Externalist Thoughts and the Scope of Linguistics

A common assumption in metaphysics and the philosophy of language is that the general structure of language displays the general metaphysical structure of the things we talk about. But expressions can easily be imperfect representations of what they are about. After clarifying this general point, I make a case study of a recent attempt to semantically analyze the nature of knowledge-how. This attempt fails because there appears to be no plausible bridge from the linguistic structure of knowledge-how reports to knowledge-how itself. I then gesture at some other places where the connection between linguistics and metaphysics is commonly, but illegitimately, assumed. 1 An Alleged Connection between Semantics and Metaphysics What is the relation between philosophy and linguistics? Many philosophers endorse, tacitly or explicitly, the view that linguistic theories can support philosophical theses in various substantial ways. In particular, there’s a long tradition in philosophy of supposing that questions about the metaphysical nature of certain phenomena can be uncovered by investigating the semantics of the expressions we use to talk about them. This has been one of the central underlying components of the “linguistic turn” and of various philosophical programs of “semantic analysis”. In short, the thought has been that for many philosophical topics, if you can discern the semantics, you’ll uncover the metaphysics. The assumption that linguistics can supply evidence for metaphysical theses has been endorsed by many, including those who are well aware of the finer details of both philosophy and of contemporary linguistic theory. For instance, James Higginbotham, who has made numerous substantial contributions to both fields, has endorsed such an assumption in numerous places (e.g., Higginbotham 1989, 1992, 2001, 2004). Higginbotham suggests that linguistic theory can be a substantial aid in “the clarification of the nature of our thoughts, what we actually express when we understand one another” (Higginbotham 2004, 575). Initially, this might seem like a primarily psychological claim. However, Higginbotham then goes on to individuate (what he calls) “thoughts” in a Kent Johnson 20 21 Externalist Thoughts and the Scope of Linguistics manner that makes them of central philosophical importance. In particular, he suggests that thoughts should be individuated, at least in part, by things in the world that thoughts and their parts refer to. For instance, he writes: Assume, what is common enough although open to question, that this clarification calls first of all for the exposition of the truth conditions of sentences, as they occur as parts of total languages, and within the contexts of their potential utterance; and assume also that any correct account of what we are inclined to assert must, over a wide domain, make us pretty much right about the way things are. Then the truth conditions of much of what we believe must be such as to be actually met; and this implies that what turns up in the metaphysics of semantic investigation cannot be passed off as a mere manner of speaking, but constitutes our best conception of the way the world is... . To say this much is not to commit semantics to an all-out realism with respect to the elements invoked. It does, however, imply that the question of realism must be taken seriously, including familiar questions of possible reduction or relativization of the objects involved, as for instance whether individual events can be reduced to regions of space and time, or whether possible worlds can be modeled as maximally consistent sets of propositions (Higginbotham 2004, 575; italics added). Here we see a decidedly “externalist” view of linguistics, in the sense that that the subject matter of linguistics is not limited to the states and processes in a speakers’ head. (Higginbotham has argued for such an interpretation of linguistics in numerous other places, e.g. his 1989.) As Higginbotham’s discussion makes clear, his “thoughts” are either identical or closely analogous to what other philosophers would call “propositions”, where the latter are interpreted in a similarly externalist way. (It’s worth observing that externalism in linguistics contrasts starkly with the more common “internalist” view, whose roots lie in Chomsky’s work on syntax. According to internalism, semantic theory should be construed as a system of mental representations of the world. Thus, talk of an expression’s “reference” is really just shorthand for some kind of conceptual/intentional state or capacity of the speaker that is associated (in some linguistically relevant way) with the expression. Thus, internalism attaches much less importance to the question of realism that Higginbotham alludes to in the passage above.) It would be great to see the sort of unification between philosophy and linguistics that Higginbotham (and numerous others; cf. below) propose. It is, however, one thing to suggest such a coordination between the disciplines, and another thing for such a view to command authority. In the next section (§2), I explain my skepticism about this view by presenting a picture of linguistics and making some distinctions. These distinctions, I believe, are what really undermine Higginbotham’s two suggestions. However, in order to bring my point home, I then (§3) make a detailed case study of a recent attempt to establish a philosophical thesis by looking to current linguistic theory. We’ll see there that the distinctions drawn in §2 provide the crucial conceptual machinery for organizing an analysis of what goes wrong with this analysis. Next in §4, I gesture at a few more places where I think similar misapplications of linguistic data have appeared in philosophical theorizing. I conclude in §5. 2 Languages as representational systems Regardless of how one interprets linguistics, it is hard to deny a claim like (Rep): (Rep) Languages are representational systems, and so may produce imperfect representations. (Rep) tells us that linguistics theorizes about representational systems. Whether our sentences “express thoughts” in the internalist sense of Chomsky (e.g., 1980, 230) or depict the world, they are representations of some sort. As such, these representations may be imperfect. That representations are frequently imperfect is a ubiquitous phenomenon. For instance, authors of maps and diagrams often deliberately depict relevant objects (e.g. roads) as proportionally larger than objects less likely to be relevant (e.g. surrounding countryside) (cf. Matthews 1994 for discussion). Such representations also frequently omit much structure in what is being represented; e.g., a road that heads due north with a small bend in its middle may appear on a map as a straight line. In the linguistic case, consider that the semantic representation associated with the noun water is plausibly of a continuous, nonparticulate substance, even though water may be composed of discrete molecules. (In any case, it’s at least possible that language and the world would differ like this, which is all that really matters here.) Many further examples can be taken from the history of philosophy. For instance, a number of philosophers have echoed the sentiments of Thomas Reid, who observes that there are ‘phrases which have a distinct meaning’ other than what their overt form suggests, and that ‘there may be something 1 I take it that whatever else “expression” is, it is a form of representation. Kent Johnson 22 23 Externalist Thoughts and the Scope of Linguistics in the structure of them that disagrees with the analogy of grammar or with the principles of philosophy... . Thus, we speak of pain as if pain was something distinct from the feeling of it. We speak of pain coming and going, and removing from one place to another’ (Reid 1785, 167–8). Although we may say John has a pain, it doesn’t follow that our language represents the situation so accurately, as involving two things, John and a pain. (One’s metaphysics may lead one to disagree here, but my point is only that it’s not apriori true that linguistic structure accurately mirrors the world.) Similarly, Chomsky has noted that the proper semantic representation of The value of my watch is increasing may be that of a certain kind of object moving on a unidimensional scale, but of course this has nothing to do with what it is for something to rise in value. (These sorts of representations are quite natural from a processing perspective, since they probably recruit our abilities for spatial cognition, which appear to be quite basic to human representation (e.g., Jackendoff 2002, Burgess et al. 1999).) Actually, for present purposes, we don’t even need the claim that our sentences are representations, since if they’re not, then they still don’t provide perfect representations. All we really need to make the above points is that our sentences’ representations can be imperfect. (Indeed, in the case of human psychology, it may be important that some of our representations are imperfect. The literature on learning and memory suggests that learning complex facts or skills often crucially involves learning to ignore irrelevant information; e.g. learning to count involves ignoring whether the objects being counted are apples or oranges.) (Rep) acts as a degree of freedom lying between the structure an expression attributes to what it represents, and the actual structure of what it represents. In order to respect this degree of freedom, it will be helpful to distinguish three types of structure relevant to theories about language. First, metaphysical (or empirical) structure is the structure present in the external thing that an expression denotes. In the typical case, this sort of structure is uncovered by scientific or metaphysical investigations into some extra-linguistic phenomenon. E.g., our best chemical theories tell us that a given quantity of water is (primarily) composed of a discrete number of molecules of H2O. The fact that water has this structure is independent of any linguistic properties of the English word water, which may well treat it semantically as a continuous, non-particulate substance. Thus, there may be little or no interesting relation between an object’s metaphysical structure and the structure of the expression describing it. Second, an expression’s syntactic structure concerns how the lexical elements of a sentence are actually organized into constituents that form the target expression, regardless of how untrained users of the language tend to think about this structure. It’s well known that much syntactic structure isn’t accessible to conscious reflection. Some of this tacit structure is also relevant to semantic interpretation, which is the norm for every other cognitive ability that psychologists have investigated. It takes much experimental work, for instance, to see the kinds of structural features that are relevant to our interpretation of a visual scene (e.g. Hoffman 1998). Third, semantic structure is the structure of the meaning (in the internalist’s sense) of our expressions. An expression’s semantic structure may not be identical to the structure of what it represents (e.g., the quote by Reid above). There is also no a priori guarantee that the semantic structure of an expression will mirror its syntactic structure. The relation between syntax and semantics remains an active area of research (cf. discussion and citations below). Often, though, syntactic and semantic structure go hand in hand, so I’ll frequently use the phrase linguistic structure to lump them together. According to the strategy proposed by Higginbotham (and others), we can investigate the nature of the metaphysical structure of parts of the world by investigating relevant parts of the semantic structure of the language we use to talk about them. Using the terminology just introduced, this suggests that there should be some kind of significant connection between (certain aspects of ) metaphysical and linguistic structure. However, a moment’s thought shows that none of these kinds of structure logically entails anything about the other. Indeed, from our present perspective, it is hard to see how some of these forms of structure would have very much to do with one another at all. So inferences from one kind of structure to another require some defense. In particular, in order to have any hope of uncovering the underlying nature of some metaphysical phenomenon, we must forge some theoretical links between some aspects of linguistic structure and some aspects of metaphysical structure. The importance of arguing for such links should not be underestimated. In what follows, I want to illustrate the importance and difficulty of finding such links. I’ll do this by examining a recent attempt to use linguistic structure to uncover the true nature of a philosophically interesting phenomenon, that of knowledge-how. The attempt fails precisely because the relations between linguistic and extralinguistic structure go simply ignored. To the extent that we cannot take such relations for granted, the general attempt to relate linguistics and philosophy remains incomplete. Kent Johnson 24 25 Externalist Thoughts and the Scope of Linguistics 3 Language as revealing our “externalist” thoughts As Ryle (1946) and many others have observed, there appears to be a real difference between knowledge that something is the case (e.g., that snow is white) and knowing how to do something (e.g., how to juggle). The latter form of knowledge is distinctive, not least because it appears that one could know how to do something without being able to articulate, recognize, affirm, etc. any propositions that describe the content of that knowledge. People like Ryle have taken such data seriously, as marking a genuine theoretical distinction. Recently, however, the distinction between knowledge-how and knowledge-that has been attacked. Stanley and Williamson (2001) – hereafter SW – ‘contest the thesis that there is a fundamental distinction between knowledge-how and knowledge-that. Knowledge-how is simply a species of knowledge-that’ (SW 2001, 411). Their argument is based around the claim that our best semantic theories represent ascriptions of knowledge-how as ascriptions of knowledge of a proposition. E.g., they argue that the relevant interpretation of (9) has a semantic form similar to (10). (9) Mary knows how to juggle. (10) There is a way w such that Mary knows that w is a way that she can