1
When I began my doctoral studies at Princeton in the early 1970s, Gilbert Harman was director of graduate studies. Among his characteristically idiosyncratic pieces of advice (along with “Start writing articles, like a grown-up, as soon as you can, and get out of graduate school as fast as you can—three years at the most”) was the opinion that there is just no point in reading anything written more than five years ago: you want to find out about the current state of the discussion, and anything that hasn’t been addressed in that length of time probably isn’t important enough to bother with. The general view among my fellow students was that he represented the ne plus ultra of ahistorical approaches to philosophy. In time, I came to realize that they were dead wrong. At that stage of his career, Gil was steeped in the history of philosophy, and everything he wrote and thought was rooted in his understanding of it. The contrary impression resulted from failing to realize that he thought that, for practical purposes, philosophy had started with Quine.
I set up the story that I tell in the body of this work in terms of a reading of the classical project of philosophical analysis, as that project was developed during the twentieth century. This is a shorter historical run-up than I often require. In Tales of the Mighty Dead 1 I trace some of the principal themes I develop in Making It Explicit2 back to the early moderns. And these days I find it difficult to make any philosophical point without starting back with Kant and Hegel. In the case of the present project, though, I find that even some of those among my philosophical friends who are usually the most sympathetic to this way of proceeding question whether the enterprise I pursue here is really strengthened by framing it historically as a way of “extending the project of analysis,” as the title of my first lecture puts it. They point out, to begin with, that it need not be so construed. The usefulness and illumination provided by the metaconceptual apparatus I introduce for thinking about relations between meaning and use and the new sorts of pragmatically mediated semantic relations that apparatus brings into view—what it can teach us about discursive practices generally and the particular vocabularies I apply it to more particularly—does not at all depend on any continuity there might be between these methods and more traditional analytic ones. Given that independence, they argue, it is unwise to burden my project by associating it with what they see as a degenerating research program motivated by suspect methodological aspirations. When Richard Rorty first read drafts of these lectures, he asked, “Why in the world would you want to extend the death throes of analytic philosophy by another decade or two?” In a similar vein, John McDowell has described what I am doing here as perversely transplanting perfectly healthy pragmatist organs into the rotting corpse of analytic philosophy, so as artificially, and no doubt temporarily, to revive it as a kind of Frankenstein monster. I want to say something here about why, in the face of such strong reactions, I still want to endorse and defend not just meaning-use analysis and the idea of pragmatically mediated semantic relations, but also the further, optional characterization and commitment to these ideas as offering a potential way forward for what is recognizably a version of the classical project of philosophical analysis.
2
There are a number of different sources of discontent with that analytic project, and I think it is important to disentangle them. A significant element of the enterprise pursued in these lectures is to respond to what I take to be the weightiest, deepest, and most important sort of objection to the classical project of philosophical analysis: the battery of considerations raised by the pragmatists, and above all Wittgenstein. But there are other sources of discomfort and lack of enthusiasm for the analytic enterprise, which I do not address in the body of this work. Some people do not see, even in retrospect, any good reasons why the advent of new logical tools should have ushered in a new philosophical era. Rorty once told me that it would not have mattered a bit to him, or to anything he had ever thought philosophically, if there had never been such a thing as formal semantics. (He delighted, in a Zen-masterish sort of way, in shocking me.) And I know others who maintain that nothing of real philosophical significance ever happened within three feet in any direction of a quantifier symbol. Such views are not only extreme, I think they are symptoms of a failure to appreciate some of the astonishing achievements and real promise of the logistical semantic tradition.
Here is an example that still fires my enthusiasm, after many years. In his seminal paper “General Semantics,” David Lewis invites us to pick whatever we like, depending on our more general philosophical proclivities, as semantic interpretants for declarative sentences and singular terms.3 The semantically relevant whatsits associated with sentences might be, as Lewis prefers, sets of possible worlds, or as Michael Dummett prefers, sets of assertibility conditions, or something else. And singular terms might be interpreted by objects or by recognition conditions, or something else. Whatever we choose, it will be settled that the semantic intepretants of one-place predicates should be functions from the interpretants of singular terms to the interpretants of sentences. And it will be further settled that adverbs—which, like ‘slowly’, transform one predicate, ‘walks’, into another, ‘walks slowly’—should be semantically associated with functions from functions from the interpretants of singular terms to the interpretants of sentences to functions from the interpretants of singular terms to the interpretants of sentences. When we then notice that their inferential behavior partitions adverbs into two classes, attributive and non-attributive, depending on whether it follows from one’s having X’d φ-ly that one X’d—as one’s having buttered the toast follows from having buttered the toast in one’s kitchen, but not from having buttered the toast in one’s imagination—we are in a position to represent in set-theoretic terms exactly what the semantic difference between these two kinds of adverbs consists in. Getting this sort of algebraic grip on the meanings of non-logical expressions is a signal accomplishment. Even as we explore different approaches to semantics, we must be sure never to lose the precious comprehension it provides. It is, it seems to me, just the sort of thing that ought to inspire the philosophical imagination. At the least, I take it we are obliged to investigate just how far such methods can take us in making visible and formally tractable various other aspects of the content of the concepts we deploy in our philosophical reflections.
Another complaint that one sometimes hears is that analytic philosophy has come to focus on narrow, technical puzzles, scholastically generated by essentially self-contained literatures. An invidious contrast is then made between worrying about, say, the details of the behavior of proper names and indexicals in modal contexts, the pros and cons of four-dimensionalist views of spatio-temporal continuants, and the semantic paradoxes, on the one hand, and the best lessons to draw from the cataclysmic transition from traditional to modern culture, the essentially social character of self-consciousness, and the relative merits of art and science as revelatory of human nature, on the other. Which sets of sample concerns, it is asked, most deserves our allegiance as faithful to the august spirit of the tradition of philosophia perennis? No doubt there are a lot of toilers in the analytic groves who keep their noses very close to the ground indeed. But it does not follow, and it is not true, that, as a result, their work, for instance in the most technical reaches of the philosophy of language, relates to the philosophical tradition primarily as simply changing the subject. (Tempting as this conclusion may be. Bruce Kuklick describes the basically reactionary impulse of one of the great American philosophy departments of the first half of the twentieth century as presenting the dispiriting spectacle of “thirdrate minds defending the rights of genius against the claims of technique.”4) Providing disciplinary matrices within which the philosophical equivalent of normal science can proceed, between and alongside suggestions for paradigm shifts of various scopes, counts more in favor of the maturity of a Fach than for its degeneracy. (This fact is just one reason that a tu quoque gesture at the most philological fringes of non-analytic philosophy does not amount to a constructive response to this challenge.)
In any case, although many philosophers working in the analytic tradition do not much concern themselves with how their concerns fit into a larger framework, I think we should think of the classical project of analysis as one way of working out a vision of us as essentially discursive beings, at once creators and creatures of our linguistic practices and the states of mind that make them possible and are made possible by them. This is the key into which traditional philosophical issues are transposed. In this sense, the linguistic turn characteristic of the twentieth century is not at all limited to the analytic tradition. One may, and should, take issue with the particular form in which that vision is worked out, but it is hardly to the point to complain that traditional issues of epistemology (and, more deeply, intentionality, agency, and normativity) have been left by the wayside because they are addressed in these terms. Even the narrower (hence more committive and controversial) versions of the analytic version of the linguistic turn, which identify it with the shift from the material to the formal mode (in Carnap’s terms), that is, from concern with discursivity to specifically metalinguistic issues, ought to be understood as offering a way of working out the basically platonistic intellectualist strategy of understanding the implicit in terms of how it can be made explicit. And there have always been central analytic philosophers whose more detailed work took place in a larger, systematic philosophical context: Quine, Sellars, Davidson, and Dummett, to name just a few. (I’ll return to this point below, when I consider the charge that analytic philosophy is committed to an outmoded and disagreeable kind of metaphysical project.)
A related (but I think distinguishable) complaint is that analytic philosophy has explicitly and (so) self-consciously cut itself off from the history of the discipline. There certainly has always been an ahistorical strand of thought in this tradition. Among the founders of the analytic tradition, Russell and Moore in particular promulgated an origin story that emphasized a radical break from all previous philosophy. The new logic was at last, for the first time, to put philosophy on the sure path of a science. But like other turn-of-the-century modernists in literature and the arts, the early analysts tended to overestimate the extent to which they broke with, rather than developed, the tradition they inherited and reacted against. (Russell wrote and cared a lot about his hero Leibniz, and the continuities between Carnap’s thought and that of his neo-Kantian teachers substantially outweigh the discontinuities, for instance.) As we achieve the temporal distance necessary for a dispassionate historical perspective on the history of early analytic philosophy itself, we ought to be able to take a more balanced view. In the body of this work I point out that the core programs of analytic empiricism and naturalism are recognizable descendants of their early modern forebears. And in Tales of the Mighty Dead I argue that this is true as well of the program of functionalism in the philosophy of mind. There, and in Making It Explicit, I also emphasize the strand of thought linking analytic philosophy to its past that Stanley Cavell epitomizes in the form of his characteristically trenchant aphorism: “Kant depsychologized epistemology, Frege depsychologized logic, and Wittgenstein depsychologized psychology.”
It remains true that many later analytic philosophers continued to relegate the history of philosophy to the status of a minor sub-field, to the point of thinking of it as providing suitable employment only for those who, for one reason or another, were not capable of doing, or disposed to do, the real thing. I have already mentioned the widely shared suspicion that my teacher Harman once harbored such a view. It was his teacher, Burton Dreben, who expressed the attitude most forcefully, in his dictum that “Garbage is garbage, but the history of garbage is scholarship.” But this was certainly never a universally held opinion. One of the reasons Wilfrid Sellars is a particular hero of mine is the way he pursued analytic philosophy within a synoptic, systematic framework that was motivated and informed throughout by his intepretation of the history of philosophy, and above all of the significance of Kant’s thought within that history. Sellars’s avowed aspiration to move analytic philosophy from its Humean into its Kantian phase (and my own to move it from its incipient Kantian to a subsequent Hegelian phase) expresses a distinctive understanding of how the development of analytic philosophy fits into the larger context of the history of philosophy. Finding ourselves, as we do, in the midst of a golden age of analytic readings of Kant (initiated by Strawson and Bennett on the theoretical side, and Rawls on the practical), contemporary analytic philosophers should not be understood as bound by the propagandistic insistence that all things have been made new in analysis that was put forward by some of the founding members as an integral part of the fighting faith they crafted as a crucial weapon in the struggle to make disciplinary room for their nascent movement. (On one point, however, they might be proven to have been prescient. Recoiling from the peculiar form of Hegelianism developed by their British Absolute Idealist teachers, Russell and Moore advocated a particularly radical surgical intervention to prevent the spread of the infection to which they themselves had earlier fallen prey. They took it that the idealist rot had set in already with Kant, who accordingly must be excised along with Hegel from the expressively progressive tradition they were retrospectively reconstructing—hence Russell’s “Zurück nach Leibniz” campaign. They were confident that one could not open the door far enough to let Kant slip through and then close it quickly enough to keep Hegel out. And on this point of tactics—whatever the advisability of their overall strategic gatekeeping enterprise—they may well turn out to have been correct. We’ll see.)
3
A fourth objection to the whole enterprise of analytic philosophy is that one feature that binds together its various, otherwise disparate, phases, schools, and projects is a commitment to an objectionable scientism. I think there is indeed a telling objection that belongs under this heading, but it is important to be clear about what it is, and to distinguish it from near relatives that in one way or another miss the mark. For there are many things to mean by ‘scientism’. One broad category is what might be called Enlightenment scientism. It is a valorization of scientific activity that privileges it as the paradigmatic, highest, and most revealing expression of our humanity—by contrast to the Romantic aestheticism that defined itself by according that privilege instead to artistic activity. (Of course, we students of that great synthesizer, the Romantic rationalist Hegel, would like to think that we have the conceptual resources to evade any necessity simply to make a choice here. But that aspiration is not my current topic.) Analytic philosophy has historically endorsed a variety of views that count as scientistic in this general sense. I think the most important issue in the vicinity concerns how one thinks about the relations between two forms of understanding, and I’ll discuss that presently. But that issue is not raised by at least one very thin form that analytic philosophy’s identification with the scientific enterprise has taken.
What might be called ‘sociological scientism’ is largely a matter of style. It is the idea that the shape of the philosophical community, and of its professional discourse, should be modeled on that of natural scientists—rather than that of, say, café intellectuals, political activists, literary critics, novelists, or poets. Philosophy is thought of as professionalized, as a discipline, as instituting, through shared training, common standards for assessing arguments and accomplishments. It is thought of as a communal, co-operative, cumulative undertaking, in which progress is made by building on the agreed-upon achievements of others, and as a cognitive enterprise in the sense that it aims to produce and extend a publicly available kind of knowledge, or more broadly, understanding. Relations among its practitioners are accordingly to be respectful and collegial, personal differences and ambitions are to be submerged in the pursuit of the larger common, impersonal goal. This ideal, which in the eyes of some of its most ardent avatars in the Vienna Circle was closely associated with a progressive social and political agenda, has, I think, served philosophy rather well.5 Though its virtues are not free of associated vices, its scientism about philosophy in this sociological sense does not seem to me to provide any reason to reject the classical project of philosophical analysis.
More substantively, ‘scientism’ can just mean ‘naturalism’. It would be wrong to identify that core program of analysis with the whole analytic project, however. Not all analytic philosophers are philosophical naturalists. Moore certainly was not. Some logical empiricists were not: the Schlick/Neurath split within the Vienna Circle revolved around the issue of which of empiricism or naturalism should yield when the two programs collide (as they did, for instance, over the status of alethic modality). And contemporary analytic consciousness studies—carrying on one strand of the empiricist tradition—very often is not only not pursued in a naturalistic key, but offers a distinctive set of anti-naturalist arguments. Furthermore, philosophical naturalism about topics (vocabularies) other than philosophy itself need not be thought of as a form of scientistism about philosophy. Here Sellars’s scientia mensura dictum may serve as a paradigm: “In the dimension of describing and explaining the world, science is the measure of all things, of what is, that it is, and of what is not, that it is not.”6 But Sellars does not take it that normative vocabulary, for instance, is in the business of describing or explaining. For that reason, he does not take it that intentional vocabulary is either. And he takes the task characteristic of philosophy precisely to be offering an account of the relations between such vocabularies and the vocabularies whose job is “describing and explaining the world,” for whose use he takes it that natural science is authoritative. (In this respect he accordingly fits the definition of the classical project of analysis I offer in my first lecture.)
I think the charge of scientism begins to get a real grip when the admiration of science characteristic of Enlightenment scientism moves beyond its merely sociological and naturalistic forms and manifests itself as methodological monism. This is the view, roughly, that scientific knowledge is the form of knowledge, and scientific understanding is the only kind of understanding that deserves the name. This line of thought gives a very strong reading to the Kantian project of “putting philosophy on the sure path of a science.” Insofar as philosophy is a cognitive enterprise at all, it must be continuous with the empirical, natural sciences—destined eventually to take its place as part of the single Unified Science. (One would hardly hold this view if one were not a naturalist about many other vocabularies apart from philosophy. But the point I just tried to make is that that kind of naturalism in no way commits one to this further step.) On this view it is a profound and dangerous mistake to think that the difference in their subject matters brings with it a substantive and fundamental difference in method and aim between the Naturwissenschaften and the Geisteswissenschaften. In particular, discursive practices and practitioners, vocabularies and the concepts deployed in using them, do not in principle require methods of investigation, or support kinds of understanding, different from those that have proven so successful in addressing the non-discursive world. This methodological commitment ties together representative figures of the analytic tradition otherwise as diverse as Russell, Ramsey, Carnap, Quine, and Fodor.
Now, I reject scientism in the form of methodological monism. Here again, care is called for, to be clear about the reasons that warrant such rejection, and about just what follows from it. In spite of its popularity among analytic philosophers, I do not think a commitment to methodological monism needs to be understood as an integral part of the analytic project itself. But neither is the association simply accidental. There are considerations that are integral to that project that can easily appear to call for this methodological stance. Some distinctions are in order. One way into the issue is provided by a sort of modus tollens. One constant in Wittgenstein’s thought, early and late, is his denial of methodologically monistic scientism. “Philosophy is not one of the natural sciences,” he says in the Tractatus,7 and this view seems to be part of what lies behind the theoretical quietism of the later work. In fact, I think Wittgenstein thinks that if systematic philosophical theorizing were possible, it would mean that philosophy is an empirical science. Since it is not, philosophers must eschew theorizing, restricting themselves instead to light, local descriptions of discursive practices, where such descriptions might provide helpful reminders in freeing ourselves from the sorts of misunderstandings and puzzlements that arise precisely from the theories implicit in inherited pictures of what is going on when we think and talk. Whether or not Wittgenstein himself reasoned in this way, I take it that it is common for his admirers to see him as presenting us with a forced choice: either embrace scientism about philosophy of the methodologically monistic sort—that is, take philosophy to be an empirical, scientific discipline—or give up the idea of systematic philosophical theorizing once and for all.
I think this is a false choice. Rejecting scientism of the methodological monistic sort does not entail giving up the possibility of systematic philosophical theorizing about discursive practice. One of the most powerful methodological features of the natural sciences is the postulation of unobservable theoretical entities, and their deployment in constructions aimed at explaining what is observable. Theoretical entities are those about which we can make only theoretical, and not observational, claims. Theoretical claims are ones that we can only become entitled to as the conclusions of inferences from other claims, not non-inferentially, as the results of exercising reliable dispositions to respond differentially to environing states of affairs by making observation reports of them. A generalization of this method would have the role played by observational vocabulary played by any antecedently available vocabulary, whether observational or not. So for instance one might postulate meanings to explain proprieties of use, where the latter are expressed in a non-semantic vocabulary, whether or not our access to claims about correct usage are made observationally or themselves inferentially.8 The claim that theorizing of this sort could be legitimate in philosophy does not commit one to the claim that this method is the only legitimate method of acquiring philosophical understanding—which is what methodologically monistic scientism claims. The generalized method of postulation and construction might be one form of philosophical understanding among others. I want to claim that what is objectionable about the methodologically monistic form of scientism is its exclusivity. Rejecting that at least leaves open the question of whether, and which, features of natural scientific investigation, explanation, knowledge, and understanding ought also to be counted among those useful and appropriate in philosophy. After all, description is also a central and essential element of scientific methodology, and even the most rigorous versions of Wittgensteinian quietism allow philosophers to describe features of our linguistic practice.
4
Asking this question, and seeing that it is not settled by rejecting methodologically monistic scientism, then raises a fifth sort of objection to the project of philosophical analysis, one that is a somewhat more careful and nuanced successor to the previous one. This is that the defining aim of philosophical analysis is to achieve a kind of understanding of vocabularies and discursive practices-or-abilities to use or deploy vocabularies that is in principle not available, because that subject matter requires another, quite different sort of understanding. The kind of understanding sought by analytic philosophers is appropriate only to non-discursive activities. The reason we should not be methodological monists is that understanding talking and thinking, concept use, vocabularies, natural language utterances and texts, is a distinctive sort of achievement. This kind of understanding, what we might call hermeneutic understanding, is not expressible in explicit rules, formalizable in regimented technical or artificial languages, as analytic philosophers are committed to doing by the semantic logicism characteristic of the classical project of analysis. The mathematized mature natural sciences have had great success in achieving what we might call algebraic understanding of great swathes of the inanimate natural world. (Whether the animate biological world, including sentient-but-not-sapient creatures and their activities, itself already calls for further special sorts of understanding remains a lively and controverted question.9) But when the topic is culture rather than nature, another sort of approach is called for. Here the paradigm of understanding is that exhibited by competent native speakers of natural languages when confronted by everyday utterances expressed in familiar vocabulary. This sort of practical grasp of meanings (the medium of the cultural) is not in the most fundamental cases a matter of explicit theorizing at all. And it is not a matter of mapping or translating the utterance into some other vocabulary (perhaps with the use of auxiliary logical vocabulary) either. (In the sense that matters for this point, the language of my thought is just my language: a language I speak.) More sophisticated forms of hermeneutic understanding, of the sort exercised by the literary critic, jurisprudential interpreter, and reader of philosophical texts, are possible, but they are both rooted in the basic one and do not come closer to having the structure of algebraic understanding.
A pragmatist line of thought common to the Dewey of Experience and Nature and Art and Experience, the Heidegger of Being and Time, and the Wittgenstein of the Philosophical Investigations is that there is such a thing as hermeneutic understanding in this sense, it is a genuine and distinctive kind of understanding, and it is the most basic kind of understanding, in the sense that all other sorts of understanding are parasitic on it and develop out of it. It is the primordial sort of practical discursive know-how: the capacity to engage in an autonomous discursive practice. In particular, they are concerned to insist that the sort of algebraic understanding characteristic of mature mathematized sciences—the sort for which analytic philosophers long—is pragmatically dependent (PP-dependent) on everyday hermeneutic understanding, which accordingly cannot be replaced by, or reduced to, the more technical kind.
I accept all of these pragmatist claims about the distinctiveness and basicness of ordinary hermeneutic understanding of discursive performances and their products. Should we conclude that the analytic project is just a mistake? I don’t think so. For this pragmatist line of thought does not entail that many aspects of discursive practice might not also be susceptible to understanding of the sort I have called “algebraic.” And where it is possible, broadly algebraic understanding has distinctive virtues, which adherents of the project of analysis are right to esteem and treasure. The slogan of the analytic project is “Faith, hope, and clarity—and the greatest of these is clarity.” The clarity in question is specifically conceptual clarity. It would seem to have two dimensions: definiteness and perspicuity. From a pragmatic point of view, the significance of a speech act is definite insofar as its normative significance is settled. From the point of view of semantic inferentialism, this means that concepts are definite insofar as their circumstances and consequences of application are settled: when one is committed and entitled to apply them, and what such application commits and entitles one to. Perspicuity is then epistemic or psychological ease of access to those circumstances and consequences of application. On this line, thinking clearly is both formulating one’s claims (claimables) so as to fix what one would be committing oneself to by endorsing them and what would entitle one to do so, and being aware of those definite consequences and circumstances of application of the concepts that articulate the contents of the concepts one is applying. Writing clearly is choosing one’s words so as both to determine the inferential boundaries (or, one could equally well say, truth conditions) of one’s claims and to convey them to the reader.
What I’m calling the “algebraic” form of understanding achieves clarity along both the dimensions of definiteness and perspicuity by constructing the conceptual contents expressed by a target vocabulary. It does that by exhibiting them as complexes formed as the products of applying explicit algorithms to the conceptual contents expressed by a base vocabulary (treated for this purpose and relative to this construction, as simple). From the broadened perspective made available by thinking about pragmatically mediated semantic relations, algorithmic elaboration of one set of vocabulary-deploying practices-or-abilities into another is, as I point out in Lecture 2, a generalization of the traditional semantic logicism characteristic of the classical project of philosophical analysis. What corresponds in this semantic-analytic project to the postulation of unobservables in empirical scientific theorizing is the employment in the algorithmic construction also of some further auxiliary vocabulary, whose use is not governed by antecedent norms but is determined instead by stipulated inferential connections to both base and target vocabulary.10 This algorithmic-constructional method (building complex things by applying well-defined operations to simpler things) is a very good, perhaps superlative, way of securing clarity of understanding. I have elsewhere called it for this reason the “gold standard” of understanding generally—by which I mean that when and to the extent it is available, it is the very best sort of understanding to have, for it takes the issue of what one means (what one is committed to by a claim, what is incompatible with it, what would count as evidence for or against it, and so on) out of the hands, out from under the authority, of the one making the claims. It establishes a fact of the matter about the inferential relations that articulate the contents of the concepts expressed by the target vocabulary that swings free of the beliefs and preferences of the concept user: what she would like to be committed to or treat as evidence for those claims. If a dispute arises, those who are assessing the claim in question can say, with Leibniz, “Let us calculate.” This aspiration to develop “a general method in which all truths of reason would be reduced to a kind of calculation … and errors—except those of fact—would be mere mistakes in calculation”11 (I take it that “truths of reason” here stands in for inferential relations that articulate the contents of the concepts involved) is one of the reasons Leibniz was a hero for Russell in the latter’s attempt to develop a notion of philosophical analysis. This sort of clarity of understanding is a pearl without price—all the more to be prized where the target vocabulary it concerns is weightier and more difficult, as is the case with many of those either used or addressed by philosophers. (This sort of clarity facilitates communication—scientism in the sociological sense.)
Appreciating this cardinal virtue of the algebraic form of understanding does not require taking issue with the pragmatist point that it is in principle parasitic on and intelligible in principle only against the background of a more basic sort of practical discursive understanding that does not at all have this explicit theoretical form. It is useless—for instance, in settling disputes about what someone is committed to by a claim couched in the target vocabulary being (re)constructed—unless there is a shared base vocabulary about whose proper use all parties can agree in their practice. We are not in a position to calculate unless we can all practically go on in the same way in counting and adding—as Wittgenstein is at pains to remind us in many different ways and many different contexts. And the same is true of algebraically computing the inferential roles or truth conditions of complex expressions from those of simpler ones. Algorithmic elaboration is a way of leveraging practical agreement in the use of one vocabulary into practical agreement in the use of another. It is true that what plays the role of a base vocabulary for one such constructive enterprise may be the target vocabulary whose proper use is algorithmically reconstructed by another. But the point Wittgenstein was after here is that it cannot be algorithmic elaboration all the way down. At some point each such chain must be anchored in practical agreement about what it is and is not correct to do with a vocabulary that is not settled by being algorithmically handed off to some prior one. And that is to say that we should not make the jump from the legitimate local aspiration to be able to settle some semantic-inferential disputes in the “Calculemus” way to Leibniz’s dream of a global lingua characteristica, all of whose concepts are governed by a calculus ratiocinator that is in this sense universal.
Acknowledging the value of the unique clarity afforded by algebraic understanding accordingly does not entail commitment to this sort of understanding being available in every case, even in principle. It does not oblige one to embrace the shaky method of the drunk who looks for his keys under the streetlamp, not because they are likely to be there, but just because the light is better there. We should admit that, sometimes, algebraic understanding is not available—indeed, that every context in which it is available contains an appeal to a base vocabulary whose use is not held in place algebraically, but depends on another sort of practical mastery and understanding. Algebraic understanding can be no more legitimate than the hermeneutic understanding on which it depends and which it leverages, amplifies, and concentrates. It follows that philosophy cannot be identified with analysis, thought of as comprising the tasks of understanding algebraic understanding and applying it in semantics. Even under the broad heading of trying to understand discursive practice, there is a more basic sort of hermeneutic understanding, whose implicit, practical, everyday species and whose explicit, theoretical, sophisticated species must both be studied and exercised by philosophers. Thinking through the presuppositions of its project shows that analytic philosophy can aspire at most to being one species of the genus. (In the third, methodological, chapter of Tales of the Mighty Dead, and again in “Hermeneutic Practice and Theories of Meaning”12 I try to say something specific and systematic about how the different aspects of discourse addressed by these two sorts of understanding and their associated disciplines complement one another.)
5
There is, then, a lot more to be understood about discursiveness than can be understood algebraically. This is obviously true de facto, and I have just rehearsed an argument that it is true also de jure. But can we know in advance that the algebraic sort of understanding is not available at all for some subject matters? Might it not be the case that the very nature of discursive practice makes it unsuitable for this sort of account? Perhaps philosophical analysis must inevitably “murder to dissect,” the very method it employs making it impossible for it ever to grasp the essence of the phenomenon it addresses. The image reminds us that claims of this same general form concerning not discourse, but life, were a commonplace in the early nineteenth century, as biology broadly shifted its concern from structure to function, from anatomy to physiology. Hegel takes over the Romantic trope of organic unity, and in a characteristic and central move transposes it into a rationalist key, by treating the sort of unity of disparate elements characteristic of organisms as an image of the semantic or conceptual holism he had discovered to be a central structure of the discursive—as the way nature already implicitly expresses a fundamental structure that becomes explicit in culture. Perhaps the holistic character of meanings or conceptual contents, of thinkables and sayables, in principle resists reconstruction by bottom-up, compositional, ultimately atomistic algebraic elaboration? Maybe by its very nature it demands instead a top-down, holistic hermeneutic form of understanding.13
I think this is a substantive challenge. It demands a much fuller discussion than I can enter into here. The conditional “if discourse is semantically holistic, then in principle no algebraic reconstruction of conceptual contents or meanings is possible,” is one that can be exploited either by modus ponens or by modus tollens. One sort of critic of analysis, inspired by Hegel, adopts the former strategy. Analytic semantic theorists such as Jerry Fodor adopt the latter. My own strategy is to dispute the inference codified in the conditional. One of the satisfying results of the incompatibility semantics sketched in my fifth lecture is the algebraic construction of a counterexample to some of the most powerful arguments for that conditional that have been offered from the analytic side. And in the sixth chapter of Tales of the Mighty Dead I consider analytically how one might get around some of the most worrisome difficulties that I think Hegel, the first philosopher to think hard about the issue of semantic holism, already wrestled with. But I think it is early days in our discussion—that we have only begun seriously to come to grips with this knotty issue. I suspect that it will resist Gordian solutions, yielding only gradually to the application of insight, patience, and analytic ingenuity. We have as yet, I think, no useful algebraic representation of the seriously multi-premise, nonmonotonic, material inferences that are responsible for the phenomenon of semantic holism.
Putting aside the particular issue of semantic holism, it is at any rate important to keep in mind that the claim that there are some vocabularies, some discursive practices-or-abilities, that are by their very nature not amenable to analytic algebraic reconstruction does not follow just from the observation made above (in denying methodologically monistic scientism) that every analysis or algebraic reconstruction of a target vocabulary must make use of and so depend on the prior semantic determinateness and understanding of what is expressed by some base vocabulary. That is, it does not follow that there is some order of, as it were natural basicness among vocabularies, which must have unexplained unexplainers (base vocabularies that do not admit of analytic algebraic reconstruction in terms of others) as its most basic elements. It might well be that although each analytic-algebraic account of the use of any vocabulary must appeal to some base vocabulary whose use is not explicated in that account, every vocabulary that plays that role of base vocabulary in some analyses plays the role of target vocabulary in some other successful analysis. A claim of the form ∀x∃y [Rxy] does not entail one of the form ∃y∀x [Rxy]. (It is true that the world has a population problem because during every minute there is a woman somewhere in the world having a baby. But it is not a productive way to address the problem to look for the woman who is having all those babies and make her stop doing what she is doing.) The sense in which algebraic understanding rests de jure on hermeneutic understanding may be merely of the local, ∀x∃y sort, not the global, ∃y∀ x sort.14
Distinguishing between taking the base-vocabulary/target-vocabulary distinction, in terms of which I have characterized the classical project of philosophical analysis, to be local and relative to particular expressive-explanatory undertakings, on the one hand, and taking it to be global and absolute, on the other—between taking it to be a matter of cognitive convenience or taste, and taking it to be something we could get substantively wrong because of how things anyway are—brings into view the notion of universal base vocabularies. That notion is at the center of the sixth and final sort of objection to the whole analytic philosophical enterprise that I will consider here. That objection arises first in an ad hominem form, as depending on subsequent commitments I undertake, collateral to the strategy of framing my presentation of meaning-use analysis as a way of extending the classical project of analysis by taking into consideration pragmatically mediated semantic relations. But the concerns it raises are deep and general, and their significance extends well beyond that parochial setting. The ad hominem point is the observation that although I take the classical project of philosophical analysis to be worth extending and developing, I reject both the core project of empiricism and the core program of naturalism. (In my third lecture I present reasons to be skeptical also about the core program of AI-functionalism—an issue to which I shall return below.) One might then ask what interest the classical project of philosophical analysis retains when shorn of those programs (and any successors that share their objectionable features), and whether the same considerations that render those programs objectionable accordingly impugn the more general umbrella project that sheltered and nurtured them. These questions turn on the issue of what is wrong with the empiricist and naturalist programs. The complaint I am concerned to assess diagnoses them as metaphysical programs, and urges us to reject them for that reason. And it claims that the analytic project generally is of philosophical interest only insofar as it takes the form of a program predicated upon such disreputable and ultimately insupportable metaphysical commitments.
What is distinctive of empiricism and naturalism, considered abstractly, is that they each see some one vocabulary (or vocabulary-kind) as uniquely privileged with respect to all other vocabularies. Empiricism takes its favored vocabulary (whether it be phenomenal, secondary-quality, or observational) to be epistemologically privileged relative to all the rest. In what I think of as its most sophisticated forms, the privilege is understood more fundamentally to be semantic, and only derivatively and consequentially epistemological. Naturalism takes its favored vocabulary (whether it be that of fundamental physics, the special sciences, or just descriptive) to be ontologically privileged relative to all the rest. In both cases, what motivates and gives weight and significance to the question of whether, to what extent, and how a given target vocabulary can be logically or algorithmically elaborated from the favored base vocabulary is the philosophical argument for epistemologically, semantically, or ontologically privileging that base vocabulary. These are arguments to the effect that everything that can be known, said, or thought, every fact, must in principle be expressible in the base vocabulary in question. It is in this sense (epistemological, semantic, or ontological) a universal vocabulary. What it cannot express is fatally defective: unknowable, unintelligible, or unreal. One clear thing to mean by ‘metaphysics’ is the making of claims of this sort about the universal expressive power of some vocabulary.
The general objection can now be put in the form of two claims:
1. The philosophical interest of the analytic project depends on endorsing programs (paradigmatically empiricism and naturalism) that are metaphysical in this sense.
2. Metaphysics in this sense is a bad idea.
The conclusion is that we should reject the classical project of philosophical analysis root and branch.
An argument for the first claim might go like this. Why would we care about showing that what is expressed in some target vocabulary can or cannot be expressed in a base vocabulary (when suitably elaborated), unless we thought for independent and antecedent reasons that, by doing that, at least in some cases we were settling what could and could not be known, what is and is not intelligible, what is and is not real (or some other philosophical issue of that magnitude)? Apart from payoffs such as those, why not just acknowledge that every vocabulary is the one it is, and not some other one? Different vocabularies have different expressive powers. So what? Taking the expressive powers of some vocabularies to be authoritative for settling issues of great pith and moment for epistemology, semantics, or ontology is the recipe for producing a responsive answer to the “so what?” question. Doing that is taking those vocabularies to be universally privileged base vocabularies, with respect to some large-scale philosophically significant assessment. And that, according to the definition we are working with, is undertaking a distinctively metaphysical commitment. In fact, the history of philosophy in the twentieth century offers some empirical confirmation for this conceptual point. The first wave of analytic results was logicist in a strict sense: not only the means of semantic elaboration, but also the base vocabulary itself was logical vocabulary. (Here I have in mind Frege’s Begriffsschrift and Grundgesetze, Russell’s “On Denoting” and Principia Mathematica, and Wittgenstein’s Tractatus.) But after that, beginning already with the Cambridge analysts of the 1920s, and accelerating with Ayer and the Vienna Circle in the subsequent decades, and continuing well past mid-century for instance with Quine, the form in which philosophical analysis really took hold was indeed in connection with what for that very reason I call the “core programs” of empiricism and naturalism. Of course, there is a delicious irony here—manifest to those working outside this philosophical tradition, but often invisible to those within it. For one important, perhaps dominant, current of classical analytic philosophy defined itself by contrast to its predecessors and contemporary rivals precisely by its anti-metaphysical tenor and fervor. Granted, Bertrand Russell, Peter Strawson, and David Lewis would not have been surprised or upset to be accused of engaging in metaphysical undertakings. But Carnap, Ayer, Quine, and Hempel certainly would.
The second limb of the objection is that metaphysics in this sense is in principle a defective enterprise. I am not really sure how arguments for such a claim at this level of generality go. I have heard four sorts, and am somewhat moved by a fifth. Empiricists reject metaphysical claims because they want to make claims there can be empirical evidence for, because they take that to be a necessary condition of those claims being candidates for expressing knowledge, or, indeed, meaning anything. Of course, that this methodological principle contradicts their own empiricist metaphysical principles, including this one, was forcefully pointed out by Hempel in his masterful “Empiricist Criteria of Cognitive Significance.”15 Corresponding considerations, of course, show that one who rejects empiricism on the grounds that it is unacceptably metaphysical cannot offer these empiricist reasons for rejecting metaphysics. Naturalists often reject metaphysics on the grounds that it is not a product of natural science. When we have real physics, why should we want, and how would we be justified in adding, metaphysics? From this point of view it seems like the attempt to add an otiose layer of hyper-physics. Or, it is an attitude toward the results of science that is itself not science, but a kind of scientism that itself has no scientific credentials. Arthur Fine’s rejection of scientific realism in favor of the “natural ontological attitude” is a sophisticated version of this thought.16 Again, though, someone who rejects naturalism as objectionable metaphysics cannot do so for these naturalistic reasons.
Pragmatists such as Rorty object to the privileging of some vocabularies as universal base vocabularies characteristic of metaphysics on the grounds that it depends on a false, indeed, ultimately magical, understanding of the nature of the sorts of privilege or authority involved. All normative statuses, including these, are instituted by social practice. There are no such normative statuses apart from our practical normative attitudes of taking or treating some things as privileged or authoritative. And for instrumentalist pragmatists like Rorty’s hero Dewey (whom he sometimes follows in this respect), the ultimate source of those attitudes is our own needs, wants, and convenience. What is wrong with the metaphysical sort of privileging of vocabularies is that it requires the idea of some vocabulary being necessarily privileged by how things are—God’s vocabulary, or Nature’s, or even Mind’s, or Meaning’s vocabulary—quite apart from our contingent projects and attitudes. Once again there would seem to be an issue about the self-referential stability of this view: is the social nature of normativity, and its normative capacity to trump metaphysical projects, just a feature of how things anyway are? Rorty’s response is that this commitment, too, is indeed to be assessed relative to our needs, welfare, and projects. We learned (well, we were supposed to have all learned, though current events make it dubious that we did) from the Enlightenment that it was bad for us in our development as mature humans in charge of our lives and institutions, to understand moral normativity as simply reflecting how things were with a non-human (albeit divine) reality. And Rorty’s practical proposal for a second Enlightenment, completing the work of the first, is to extend that lesson of self-reliance from the practical to the theoretical sphere, for reasons analogous to those that warrant the first move.17 This is radical and controversial.
A more Wittgensteinian pragmatism addresses metaphysical programs more in a retail than a wholesale spirit. It addresses empiricism and naturalism, and any successor projects one by one, seeking to undermine the specific claims of privilege they put forward. (Here the various criticisms Sellars addresses to empiricism, adverted to in my opening lecture, as well as those of Quine and Austin can serve as paradigms.) But it also expresses a more general suspicion that any such program will turn out, upon examination, to have been motivated by a philosophical anxiety that can be traced to some relatively specific misleading philosophical picture of what knowledge, mind, meaning, or reality must be like—on pain of some Bad Consequence. The best anti-metaphysical strategy is then to diagnose and dissolve that underlying misconception, thereby relieving the felt pressure that had made a metaphysical response seem possible because necessary. McDowell reads Kant and Hegel as already engaged in enterprises with this diagnostic-therapeutical, anti-metaphysical shape. As far as the general issue is concerned, I think this is an anti-metaphysical attitude, and a template for arguing against metaphysical programs, rather than an argument as such.
7
What most gives me pause about the commitments underlying programs of the sort I am calling ‘metaphysical’ is that they essentially require us to quantify over all possible vocabularies. Universal base languages are base languages from which every vocabulary that is legitimate in some sense (specific to the metaphysical program) can be elaborated as a target vocabulary. I have my doubts about that notion. It is not that I am confident that no sense can be made of the notion of all possible vocabularies. It is rather that I do not think it comes with a clear sense. If it is to make sense, we must give it a sense. And I don’t know how to do that. As a graduate student, I attended a seminar offered by the great anthropologist Clifford Geertz. In the first session, in the course of defining ‘culture’, the topic of anthropology, he defined a language as “a symbol system in which everything can be said.” In a question I expressed my natural curiosity as to what he meant by that “everything.” He declined further comment—wisely, perhaps. I’d still like to know what one might mean by it. The worry is that no definite or determinate totality is being delineated. Maybe new vocabularies become possible all the time. This issue arises equally, and for the same reason for talk of all possible facts. Following Frege, I understand facts as true thoughts—in the sense of true thinkables, rather than true thinkings, of course. And I understand what is thinkable to be what is claimable, what is expressible in some vocabulary. So there are as many facts as there are true statements in any vocabulary. (I don’t object to people using ‘fact’ so as to allow for the possibility of facts not expressible in any vocabulary. But they must undertake the labor of making sense of that more capacious notion, and showing that there is real work for it to do.) So talking about “all the facts” and talking about “all possible vocabularies” involves a common set of commitments, ones about which I am uneasy. (I am equally uncomfortable with quantification over all “objects.” objects : sortals :: facts : claims, all depending on what possible vocabularies there are.)
One might think in this connection about the third proposition of the Tractatus: “The world is determined by the facts, and by their being all the facts” (1.11). Wittgenstein rejects the idea of facts that cannot be expressed: for any fact, there is some vocabulary in which it can be stated (clearly). To talk about “all the facts” is then to talk about everything expressible in any vocabulary. The Tractatus is predicated on there being some one vocabulary in which everything sayable at all can be said. It is in this sense a work of metaphysics. And the claim I’ve cited is a metaphysical claim. Nothing but embarrassment results if we ask about the official status of the second conjunct of this proposition in the botanization of the Tractatus. That a specified collection of facts is all the facts is not itself one of those facts, nor is it a fact of that kind. But facts are what can be said. Wittgenstein officially denies that we can even say that there are facts, or how many there are (4.1272). Is it, then, something that is shown? But how? The awkwardness here shows how the issue of the intelligibility of quantifying over all vocabularies arises in this particular setting. But cognate difficulties attend claims such as “All the facts are physical facts.” Putting aside worries about the boundaries of “physical facts” (it is a very implausible claim if one means contemporary physics, and it is hard to say what one means by something like “eventual” or “ideal” physics in a way that remains plausible without circularity) and the sense of “are,” the question of how to give a definite sense to “all the facts” in such a way as to make the claim plausible without being question-begging remains. I just don’t know what we are saying when we talk this way.
Happily, I do not need to assess the success of an objection along these lines to the viability of metaphysical projects in order to address the objection to the classical project of analysis that stems from noting the metaphysical character of its core programs of empiricism and naturalism. For I reject the other limb of the objection: the conditional claim that if the core programs are defective, because metaphysical, then the analytic enterprise is revealed as defective or pointless. For one need not endorse any commitment to globally privileging some vocabularies as base vocabularies with respect to all vocabularies—which are legitimate in one sense or another only if and insofar as they can be reconstructed as target vocabularies elaborated from those base vocabularies—in order to vindicate the project of locally identifying particular cases where the base-vocabulary/target-vocabulary relation can be shown to obtain. For there is a distinctive kind of philosophical understanding that consists in practical mastery of the relations among vocabularies that become visible when we trace various possible base→target relations through the (admittedly, in general, ill-defined) field of vocabularies. As an example, functionalism, and its specifically computational AI species, is an attempted local privileging of one vocabulary over another: functional vocabulary over intentional vocabulary. It has no pretensions at all to global privileging of functional vocabulary, to its being a universal base vocabulary. Though I have offered some reasons for skepticism about even the most plausible, pragmatist, version of AI-functionalism about intentionality, I think that a suitably broadly construed functionalism is something like the only straw floating as a prospect for an account, in other terms, of intentionality.18 Of course, it may well be, as some Wittgensteinians, for instance, think, that intentional discourse can only be understood in its own terms, and not in terms of some other vocabulary. For reasons I discuss above, that does not mean that understanding would totter. Analytic (algorithmic, algebraic) understanding is not the only kind of genuine philosophical understanding, and it is not always available.
David Lewis propounded a view of philosophy that was inspiring to me when I was his student, and inspires me still. He thought that what philosophers should do is lay down a set of premises concerning some topic of interest as clearly as possible, and extract consequences from them as rigorously as possible. Having done that, one should lay down another, perhaps quite different set of premises, and extract consequences from them as rigorously as possible. The point was not in the first instance to endorse the conclusions of any of these chains of reasoning, but to learn our way about in the inferential field they all defined, by tracing many overlapping, intersecting, and diverging paths through the terrain. That is how we would learn what difference it would make, in various contexts, if we were to endorse some claim that figures as a premise in many of the inferences, and what might entitle us to a claim that shows up as a consequence in many of the inferences. Actually plumping for and defending any of these theses is then a subsequent, parasitic, and substantially less important stage of the process. The principal aim is not belief, but understanding.
One thing that was liberating and exhilarating about this metaphilosophical attitude is that Lewis accordingly didn’t care much what reasons one had for starting with one set of premises rather than another. He was entirely open to, and indeed eager to, turn his awe-inspiring intellect to following out the consequences of even the wackiest of claims. Pavel Tichy visited while I was still in graduate school at Princeton, and he was then gripped by just such an obsession. He had somehow gotten the idea that there was some finite number n such that it was a necessary truth that there were exactly n things in each possible world, and that there were no further constraints on transworld identification of objects or counterparts. I don’t know how he thought he could count objects (Lewis later had what is at least a responsive answer to this question from a physicalist point of view). And I can’t imagine what would make one think his axiom true. But Lewis was perfectly willing to figure out the sort of modal logic and metaphysics one would get on that assumption. You never know where you might learn something.
I think this is the spirit in which we should think about semantic relations between different vocabularies. It is worth seeing how and to what extent different target vocabularies can be elaborated from various base vocabularies—including, and perhaps especially, with pragmatic detours through the specifications of practices-or-abilities necessary or sufficient to deploy those vocabularies—because that is a way of coming practically to know our way around those vocabularies, our discursive practices, and the subject matters they make it possible for us to talk and think about. Exploring all the available paths between landmarks is a way, perhaps the only way, of learning to find our way around these woods, acquiring a practical conceptual mastery of the many aspects of discursive practice, and their relations to one another. Let me repeat that I am not claiming this is all there is to philosophy, or that this is the only way to do philosophy. But this sort of semantic analysis yields one valuable kind of understanding that is potentially of great philosophical value.
Notice that, on this view, one’s exploration of semantic relations (including pragmatically mediated ones) among vocabularies of antecedent philosophical interest need not be motivated by some global, monolithic program, such as empiricism or naturalism. The merit or benefit of the analytic project is not hostage to such programs, for the distinctive kind of understanding it aims at is not. That understanding can be well served by accumulating particular, local connections that support no antecedent global program and perhaps could be predicted by none (both of which are true of most of the applications of meaning-use analysis I consider in these lectures). Nor must the search for such semantic relations among vocabularies and the discursive practices-or-abilities they specify or that deploy them be motivated by some deep-seated philosophical anxiety or puzzlement, the proper deflating diagnosis of which then exhibits or renders the task of exploring those relations otiose. Simple curiosity, the desire to deepen our understanding, can suffice as much for this sort of philosophical theorizing as for the empirical scientific variety. Indeed, as Kuhn has taught us in the latter case, it really does not matter why the scientists do what they do, since the institution can ensure that so long as they act professionally, the result will be to extend our knowledge and deepen our understanding. So we might strive to make it be in philosophy—a light and harmless sort of motivational scientism.
8
With this thought on the table, I want to return briefly to the issue of metaphysics. Taking our cue from Geertz, we might think of metaphysics as the enterprise of crafting a vocabulary in which everything can be said. Now, he might be right that, in the sense I am after, natural languages are autonomous vocabularies in which everything can be said. But “craft” is doing some work in this definition. The metaphysician aims to construct a technical, artificial vocabulary with that same expressive power. Why? The greater control that regimentation gives vocabularies whose basic semantics is stipulated—in some other vocabulary, perhaps a natural language (no escaping the need for hermeneutic understanding)—and the rest of whose semantics is computed algorithmically. For we have a distinctively clear sort of understanding of whatever other vocabularies can be elaborated as target vocabularies from a base vocabulary constructed so as to exhibit this structure. We can get around my earlier worries about the concept of all possible vocabularies by understanding the “everything” regulatively, rather than constitutively. That is, for every vocabulary anyone comes up with, the metaphysician is committed to the favored base vocabulary being adequate, when suitably elaborated, to express what it expresses. We start by trying to codify the vocabularies we have, but acknowledge the commitment to address any more that may come along.
I think metaphysics in this sense is a perfectly reasonable undertaking, and that we potentially have a lot to learn from pursuing it. It is, perhaps, somewhat quixotic—but that is a practical, not a theoretical drawback. If we are to reap the rewards in understanding that engaging in this kind of metaphysical enterprise promises, however, I think it is crucial that it be pursued in the open-minded, pluralistic spirit of Lewis, and not in a more small-minded and exclusionary one. The distinction arises when the metaphysician fails to reconstruct in the favored terms all the antecedent uses of all the vocabularies it aspires to codify. I take it that such partial failures are inevitable. The task is just too hard, both for practical reasons and for principled ones. It is probably too much to ask even that for every target vocabulary one find some other regimented base vocabulary from which it can be elaborated. Finding some one regimented base vocabulary in terms of which every such target vocabulary can be reconstructed is far more difficult and unlikely. That is why it is to some extent a quixotic quest. But for the same sort of reasons that led to Popper’s methodological recommendation to endorse the strongest, most easily falsifiable theory not already falsified by the evidence, it can make sense to pursue the quantificationally more difficult goal. This is because, to the limited extent that one does succeed, one finds out more both about the metaphysical base vocabulary and about the target vocabularies to which it turns out either to be expressively adequate, or not to be expressively adequate. Traditional metaphysics treats the distinction as invidious. It denigrates and dismisses what resists formulation in its favored terms as ontologically second class: as unreal, as mere appearance. So, for Leibniz, relations and evil are unreal, relegated to the phenomenal realm of appearance. Later metaphysicians found themselves similarly rejecting as unreal such phenomena as time. (One of my favorite mind-benders is the attempt of some British idealists to dismiss finitude as an illusion, the effect produced by the infinite Absolute—what there really is—on poor incapable … merely finite minds.) In a more contemporary semantic key, the term of disapprobation may be “unintelligible” rather than the ontological “unreal.” That, at any rate, is the way the logical empiricists talk, and admirers of theirs such as Quine continued the practice (modal logic merely “engenders an illusion of understanding” of modal vocabulary, and intentional vocabulary is merely apparently coherent). (Nineteenth-century idealism, which more or less equates the unreal with the unintelligible, may be thought of in this respect as a transitional phase.)
But such a mean-spirited, suspicious, begrudging, exclusionary attitude is not the only one possible. One might instead take the great positive payoff of a particular metaphysical effort to consist in no small part in the particular line it draws between what target vocabularies (and which bits of each) can, and which cannot, be captured with the expressive resources of the base vocabulary whose metaphysical credentials are being explored. Each regimented base vocabulary, we might suppose, will determine a different boundary between the (relatively) expressible and inexpressible. Suppose we found out (it wouldn’t surprise me) that there is simply no way to say in the language of fundamental physics, no matter how it is elaborated with the resources of the most advanced logic, what Samnel Beckett said when he said, “I can’t go on. … I’ll go on,” or what Richard Nixon said when he said, “I am not a crook,” or what the Buddha said when he told the hot-dog man, “Make me one with everything.” We don’t need to say that they didn’t say anything, or to pretend that we can’t understand what they did say, we don’t need to deny that there is such a thing as going on, being a crook, and so on, to learn something about saying things from the relation between their vocabularies and that of fundamental physics that shows up in this expressive mismatch. Indeed, I think we learn more if we do not go on to adopt the wholly optional dismissive attitude. Our slogan should be “Metaphysical discrimination without denigration.” And just as Lewis thought it essential that we draw consequences rigorously from many sets of premises, so as to learn our way around by taking many different paths through the terrain, so the virtues of the metaphysical enterprise will manifest themselves most fully if we try out many different possible metaphysical base vocabularies. (A side benefit of adopting this plan is that we then need not resign ourselves to living out our lives oppressed by the steady drip, drip, drip of naturalistic semantics and ontology.)
My characterization of metaphysics transposes what is normally thought of in ontological terms into a semantic key. These versions can be thought of as related to one another on the model of Carnap’s material and formal modes. I’ve described the metaphysical project in metalinguistic terms. It might seem that the translation back and forth between these two ways of talking is so straightforward that it is perverse to flout ordinary philosophical usage by insisting on the metalinguistic version here. But in this case there is a significant asymmetry between them. Indeed, I think the asymmetry here reveals something important about the Carnapian dyad that we might otherwise not have seen: a new justification, from the side of pragmatism, for the characteristically analytic preference for the formal mode. For thinking of the metaphysical enterprise in semantic terms, as seeking to establish distinctive sorts of relations among vocabularies, opens up the possibility of considering in this case, too, pragmatically mediated semantic relations between vocabularies, in addition to the traditional kind. In particular, we can lay alongside the aspiration to find a vocabulary in which everything can be said, the aspiration to find one in which one can say everything one must be able to do in order to say anything, that is to use any vocabulary whatsoever. This is just the idea of a universal pragmatic metavocabulary.
The sort of illumination one would get from succeeding at the task of constructing a regimented de facto universal pragmatic metavocabulary is not exactly the same as that which one would get from succeeding at the task of constructing a regimented vocabulary whose expressive power encompassed that of all the vocabularies we could come up with to try it out on. But they would evidently be complementary forms of understanding: one telling us what we can say, and the other what we must do to say it. Further, any adequate pragmatic metavocabulary for a semantically adequate metaphysical vocabulary would be a universal pragmatic metavocabulary. I have already suggested, however, that the real payoff from the metaphysical enterprise should not be thought of as consequent upon the anticipation of complete success at producing a regimented semantically expressively universal vocabulary. In place of such a wholesale cognitive reward, we should think of the accumulation of retail rewards. Each only partially successful try at a universal metaphysical vocabulary draws a line between those antecedent vocabularies it can reconstruct, and those it cannot. And each such endeavor will draw a different line. The lesson I drew from the young David Lewis’s methodologically principled polymorphous theoretical promiscuity is that a valuable kind of understanding consists in the sort of knowing our way about secured only by multiplying the crisscrossing of concrete ways of drawing the boundary between the expressible and the inexpressible, not globally and absolutely, but locally and relative to specific base vocabularies. The same will hold true of attempts to construct regimented universal pragmatic metavocabularies: their value lies in the details of their only partial successes, in where, specifically, they fail, and in how the line between partial success and partial failure varies as we try out quite different candidate base vocabularies. Here one thinks of the parable with which Hempel closes “Empiricist Criteria of Cognitive Significance,” in which a dying father tells his sons a vast treasure is buried in their vineyard. Only many years later do they realize that their energetic but fruitless digging in search of the imaginary trove has led to their turning over the soil in just the way needed to ensure that their vines flourish. The romantic dream of total transformation by a single, magical find inspires the hard, unexciting daily work that gradually lays up the real treasure.
The parallel between the metaphysical goal of crafting a regimented universally expressive vocabulary and that of constructing a universal pragmatic metavocabulary, the genus of which these two tasks are species, is invisible if we think of metaphysics exclusively in ontological terms. The wider perspective is available only if we construe it semantically, in terms of relations between vocabularies. So viewed, they show up as complementary, corresponding to elements not only of the dimension defined by the semantic/pragmatic distinction, but also of the objective/subjective dimension: what is talked about and talking about it. They address objective-ontological and subjective-practical sides of the coin of discursiveness. Although both these sorts of totalizing project come into view from the vantage point of meaning-use analysis, what I am doing in the body of these lectures is neither metaphysics nor an attempt to construct a universal pragmatic metavocabulary. Rather, I am looking for a different kind of metavocabulary. It is at a higher level, making it possible to express crucial structural features of the relations between the dimensions of discursiveness they take as their targets, relations between what is said and what is done.
I have devoted this Afterword to responding to the suggestion that it detracts from my enterprise to frame it as a way of continuing and extending the classical twentieth-century project of philosophical analysis, because that project is a degenerating research program whose basic orienting commitments no longer deserve our allegiance. Although, as I acknowledged at the outset, that framing is not an essential element of the meaning-use analysis which it is my primary purpose to introduce here, there is another sense in which associating it with the analytic tradition is not merely optional. For, even apart from that way of motivating it, features that are intrinsic to my project place it squarely within the analytic tradition. (Sometimes you get to choose your tradition, and sometimes it chooses you.) I am, after all, principally concerned with developing a regimented vocabulary for expressing and discussing semantic relations among vocabularies, including pragmatically mediated ones. That is an analytic project, at least relative to one way of distilling an essence out of that multifarious tradition, one way of retrospectively rationally reconstructing it so as to make or find a common project that then becomes visible as having been implicit in it all along. My own philosophical interests focus on discursiveness, and in particular its semantic dimension. So I tend to view the history of philosophy—not just of analytic philosophy, but also Kant and Hegel and their early modern ancestors—through semantic spectacles, whose lenses throw into highest relief philosophers’ accounts of the nature of conceptual content and its place in discursive cognitive and practical activity. I think there is much to recommend this perspective on the philosophical tradition.19 But I am aware that it may appear to many as just another instance of the methodological carpenter’s rule: to the man who only has a hammer, the whole world looks like a nail. (More gender-neutral versions of this slogan seem somehow to miss something central about it.) Be that as it may, in a straightforward sense, to construe philosophy’s task in semantic terms (including, as I have just done, the traditionally ontological task of metaphysics) is to adopt a characteristically analytic perspective.
The project pursued here also places itself squarely in the analytic tradition in the centrality and importance it accords to logical vocabulary. Trying to get clear about the expressive role distinctive of logical vocabulary is one of the cardinal applications of the general metaconceptual apparatus deployed here. It provides the template on which the analyses of alethic modal and deontic normative vocabulary are modeled. And one of the principal insights gleaned from thinking about pragmatically mediated semantic relations is the new vindication of the semantic logicism characteristic of the classical project of philosophical analysis.
Further, the core idea I am pursuing in these lectures is itself a new kind of analysis: meaning-use analysis. It is generically like the original, narrower, logicist notion, in that it permits the exact, algebraic specification of relations among vocabularies, including those that have traditionally been of interest to philosophers. Complex meaning-use relations are built up from the basic ones combinatorially and recursively, in a way vividly and perspicuously represented by meaning-use diagrams. So it can be settled exactly what one is committed to by, and what one needs to show in order to entitle oneself to, the assertion that various vocabularies and associated sets of discursive practices-or-abilities stand in some specific complex meaning-use relation. Such a regimented, algebraic meta(meta)vocabulary evidently is a generalization of what classical philosophical analysis aimed at, widening the purely semantic scope of the tradition to incorporate relations between the meanings expressed by various vocabularies and the practices-or-abilities of deploying and specifying them.
Finally, among the senses in which a work can belong to the tradition of analytic philosophy is a stylistic one. Analytic philosophers do things such as distinguish six sorts of objections one might have to the analytic project, five senses of ‘scientism’, four ways in which a project belongs to the analytic tradition, two sorts of conclusion one could draw from the only partial success of a metaphysical project, and so on. In this sense, too, the current project reveals itself throughout as belonging to the analytic tradition. But this is a style of philosophical writing and thinking that long antedated twentieth-century philosophical analysis, for it is fully on display already in Aristotle, in Kant, in Husserl, and in the Heidegger of Sein und Zeit and the Grundprobleme.
I want to close by briefly addressing another ad hominem question I have found is often raised by this material: what is the relation between meaning-use analysis and the view developed in Making It Explicit? The short answer is that these are distinct projects. Meaning-use analysis does not depend at all on any of the principle theses of Making It Explicit: not its normative pragmatics, not its inferentialist semantics, not its account, in terms of those, of what is expressed by representational locutions. That having been said, there are two sorts of connection. First, the theoretical apparatus I present here did grow out of my thinking about what I was doing in Making It Explicit—in particular, about how semantic considerations were entangled, often fruitfully, with pragmatic ones. Again and again I found myself needing sharper analytic tools in order to make clear the space of theoretical and explanatory options, and the considerations that told in favor of one or another of them. Meaning-use analysis strives to be neutral as far as the more detailed accounts of meaning, use, and the relations between them, to which it is applied. It aims to be adequate to express Dummett’s views, Davidson’s, David Lewis’s, or Stalnaker’s—those, indeed, of any of the relatively small number of philosophers who are explicitly concerned with what it is thinkers and talkers must do (the practices they must engage in or the abilities they must exercise) in order thereby to count as associating semantically relevant whatsises (whatever semantic interpretants the semantic portion of the theory assigns) with their locutions and performances. But its home proving-ground was the account of discursive practice and conceptual content presented in Making It Explicit.
Second, the examples to which meaning-use analysis is applied in these lectures, and so the specific lessons about philosophically interesting vocabularies it is used to extract, are often ideas that are rooted in or familiar from my earlier work. The discussion in Lecture 2 of the expressive role characteristic of logical vocabulary, for instance, will be immediately recognizable to readers of Making It Explicit and Articulating Reasons.20 The apparatus of meaning-use analysis makes it possible to develop and clarify with hitherto impossible precision the inferentialist-expressivist line of thought about logic presented there. Again, the two-sorted deontic normative pragmatics, in terms of commitment and entitlement, and the three-sorted botanization of consequential-inferential semantic relations in terms of which the semantically primitive notion of incompatibility is introduced in Lecture 5 are just those motivated and discussed in Making It Explicit and Articulating Reasons. The subsequent exposition of incompatibility semantics for alethic modality redeems a promissory note issued in those works. And further examples are not far to seek.
This is not at all because the metatheoretic machinery of these lectures makes sense only in the context of those more particular ground-level views. My conception of the rhetorical shape I wanted the lectures to take is that I would move back and forth between introducing more of the general concepts of meaning-use analysis and making them more definite and showing their importance by applying them to particular cases—to issues concerning vocabularies of antecedent philosophical interest. The more suggestive and illuminating the results of such applications turn out to be, the more reason there is to think that the meta-metavocabulary of meaning-use analysis is cutting at important joints, making visible structures of general significance. The value of the result was supposed to be something like the vector product of what we learned from the applications and what we learn about the apparatus by applying it. Even those with no antecedent interest in relations between meaning and use might find that they learned a lot about issues they did care about (logic, AI, modality, intentionality …) by broadening their horizons to consider pragmatically mediated semantic relations among vocabularies. In looking for such payoffs, I naturally began with topics I already had something to say about, described in a vocabulary articulating a conceptual scheme I had already developed and deployed to some good effect elsewhere, where it seemed that additional analytic power might yield substantial further insights. The real test, of course, will be the extent to which others can usefully apply the ideas introduced here to further problems, vocabularies, and discursive phenomena, described in other terms, to produce promising and fruitful philosophical perspectives, approaches, ideas, and understandings.