LOGICAL POSITIVISM, LANGUAGE, AND THE RECONSTRUCTION OF METAPHYSICS
by Gustav Bergmann1. Introduction. A philosophical movement is a group of philosophers, active over at least one or two generations, who more or less share a style, or an intellectual origin, and who have learned more from each other than they have from others, though they may, and often do, quite vigorously disagree among themselves. Logical positivism is the current name of what is no doubt a movement. The common source is the writings and teachings of G. E. Moore, Russell, and Wittgenstein during the first quarter of the century. However, two of these founding fathers, Moore and Russell, do not themselves belong to the movement. The logical positivists have also greatly influenced each other; they still do, albeit less so as the disagreements among them become more pronounced. There is indeed vigorous disagreement, even on such fundamentals as the nature of the philosophical enterprise itself. The very name, logical positivist, is by now unwelcome to some, though it is still and quite reasonably applied to all, particularly from the outside. Reasonably, because they unmistakably share a philosophical style. They
*** * This essay appeared in English in Rivista Critica di Storia della Filosofia 8. (1953),
453-81. Reprinted by permission.
30
all accept the linguistic turn Wittgenstein initiated in the Tractatus. To be sure, they interpret and develop it in their several ways, hence the disagreements; yet they are all under its spell, hence the common style. Thus, if names in themselves were important, it might be better to choose linguistic philosophy or philosophy of language. In fact, these tags are now coming into use. But they, too, like most labels, are misleading. For one, the concern with language is nothing new in first philosophy or, if you please, epistemology and metaphysics. Certainly all "minute philosophers" have shared it. For another, there is strictly speaking no such thing as the philosophy of language. Language may be studied by philologists, aestheticians, and scientists such as psychologists or sociologists. To bring these studies thoughtfully together is well worth while. Customarily, such synoptic efforts are called philosophy. There is no harm in this provided they are not mistaken for what they are not, namely technical philosophy. Rather than being philosophers of language, the positivists, who are all technical philosophers, are therefore philosophers through language; they philosophize by means of it. But then, everybody who speaks uses language as a means or tool. The point is that the positivists, newly conscious of it, use it in a new way.
The novelty is, I believe, radical. Even the greatest innovators never do more, can do no more, than add one or two features to the tradition perhaps submerge one or two others. The tradition as a whole persists. Features is a vague word. I had better speak of new questions and methods, for they, not the answers we give, matter. The logical positivists neither added nor submerged a single major question. Their characteristic contribution is a method. This may mean radical novelty; it does, I believe, in their31
case. There is a sense, though, in which the linguistic turn has not even produced startling new answers. The answers the positivists give to the old questions, or those which most of them give to most, are in some respects very similar to what has been said before within the empiricist stream of the great tradition. On the other hand, both questions and answers are so reinterpreted that they have changed almost beyond recognition. At least, alas, beyond the recognition of many. Many of the logical positivists themselves, like other innovators before, even thought that they had disposed of the tradition. Some still believe it. I think there is merely a new method, though one that is radically new, of approaching the old questions.
This is not a historical paper. I wish to speak as a philosopher. Thus, while I am aware of how much I owe to others. I can only speak for myself. Nor is my intent primarily critical. Yet, such is the dialectical nature of philosophy that we cannot either in thinking or in writing do without that foil the ideas of others provide. This makes us all critics as well as, in a structural sense, historians. Thus, while it is my main purpose, or very nearly so, to explain one kind of logical positivism. I shall almost of necessity, discuss all others. They fall into two main divisions. The one is made up by the ideal linguists, the other by the analysts of usage, more fully, of correct or ordinary usage. The ideal linguists are either formalists or reconstructionists. The outstanding formalist is Carnap. What the reconstructionists hope to reconstruct in the new style is the old metaphysics. Clearly, from what has been said, I am a reconstructionist. There is, third, the pragmatist variety. These writers, we shall presently see, are best counted with the ideal linguists. Usage analysis flourishes above all at Oxford and Cambridge. These
32
philosophers are also known as, fourth, the therapeutic positivists or casuists. One variant of this view deserves to be distinguished. For want of a better term I shall, with a new meaning, resuscitate an old one, calling this view, fifth, conventionalist. This wing is led by Ryle.
The expositor's position determines, as always, his strategy. The argument will center around reconstructionism. But since I believe the method to be neutral in that it may be used by all and any, I shall set it off as clearly as I can from the specific conclusions to which it has led me. Not surprisingly, these conclusions, or answers to the old questions, lie within the empiricist tradition, if it is conceived broadly enough to include the act philosophies of Moore and Brentano. The debt to Hume and the phenomenalists in general is, naturally, tremendous. One clever Englishman recently proposed the equation: Logical Positivism is Hume plus mathematical logic. He has a point, though by far not the whole story. But whatever these specific conclusions may be, I can hardly do more than hint at a few of them. This must be kept in mind throughout. I have, of course, discussed them elsewhere. Here, however, they serve mainly as illustrations, pour fixer les idees, for even in philosophy abstractness cannot without disadvantage be pushed beyond certain limits.
2. The linguistic turn. What precisely the linguistic turn is or, to stay with the metaphor, how to execute it properly is controversial. That it must be executed, somehow or other, is common doctrine, flowing from the shared belief that the relation between language and philosophy is closer than, as well as essentially different from, that between language and any other discipline. What are the grounds of this belief and how did it arise?
33
First. There is no experiment on whose outcome the predictions of two physicists would differ solely because the one is a phenomenalist, the other a realist. Generally, no philosophical question worthy of the name is ever settled by experimental or, for that matter, experiential evidence. Things are what they are. In some sense philosophy is, therefore, verbal or linguistic. But this is not necessarily a bad sense. One must not hastily conclude that all philosophers always deal with pseudoproblems. Those who thus stretch a point which is telling enough as far as it goes, are overly impressed with the naive "empiricism" of the laboratory. Most of them are formalists. Scientism and formalism, we shall see, tend to go together. Second. Philosophers maintain in all seriousness such propositions as that time is not real or that there are no physical objects. But they also assure us that we do not in the ordinary sense err when, using language as we ordinarily do, we say, for instance, that some event preceded some other in time or that we are perceiving physical objects such as stones and trees. Outside their studies, philosophers themselves say such things. Thus they use language in two ways, in its ordinary sense and in one that is puzzling to say the least. To decide whether what they say as philosophers is true one must, therefore, first discover what they say, that is, precisely what the peculiar sense is. The inquiry is linguistic. It starts from common sense, for what else is there to start from. These points were pressed by G. E. Moore. His emphasis on ordinary usage and common sense reappears, of course, in the British branches of the movement. The common-sense doctrine also influenced the reconstructionists. It is worth noticing, though, that in the form in which all these positivists have adopted it, the doctrine is not itself a philosophical proposition. Rather, it helps
34
to set their style, assigning to philosophy the task of elucidating common sense, not of either proving or disproving it. In this form the common-sense doctrine also represents at least part of what could be meant by saying, as both Husserl and Wittgenstein do, that philosophy is descriptive. Third. This point stands to the second in a relation similar to that between morphology and physiology or, perhaps, pathology. We have seen that philosophers, using language in their peculiar sort of discourse, arrive at such propositions as that there are no physical objects. Taken in their ordinary sense, these propositions are absurd. The man on the street, however, who uses the same language never ends up with this kind of absurdity. We also know that the conclusions one draws depend on the grammatical form of the statements that express the premises. We notice, finally, that sometimes two statements, such as 'Peter is not tall' and 'Cereberus is not real', exemplify the same grammatical form though they say really quite different things. We conclude that philosophers come to grief because they rely on grammatical form. What they should trust instead is the logical form of statements such as, in our illustration, 'Peter is not tall' and 'There is no dog that is three-headed, etc.'. Consistently pursued, the notion of logical form leads to that of an ideal language in which logical and grammatical form coincide completely. Both notions took shape when Russell answered several philosophical questions, some about arithmetic, some about just such entities as Cerebus, by means of a symbolism. There is one more suggestion in all this, namely, that in an ideal language the philosopher's propositions could no longer be stated so that he would find himself left without anything to say at all. 'Peter exists', for instance, has no equivalent in Russell's symbolism, Peter's existence showing itself, as it
35
were by the occurrence of a proper name for him. Ontology is, perhaps, but an illusion spawned by language. So one may again be led to think that all philosophy is verbal in a bad sense. The suggestion seduced the formalists as well as those who later became usage analysts. It even seduced Wittgenstein. The reconstructionists reject it. According to them, philosophical discourse is peculiar only in that it is ordinary or, if you please, commonsensical discourse about an ideal language.
Ordinary disourse about an ideal language is, indeed, the reconstructiont version of the linguistic turn. But a statement so succinct needs unpacking. Precisely what is an ideal language? I cannot answer without first explaining what syntax is.
3. Syntax Signs or symbols may be artificial, that is, expressly devised, or they may have grown naturally. In either case they do not stand for anything by themselves. We speak by means of them; we "interpret" them; having been interpreted, they "refer." Syntax deals only with some properties of the signs themselves and of the patterns in which they are arranged. This, and nothing else, is what is meant by calling syntax formal and schemata syntactically constructed formal languages. It would be safer to avoid any term that suggests interpretation, such as 'language', 'sign', or 'symbol'. I shall simply speak of syntactical schemata and their elements. Or one could use a prefix to guard against confusion, calling the elements f-signs, for instance, 'f' standing for 'formal'. In this section, where I discuss only f-notions, I shall suppress the prefix. Later on I shall occasionally take this precaution. In themselves, signs are physical objects or events. Written signs, and we need not for our purpose consider others, are instances of geometrical shapes. Syntax
36
is thus quite commonsensical business. It is, so to speak, a study geometrical design. But philosophers are not geometricians. They do not invent and investigate these schemata for their own sake, as mathematical logicians often do, but with an eye upon their suitability for serving, upon interpretation, as the ideal language. Making this claim for any one schema, the geometrician turns philosopher, committing himself to a philosophical position. This is why I insisted that the method as such is neutral. Yet, to introduce neutrally the syntactical notions or categories (f-categories!) which I shall need would be tediously abstract and is, at any rate, quite unnecessary for my purpose. So I shall, instead, introduce them by describing that particular schema of Russell's Principia Mathematica. Very broadly indeed; and I shall have to speak broadly throughout the rest of this section, simplifying so sweepingly that it amounts almost to distortion, though not, of course, as I judge it, to essential distortion.
The construction of the schema proceeds in three steps. First one selects certain shapes and kinds of such as its elements or signs. Then certain sequences of shapes are selected or, if you please, defined as its sentences. Order, as the term sequence implies, enters the definition. Finally a certain subclass of sentences, called analytic, is selected. Turning to some detail, relatively speaking, I shall, in order to fix the ideas, add in parentheses some prospective interpretations from our natural language. First. The elements are divided into categories. Though based on shape and nothing else, the divisions are not nominal in that the definitions of sentence and analyticity are stated in their terms. Signs are either
37
logical or descriptive. Descriptive signs are either proper names ('Peter'), or predicates and relations of the first order ('green', 'louder than'), or predicates and relations of higher orders ('color'). Logical signs are of two main kinds. Either they are individually specified signs, connectives ('not', 'and''if then') and quantifiers ('all', 'there is something such that'). Or they are variables. To each descriptive category corresponds one of variables, though not necessarily conversely; to proper names so-called individual variables (such phrases as 'a certain particular'), and so on. Second Sentences are either atomic or complex. Atomic sentences are sequences of descriptive signs of appropriate categories ('Peter (is) green', 'John (is) taller than James'). Complex sentences contain logical signs ('John (is) tall and James (is) short', 'There is something such that it is green'). Third. In defining analyticity arithmetical technics are used; in the sense in which one may be said to use such technics who, having assigned numbers to people on the basis of their shapes, called a company unlucky (f-unlucky!) if the sum of the numbers of its members is divisible by 13. A sentence is said to follow deductively from another if and only if a third, compounded of the two in a certain manner, is analytic. ('p' implies 'q' if and only 'if p then q' is analytic.) The definition of analyticity is so designed that when a descriptive sign occurs in an analytic sentence, the sentence obtained by replacing it with another descriptive sign of the same category is also analytic. (In 'Either John is tall or John is not tall', the terms 'John' and 'tall' occur vacuously.) Two such sentences are said to be of the same "logical form"; analyticity itself is said to depend on "form" only, which is but another way of saying that it can be characterized by means
38
of sentences which contain none but logical signs. This feature is important. Because of it, among others, f-analyticity can, as we shall see, be used to explicate or reconstruct the philosophical notions of analyticity which unfortunately, also goes by the name of formal truth. Unfortunately, because the f-notion of logical form which I just defined needs no explication. The philosophical notion, like all philosophical ones, does. To identify the two inadvertently, as I believe Wittgenstein did, leads therefore to disaster. But of this later.
The shapes originally selected are called the undefined signs of the schema. The reason for setting them apart is that many schemata, including the one I am considering provide machinery for adding new signs. To each sign added corresponds one special sentence, called its definition, the whole construction being so arranged that this sentence is analytic. This has two consequences. For one, the definitions of the language which, in some sense, the schema becomes upon interpretation, are all nominal. For another, interpretation of the undefined signs automatically interprets all others. Defined signs whose definitions contain undefined descriptive signs are themselves classified as descriptive.
4. Ideal language and reconstruction. To interpret a syntactical schema is to pair its undefined signs one by one with words or expressions of our natural language, making them "name" the same things or, if you please, "refer" equally. An interpreted schema is in principle a language. In principle only, because we could not speak it instead of a natural language; it is neither rich nor flexible enough. Its lack of flexibility is obvious; it lacks richness in that we need not specify it beyond, say, stipulating that it contains color predicates, without bothering which or how many. Thus even an interpreted schema is merely, to use the term in a
39
different sense, the "schema" of a language, an architect's drawing rather than a builder's blueprint. The ideal language is an interpreted syntactical schema. But not every such schema is an ideal language. To qualify it must fulfill two conditions. First, it must be complete, that is, it must no matter how schematically, account for all areas of our experience. For instance, it is not enough that it contain schematically the way in which scientific behaviorists, quite adequately for their purpose, speak about mental contents. It must also reflect the different way in which one speaks about his own experience and, because of it, of that of others; and it must show how these two ways jibe. Second, it must permit, by means of ordinary discourse about it, the solution of all philosophical problems. This discourse, the heart of the philosophical enterprise, is the reconstruction of metaphysics. So I must next explain how to state, or restate, the classical questions in this manner and, if they can be so stated, why I insist that this discourse is, nevertheless, quite ordinary or commonsensical though, admittedly, not about the sort of thing the man on the street talks about. Making the range of his interests the criterion of "common sense" is, for my taste a bit too John Bullish.
Consider the thesis of classical nominalism that there are no universals. Given the linguistic turn it becomes the assertion that the ideal language contains no undefined descriptive signs except proper names. Again, take classical sensationism. Transformed it asserts that the ideal language contains no undefined descriptive predicates except nonrelational ones of the first order, referring to characters exemplified by sense data which are, some ultrapositivists to the contrary notwithstanding, quite commonsensical things. I reject both nominalism and sensationalism. But this is not the
40
point. The point is that the two corresponding assertions, though surely false, are yet not absurd, as so many of the classical theses are, as it is for instance absurd to say, as the sensationalists must, that a physical object is a bundle of sense data. Obvious as they are, these two illustrations provide a basis for some comments about the reconstruction in general.
First. I did not, either affirmatively or negatively, state either of the two classical propositions. I merely mentioned them in order to explicate them, that is, to suggest what they could plausibly be taken to assert in terms of the ideal language. For the tact and imagination such explication sometimes requires the method provides no guarantee. No method does. But there is no doubt that this kind of explication, considering as it does languages, is quite ordinary discourse. Yet it does not, by this token alone, lose anything of what it explicates. To say that a picture, to be a picture, must have certain features is clearly, to say something about what it is a picture of. I know no other way to speak of the world's categorial features without falling into the snares the linguistic turn avoids. These features are as elusive as they are pervasive. Yet they are our only concern; that is why the ideal language need be no more than a "schema." I just used the picture metaphor, quite commonsensically I think, yet deliberately. For it has itself become a snare into which some positivists fell, not surprisingly, since it is after all a metaphor. Of this later. Second. A critic may say: "Your vaunted new method either is circular or produces an infinite regress. Did you not yourself, in what you insist is ordinary discourse, use such words as 'naming' and 'referring'? Surely you know that they are eminently philosophical?" I have guarded against the objection by putting quotation marks around these words when I first used them. The point is that
41
I did use them commonsensically, that is, in a way and on an occasion where they do not give trouble. So I can without circularity clarify those uses that do give rise to philosophical problems, either by locating them in the ideal language, or when I encounter them in a philosophical proposition which I merely mention in order to explicate it, or both, as the case may be. But the critic continues: "You admit then, at least, that you do not, to use one of your favorite words, explicate common sense?" I admit nothing of the sort. The explication of common sense is circular only as it is circular to ask, as Moore might put it, how we know what in fact we do know, knowing also that we know it. Third. The critic presses on: "Granting that you can without circularity explicate the various philosophical positions, say, realism and phenomenalism. I still fail to see how this reconstruction, as you probably call it, helps you to choose among them." I discover with considerable relief that I need no longer make such choices. With relief, because each of the classical answers to each of the classical questions has a common-sense core. The realist, for instance, grasped some fundamental features of experience or, as he would probably prefer to say, of the world. The phenomenalist grasped some others. Each, anxious not to lose hold of his, was driven to deny or distort the others. From this squirrel cage the linguistic turn happily frees us. Stated in the new manner, the several "cores" are no longer incompatible. This is that surprising turn within the turn which I had in mind when I observed that the old questions, though preserved in one sense, are yet in another changed almost beyond recognition. To insist on this transformation is one thing. To dismiss the classical question out of hand as some positivists unfortunately do, is quite another thing. Fourth. The method realizes the old ideal of a philosophy
42
without presuppositions. Part of this ideal is an illusion, for we cannot step outside of ourselves or of the world. The part that makes sense is realized by constructing the schema formally, without any reference to its prospective use, strict syntacticism at this stage forcing attention upon what may otherwise go unnoticed. But the critic persists: "Even though you start formally, when you choose a schema as the ideal language you do impose its "categories" upon the world, thus prejudging the world's form. Are you then not at this point yourself trading on the ambiguity of 'form' as you just said others sometimes do?" One does not, in any intelligible sense, choose the ideal language. One finds or discovers, empirically if you please, within the ordinary limits of human error and dullness, that a schema can be so used. Should there be more than one ideal language, then this fact itself will probably be needed somewhere in the reconstruction; equally likely and equally enlightening, some traits of each would then be as "incidental" as are some of Finnish grammar. More important, all this goes to show that the reconstructionist's philosophy is, as I believe all good philosophy must be, descriptive. But it is time to relieve the abstractness by showing, however sketchily, the method at work.
5. Three issues. The common-sense core of phenomenalism is wholly recovered by what is known as the principle of acquaintance. (Later on I shall restore the balance by reconstructing what I think is the deepest root of realism. Realism, to be sure, has others, such as the indispensibility of the quantifiers which permit us to speak of what is not in front of our noses. But these roots run closer to the surface.) The word principle is unfortunate; for description knows no favorites. The feature in question is indeed a principle only in that quite a few other explications are found
43
to depend on it. What is asserts is that all undefined descriptive signs of the ideal language refer to entities with which we are directly or, as one also says, phenomenally acquainted. Notice the difference from sensationism. Relational and higher-order undefined predicates are not excluded. The indispensibility of at least one of these two categories is beyond reasonable doubt. Nor does the principle exclude undefined descriptive signs that refer to ingredients of moral and aesthetic experience. If ethical naturalism is explicated as the rejection of such terms, then one sees that a reconstructionist need not be an ethical naturalist. I, for one, am not.
The ideal language contains proper names, the sort of thing to which they refer being exemplified by sense data; 'tree' and 'stone' and 'physical object' itself are, broadly speaking, defined predicates, closer analysis revealing that the "subjects" of these predicates, do not refer to individual trees and stones. That this amounts to a partial explication of the substantialist thesis, accepting a small part of it and rejecting the rest, is fairly obvious. Another aspect of the matter raises two questions. Definitions are linguistic constructions, more precisely, constructions within a language. How detailed need they be? What are the criteria for their success? To begin with the second question, consider the generality 'No physical object is at the same time at two different places'. Call it S and the sentence that corresponds to it in the ideal language S'. Since 'time' and 'place' is S refer to physical time and place, the descriptive signs in S' are all defined. Their construction is successful if and only if S' and a few other such truths, equally crucial for the solution of philosophical problems, follow deductively from the definitions proposed for them in conjunction with some other generalities containing only undefined descriptive signs, which
44
we also know to be true, such as, for example, the sentence of the ideal language expressing the transitivity of being phenomenally later. The construction is thus merely schematic, in the sense in which the ideal language itself is merely a schema. The building stones from which it starts in order to recover the sense in phenomenalism are so minute that anything else is patently beyond our strength. Nor, fortunately, is it needed to solve the philosophical problems. To strive for more is either scientism or psychologism, scientism if one insists on definitions as "complete" as in the axiomatization of a scientific discipline, psychologism if one expects them to reflect all the subtlety and ambiguity of introspective analysis. Formalists tend to scientism; usage analysts to psychologism.
Analyticity is not a common-sense notion. However, the differences that led philosophers to distinguish between analytic and synthetic propositons are clearly felt upon a little reflection. There is, first, a difference in certainty, one of kind as one says, not merely of degree. Or, as it is also put, analytic truth is necessary, synthetic truth contingent. Certainty is a clear notion only if applied to beliefs. Besides, what is sought is a structured or objective difference between two kinds of contents of belief. There is only this connection that, once discovered, such a structural difference will be useful in explicating the philosophical idea of certainty. Second, analytic (tautological) truths are empty in that they say nothing about the world, as 'John is either tall or not tall' says nothing. Third, there is even in natural languages the difference, often though not always clear-cut, between descriptive (not f-descriptive!) words such as 'green' and logical (not f-logical!) ones such as 'or'. Analyticity depends only on the logical words and on grammatical "form." Fourth, descriptive
45
words seem to refer to "content," to name the world's furniture, in a sense in which logical words do not. These, I believe, are the four felt differences which philosophers, including many positivists, express by calling analytical truths necessary, or formal, or syntactical, or linguistic. Without explication the formula courts disaster; its explication has four parts, all equally important. First, our knowledge that all "content" variations of analytic "form" ('George is either tall or not tall', 'James is either blond or not blond', etc.) are true is, in the ordinary sense, very certain. But no claim of a philosophical kind for the certinty of this knowledge can be the basis of our explciation; it can only be one of its results. Second, the notions of analyticity and of logical and descriptive words correspond to perfectly clear-cut f-notions of the ideal language. Third, the specific arithmetical definition of f-analyticity in the ideal language (that is, in the simplest cases the well-known truth tables) shows in what reasonable sense analytical truth is combinatorial, compositional, or linguistic. Fourth, arithmetic, the key to this definition, is itself analytic upon it. Taken together these four features amply justify the philosophers' distinction between what is either factual or possible (synthetic) and what is necessary (analytic), between the world's "form" and its "content." But if they are taken absolutely, that is, independently of this explication, then the phrases remain dangerously obscure. Greatest perhaps is the danger of an absolute notion of form as a verbal bridge to an abosolute notion of certainty. Nothing is simpler, for instance, than to set aside syntactically a special class of first-order predicates, subsequently to be interpreted by color adjectives, and so to define f-analyticity that 'Nothing is (at the same time all over) both green and and red' becomes analytic. Only, this kind of f-analyticity would
46
no longer explicate the philosophical notion. Ours does. But that it does this is not itself a formal or linguistic truth.
Ontology has long been a favorite target of the positivistic attack. So I shall, for the sake of contrast, reconstruct the philosophical query for what there is. The early attacks were not without grounds. There is, for one, the absurdity of the classical formulations and, for another, the insight, usually associated with the name of Kant, that existence is not a property. In Russell's thought, this seed bore double fruit. On the one hand, when 'Peter' is taken to refer to a particular, 'Peter exists' cannot even be stated in the ideal language; his "existence" merely shows itself by the occurrence of a proper name in the schema. On the other hand, such statments as 'Ther are no centaurs (centaurs do not exist)' or "There are coffeehouses in Venice' can be expressed in the ideal language, in a way that does not lead to absurdity, by means of quantifiers, which are logical signs, and of defined predicates, whose definitions do not involve the "existence" of the kinds defined. This is as it should be. Ontological statements are not ordinary statements to be located within the ideal language; they are philosophical propositions to be explicated by our method. Logical signs, we remember, are felt not to refer as descriptive ones do. This reconstructs the clasical distinction between existence and subsistence. Ontology proper asks what exists rather than subsists. So the answer to which we are led by our method seems to be a catalogue of all descriptive signs. Literally, there can be no such catalogue; but one would settle for a list of categories, that is, of the kinds of entities to which we refer or might have occasion to refer. But then, every serious philosopher claims that he can in his fashion talk about everything. So one could no hope to reconstruct the various ontological
47
theses by means of a list of all descriptive signs. The equivalent of the classical problem is, rather, the search for the undefined descriptive signs of the ideal language.
1 I used the idea implicitly when I explicated nominalism and phenomenalism. To show that it is reasonable, also historically, consider two more examples. Take first materialism or, as it now styles itself, physicalism or philosophical behaviorism. Interpreted fairly, even this silliest of all philosophies asserts no more than that all mental terms can be defined in a schema whose undefined descriptive predicates refer to characters exemplified by physical objects. Quite so. I, too, am a scientific behaviorist. Only, the materialist's schema is, rather obviously, incomplete and therefore not, as he would have to assert, the ideal language. Russell, on the other hand, when he denied the existence of classes, meant, not at all either obviously or sillily, no more than that class names are defined signs of the ideal language.
Wittgenstein. Historically, Wittgenstein's Tractatus is the source of reconstructionism. Systematically, having at least four grave defects, it is merely an anticipation. Each of these defects has influenced the course of the movement.
There is first, Wittgenstein's famous ineffability thesis. Taken by itself, even an interpreted syntactical schema is mute. Literally, it does not speak about or refer to anything; we speak or refer by means of it. Of course, one may guess as Fancois Champollion once did or very nearly did, what a syntactical schema could be made to say; but this is a different story. Yet, Wittgentein's formula that language cannot speak about itself is misleading in two ways. For one, it
***** 1 One could argue that this conception of ontology is anticipated in the Tractatus (2.01, 2.02, 2.027). But I was not aware of that when I first proposed it.48
may be taken to assert that one cannot or must not construct syntactical schemata which upon interpretation refer to themselves. Natural languages do in this perfectly harmless sense speak about themselves. It is true, though, and this is probably one of the contributory causes of many confusions, that in natural languages such discourse sometimes produces paradoxes, particularly and not surprisingly if one neglects the distinction between an expression and its name. In the ideal langauge this need not be the case. The precautions its syntactical construction requires in order to avoid those neat paradoxes to which the mathematicians have first called our attention may be safely left to the mathematicians. This is not a philosophical problem. Nor was Wittgenstein, I believe, overly impressed by it. He used his formula to reject all philosophy as a futile attempt to talk about the ineffable. This is his version, well suited to the streak of mysticism that was in him, of the linguistic turn. But then again, it shows how close to the reconstructionist idea he really was. He did in fact propose an ideal language; only, he forbade de jure all discourse about it, that Moorean discourse, if I may so call it, which is the heart of the philosophical enterprise. De facto, as Russell observed, he managed to say a good deal about his ideal language, in a style as terse as it is often obscure, and certainly anything but Moorean.
Second, Wittgenstein's notions of form and analyticity are absolute, in the sense I explained and rejected. Consider, for instance, his assertion that every "possible world," whatever its "content," would yet be of the same "logical form" as ours, which implies, among other things that 'p or not-p' is in all possible world's "necessarily true." The point is that without some such explication as I have given the expressions within double quotes remain problematic. Or, to put the same thing
49
differently, Wittgenstein uses them, in what should be ordinary discourse about the ideal language, philosophically. Take the relevant explication of 'possible'. What is thus possible is what is expressed by a sentence which is synthetic upon that notion of analyticity that fits our world. The circularity is obvious. Even worse, one can so explicate the phrase 'possible world' that Wittgenstein is seen to be outright mistaken. Nothing keeps one from constructing a schema, in all other respects like ours, whose "logic" is, say, three-valued. To be sure this is tediously abstract; conversely, to call such a schema a possible world is a very exuberant metaphor. For one would not know how to interpret it so that it becomes the ideal language. But I, for one, do not wish to argue with those who insist that the world "could" be so, or could so change tomorrow, that this and ony this schema could serve as ideal language. I never claimed to do more than to describe what I find, empirically, as it were.
In contrast to this hypostatization of form, a strange weakening of the notion occurs at two places in the Tractatus. This is its third major defect. The dominant and, I think, correct doctrine is that what is formal or syntactical about such predicates as 'green', 'red' and 'square' is merely that they are predicates, more precisely, nonrelational predicates of the first order. Yet, Wittgenstein also maintains that some such sentences as 'nothing is (all over at the same time) both green and red' is analytic. 'Nothing is both green and square' is not even true. Since Wittgenstein himself exploits to the utmost the insensitivity of analytic form to content variation, the insensitivity of analytic form to content variation, the inconsistency is only too manifest. One of the causes of this curious slip is, perhaps, the confusion to which I have called attention before of the purely structural notion
50
of analyticity with the essentially psychological one of certainty. It is indeed mere quibbling to insist that there is a difference in certainty among the two beliefs that nothing is both green and not green and that nothing is both green and red. But if one wishes to maintain that the second of these two propositions is analytic, then one is virtually forced to argue as the casuists argue. In his later life Wittgenstein became indeed the founding father of casuism. But of this presently. In the Tractatus the weakened notion of form appears once more, in a passage where space and time are counted part of the world's form, not with its content. This is a curious Kantian echo. Some spatial and temporal predicates must appear among the undefined descriptive signs of the ideal language. This is the core of the so-called relational theory of space and time. To put it succinctly, the world is not in space and time; space and time are in the world. Nor is it without interest in this connection that Wittgenstein was very fond of the metaphor of logical space, for the world is not suspended in logical space either. Or, if you please, everything is a fact, or merely a fact. Only, to say this is to say no more than that at night all cows are black. Philosophy is still the art of distinction.
There are, finally, what seem to be the grounds Wittgenstein's ontology. I say seem because I believe that he himself was not aware, or not fully aware, of these grounds. An unexamined metaphysics, that is, one implicitly held, is for a philosopher the worst metaphysics of all. This is therefore where the fourth trouble lies. I, for one, share what I take to be Wittgenstein's none too explicit ontological views. But I also believe that if these views are held explicitly and on the right grounds then they do not occupy the strategic position which implicitly they have in his thought. In other words, if
51
I discovered that the reasoning that led me to them was fallacious or if the world were in this one respect different, I would not by this token alone have to modify or abandon any other of my opinions. Stated as I would state it, the thesis is that the undefined descriptive signs of the ideal language are all either proper names or first-order predicates including relational ones. I have called this view elementarism in order to indicate that next to nominalism it is the least crowded of all ontologies. The hidden grounds on which Wittgenstein hold it are two. The first is not hard to find, particularly if one examines the traditional version which says that all there is are particulars and the characters they exemplify. Replace in this formula the deliberately jejune 'there is' by the conventionally unanalyzed 'exists', replace 'particulars' by 'particulars in space and time', and you will recognize the old Democritean suspicion against all "abstract" or "Platonic" entities, the rejection of anything not "in" space and time. Of course, 'in' is here merely a rudimentary metaphor, a particular being no more "in" the web of spatial and temporal relations, some of which it exemplifies, than it is "in" greenness if it is green or "in" pitch if it is a tone. Clearly, there is some connection between this root of Wittgenstein's elementarism and his views on space and time; and there is at least some vague affinity to the Democritean physicalism of most formalists. Nor is it without interest that, as I hope to show, elementarism plays a strategic role in Ryle's thought. Wittgenstein's second ground is more deeply hidden. Let me explain. So far I have not distinguished between naming and referring. Yet, Wittgenstein was impressed with the difference between the ways in which sentences and terms "refer." According to him, a sentence refers to a state of affairs; it does not name it, for what is there to be named in the case
52
of, say, a false sentence? Proper names and undefined predicates, on the other hand, if they are introduced in accordance with the principle of acquaintance, do name something that is there. Adopting this usage, an elementarist might say that what exists is what can be named. Surely this is a very roundabout way of speaking and, like all such ways, may lead to perdition. At the moment this is not my point. The point is that this is where the picture metaphor, taken literally, creates pseudoproblems. A picture depicts trait by trait what it is the picture of. So do names. Language does not, otherwise it would be of no use. Yet, according to the metaphor which I believe guided Wittgenstein, language is a picture of the world. Of course, he was not so foolish as to do away with sentences. Nor was he, I think, overly impressed by the classical poser of negative propositions. But the fact remains that the only sentences that did not puzzle him were atomic ones such as 'this is green' or 'this is to the left of that'. Defined predicates and complex sentences he really trusted were connective combinations, such as conjunctions and disjunctions, of atomic ones, because the truth tables had convinced him that connectives do not, as it were, pretend to name anything. So they are harmless and may pass. This, I believe, is the second ground of his elementarism. The overt motive was, and with some still is, to find a so-called meaning criterion, that is, to circumscribe what may be said in the ideal language. The picture metaphor is a poor guide in this undertaking. Nor is this really a new a separate problem. All we need to observe if we want to make sure that we understand what we say is syntax and the principle of acquaintance, the one when we form our sentences, the other when we choose their descriptive constituents.
53
7. Further issues The ideal language contains such statements as "There are coffeehouses in Venice' and 'There are physical objects'. This answers the criticism that whoever accepts the principle of acquaintance, since he is not able to say these things, cannot escape the absurdities of phenomenalism. Such critics hanker after the philosophical assertion that physical objects exist, that is, as I explicate it, they insist that the ideal language contains undefined descriptive signs referring to physical objects and their characters. This is a far cry from common-sense realism. Yet, realism has a deeper root, which we must first explicate and then reconstruct. Common sense rightly insists on the difference between 'This is a tree' and 'I know (see) that this is a tree'. Neither of these two statements being philosophical, they must both be located in the ideal language. The second raises the issue of acts or, as some positivists call them, propositional attitudes, knowing, remembering, doubting, and so on, in brief, the issue of awareness. This is essentially not a psychological question, though it is also psychologically undeniable that, first, an awareness, say, a knowing, is different from its content, say, that this is a tree; and, second that some awarenesses, say, a knowing, is different from its content, say, that this is a tree; and, second, that some awarenesses, say, that this is a tree, are the contents of others. The specter of an infinite regress which Hobbes raised at this point against Descartes and which Brentano tried to lay by his notion of an eignentuemliche Verflechtung need not worry us. This is indeed a psychological affair; all we are called upon to produce is a schema of the act pattern, that is, of such sentences as 'I know that this is a tree'. Now I know no way of introducing the act verbs by definition, without using in these definitions at least one of them, that does not run into difficulties somewhere. Taking a hint from the psychologically inclined among the classical
54
phenomenalists, one might try, for instance, to define these characters in terms of some minute kinaesthetic qualities and similar fictions of a doctrinaire introspectionism. The trouble is that since these are fictions, the proposal violates the principle of acquaintance. Taken philosophically, it thus raises more questions than it can answer. It follows that the ideal language contains at least one undefined descriptive sign naming an act, say, awareness. What is the syntactical status of this sign? If one follows the lead of ordinary grammar, then what is needed is not only a new sign but a new syntactical category, since as they are used in our natural languages the act verbs correspond neither to predicates nor to proper names of the ideal language, for in our natural languages 'knowing that' is, without being a connective, a modifier of a whole sentence, the one which is called the dependent clause. To choose this course is to reintroduce many of the classical absurdities; also, it amounts to abandoning some of the clarifications we have achieved, for instance, as I cannot here explain in detail, the explication of analyticity. Fortunately, this course need not, nay, must not be chosen. The decisive insight is that an awareness, which is never identical with its content, is itself a particular. (This is indeed the deepest and most significant core of Berkeley's "nominalism.") The name of a particular, however, is a proper name and this makes the act verbs in the ideal langauge ordinary predicates. Are they relational predicates? Or, to put the same thing differently does the sentence which corresponds, on the ground floor of the ideal language, to our paradigm, read more nearly like 'I know that this is a tree' or like 'Knowing that this is a tree'? This is, of course, the issue of the Self. The second alternative is, I believe, correct. All I can say here is that while when I direct my gaze
55
inwards I do find knowings. I do, like Hume, find no such thing as a knower. This sort of Self is an illusion spawned by ordinary grammar. The empirical Self or, with the linguistic turn, the common-sense usage of the personal pronouns must and in principle can be reconstructed in a pattern that was invented by Locke. For the rest, I believe that the inability to cope with the act which most positivists have inherited from the classical phenomenalists is a very grave shortcoming. To have called attention to it is the other decisive contribution of Moore who, in this respect, owes a great debt to Brentano and, through him, to Locke. The issue lies indeed deeper than the realism-phenomenalism controversy; it appears already within the "phenomenally given" in the difference between "This is a tree percept' and 'I am having a tree percept'. Sometimes one should speak metaphorically or metaphysically, lest one's soul shrivel; sometimes one must, if one wishes to be understood. Let me say, then, that without this correction the positivist's universe remains flat, with subject and object squeezed into one plane. This is a taint of either phenomenalism or materialism which the linguistic turn as such does not remove. But if the correspondence is made the way I propose then one does not need to abandon either elementarism or the syntactical schema of Principia Mathematica, for syntactically 'being aware' is just a predicate among predicates and the name of an awareness just a particular among particulars.
With respect to the act verbs this is as it ought to be, reflectinng as it does the place which minds do have in the world. About the names of contents or, more precisely, kinds of contents, something more can be done provided one is not afraid, as we have seen one need not be, of languages which do in a sense speak about themselves. Let me explain. To say
56
that the sentences of the ideal language correspond to actual or potential contents of awareness is but a reaffirmation of the principle of acquaintance which I express by calling these sentences the texts of mental states. To be a mental state of a certain kind is indeed the same as, or "means," to have a certain text. With this in mind, let us take our next cue from a device used in ordinary language whenever one wants to distinguish an expression from its name and ask whether we could not use ''This tree is green''2 as a predicate that names in the ideal language the kind any particular awareness with this text exemplifies. Two things should be noticed. First, the device must not be introduced uncritically, since in this case the sign is not quite innocent of what it names. Second, if we introduce it at all, the ideal language should also contain the truism ''This is a tree' means this is a tree' and, at the same time, show that it is a truism. For, I repeat, to be an awareness of a certain kind and to have a certain text are not two things, but one. I believe that one can, not uncritically, add the quotes and a sign to be interpreted as 'means' to our schema; not uncritically, because a satisfactory argument can be made on purely syntactical grounds that the two new signs are logical and that the truism is analytic. This, by the way, is the addition to the syntactical schema of Principia which I mentioned much earlier. But let it be noticed that the inclusion of at least one undefined act predicate among its descriptive vocabulary, or the correction as I just call it, it quite independent of this addition to its logical apparatus.
Up to this point I have not once used either 'means' or 'meaning'. The cause of this is, of course, the ambiguity
***** 2 This is not a double quote, which is merely a stylistic device, but one "semantical" or single quote within another.
57
of the term and the conviction that it has been a rich source of confusion among positivists as well as among others. One of the connotations, the one with which I just used the term and the only one with which I shall use it, is purely linguistic; it does not strictly speaking refer to anything any more than does either 'or' or 'all'. Loosely speaking, it marks the "relation" between an expression and its name. This loose usage, we shall see, was very misleading. In another connotation or, if you please, in two others, I avoid the term, speaking instead of naming and referring. These I shall discuss presently. As for our addition to the syntactical schema, it also yields an explication of the philosophical notion of truth. By means of 'means' one can within the ideal schema define a predicate, to be interpreted as 'true', such that to state that a sentence is true is analytically equivalent to stating first what it means and then this meaning. This is the small nourishing kernel of the correspondence theory of truth, one of the strongholds of philosophical realism. I call it a small because this truism, too, is a verbal affair. Aside from the help these explications afford in avoiding some errors, meaning and truth are indeed philosophically arid notions. Nor is this surprising. What could one possibly learn about the world from those sentences of the ideal language which since they contain one or both of these terms, are essentially about language? Concerning further connotations of truth two comments will suffice. In ordinary discourse the term and its derivatives serve no purpose besides style and emphasis. Philosophically, there is also the unclear idea of verification. Some even consider a so-called verification principle a cornerstone of positivism. They argue that we know only what we have verified and, in a blend with the equally hazy notions of a meaning criterion, that we must not say
58
what cannot at least in principle be verified. The trouble is, as with the meaning criterion, that this is not a new or separate problem. Having verified something is the same thing as to know it, or something else from which one infers it, as one knows what he sees or hears or is otherwise directly acquainted with.
The pragmatists. The pragmatist writers study language as it names or refers. Theirs is a scientific investigation, philosophical only in the peripheral sense of being for the most part speculative and anticipatory. A really worth-while science of linguistic behavior does as yet not exist, could not possibly exist as long as scientific psychology and sociology are in their present rudimentary state. Why, then, mention these studies at all? Why tag their authors with the name of a philosophical movement? Science sometimes aids the philosopher indirectly. It may help us to see that a certain question is scientific, not philosophical. (To extend this claim dogmatically to all philosophical questions is scientism.) Or it may help us to recognize that what we thought was one question of a philosophical nature is a tangle of several, one of them scientific. (The others may be philosophical or moral.) The tangle we must eventually unravel is meaning, its scientific component what I call naming or referring. This explains our interest. The label I use can also be justified. The authors who write on this subject and are also philosophers are either pragmatists or were at some time under the influence of pragmatism, particularly in its Deweyan or instrumentalist variety. Another influence upon them is formalism. The link between instrumentalism and formalism is scientism, that is, as I just explained, the belief that the sound core or any philosophical problem is always a scientific one and nothing else. This version of the rejection of all
59
philosophy is the very essence of instrumentalism, with psychology and sociology as the sciences destined to supplant metaphysics. Dewey is a philosopher only in that he supports this doctrine with an elaborate rationale which is in part genuinely philosophical and, curiously enough, in the Hegelian pattern.3 This is why I think that the tag is not only justifiable but enlightening.
Imagine an anthropologist who just arrived on an inhabited island whose names begins with 'O'. He will soon discover that the sounds the natives make are of two kinds, one being distinguished by its causes as well as its effects. This interesting kind he calls o-words. Or, if you please, this is his definition of 'o-word'. But one must not forget that he does not proceed as arbitrarily as the syntacticist who makes "by definition" certain shapes the elements of his game. Only the name, if anything, of this class of sounds is arbitrary; membership in it depends on the causes and effects a sound has. Among these our anthropologist will, if he is wise, include mental states. (I mention this because most pragmatists make so much of their behaviorism. But the sound core of this doctrine is merely a way of speaking about minds. For scientific purposes though, I would be the first to insist, this is the only proper way, at least in principle.) The pattern of causal relations among o-words and other things is intricate and subtle; yet its essence can be stated very simply. The occurrence of a particular o-word, for instance, is in a characteristic fashion causally connected with thought and over behaviour concerning trees. To this connection our anthropologist refers when he says that the o-word o-means tree, or,
***** 2 For a brilliant analysis of the structure of pragmatism as well as of its influence and affiliations see the long essay by May Brodbeck in M. Brodbeck, J. Gray, W. Metzger, American Non-Fiction 1900-1950 (Chicago, 1952).
60
synonymously, that it o-names or o-refers to trees. I, for one, shall again shun 'meaning', even with a prefix. O-reference, then, is nothing "logical" or "definitional"; it is a natural or empirical relation among o-words on the one hand and events, states of affairs, and all sorts of things on the other, in exactly the same sense in which being-to-the-left-of and being-louder-than are natural relations, only much more complex. In the ideal language 'o-refers' is a descriptive expressions; this is why I call the relation natural or empirical. The expansion of this expression which contains only undefined descriptive signs is very long, this is why I call the relation complex. There are several philosophical lessons in this scientific fairy tale. They are all negative, as one would expect; yet they are important.
Much of our talk about language is not one bit different from the anthropologist's, though we do not, of course, use prefixes. Such discourse is, therefore, of the same kind as scientific discourse, which is not philosophical but merely an extension of common sense. It moves, to speak metaphorically, in a constituted universe in which no philosophical question ever needs to be asked and in which none can be answered. Let me explain why I again insist on this point. I have said before that I always use 'naming' and 'referring' commonsensically. If I wanted to express this in a manner which is now, after my story has served its purpose, merely ridiculous, I could say that proper explication of the two words is always that of 'o-naming' and 'o-referring'. But it is far from ridiculous to point out that if one says, as I did, that language refers to events, states of affairs, and all sorts of things, one does not thereby embrace an ontology of events states of affairs, and all sorts of things. For obviously, this is a commonsensical or if you insist, scientific use of the
61
crucial phrases. One more comment should drive home the point that such discourse about language, useful though it may be heuristically, is yet philosophically irrevelant. Imagine that our anthropologist, being of a philosophical bent, attempts a classification of o-words that might not occur to everyone. He proposes to divide them into o-descriptive and o-logical ones. 'Descriptive' and 'logical' (not 'o-descriptive' and 'o-logical'!) are philosophical words that need explication. To learn or to discover for himself this explication our scientific friend must study philosophy. Having done this he may, if he wishes, by a simple process of translation, transfer the distinction, with all the roughness it has in natural languages, from his own to the native idiom. Frankly, I do not see much point in the exercise. What I do see is that if he does no know the explication, anthropology will not help him to find it.
It is time that explain my preference for the letter 'O'. It stands for the first letter in the 'object-language'; for this is what the islanders' idiom is often called while the anthropologist is said to speak a metalanguage. Both terms are clumsy and unnecessary; but they are widely used; so I shall say a few words because of the bewilderment and the confusion they have caused. The plurality of natural languages is not relevant for our purposes. If one disregards it, then there is one and only one language, the natural language we all speak and understand. Using it, one may either speak "scientifically" about language or one may constructand interpret artificial sign systems. Cryptographers design such codes, as Ryle contemptuously calls them, and we have, for our very special purpose, designed the schema of one. Thus, if those two clumsy terms are to be used at all, then our natural language is the only metalanguage; the so-called
62
ideal language is the object language of philosophical discourse. Also, language is, metaphorically speaking, a closed universe; it cannot literally contain another language. If for instance, a German teacher of French says ''chien' bedeutet Hund', then ''chien'' is literally the German name of the French word 'chien'. That is, of course, the origin of the quoting device. But to say all this is not to say that syntacticists cannot construct several sign systems and, still syntactically, relate them to each other in all sorts of manners. One must not be confused by the ways they have of speaking about their mathematical games.
Now, finally, the meanings of meaning, of which there are at least four. There is, first, the one I call meaning, a logical or linguistic relation between words and words. There is, second, the one I call naming or referring, a natural relation between words and things. Some of the classical philosophers use the term in a third sense for the contents of acts which are said to inexist intentionally in them. This is, so to speak, a metaphysical kind of referring, the appropriate metaphysics being that of direct realism. Explicating as we did the act and referring, we have, I submit, reconstructed the sound core of this classical position. Again, psychologists who inquire how the meaning is carried, either associatively or by some conscious representative such as an image when, say, a man hears a word and knows what it means, ask a purely scientific question. This is the fourth theoretically important meaning of meaning. The possibility of confusions among the four is only too obvious. Formalism, we shall see, fails to distinguish between the first two.
Formalism Formalism is the end product of a combination of three attitudes. One is rejection of classical philosophy common to virtually all early positivists, the other de-
63
votion to science, the third a great love for artificial sign systems. The three go well together. The argument that philosophical controversies cannot be decided experimentally, for instance, serves as a bridge between the first two; again, admiration of the physical sciences may tempt one to put too high a value on the new symbolic technics and thus in the end to pay for their coveted neatness and precision with vacuity. Carnap, the outstanding formalist, has always lived in this intellectual climate. Less than twenty-five years ago he was, nevertheless, as close to reconstructionism as anybody had ever been before except, perhaps, Wittgenstein. He was also the first to recognize the significance of that bit of arithmetic that is so important in the explication of analyiticity. Later on, soon after he had moved to America, his formalism became more pronounced. Fully developed, formalism is an uneasy marriage between two extremes. The one, from which I take the name is purely "formalistic." The other is a crude metaphysics implicitely held; this, I believe, is the price every philosophy that explicitly rejects metaphysics must pay.
To understand the formalistic side of formalism, consider the old question whether numbers exist. Russell both explicated and answered it admirably by intepreting number signs as defined logical expression of the ideal language. This corresponds to the classical answer that numbers subsist. But the formalist now tells us again that the question itself is not a good question, that the only one we can and need ask and answer is whether a certain symboism contains number signs; if it does, then numbers exist in this symbolism; if not, not. If this were all there is to philosophy, it would be about words and nothing but words. This is indeed the epiphany of the word.
64
The implicit metaphysics is an uncritical realism which manifested itself early in the formalists' concentration on artificial languages whose undefined descriptive terms are, broadly speaking, the names of physical objects and of some of characters they exemplify.3a Except for mental terms and the more abstract ones of science and common sense, this is virtually the language of what I called a constituted universe. With respect to the concepts of science, the formalists' analytical achievement is brilliant and by now virtually noncontroversial. It is, of course, also quite independent of their specifically philosophical views. The mental terms they are content to define as the behavioristic psychologists do, within the sort of artificial language they prefer. This is their physicalism and their behaviorism. Thus they prevent themselves, by a typically scientistic foreshortening, from coming to grips with at least two large groups of philosophical issues, those raised by mind and those centering around the realism-phenomenalism controversy. Obviously, their artificial language is therefore not the ideal language. These, I hinted, are old weaknesses. In its most recent state formalism expands its artificial language into a "semantical system." All it gains is the opportunity for new errors and confusion.
Generously interpreted, a semantical system is the syntactical construction, alswys within the limits of physicalism and behaviorism, of what our anthropologist had to tell us. So let us consider one of his characteristic sentences, say, ''tree' o-names tree'. The formalist transcribes it by ''tree' designates tree'. 'Designates', it will be noticed, is another word I have so far avoided. The formalist collapses into it the two radically different ideas I distinguish as mean-
***** 3a For a detailed demonstration by one of my students of the realism implicit in Carnap's as well as in Schlick's views see L. Pinsky, "Positivism and Realism," to appear in Mind.
65
ing and referring. He opts for the linguistic connotation, meaning, when he says, as he does, what I would express by calling his artificial language syntactically constructed and 'designates' one of its logical signs. This latter illusion is probably enhanced by the "definitional" freedom we have in devising codes. He opts for the empirical connotation, referring, when he insists, as he also does, that in the paradigm ''tree'' and 'tree' refer to the word 'tree' and to trees respectively. So interpreted, a semantical system is merely an "axiomatization" of behavior science and therefore, as I explained, without philosophical significance. As far as this future science is concerned, is it really likely to benefit from an axiomatization which axiomatizes nothing but, and that barely, the crude common sense any science must try to go beyond? One does not, after all, cut butter with a razor. There are besides, the very real possibilities of present philosophical confusion. I select two, one a formalistic excess, the other a curiously naive excursion into ontology.
One of the formalists' illusions seems to be that "semantics" avoids the "formalism" of "syntax." Of course, this is but the "formalistic" version of the old error that a sign system as such says anything. Analyticity is a syntactical notion. So it is in the following manner replaced by one supposedly superior. In terms of 'means', we remember, one can define the equally linguistic 'true'. The formalist, by means of the ambiguous 'designates', defines in his semantical system L an equally ambiguous 'true'. This gives him in L a sentence which I transcribe, a little cavalierly, as follows: 'This tree is green' is true=Def. ('this tree is green' designates this tree is green) and (this tree is green). The expression in the first parenthesis on the right side he calls a "semantical rule" which he knows to be true provided only he under-
66
stands L. The expression in the second parenthesis being a synthetic sentence, he does not, merely by understanding L, know whether it is true or false. Yet, he discovers, there are some cases where the second parenthesis does not bother. In this happy event he calls the quoted sentence on the left side L-true, this being the notion which is to supplant analyticity. Of course, these happy cases are exactly those in which the expression in the second parenthesis is analytic in L. It follows that to have a significant notion of L-truth one must first have a significant notion of analyticity in L. Far from having been disposed of, the problem has merely been pushed back one trivial step. Of course, the thing is not done quite that transparently, otherwise it would not have been done at all. But this is what it amounts to.4
Now for the excursion into ontology. Meaning, naming or referring, and analyticity are connected in the following three ways. First, construing 'meaning' as strictly as I think one must one cannot in the ideal language truly say ''Peter loves Mary' means Mary is loved by Peter' is analytic. Second, one will wish to say, as one can without difficulty, that ''Morning star' means Evening star' is false but that 'Morning star' and 'Evening star' refer to the same thing. Third, one will wish to say, as one again can without difficulty, that ''Man' means featherless biped' is false because 'A thing is a man if and only if it is a featherless biped' is not analytic. If, however, one wants to express all these ideas by axiomatizing an ambiguous 'designates' and without even
***** 4 I argued this point quite technically by analyzing an error in Carnap's Meaning and Necessity. Unimportant in itself, this slip illustrates perfectly the basic misconception and could hardly have occurred without it. See Mind 57 (1948), 494-95)
67
the notion of analyticity to fall back upon, then one has quite a mathematical problem in hand. So one may overestimate its solution and, guided by habits not wholly extinguished, interpret this solution philosophically. This Carnap has lately done by suggesting that what "exists" is what is designated. We remember the pattern, with a different word, from Wittgenstein. But we also remember that 'designates' stands ambiguously for 'o-refers' and that while our anthropologist did not make any ontological commitments by making words refer to all sorts of things Carnap now does. This curious result is that propositions in the classical sense, the golden mountains once thought about, and all sorts of things are being returned to ontological status through this semantical back door. It was, I believe, Lord Acton who once made a penetrating remark to the effect that those who ignore the past may, at its own risk, have to relive it.
10. Causism Philosophy is the disease of which it is the cure.5 This is the essence of casuism distilled into one sentence. The disease is metaphysics. The cure is ordinary discourse about ordinary language, the casuist's version of the linguistic turn. But this discourse is critical, not merely descriptive as the anthropologist's. One could even call a therapeutic, comparing one whom the casuist involves in philosophical conversation about the classical problem with a patient who has entered upon a cure. For if the treatment is successful then the patient will cease to worry about the problem because he has recognized that it is merely a puzzle, a pseudoproblem created by bad grammar. 'Worry' and 'puzzle' are indeed key words of this linguistic existentialism
***** 5 This phrase is Feigl's who is not a casuist but a disciple of Carnap and who used it in another context.
68
with happy ending. This is where the other tag, therapeutic positivism, comes from.
The casuist, we see, shares the belief common to all positivists that philosophical difficulties stem from bad grammar. Only, he embraces the extreme view that good grammar dissolves, rather than solves all classical problems and thus completely disposes of philosophy. In this respect casuist and formalist are alike. They part ways when the former rejects the artificial languages which the latter uses to spot bad grammar. How, then, does the casuist know good grammar from bad? This is where his casuism comes in. I shall state this part of the doctrine in four steps, more or less as if it were an argument. First, the casuist is greatly impressed with a fact I mentioned when I sketched the background of the linguistic turn. The man on the street who uses ordinary language to talk about what he ordinarily talks about, though he may of course get into trouble, never gets into the sort of trouble we know so well as philosophical perplexity. Second, the meanings of words (I think I can now safely speak of meaning) vary with the situations in which we apply them and, which is perhaps the same thing, with the verbal contexts into which we put them. The good grammar or correct usage of a word, or of a connotation of a word, consists of all the ways in which it can be combined with others into sentences that do not yield philosophical perplexity. This is the casuist's implicit conception of good grammar. Usually, though, a few strategic sentences will suffice to give us the right idea. Third, if a word has two more or less distinct contexts, one may inadvertently transfer the grammar of the one to the other. This is bad grammar of the sort that can lead to philosophical perplexity. In the case of a gradual transition between contexts trouble may
69
arise when one forgets that while they are similar in some respects they are also in varying degrees unlike each other. Fourth, the cure is performed by exhibiting all these differences and transitions, gradual or abrupt as the case may be, and by pointing at the exact spot where the fateful slip occurred. We understand now why artificial languages don't appeal to the casuist. In constructing them, or at least in constructing them well, one does what he wants to do in his own way. Nor does he believe that there are any traffic rules or general philosophical propositions to keep us on the straight and narrow path of correct usage. All one can do is to do the job in each case exactly as the circumstances of the case require. This is where the tag, casuism, comes from. This stress on philosophy as an activity, this horror of philosophical propositions, is probably, as Ryle suggested, an attenuation of Wittgenstein's ineffability thesis. This, then, is the casuist doctrine. What are its weaknesses and its strengths?
There are some strengths. A casuist's distinctions may be part of the required explication. Induction is, I think, a case in point. From one who worries about the grounds of induction the casuist inquires whether he does not perhaps wonder how any number of instances can ever furnish the adequate reasons for a generalization. If the worrier agrees that this is what he is really worrying about, then he is told that one also speaks of adequate reasons where a deductive proof is in order but that this is surely not the sort of thing he is looking for. As it happens and as we have all along known this is indeed a part of the required explication. Or take the metaphor that calls percepts images. The "grammar" of "image" helps one to understand a good deal of what has been said for or against representative realism. In this case,
70
though, the service is very modest, really not more than an old and familiar way of focusing the problem. Generally, the casuist's contribution may be, and often is, of some propaedeutic value. His basic weakness is that he stops too soon. Comparing the philosopher to a surgeon, one might say that the casuist sometimes exposes the delicate inner organ but that he never operates on it. I shall give two schematic illustrations, calling them schematic because I do not care to attribute them to anybody in particular; one to demonstrate how the casuist convinces himself that no operation is necessary; one to show how he, too, may become the victim of an implicit metaphysics. Some doubt that we can ever know what is in another's mind; some believe that we can and sometimes do know such things. My schematic casuist first tells the believer that he believes as he does because he has all the evidence he can possibly have, namely, the other's behavior. Next he tells the doubter that he doubts because, since he does not share the other's awareness, he does not have all the evidence anyone (which includes the other) could possibly have. Then he disposes of the problem of other minds by pointing out that the two argue at cross purposes since they do not use 'knowing' with the same grammar. I notice, first, the guileless use of 'possible' where it carries an eminently philosophical meaning. I notice, second, that the little play on 'he' and 'anyone' presupposes other minds, which in this case entails circularity. And I notice, third, that while we all without circularity agree on what we are directly acquainted with and that we are not so acquainted with other minds, it takes some doing to explicate on the basis what we mean when we say truly that we do sometimes know not only how the other behaves but, in a sense, also what is in his mind. This is of course
71
the problem or, if you please, the operation the surgeon must perform. Now for the illustration with the implicit metaphysics. Some casuists think they know from correct usage that we know something if we have a perfect reason to believe it and, also, that we know whatever we do know indubitably or for certain. My schematic casuist uses this piece of "grammar" to attack those who insist that they are never completely certain whether such statements as 'this is a book' are true. If I see a book on my shelf, so the argument goes, and if I make a few tests to convince myself that I am not the victim of a perceptual error or illusion, then I have a perfect reason for believing the proposition. Thus I do not merely believe it; I know it. And since I know it I know it for certain. Clearly, the argument is a barely veiled petitio principii. Clearly, the very difficult philosophical notion of certainty has here under the cover of grammar entered into supposedly ordinary discourse. Clearly, the implicit metaphysics is some sort of realism.
The shortcomings of all this are so grave and so obvious that one must ask why so many clever people have not seen them. Is there, perhaps, a core of central belief which makes all these strange things plausible and is yet at first sight not too implausible in itself? The core of casuism is confused notion of grammar, derived I think from Wittgenstein's unduly absolute and unduly weakened ideas of form and analyticity. The typical casuist accepts these two notions, form and analyticity, though he doesn't worry much about them, probably because he thinks that Wittgenstein has done that once and for all. An analytic truth, then, is for him a proposition that is true because of the form it has or, as he also says, by virtue of its meaning. Being linguistic and not factual, it could not possibly be false and is, therefore, be
72
yond controversy. Some even call it truth by definition, a bad slip due to the circumstance that within a language definitions are indeed analytic and, in an obvious sense, beyond controversy. (This is also the source of a line of thought according to which all philosophical propositions are definitions. Philosophical explications are, in a reasonable sense of the term, real definitions; and there is nothing arbitrary about these.) This part of the core stems from Wittgenstein's absolutism. To grasp the other part one merely needs to think of "form" as grammatical form, in some vague sense of grammar or usage which remains completely obscure except that it is always invoked to prevent one from asking philosophical questions. In this way one may come to believe, as Wittgenstein did and the casuists do, that 'everything that is colored is is extended' is analytic or what amounts to the same thing, that to be extended is part of what it "means" to be colored. I do not at all understand this meaning of 'means'.6 But I understand very well that it can be used to make everything a question of grammar or usage. This is patently absurd.
11. Conventionalism. Ryle shares the casuists' views on language. Like them, he tenderly probes the idiom; like them, he rejects artificial sign systems. Unlike most of them, he is by now virtually a philosophical behaviorist. But this is not the most important difference. What sets him apart is that he does not shun philosophical propositions. Those he actually propounds may be, and in my opinion are, utterly futile.
***** 6 Historically speaking. I have a hunch. If one accepts this meaning of 'means' then one can convince oneself that all synthetic propositions are really analytic because they are either "implicit definitions" or the deductive consequences of such. This is rationalism of the Hegelian or Deweyan variety. We stand here at the source of some pragmatist undercurrents in recent positivism which we are not of the rather trivial scientific variety.
73
Yet, the intellectual motives behind them are interesting and genuinely philosophical. Fortunately for all of us, a philosopher's stature is not measured by his beliefs alone. The motives appear most clearly in the papers Ryle wrote before he attained the influence he now has in England. Two stand out.
Ryle is committed to an elementarism even more radical than Wittgenstein's and, equally important, to direct realism. (The Wittgenstein of the Tractatus was not very interested in the realism-phenomenalism issue.) The second commitment is admirable in one who is wedded to some sort of philosophical realism, as Ryle always was, probably on common-sense grounds as, say, Reid understood common sense. Such a one sees clearly that the indirect or representative realist of the Lockean or any later variety achieves nothing, merely pushes the difficulty one step back, as it were, since he cannot bridge the gap between the mental object he knows and the physical object that causes it but which he does not know. This, I suggest, is how Ryle came to dislike all mental objects. This, I suggest, is how Ryle came to dislike all mental objects. The dislike is not only the driving motive behind his behaviorism; it also provides the bridge to his elementarism. For the overt ground on which he attacks mental objects, whether they be humble sense data, feelings, percepts, or acts, is that they are all abstract, nonempirical or Platonic entities (not, we remember, in space in time). They are thus in the same company with simple higher order characters, with propositions in the classical sense, with golden mountains supposed to exist or subsist merely because they were once thought of, and with any fanciful entity that ever inhabited a philosopher's ontology. Some may think the company mixed and, therefore, the judgment that dooms all its members harsh, not to say, rash. I cer
74
tainly think so. Yet this is, for better or worse, Ryle's brand of elementarism. With it goes, as in Wittgenstein's case, an implicit reliance on the picture metaphor and that strict use of 'naming' which we recognized as its symptom. Probably because he was so sensitive to this cue, Ryle spotted immediately the overcrowded ontology which Carnap now propounds in that curious roundabout way I discussed. At this point Ryle is very consistent. He takes his own medicine, worrying about what any but an atomic sentence could possibly name or, if you please, refer to. The range of his worry is wide indeed. It comprehends not only tautologies but also generalities such as 'everything that is green is extended', moral judgments such as 'killing is evil', and still other kinds of sentences.
The solution we are offered is surprising. I shall take my cue from the rejection of Wittgenstein's doctrine that tautologies are ordinary sentences, peculiar only in that they say nothing. Ryle teaches, more radically, that they are not sentences. This, I submit, is his particular version of the linguistic turn. The implicit notion of a sentence is, clearly, the sort of thing that refers to what "exists" or could exist. Of course, Ryle is not so foolish as to reject outright all those suspect kinds of sentences as nonsense. Yet he insists, what amounts practically to the same thing, that they differ from real sentences in two respects. First, they are about language or, rather about linguistic behavior. Second, they are rules about the use of language and, therefore, like all rules neither true nor false. Furthermore, they are ex post facto rules, abstracted from linguistic behavior by one who watches people speak as a coach watches his players, not rules which either explicitly guide the players or justify what they do while on the playing field. This latter emphasis
75
has two reasons. First, Ryle wishes wishes to reject the naive "rationalism" which fancies that, say, deductive inference consists psychologically in the conscious application of a syllogistic rule. Second, he wishes to point out that if such a rule were without further comment offered as a justification of inference one could immediately ask what justifies the rule. As far as they go, the two points are sound. Only, they do not go very far, not beyond exposing once more certain old errors which hardly anybody makes nowadays. Nor does Ryle see that the second argument can be turned against himself. There is indeed nothing in his interpretation of these sentences as ex post facto rules of linguistic behavior that would help him to answer a similar question. His interpretation makes the question only more obvious, an answer more urgent. The question is, of course, why just these rules and not others; or would others do as well, say, for instance, those of a three-valued logic.7 Ryle has no answer; all he tells us is, in effect, "what is being done." This makes his philosophy a form of conventionalism, with the peculiar twist, the standard casuist unmasking of all classical problems as artifacts of bad grammar.
12. Conclusion Incidentally this essay presents a structural analysis, and in a sense even a historical one, of the several branches of the positivistic movement. Primarily it is an exposition of my own reconstructionism. But I have refrained from either advocating or appraising this position. At least I have tried. In conclusion I shall permit myself a few such remarks, theoretical as well as cultural. Theoretically, I
***** 7 For a detailed analysis of Ryle's logic by one of my students, see N. H. Colburn "Logic and Professor Ryle," Philosophy of Science, 21 (1954)
76
should like to say that in the principle of acquaintance this philosophy preserves the core of classical empiricism as in the struture of its ideal language it contains the doctrine Russell called logical atomism. Yet it is neither phenomenalism nor realism. It takes awareness seriously and it does not imply ethical naturalism. I hurry to add that I have found all its major pieces ready made. I merely polished their rough edges and fitted them together. If I have added anything at all, it was the effort to bend the philosophies of the act back into the main stream of empiricist thought. This is a contribution that was long overdue, building as it does on the work of Moore and Brentano. Culturally, there is nothing left in this philosophy of science worship, the implicit materialism, the shallow hedonism, and the frivolous social optimism that have often marred the Continental branch of the movement, though surely not the spirit of Wittgenstein. Nor does it, I think, share the futiliatarianism into which the British branch has so unhappily degenerated. Finally, unlike either branch, it is firmly committed to the classical tradition. To use a bad metaphysical word in a good human sense, this is indeed its one commitment that is absolute.
77
Home