have weight). Now, the point here is not that the existence of such a thing is impossible or contradict- ory. The point, rather, is that understanding (12) hardly commits one to the possibility of such a thing, which flies in the face of our common- sense metaphysics that separates concrete from abstract objects. Of course, it would be madness to conclude that book does not refer to books (what else would it refer to? Cats?). The point of the example is only to highlight that it would be a mistake (a mistake we never make) to project from the semantic aspects of a word to a kind of object that independently exists. This being so, it is equally misguided to think that the semantic properties of our words are somehow answerable to or derivable from independently existing kinds of things. None of us really believe in such things as books, if by book one means things that realize all the properties book may be used to speak about.
Similar remarks apply to any noun one may care to consider (to say nothing of the abstractness of verbs and adjectives). Consider proper names.
It is standard in philosophical discussions to specify the meaning of a name disquotationally, which simply means that one specifies the refer- ent of the name by using the name. For example:
(Dis-Name) ‘Mary’ refers to Mary
There is something right about such specifications, of course, but only in that the arbitrariness of the name–world relation is captured. What the disquotational method cannot explain, at least not without further ado, is how proper names may designate persons, concrete things, stuff or abstract objects, depending on the choice of verb, independently of whether we take the name to refer to any existent thing or not. Consider:
(13) a. After the steamroller was finished, Father Christmas was smeared across the road
b. What this drama series needs is a Shakespeare
It doesn’t matter if Father Christmas and Shakespeare are fiction names; they behave just like referring proper names in being able to serve as both mass nouns and abstract nouns. It is hardly for linguistics to decide or even care if there really was a playwright from Stratford, rather than one or more secret authors.
The problem for externalism in the cases rehearsed is not that names cannot be construed as labels for things in the world, and so cannot possess meanings specifiable merely by ‘disquoting’. The problem is that such a conception of names does not appear to explain much of their content. Why on earth should names have such varied semantic aspects invariant over the existence and inexistence of their putative referents, if they are mere labels for things?
The bottom line, here, is that a name, just like any other lexical item, offers a complex set of perspectives with which we can speak about the world and fictions indifferently, but no one imagines that the perspectives at issue all reside in things out there anyhow. Chomsky (2004, p. 391) expresses the general point:
We can, if we like, say that the word ‘book’ refers to books, ‘sky’ to the sky, ‘health’ to health, and so on. Such conventions basically express lack of interest in the semantic properties of words and how they are used to talk about things.
5.2 Going right and wrong
We can, it seems, go wrong about language, both semantically and syntactically. If, however, properties of language were in fact entirely determined by individual cognitive states, then how could we go wrong about language? Anything we thought was right, would be right, given that there would be, by hypothesis, no external language to which we are answerable. So, the very possibility of error appears to require externalism.
First off, internalism is not inconsistent with the possibility of linguis- tic error. All that internalism entails is that speakers do not go wrong about an external language/meaning. Internalism is perfectly consistent with the bare idea of speakers going wrong, deviating from their own criteria or standards, which might even be conceived by speakers as involving a public language or ways of speaking given a particular audi- ence, say. So, what externalism requires is an argument that linguistic error actually involves the appropriate externalia, not the mere presump- tion that it does. It is not so easy to offer a relevant case in favour of externalism. Speakers are interested in converging, perhaps for general psychological reasons that have little intrinsic connection with language, but the convergence doesn’t require or create anything that has an
independent life to which speakers can go right or wrong about – they are interested in converging with each other, not on a language.
Secondly, as intimated, to grant the phenomenon of linguistic error is not to grant that a theory of language must explain it. Speakers’ sen- sitivity to error appears to be a massive interaction effect, involving many cognitive capacities, including self-consciousness, above and beyond language. Acknowledging error is not constitutive of being linguistically competent, no more than not being crass or impolite is. We perhaps unreflectively consider being susceptible to correction to be central to language because that is how we learn language in the first place (Dum- mett, 1978; Quine, 1960; Wiggins, 1997; among many others). As it is, no empirical study has supported such ‘common sense’: children learn language with remarkably little negative feedback of any kind; they tend to ignore what feedback they do receive; and they make practically no mistakes, anyhow (Crain and Thornton, 1998; Roeper, 2007). As for adults, we might consider a speaker who persists in what we regard as an ‘error’ to be ignorant or obtuse, but we would be leery of describing him as linguistically incompetent.
5.3 Thought experiments
A pair of thought experiments has convinced many a philosopher of the truth of externalism as regards both linguistic and mental content. The thought experiments, due to Putnam (1975) and Burge (1979; 2007), are designed to show that we individuate a subject’s thought or mean- ing relative to their environs, physical or social, not merely how things locally (internally) stand for the subject. Thus, imagine a Twin Earth exactly like Earth in every detail save that the stuff they call water has a chemical composition of XYZ, not H2O. XYZ falls from the sky, quenches thirst, comes in snow and ice at winter time, and so on, but is not the same stuff as H2O. Putnam (1975) asks us to reflect on Oscar (on Earth) and Twin-Oscar (on Twin-earth), who are physical Doppelgangers. Now, the intuition we are invited to accept is that while both may be equally ignorant of the respective chemistry of their planets, they possess dis- tinct concepts, albeit homophonously expressed. The reason is that Oscar’s thoughts are about (made true/false by) water (H2O), whereas Twin-Oscar’s thoughts are about some other stuff (XYZ). Burge offers a social variant of the same externalist intuition, where it is community
agreement about word meaning (as opposed to the facts of chemistry) that is individuative of the concepts we linguistically express. The point of both ‘experiments’ is to trigger the intuition that we attribute thoughts partly on the basis of how things are in the world independ- ently of how the thinker conceives of the world.
The issues arising from this pair of ‘thought experiments’ take up a good deal of the philosophy of language and mind of the past two decades. Space demands a brevity that I hope is not glib.
Putnam and Burge are clearly right that we individuate certain con- cepts in terms of ‘essences’ that are not exhausted by the surface prop- erties of typical samples of the kind. So much defeats a traditional empiricist view of concepts and is in fact supported by much empirical work in developmental psychology (e.g. Carey, 2009; Keil, 1989). What is much less clear is why such anti-empiricist essentialism supports externalism.
The Twin Earth ‘experiment’ turns, not on water being H20, but only upon our meaning by ‘water’ some stuff reference to which is not fixed by its apparent properties. Satisfying this condition does not involve there being any actual kinds at all. For example, we have lots of kind concepts that have no corresponding kinds (earth, air, fire, etc.); as far as cognition is concerned, it would be an accident that a given kind term corresponds to an actual kind (Chomsky, 2000, p. 149). In other words, our conceptions of word meanings might be externalist, but that is no reason to think that externalism is true, only that it is an intuition that needs explaining.
Apart from such a general complaint, the intuitions themselves are less than clear-cut. We happily call stuff ‘water’, and would resist correc- tion, even though we know what we happen to be referring to it isn’t water by chemical standards (the contents of practically any river would do). So, even when we know the relevant chemistry, it appears not to anchor our referential use of language. Equally, we certainly don’t call tea or coffee ‘water’, even if reflection would inform us that the cup of cof- fee is more water than the same amount of liquid drawn from the aver- age river or lake. Burge (2003, p. 457) responds to such complaints by suggesting that ‘We all recognize that tea, coffee, and Lake Erie are mostly water’. Well, ‘we’ who read Burge no doubt do, but externalism is supposed to be a thesis about human conceptuality, precisely not a
thesis about the general knowledge of this or that population. Milk, orange juice, cucumbers, most of the organic world, and comets too, are ‘mostly water’, but not even someone included in Burge’s ‘we’ is likely to think of themselves, qua being ‘mostly water’, to fall under the extension of ‘water’. Coffee made with milk is still coffee, not water, even though milk is ‘mostly water’. I do not think that any orange-coloured, citrus- flavoured sweet liquid is orange juice, but I don’t think orange juice is water either. Water, water everywhere, but not a drop to refer to.
As it stands, inquiry into the constitutive, metaphysical conditions of the individuation of this or that property appears to be an inquiry into the distinctive human cognitive profile, which does not presuppose any external kinds ‘out there’ anyhow.