Analytic metaphysicians assume that philosophical — especially ontological — questions can (and perhaps should) be taken at face value, as Timothy Williamson urges. For Williamson, taking philosophical problems at face value and accepting “the” linguistic turn are mutually exlusive (and, it seems, collectively exhaustive) choices. The expression “linguistic turn” Williamson associates with Michael Dummett, and his Frege-centered version of the history of analytic philosophy:
Only with Frege was the proper object of philosophy finally established: namely, first, that the goal of philosophy is the analysis of the structure of thought; secondly, that the study of thought is to be sharply distinguished from the study of the psychological processes of thinking; and, finally, that the only proper method for analysing thought consists in the analysis of language. . . (quoted by Williamson, Philosophy of Philosophy, pp. 11-12)
Dummett also thought that Frege had shifted the focus of philosophy from epistemology, which had dominated since Descartes, to logic as the structure underlying language. So the primary agenda of philosophy shifted from theory of knowledge to the philosophy of language. But despite his lifelong attention to this question, Dummett was too vague. His criteria both include too much and exclude too much. As Williamson (pp. 12-3) points out, by Dummett’s criteria Derrida could be counted among those who took the linguistic turn, while Russell can’t. And it does seem odd to define the linguistic turn in such a way that the Oxford ordinary language philosophers following Austin turn out not to qualify. From their point of view, it’s hard to consider Frege as even having had a “philosophy of language” (quite apart from Joan Weiner‘s point). The “linguistic turn” in the tradition of Locke (to which they belonged) is very different from that in the tradition of Leibniz (to which Frege belonged). Both proponents and detractors of “the” linguistic turn, then, need to recognize that this term can be used only with an indefinite article, with an adjectival qualifer, or in the plural (Losonsky’s book on Linguistic Turns is good on this).
Carnap’s linguistic turn certainly differed from Frege’s. But, you will object, surely a specifically Carnapian linguistic turn, if there is one, has to stand or fall with the Fregean one? Carnap had, after all, been Frege’s student, and the continuities between them are well known (Gabriel, Ricketts, etc.). Even in this literature, though, the radical discontinuities are also fully acknowledged. As Ricketts points out, Carnap rejected Frege’s universalist conception of logic “and with it Frege’s identification of logical generality with quantificational generality.” Carnap thought “that no attempt to formulate the principles of demonstrative inference [so as] to show that alternative formulations are either notational variants or incorrect can be successful.” Moreover, in repudiating this ideal of a unique, science-embracing logic, “Carnap rejected Frege’s assumption of a common store of logically interrelated thoughts expressed by the sentences of colloquial language and perspicuously expressible by sentences couched in the framework of [Begriffsschrift].” (Ricketts in Reading Putnam , pp. 190-1)
Still you might think that a Carnapian linguistic turn must be dependent somehow on the specifically Fregean one. And yes, this was true of the early Carnap, but the principle-of-tolerance Carnap is a different story. Tolerance represents a linguistic turn of a completely different sort from the Fregean one; it began, in fact, with a complete rejection of that earlier turn (see my 2007 and 2009 papers with Steve Awodey).
Behind this rejection was a new conception of philosophy, which broke decisively with the traditional conception of philosophy as a kind of knowledge, resulting from analysis. Russell had said “That all sound philosophy should begin with an analysis of propositions is a truth too evident, perhaps, to demand a proof.” In remarks like this, Peter Hylton points out, analysis is regarded as “quite uncontroversial, no more than common sense,” though in fact
Any given conception of . . . analysis is in fact inextricably tangled in metaphysics. The idea of “finding and analysing the proposition expressed” by a given sentence is one that makes sense only within a given philosophical context, which imposes constraints on the process; the philosophical context cannot itself, therefore, be based on a neutral or uncontroversial notion of analysis. (Hylton’s 2005 paper “Beginning with Analysis,” p. 30 in reprint)
Likewise Frege when he “finally established” — Dummett’s claim — the “proper object of philosophy” as “the analysis of the structure of thought.” And “the only proper method” for the analysis of thought, in this view, “consists in the analysis of language.” As Hylton reminds us, though, there is a basic problem with all this talk of analysis:
The obvious problem is that what we actually have to deal with are not propositions but utterances of sentences. If we cannot assume that the proposition expressed by a sentence has the same structure as the sentence itself, then we need some other guide to the structure of the proposition. If two philosophers consider a given sentence, and one says that its analysis is so-and-so, and the other says it is such-and-such, how can this dispute be settled? Saying that analysis is a process of finding a sentence which accurately reflects the structure of the underlying proposition is of no help, for each philosopher can claim to have done that. . . (same paper, pp. 43-4)
So there is no way to resolve this dispute other than invoking (metaphysical) principles to constrain the process of discovering “deep” structure or content below the surface of what people say and write. Carnap left behind this entire program of trying to identify any such underlying structures. What we can unambiguously find in a language is put there by us, by deliberate stipulation; it is there because we put it there, in a language we specify. And there are no criteria whereby any of these stipulated set-ups could be singled out as “correct” or “true”; they can only be more or less suitable for this or that purpose. They were candidates not for truth (which could only be defined within one of them, relatively to it) but rather for usefulness. Philosophy became not the discovery of something under the surface of ordinary language, but the creation of something new — not a pursuit of truth but a creative engineering task, of locating strategic improvement opportunities and devising more precise replacements for the key parts of the vague and confusing conceptual environment we live in.
Carnap’s linguistic turn isn’t the Fregean one, then, and doesn’t depend on the Fregean one. It makes no Fregean claims about what really is, about the “true” logic, or about a “third realm.” There is nothing to find out; there are decisions to be made, about which language to use for what, or which explication for a troublesome concept works best.
But this Carnapian linguistic turn utterly subverts Williamson’s pre-Fregean program of “taking philosophical problems at face value.” The Sorites problems, for instance, that Williamson devotes so much ingenuity to and holds up as exemplary for philosophical study (Philosophy of Philosophy Ch. 2), simply evaporate under the Carnapian linguistic turn. The vague predicates of ordinary language present no problem of explication. In most contexts the imprecision of ordinary language is no defect, but appears well-adapted to the evolutionary demands of learnability and of usability in a broad and flexible variety of action contexts. If more precision is needed in a particular context, an explicated concept (a replacement for the vague one) is used instead; the explicandum itself has no particular relevance to, let alone authority over, the use of its explicata. We don’t after all think it’s a defect of a magnifying glass or a human eye that it isn’t a microscope, but neither do the properties of the eye constrain the design of microscopes.
So Williamson’s paradigm of a philosophical question, “Was Mars always either dry or not dry?” (p. 24) is of no explicatory interest. It has no face value, for Carnap, since the ordinary-language predicate “dry” is obviously vague. The question only works, then, within ordinary-language contexts, not in contexts where more precision is required, as in discussions of Mars. To take it at face value is to take an external question literally, rather than recasting it as a question about the choice of more precise explicatum. Williamson argues that his Mars question is not “about language” (pp. 26-31), and therefore has to be taken at face value. But from a Carnapian viewpoint it’s simply a misapplied tool — a case where ordinary language is being applied outside the domain it was evolved to apply to. It’s a case where a more precise instrument is needed. There is no face value in the sense that the tool being considered has no application in the proposed context.
I can use the vague word “dry” in ordinary contexts without knowing anything about its limits of appropriate application. I can distinguish dry from wet laundry in an instant without worrying about the remaining water molecules that surround us in the driest air. But what would be the point in arguing over whether Mars or the Mojave is “really” dry? “Dry” is not the kind of concept that would ever occur where the distribution of water molecules is under serious consideration, since the conceptual tools our species has developed for such occasions purposely avoid vagueness.
Vagueness is simply not much of an issue in science, then, or at least in pure science. In applied languages developed for practical communication, such as the semi-precise languages of navigation or accounting, practical requirements obtrude, e.g. the trade-offs between precision and usability. In those cases, vagueness becomes a political problem, though still not an explicatory one. To make the language of accounting perfectly precise would make it unusable, but the flexibility allowed by a range of vagueness creates opportunities for distortion or misrepresentation. These trade-offs can only be resolved by political interaction among the various user constituencies of accounting — managers, lenders, investors, tax authorities, and so on. No one would think of resolving such disagreements by arguing over the “true” meaning of, say, depreciation, or by attempting to establish the “underlying” logic of its use beneath the surface structure of ordinary business talk. Depreciation rates differ in different countries, since the balance of power among (indeed the identities of) the constituencies differs from place to place. No “analysis” of the use of “depreciation” in business discourse would yield anything of the slightest relevance to these negotiations. Why, then, should we expect it to yield anything in the case of ordinary language, with its even looser constraints on usage and its vast multitude of contending constituencies, forcing it to accommodate even more exorbitant demands for flexibility?
Of course Carnap realized that the ordinary language we use for everyday transactions, political argument, interpersonal relations, and a thousand other purposes could not simply be replaced by something more precise; he was willing to live with piecemeal improvement. He was sympathetic, for practical purposes, to Neurath’s idea of a “universal vernacular” in which ordinary language defers to all available explicata. But in such a language a word such as “dry” with obvious scientific explicata (along the lines “concentration of H2O below a certain threshold,” where the threshold varies by context) would simply defer to that scientific precision. Its face value is of no conceivable interest.